Fred Wilson

1. Snapshot and testable edge thesis

Predictions — [researcher inference | moderate]. If the edge is real, unseen decision files should describe the participant behavior, specific network mechanism, formation remaining, capital efficiency, and next falsifiable milestone before commitment. Relationship-sourced cases should state what changed rather than treating familiarity as a quality score. Follow-ons should follow improved evidence, and Wilson-attributed network cases should outperform vintage-, stage-, price-, vehicle-, access-, and partner-matched cases on net cash outcomes profile predictions.

Falsifiers — [researcher inference | strong]. The thesis weakens if investment memos omit those mechanisms, narratives were added after success, retention did not predict follow-ons, reserves mainly protected sunk cost, or matched non-thesis cases perform similarly. It fails as a claim of personal economic alpha if complete net cash flows show no Wilson-attributed advantage after controlling for team, vehicle, vintage, access, fees, and a dominant outlier skill/team/vehicle tests.

2. Core philosophy

Wilson's own words first

• Wilson: “The entrepreneur is the customer and the LP is the shareholder.” (Our Customer Is the Entrepreneur, ¶3, 2005; 11 words).
• Wilson: “We plan to stay small, do our own work, help entrepreneurs build great companies” (Union Square Ventures, ¶13, 2005; 14 words).
• Wilson: “raise only enough money to achieve a set of near term milestones” (Deal Size, ¶13, 2006; 12 words).
• Wilson: “Because you never know how much money you’ll need for some of your best investments.” (Reserves, ¶15, 2007; 15 words).
• Wilson: “To be completely and totally honest, we don’t yet know.” (Twitter, ¶16, 2007; 10 words).

[researcher inference | strong] Read together, the direct voice links stakeholder orientation, small-team execution, milestone-matched capital, reserve optionality, and willingness to invest before monetization is known. The synthesis is ours: the sentences were written in different contexts and are not a recovered checklist.

Wilson's system is best read as linked decisions, not a collection of slogans.

1. Invest while consequential formation remains. [investor-stated | strong] Stage is defined by how much of the product, team, market, and organization remains to be built and whether the investor can materially help—not by a financing label alone Early Stage Investing, stage/ownership analysis. Decision use: write the remaining-formation ledger and the investor's concrete contribution before labeling a round. Boundary: “early” does not make an immature market large, a team capable, or a price attractive.
2. Finance the next learning milestone. [researcher inference | strong] Wilson's early-stage, check-size, reserve, and failure writings together imply that the first check should buy enough time to answer the next important uncertainty without financing speculative scale written synthesis. Decision use: name the milestone, amount, runway, and stop/change rule. Boundary: excessive frugality can prevent a company from reaching a real threshold; the rule is uncertainty-matched capital, not the smallest possible check.
3. Reserve for changed evidence, not familiarity. [investor-stated | strong] Reserves preserve ownership and let a fund invest more after learning Reserves note. Decision use: require a fresh milestone and update the full expected-value case at every follow-on. Boundary: ownership targets and prior work create sunk-cost bias; pro rata is not evidence.
4. Underwrite the behavior before the business model settles. [investor-stated | strong] Twitter's public/asynchronous graph, Zynga's use of social distribution, and Coinbase's protocol infrastructure show the method at decision time Twitter · Zynga · Coinbase. Decision use: map participant, repeated action, contribution, distribution, retention, and value capture. Boundary: usage without durable retention, governance, or monetizable control can be socially important and still be a poor venture investment.
5. Use relationships as a longitudinal sensor. [investor-stated | strong] Wilson treats connectors, public writing, founder contact, and direct product use as ways to see change over time Connectors note · Boxee lab. Decision use: record the prior thesis, new observation, and independent check. Boundary: warmth, fame, and repeated contact are access advantages and confirmation-bias risks, not quality scores.
6. Build a fund that can execute the thesis. [investor-stated | strong] USV deliberately linked small team, partner labor, fund size, ownership, reserves, and sector focus formation note. Decision use: reject attractive companies when stage, check, ownership, time, or vehicle cannot fit. Boundary: a small focused vehicle can miss capital-intensive breakthroughs, and opportunity funds can change incentives.
7. Version the thesis and expose corrections. [documented behavior | strong] Wilson revised venture math after criticism, changed his view on social-network applications, published the Airbnb miss, corrected GeoCities credit, and separated marks from liquidity intellectual chronology. Decision use: preserve dated belief, disconfirming observation, and resulting action. Boundary: public correction proves transparency, not that the correction was timely or improved returns.
8. Serve founders, owe duty to the company. [investor-stated | strong] The entrepreneur-as-customer metaphor emphasizes responsiveness; StarMedia and later board doctrine impose a harder company-level duty customer note · Reboot note. Decision use: clarify whether the investor is advising, approving, monitoring, or acting as fiduciary. Boundary: loyalty to a founder can conflict with leadership change, employees, investors, or legal obligations.
9. Keep narrative, labor, ownership, and return separate. [researcher inference | strong] Wilson authored Twitter, Zynga, and Coinbase posts, but filings show vehicle ownership and shared power; Colonna led GeoCities, Wenger led several signature deals, and Burnham narrated Tumblr role matrix. Decision use: build a role ledger before assigning credit. Boundary: a filing stake is not sourcing, a byline is not a vote, and an acquisition price is not fund proceeds.

The system's coherence is valuable; it is also a source of overconfidence. [researcher inference | strong] A story that joins behavior, small checks, ownership, reserves, and trust can explain many choices after the fact. Only dated decisions, rejects, ordinary cases, and capital-weighted cash flows can distinguish an operating system from an elegant retrospective narrative skeptical tests.

3. Sourcing

Channels and evidence

Reusable sourcing checklist

• [researcher inference | strong] Record first contact, date, source, any economic relationship, named internal owner, and whether the relevant thesis existed beforehand.
• [researcher inference | strong] Write what has changed since the prior contact; do not let “known for years” substitute for product, team, or market evidence.
• [researcher inference | strong] Obtain one neutral user, one churned or failed user, and one informed skeptic when portfolio or connector references dominate.
• [researcher inference | strong] Log passes with the same taxonomy as investments: market frame, behavior, retention, value capture, team, price, vehicle fit, and kill criteria.
• [researcher inference | moderate] Test whether the channel yields better capital-weighted net outcomes than cold or non-thesis sourcing after controlling for access, vintage, and stage. The current public record cannot perform that test.

4. Picking

Selection hierarchy

1. Formation remaining and ability to matter. [investor-stated | strong] What remains in product, team, distribution, market, and organization; which parts can this partner and vehicle materially affect? Early Stage Investing
2. Participant behavior. [researcher inference | strong] Who participates, what repeated action creates value, why do they return, what contribution improves the network, and what cohort evidence distinguishes novelty from durable use? Twitter lab
3. Market expansion and frame. [researcher inference | strong] Is the team describing today's niche or a behavior that can expand it? Airbnb is the required false-negative control Airbnb lab.
4. Founder and team. [investor-stated | strong] Wilson emphasizes intelligence, energy, integrity, product instinct, recruiting, and learning; the corpus does not provide a validated scorecard or false-positive rate MBA Mondays note.
5. Value capture and law. [researcher inference | strong] Distinguish protocol adoption, application position, governance, issuer financing, token structure, security/compliance, and enforceable rights. Coinbase and Kik must be analyzed together as divergent boundary cases protocol cases.
6. Price, ownership, and vehicle fit. [investor-stated | strong] A company can be excellent and still fail the fund if the check, stake, dilution, reserve burden, or time demand cannot matter Deal Size · venture-math sequence.

Diligence record

No published weighting formula survives. [researcher inference | strong] The practical safeguard is to force each gate into a dated memo, let one independent reviewer argue the opposite case, and state which observation will change “yes,” price, check, reserves, or governance.

5. Deal and portfolio mechanics

Check, ownership, and reserves

Wilson's mid-2000s doctrine used roughly $250,000–$2.5 million initial checks and an ownership target large enough that partner work and a breakout could matter; those are historical examples, not current USV policy [investor-stated | strong] Deal Size · Early Stage Investing. The operating logic is:

1. [researcher inference | strong] Size the first check to reach a named learning or replication milestone with an error margin.
2. [researcher inference | strong] Model ownership after the next two rounds, option pool, insider participation, and likely vehicle allocation; do not underwrite only the entry stake.
3. [investor-stated | strong] Set reserves at portfolio construction, then re-underwrite every company from zero when new capital is requested Reserves.
4. [researcher inference | strong] Separate a core-fund follow-on from an opportunity-vehicle investment: mandate, risk, price, ownership, conflicts, and economics can differ vehicle chronology.
5. [researcher inference | strong] Test the fund with and without the largest outcome; the public record cannot show company contribution or Wilson-specific net alpha Pareto boundary.

Return and valuation discipline

[investor-stated | strong] Wilson distinguishes a financing mark from liquid value and suggests haircut thinking; the lesson is to expose, not canonize, the assumption Valuing a Venture Capital Portfolio. [researcher inference | strong] Every claim should identify unit, vehicle, gross/net, realized/unrealized, as-of date, source, and denominator. Acquisition consideration is not investor proceeds; filing ownership is not a return; affiliate sales are not Wilson income.

The only located LP-reported fund signal is one LACERA commitment to Union Square Ventures 2012 Fund, L.P.: reported: 23.75x TVPI (LACERA, PDF/printed p. 40, “Top Performing Investments (since inception),” USV 2012 row, 2026), reported: 23.15x DPI (same table and row, 2026), and reported: 53.4% net IRR (same table and row, 2026), all as of 2026-03-31 [contemporaneous record | strong]. TVPI includes residual value, DPI reflects distributions, and LP cash-flow timing/class terms can differ. This is not a firm composite, gross return, company attribution, or Fred-only result.

Portfolio review checklist

• [researcher inference | strong] Recalculate expected value and next milestone from current facts; hide original cost and partner name during the first pass.
• [researcher inference | strong] Show cohort retention, burn/runway, leadership constraint, regulatory/security status, and a no-new-capital plan.
• [researcher inference | strong] Compare follow-on with every new opportunity on absolute expected return, not only dilution protection.
• [researcher inference | strong] Report cash distributions, residual mark, write-offs, and exposure by vehicle and company; never merge TVPI, DPI, IRR, or paper valuation.
• [researcher inference | moderate] Record whether board work changed an observable decision. Meeting frequency or partner time alone is not value creation.

6. Vehicle and incentive fit

Vehicle-fit rule — [researcher inference | strong]. Before diligence, state the eligible entity, mandate, stage, target check, minimum useful ownership, reserve burden, time horizon, conflicts, decision authority, and economics. Decision use: a “no” caused by vehicle mismatch should be recorded separately from a company-quality pass. Boundary: vehicle discipline can look like selection wisdom after a missed winner; it can also protect a coherent portfolio. The record needs both false positives and false negatives.

7. Team attribution and work with founders

The investment is a team production function

Attribution protocol — [researcher inference | strong]. For each company, maintain separate fields for introduction, thesis author, diligence, deal lead, negotiation, approval, vehicle, beneficial ownership, board role, reserves, exit decision, public narration, and economics. Filing “deemed shared power” is not personal stock; a byline is not a sole decision; a director title is not value-add proof. Decision use: assign credit only to the evidenced function. Boundary: private collaborative processes will often leave “unknown,” and equal-split assumptions are no better than celebrity attribution.

Founder service and company duty

[investor-stated | strong] Wilson's “entrepreneur as customer” doctrine asks the VC to earn trust through useful work, responsiveness, introductions, financing, recruiting, and candor Our Customer. Reboot adds the relational mechanism: trust compounds through repeated acts rather than titles Reboot. But StarMedia supplies the boundary: Wilson says the board delayed a founder transition and failed the company [retrospective witness | strong] StarMedia postmortem.

Decision use — [researcher inference | strong]. At every material intervention, label the role: founder coach, board fiduciary, financing participant, conflict-holder, or public advocate. Put the company's constraint and decision on the agenda, assign an owner and date, and record dissent. Boundary: helpfulness can become dependency, loyalty can delay leadership change, a large reserve can bias governance, and public advocacy can create conflicts. Founder satisfaction, meeting frequency, and company outcome are different measures.

8. Views on the craft

What Wilson appears to believe

• Venture is a power-law, ownership-sensitive craft. [investor-stated | strong] A few companies must matter to a fund, so initial ownership, reserves, dilution, and fund size belong in the decision venture-math note. Decision use: calculate fund contribution under explicit exit and dilution cases before commitment. Boundary: stylized math is not an observed distribution; it can rationalize overpaying or concentration.
• Early-stage work is formation work. [investor-stated | strong] Investors earn relevance while product, market, team, and organization remain unsettled Early Stage Investing. Decision use: specify the formation task and who owns it. Boundary: board advice is not founder execution, and “ability to help” is not outcome evidence.
• Domain focus improves judgment and access. [investor-stated | strong] USV's original narrow focus and Wilson's long public study aim to compound pattern recognition formation note. Decision use: maintain a dated market map, rejected cases, and unknowns. Boundary: focus becomes style drift if its mechanism changes silently, and expertise can become frame lock.
• Great investors change their mind without changing style accidentally. [documented behavior | strong] Zynga and venture math show public revision; Flatiron shows what Wilson calls drift intellectual chronology. Decision use: version both thesis and fund mandate. Boundary: a revised story without a dated action is commentary, not learning.
• Boards should spend live time on decisions and constraints. [investor-stated | strong] Reporting belongs in prereads; unresolved issues need owners and follow-through Board Meeting Suggestions. Decision use: track decision latency and completion. Boundary: concise meetings can suppress weak signals; governance quality also depends on information rights, independence, and conflict management.
• Failure must remain in the teaching set. [documented behavior | strong] Wilson published Kozmo, Airbnb, Turntable, StarMedia, and other adverse accounts failure record. Decision use: review false positives, false negatives, and ordinary outcomes with the same template. Boundary: published examples are selected and retrospective; missing losses and passes may change the lesson.

Greatness standard

[researcher inference | strong] A great investor should demonstrate more than famous companies and coherent doctrine: a repeatable pre-decision mechanism, favorable capital-weighted outcomes across a defined denominator, accurate self-correction, appropriate team credit, aligned vehicle construction, durable founder/company contribution, and performance that survives removal of the dominant outlier. Wilson's corpus strongly supports doctrine and transparency; public evidence is insufficient for the personal net-return and counterfactual value-add tests Pareto assessment.

9. Signature bets and decision process

The eight labs are replayable because each separates contemporaneous evidence, decision, unknowns, and later outcome. They are not a ranking of personal returns.

Lab protocol — [researcher inference | strong]. Freeze the decision-date packet; write one-page answers to original thesis, strongest objection, unknowns, and terms/vehicle; decide without outcome knowledge; then reveal the actual record. Boundary: no lab should infer missing terms or pretend a later explanation was available at the time. Rank the quality of reasoning separately from company outcome and fund return.

10. Mental models and maxims

The companion mental-model field guide retains 17 model families in this exact order. The index below is deliberately literal so this philosophy and the field guide remain one auditable system.

1. Build domain expertise before committing capital. [researcher inference | strong] Use a dated market map and explicit unknowns; expertise is a prior, not a veto on disconfirming founder evidence.
2. Use relationships as longitudinal evidence, not as a quality score. [researcher inference | strong] Record what changed; warmth without changed evidence is not diligence.
3. Define stage by formation remaining and the investor’s ability to matter. [researcher inference | strong] Name the formation task and contribution; do not confuse an early label with attractive risk.
4. Finance the next falsifiable learning or replication milestone. [researcher inference | strong] Match capital to uncertainty; allow enough error margin to reach valid evidence.
5. Reserve for changed evidence, not familiarity or sunk cost. [researcher inference | strong] Re-underwrite from zero; pro rata and relationship are not milestones.
6. Match fund, vehicle, team, and succession to the opportunity set. [researcher inference | strong] Make mandate and authority explicit; reject vehicle-mismatched deals without calling them bad companies.
7. Underwrite networks at the participant-behavior level. [researcher inference | strong] Test repeated behavior, contribution, retention, and distribution before generic “network effects.”
8. Map enabling disruption and incumbent pressure before calling a market an “edge”. [researcher inference | strong] Identify what changed and who captures value; a large trend is not a company moat.
9. Separate protocol resilience from application capture, token structure, and law. [researcher inference | strong] Use separate adoption, economics, governance, and legal kill tests.
10. Version the thesis; distinguish disciplined adaptation from style drift. [researcher inference | strong] Preserve old belief, new evidence, action, and mandate impact.
11. Turn conviction into falsifiable patience. [researcher inference | strong] State time horizon, milestones, stop rules, and evidence that would change the view.
12. Serve founders while preserving duty to the company and stakeholders. [researcher inference | strong] Separate coaching from fiduciary approval and conflicts.
13. Design boards around decisions, CEO constraints, and traceable follow-through. [researcher inference | strong] Track decisions and owners; meeting cadence is not governance quality.
14. Separate narrator, partner labor, decision rights, vehicle ownership, and economics. [researcher inference | strong] Use a role ledger; never convert publicity or a filing stake into personal credit.
15. Externalize reasoning; preserve assumptions, corrections, and misses. [researcher inference | strong] Version the record; transparency still requires outcome testing.
16. Keep marks, mortality, multiples, and net cash returns separate. [researcher inference | strong] Label unit, vehicle, gross/net, realized state, date, and denominator.
17. Convert institutional commitments into comparable funnel and outcome data. [researcher inference | strong] Publish definitions and cohort outcomes; a promise or raw count is not progress.

These are analytical reconstructions, not a recovered USV manual. [researcher inference | strong] The field guide contains the derivation, misuse, diagnostics, failure modes, and transfer conditions; the list above is only the decision index.

11. Evolution and contradictions

The central contradiction is productive but unresolved. [researcher inference | strong] Wilson's edge story depends on stable focus and patient conviction; his learning story depends on changing frames. The solution is not “hold longer” or “change faster.” It is versioned conviction: specify the mechanism, milestone, horizon, and evidence that changes thesis, terms, reserve, governance, or exit.

12. Failures and limits

Case-based failure ledger

What cannot be concluded

• [researcher inference | strong] The public record cannot compute a personal Wilson hit rate, loss rate, gross multiple, net multiple, IRR, or alpha. The 32, roughly 55, 21, and 230+ counts are different, overlapping universes denominator proof.
• [researcher inference | strong] One LACERA commitment cannot attribute fund results to Wilson, a company, or the entire firm; current ADV entities cannot reconstruct historical funds or returns vehicle boundary.
• [researcher inference | strong] Filings can establish issuer, vehicle stake, shared voting/beneficial-ownership treatment, governance dates, and some affiliate transactions. They generally cannot establish sourcing, vote, carry, value-add, or personal proceeds filing boundaries.
• [researcher inference | strong] Public failure writing is positively and pedagogically selected. Missing quiet losses, passes, ordinary exits, and partner-led cases may change every apparent lesson.

13. Historical operating environment

[researcher inference | strong] Historical translation must control for vintage, capital supply, build cost, regulation, platform structure, and exit market. A rule is transferable only if its causal mechanism—not merely its language—survives those changes.

14. Modern VC translation (as of 2026-08-21)

Each row is an analyst-created transfer test. No row asserts that Wilson or USV currently uses the proposed implementation.

15. Comparison with completed peers

Only completed Canon profiles are used. Martin Casado and Trae Stephens are modern controls for public doctrine, team attribution, and vehicle opacity. They are not era, strategy, sector, access, or performance peers, and no ranking of returns is defensible.

[researcher inference | strong] The correct comparative question is not “who is best?” It is which pre-decision mechanism, in which domain and vehicle, survives adverse cases and a complete denominator. The three public records show sophisticated doctrine and severe person-level economic opacity. Using either modern investor as a performance benchmark would invent comparability.

16. The strongest case against greatness

The strongest skeptical case is cumulative:

1. The personal denominator is missing. [researcher inference | strong] Wilson's 2007 “32” snapshot contains anonymous positions; Flatiron's roughly 55, a first-USV-fund 21, and USV's 230+ are noncomparable. The 41 named-case ledger cannot calculate hit rate or capital-weighted outcome denominator.
2. Team and vehicle explanations are powerful. [researcher inference | strong] Colonna led GeoCities, Wenger led multiple signature companies, Burnham co-built the firm and narrates Tumblr, founders created the outcomes, and filings show shared entity power. Wilson's visibility can absorb other people's labor role map.
3. Era and access can mimic skill. [researcher inference | strong] Flatiron entered an underfunded regional internet market; USV formed after the crash as application-layer costs fell; Euclid, institutional LPs, connectors, and AVC improved access. Vintage-matched net evidence is absent historical context.
4. A few outliers may dominate. [researcher inference | strong] Public evidence cannot calculate results excluding the largest company. Coinbase affiliate transactions and one LP's USV 2012 metrics are impressive signals but cannot identify Wilson's contribution or ex-outlier result Pareto view.
5. Doctrine can be retrospective. [researcher inference | strong] The corpus contains decision-near writings, but many origin, governance, and failure accounts are years later. A coherent framework can be fitted to winners unless contemporaneous memos, passes, and follow-on decisions align.
6. The method has visible false positives and negatives. [researcher inference | strong] Kozmo, StarMedia, Turntable, Kik, and selection inequity test capital, governance, retention, law, and access; Airbnb shows that focus can become frame lock. Ordinary outcomes are less visible failures.
7. Public commitments are not outcome proof. [researcher inference | strong] Writing candidly, supporting founders, and acknowledging institutional failure are valuable behaviors. They do not prove superior selection, value-add, inclusion improvement, or net returns.

What would change the skeptical conclusion — [researcher inference | strong]. Recover an opportunity-level funnel with dated memos and passes; checks, ownership, reserves, board interventions, exits, and net cash flows by vehicle; a function-level partner ledger; comparable failed/ordinary cases; fund cash flows across vintages; and results with and without dominant outcomes. If Wilson-attributed cases then show superior net performance and better decision calibration after controls, the greatness claim strengthens materially. Until then, the defensible conclusion is narrower: an unusually transparent and influential venture operating doctrine, several well-supported decision contributions, important partner/team effects, and incomplete proof of personal economic greatness.

17. Analyst study guide, glossary, and go deeper

Study path

1. Start with the profile for chronology, people, edge hypothesis, and falsifiers.
2. Read the written corpus, beginning with USV formation, customer, early stage, deal size, reserves, failure, Twitter, GeoCities, Airbnb, Coinbase, venture math, appraisal, and valuation.
3. Use the spoken corpus for biographical recollection, trust/board doctrine, partner attribution, and the limits of edited or partial transcripts.
4. Replay the eight no-hindsight labs before reading the outcome block.
5. Work through the 17-model field guide, then return to sections 10 and 14 here to test transfer.
6. Audit the source map whenever a claim needs the underlying primary artifact, access boundary, or locator.

Ranked primary learning set

1. USV formation note — vehicle, partnership, stage, AVC boundary.
2. Early Stage Investing, Deal Size, and Reserves — the linked portfolio architecture.
3. Twitter, GeoCities, and Airbnb — decision-near win, team-credit correction, and false negative.
4. Reboot and Board Meeting Suggestions — trust, leadership, and governance.
5. Coinbase, Kin, and DefendCrypto — protocol thesis, conflict, and legal boundary.
6. Business Insider interview — unusually useful partner/deal/failure attribution, still retrospective.
7. Venture-math sequence and Valuing a Venture Capital Portfolio — correction and return-unit discipline.

Ranked secondary and contextual controls

1. Wired — “Flatiron VC Fund Pins Hopes on Silicon Alley” (1996). [contemporaneous record | moderate] Best independent near-launch control for Flatiron's stated capital structure, regional market, and first investments. It is reported journalism, not a partnership agreement, portfolio denominator, or audited performance record.
2. Forbes — “Flatiron Partners Called Back Home” (2001). [contemporaneous record | provisional] Useful adverse, near-period route on post-crash Flatiron and the Chase relationship. Access is partial/paywalled, so use it as a retrieval/context lead rather than proof for any unrecovered sentence.
3. Tren Griffin / 25iq — “A Dozen Things I've Learned from Fred Wilson” (2014). [researcher inference | provisional] A potentially useful secondary map of Wilson maxims and underlying-source leads. The current route is a cache miss/partial-access artifact; no quotation or doctrine should be accepted without returning to Wilson's original work.

No other secondary route was elevated above these three. [researcher inference | strong] The final analysis relies on Wilson/USV writings, full or bounded transcripts, issuer/regulatory records, and dated filings wherever available; secondary sources are controls and retrieval aids, not substitutes for underlying evidence.

Glossary

Reusable analyst checklists

Sourcing card

• [researcher inference | strong] Source/date/relationship; pre-existing market map; new evidence; independent skeptic; pass category; later outcome.

Investment memo card

• [researcher inference | strong] Participant behavior; expanding market; formation remaining; founder learning; enabling change; capture; incumbent pressure; law/security; terms; vehicle; next milestone; strongest opposite case; kill/change criteria.

Follow-on card

• [researcher inference | strong] Evidence changed since prior check; cohort/economics; milestone; leadership/governance; no-new-capital case; ownership/dilution; absolute expected return versus new opportunities; conflict/vehicle allocation.

Board card

• [researcher inference | strong] CEO constraint; decision required; preread; dissent; role/fiduciary status; owner/date; follow-through; stakeholder/security/regulatory consequences.

Attribution and return card

• [researcher inference | strong] Introducer; thesis author; deal lead; diligence; negotiation; approval; board; reserve; exit; public narrator; vehicle; beneficial owner; cash recipient; gross/net; realized/unrealized; as-of date; denominator.

The best way to study Wilson is to imitate the audit trail, not the aphorism. [researcher inference | strong] Write the thesis before the outcome, preserve the pass, label every role and return unit, and make the evidence that would change your mind visible to the next analyst.

Pareto 80/20 — what drove the record

[researcher inference | strong] Public evidence does not support a literal contribution analysis of Fred Wilson's lifetime record. It supplies company stakes at selected filing dates, a few participant-reported outcomes, and one limited partner's net metrics for one USV fund; it does not supply a deduplicated personal portfolio, company-level cost bases, complete cash flows, carry allocation, or Wilson-only proceeds. The three rankings below therefore answer three different questions and must not be combined into a synthetic score.

Ranked view 1 — economic contribution signal

[researcher inference | provisional] This ranks the strength of public evidence that a company or vehicle mattered economically, not verified dollars or percentage contribution to Wilson's record.

[researcher inference | strong] Concentration test: no public source permits reconstruction of company-by-company invested capital and distributed value, so neither top-five contribution nor “record excluding the dominant outlier” can be calculated. Wilson's 2007 aggregate is count-based and unnamed; the LACERA schedule is one LP's fund-level result.

Ranked view 2 — historical importance

1. [researcher inference | strong] GeoCities is the bridge from New York's early web ecosystem to Flatiron's public identity, but Jerry Colonna—not Wilson alone—brought and led it (Wilson's credit correction; GeoCities, ¶¶1–31).
2. [researcher inference | strong] Twitter is the clearest early embodiment of USV's network thesis: emergent user invention and an open API mattered while the business model remained unknown (Twitter investment announcement, ¶¶1–8).
3. [researcher inference | strong] Coinbase marks the transition from consumer networks to open financial protocols and adds an unusually visible Wilson board and public-advocacy trail (Coinbase announcement; Coinbase S-1, Wilson biography).
4. [researcher inference | strong] Etsy matters institutionally because Wenger found it and took the first USV board seat while Wilson later held long governance roles, demonstrating succession rather than sole-star attribution (Mixergy, Wenger exchange; Etsy 2026 proxy, board biography).
5. [researcher inference | moderate] Zynga, MongoDB, Tumblr, Foursquare, and Boxee are a ranked cluster of thesis expansion: social-platform distribution, developer infrastructure, creator networks, location, and connected media. Their importance is clearer than their Wilson-specific economics (USV company directory, 2007–2009 entries; Business Insider, partner and Tumblr passages).

Ranked view 3 — decision-learning value

1. [researcher inference | strong] Airbnb pass: the consented 2009 email record and Wilson's later reflection show that evaluating the current wedge instead of founder adaptability and market expansion can create a false negative (Paul Graham, “Subject: Airbnb,” emails dated 2009-01-23–02-21; Airbnb, later reflection).
2. [researcher inference | strong] Kozmo: success in one city did not justify simultaneous multi-city scale; financing escalation converted an experiment into a total loss (Worthless Stock Certificates, Kozmo section and Wilson reply).
3. [researcher inference | strong] StarMedia: loyalty to a founder can conflict with duty to the company when leadership fit changes; the board's delayed action is the claimed failure (Reboot transcript, PDF pp. 5–7).
4. [researcher inference | strong] Turntable: aggregate growth hid six-to-eight-week cohort churn, illustrating why retention evidence must outrank headline acquisition (Business Insider, Turntable exchange).
5. [researcher inference | strong] GeoCities: a positive case demonstrates role-specialized syndication, founder-incentive repair, selective secondary liquidity, and the danger of compressing a team outcome into “my deal” (GeoCities, ¶¶4–34).
6. [researcher inference | moderate] Kik/Kin: portfolio advocacy, governance, token funding, and legal classification became entangled; the later issuer judgment is a falsifier for the asserted legal outcome, not a finding against Wilson (DefendCrypto.org; SEC Kik case page).

Evidence boundary and searched denominator

coverage fraction: not computable

[researcher inference | strong] There is no defensible deduplicated lifetime denominator for Wilson. The public counts below overlap across person, firm, fund, vintage, and time; “investment” can mean company, financing, token/project, or fund position. The analysis therefore reports each universe separately and does not divide named winners by 32, 55, 21, or 230+.

[researcher inference | strong] Search scope: the work queried the USV company directory, AVC and USV posts, four long-form interviews, Flatiron-era press, fund announcements, issuer filings for GeoCities, Twitter, Zynga, Etsy, MongoDB, Coinbase, Tumblr, and Airbnb (Airbnb S-1/A, filed 2020-12-07), the SEC Kik record, LACERA, Form ADV, and current legal-case routes. The live company directory is a firm marketing/portfolio index with first-investment labels; it is neither a historical snapshot nor a Wilson role register.

[researcher inference | strong] Reconstruction threshold: “sufficiently reconstructed” requires a dated decision or near-decision source, a supported role classification, independently anchored vehicle/transaction or outcome evidence when available, and an explicit unknowns list. A company name in a directory or interview is “researched” but not sufficient.

[researcher inference | strong] Attribution classes: “Wilson-attributed” requires Wilson to claim or be assigned a sourcing/public-narrator/board role; “named-partner” preserves another person's role; “shared/collective” reflects firm decision or multiple partner power; “vehicle-only” is a legal-owner fact; “firm-only” has no person evidence; “pass” has no investment; “outside-denominator” is personal/angel or otherwise outside a managed fund. Narrative prominence is never used as ownership evidence.

Historical market, regulatory, and base-rate context

1996–2000 — Silicon Alley, concentrated capital, and public-market pull

[contemporaneous record | strong] Flatiron launched into an undercapitalized New York internet market. A December 1996 report described a $50 million fund—$25 million each from SoftBank and Chase—plus options for each backer to invest another $50 million directly, identified more than 700 internet-related companies in Silicon Alley, and named Yoyodyne as the first investment (Wired, “Flatiron VC Fund Pins Hopes on Silicon Alley”). These near-period figures support capital/geography context, not Flatiron performance.

[researcher inference | strong] GeoCities' top-ten traffic, scarce local growth capital, and syndicate specialization were sensible 1996 evidence; a 1999 stock-for-stock exit during an internet boom was a different market state. Yahoo's filing records a May 28, 1999 closing with 21,640,342 Yahoo common shares exchanged for GeoCities common; separately, 8.432 million GeoCities options converted into approximately 5.707 million Yahoo options. It does not establish Flatiron ownership or proceeds (Yahoo GeoCities filing, MD&A and Note 11).

[retrospective witness | moderate] Flatiron then overextended. Wilson describes the fund/team expanding after early success, too much capital changing behavior, and new investing ending around 2000 (VC Cliche of the Week, Flatiron postmortem; Mixergy, opening Flatiron exchange). That diagnosis is plausible but cannot substitute for vintage cash flows or a contemporaneous portfolio schedule.

2003–2008 — post-crash application-layer economics

[contemporaneous record | strong] USV began deploying in 2004 after the dot-com infrastructure build-out, when modern social media and smartphones did not yet exist (USV 2024 retrospective, opening history). Brad Burnham's 2008 fund notice says the first fund began as the economy emerged from the crash and lightweight web services were not broadly understood; by 2008 the category was crowded and recession risk was visible (New Fund — Same Focus, market-context paragraphs).

[investor-stated | strong] The decision-time doctrine responded with small initial checks, low burn, material ownership, and networks as defensibility. Wilson described $250,000–$2.5 million initial investments and 15–20% target ownership, with more capital reserved for evidence-backed follow-ons (Early Stage Investing, capital and ownership paragraphs; Deal Size, check-size and ownership paragraphs). This is stated policy, not proof of every transaction.

[investor-stated | moderate] Wilson's 2007 personal snapshot observed five failures among 26 realized/effectively realized investments and projected that the six then-unrealized positions could raise the failure share; it is a selected personal sample, not an industry base rate (Why Early Stage Venture Investments Fail, outcome table). No comparable contemporaneous, audited base-rate dataset was recovered for the named portfolio, so this report does not import Wilson's later one-third heuristic backward.

2009–2013 — network saturation, incubators, mobile, and crypto

[contemporaneous record | strong] By 2011 USV said social-network value and venture attention had expanded dramatically since 2004, making new network launches harder. The firm created a separate opportunity vehicle so larger follow-ons would not force the core strategy to change (The Opportunity Fund, ¶¶1–8).

[contemporaneous record | strong] Airbnb arrived through Y Combinator in early 2009, adding accelerator selection and small-seed evidence as alternatives to a conventional Series A process; USV met the founders but did not pursue (Paul Graham, “Subject: Airbnb,” 2009 email thread; Airbnb, later reflection). [researcher inference | strong] Twitter showed that user behavior could develop before a business model, whereas Turntable later showed that acquisition without retention was not durable (Twitter, rationale and business-model paragraph; Business Insider, Turntable exchange).

[investor-stated | strong] Coinbase was the firm's first disclosed Bitcoin-sector investment after roughly a year of Wilson/Wenger study. The May 2013 thesis framed Bitcoin as an open value-transfer protocol and Coinbase as wallet, merchant, and exchange infrastructure (Coinbase announcement; Business Insider, Bitcoin exchange). The post's optimism predates later U.S. regulatory litigation and must not be read through 2021 liquidity.

2017–2026 — tokens, enforcement, institutional succession, and expanded remit

[investor-stated | strong] Wilson publicly favored Kik's Kin plan in 2017, later disclosed a Kik board seat and personal token contribution while advocating against the SEC's approach (Kin; DefendCrypto.org). [documented behavior | strong] A federal court later found Kik's 2017 offering unregistered, and the issuer paid a $5 million civil penalty (SEC case page, complaint/judgment summary). The finding is against Kik, not Wilson or USV personally.

[contemporaneous record | strong] By 2024 USV described an expanded portfolio across networks, marketplaces, developer tools, health, fintech, decentralized systems, climate, energy, and AI, tied together by “edge” entry into large markets under pressure (USV 2024 retrospective, history and thesis). This current framing helps classify later firm investments; it does not prove the same rubric governed a 1996 or 2007 decision.

Vehicle economics and incentives

[investor-stated | strong] Reserve doctrine: Wilson says some companies require unforeseeable later capital and that reserves preserve the right to participate after evidence improves (Reserves, ¶¶1–10). [researcher inference | moderate] The incentive tension is real: pro-rata protection can preserve ownership in winners but also encourage escalation in weak companies. The public record does not disclose USV's company-level stop rules, write-off authority, recycling caps, or reserve forecast errors.

[contemporaneous record | strong] Only public-LP performance signal located: for LACERA's commitment to Union Square Ventures 2012 Fund, L.P., LACERA reported: 23.75x TVPI (LACERA, PDF p. 40 / printed p. 40, “Top Performing Investments (since inception),” USV 2012 row, 2026), reported: 23.15x DPI (same table and row, 2026), and reported: 53.4% since-inception net IRR (same table and row, 2026), as of March 31, 2026. TVPI includes residual value; DPI is distributed capital; LP cash-flow timing and fee/class terms may differ. This is one LACERA commitment, not a current whole-fund mark, firm composite, gross return, company result, or Fred-only result.

People, capital, and influence map

Capital flow and authority

1. [contemporaneous record | strong] Capital providers: SoftBank and Chase supplied Flatiron capital and direct-investment capacity in 1996; USV fund notices refer to limited partners without naming a complete LP roster (Wired, 1996 capital structure; USV 2022 funds, closing thanks).
2. [researcher inference | strong] Vehicles and adviser: capital sits in legal vehicles, while an adviser/GP directs investments subject to governing documents. The 2026 Form ADV, Schedule D §7.B(1) names entities but not the full internal vote, allocation, or carry system.
3. [contemporaneous record | strong] Partnership: Wilson and Burnham founded USV; Wenger and Weissman later assumed greater leadership; Buttrick, Kaden, Munson, Grossman, Mesele, Rachlin, and others entered differentiated investing, legal, finance, network, and operating roles (USV 2016; USV 2019; USV 2022).
4. [researcher inference | strong] Portfolio governance: board seats and shared voting power connect partners to companies, but board authority is fiduciary to the company and is not identical to fund ownership, sourcing, or personal economics. The distinction is visible in Etsy's 2026 proxy and issuer ownership footnotes.
5. [researcher inference | strong] Return path: exits and distributions flow from company securities to vehicles and then to LPs/GPs under private agreements. Public acquisition price or filing-date ownership is therefore not a personal return.

Deal-role matrix

Comprehensive deal ledger

Material case ledger

Additional discovered-company ledger

[researcher inference | strong] These 27 rows, together with the 14 material cases, complete the named 41-case research ledger. “Firm-only” means the available source supports organization-level association but not a Wilson role; absence of a reconstructed outcome is not a failure classification.

Ex-ante forensic case studies and decision labs

[researcher inference | strong] Each case freezes the evidence at the stated decision date. The exercise appears before the actual decision and outcome. Exercise answers are analyst-created, not claims about what Wilson or a committee actually wrote.

GeoCities — fall 1996

• [retrospective witness | moderate] Known then: GeoCities was a top-ten Media Metrix property; CMGI already owned half; founder David Bohnett had received little prior financing; Colonna knew the company; a $10 million pre-money proposal was on the table (GeoCities, ¶¶4–15).
• [researcher inference | strong] Unknown / believed / disputed: revenue quality, burn control, retention, advertiser economics, and exact cap table were not published. Colonna/Flatiron believed community and self-expression could compound. CMGI disputed the initial ownership/financing structure; founder incentives were visibly impaired.
• [researcher inference | moderate] Alternatives and context: decline; finance alone; syndicate; smaller bridge; restructure founder equity. Scarce East Coast internet capital and strategic LP capacity improved access but could also encourage overfunding (Wired, 1996).
• [researcher inference | moderate] Risks / kill criteria: unresolved founder alignment; inability to install financial controls; traffic without durable engagement or revenue; syndicate conflict; burn beyond milestones.
• [retrospective witness | moderate] Terms / follow-ons / governance: proposed Flatiron/SoftBank/Chase $6 million for about 30%, with CMGI $2 million; eventual round $9 million with Innocal/Intel. Exact liquidation preferences and board votes remain unknown (GeoCities, ¶¶14–27).
• [researcher inference | moderate] Milestone test: require traffic retention, financial controls, and founder/CEO alignment before a larger insider round.

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: yes, conditionally, because top-tier usage plus scarce category capital creates asymmetric upside, but only after cap-table and control repair.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: CMGI/founder misalignment that cannot be contractually repaired, unauditable cash controls, or engagement that is mostly undifferentiated portal traffic.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: meaningful syndicate ownership with founder-option refresh, staged funding, information rights, board representation, and explicit budget controls; the reported 30% syndicate proposal is a reference, not an optimality proof.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: retained/cohort usage, advertising demand, clean monthly reporting, controlled burn, and a functioning founder/CEO/board relationship.

[retrospective witness | moderate] Actual decision and outcome: Colonna brought and led the deal; Wilson devised the option intervention; multiple investors funded $9 million. A 1997 insider round followed at a reportedly fourfold valuation, Flatiron/Chase later sold one-third, and Yahoo closed the stock acquisition in 1999 (GeoCities, ¶¶14–34; Yahoo filing). The public record does not show the original memo, full terms, or final vehicle return.

Twitter — July 2007

• [investor-stated | strong] Known then: people were posting publicly, followers could opt in, users had invented @replies, and an open API supported third-party applications; scale/reliability needed capital (Twitter, investment-rationale paragraphs).
• [investor-stated | strong] Unknown / believed: Wilson explicitly says the business model was unknown and presents emergent behavior plus an open platform as the network thesis (Twitter, business-model and platform paragraphs).
• [researcher inference | moderate] Disputed / untested: monetization, abuse, infrastructure cost, governance, and platform capture remained open tests.
• [researcher inference | moderate] Alternatives and context: invest in the spinout; wait for monetization; back a competing publishing/network layer; pass because the service was feature-like. The application-layer web and API ecosystem were growing, but mobile distribution was immature.
• [researcher inference | moderate] Risks / kill criteria: retention limited to a novelty cohort; reliability collapse; closed API; founder/control deadlock; no route to revenue that preserves network growth.
• [researcher inference | insufficient public record] Terms / milestones / follow-ons / governance: original check, price, ownership, board, liquidation rights, and reserve plan are absent. Appropriate milestones would include active-user retention, API ecosystem quality, reliability, and evidence of repeat use beyond early adopters.

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: yes, as an option on a new communication graph, accepting business-model uncertainty if ownership and capital needs fit a small core fund.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: weak repeat use, continued outages without an engineering plan, or a governance structure unable to choose product leadership.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: staged early-stage exposure with meaningful but noncontrolling ownership, board/information rights, and reserves tied to reliability and retention.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: durable cohorts, organic user-created behaviors, a healthy developer ecosystem, and measurable reliability improvement—not a premature revenue plan alone.

[contemporaneous record | strong] Actual decision and outcome: USV financed Twitter and publicly accepted business-model uncertainty. The later S-1/A shows a 5.9% pre-offering USV position with shared partner power (Twitter S-1/A, principal-stockholder footnote 16). It does not show initial terms or realized USV/Wilson returns.

Coinbase — May 2013

• [investor-stated | strong] Known then: Bitcoin offered open, global value transfer; Coinbase provided wallet, merchant, and fiat-exchange functions; USV called the product/team very early and invested alongside Ribbit, SV Angel, and FundersClub (Coinbase).
• [retrospective witness | moderate] Believed: Wilson and Wenger had studied Bitcoin for about a year and believed a protocol layer could support a major financial network (Business Insider, Bitcoin exchange).
• [researcher inference | moderate] Unknown / disputed: custody security, regulatory classification, bank access, liquidity, consumer adoption, and value capture remained uncertain.
• [researcher inference | moderate] Alternatives and context: buy/provide protocol exposure; fund mining or payments; back a noncustodial tool; wait for regulation; pass on price/volatility. The seed round was reportedly too small for USV, so waiting for Series A preserved check fit but increased price/competition risk.
• [researcher inference | moderate] Risks / kill criteria: security failure, inability to maintain bank/payment access, poor custody controls, regulation barring the model, or usage driven only by speculation rather than repeat transaction demand.
• [researcher inference | insufficient public record] Terms / milestones / follow-ons / governance: announcement omits check, price, ownership, rights, and milestones. S-1 later identifies Series A holder board designation and Wilson joining in 2017, but cannot be projected backward (Coinbase S-1, Wilson biography and charter rights).

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: yes, in a staged position sized for regulatory/custody tail risk, because exchange/wallet infrastructure could capture adoption without predicting Bitcoin's final use.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: compromised customer assets, no durable banking relationships, regulators prohibiting core flows, or no repeat usage outside price speculation.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: material minority ownership, security/audit information rights, board-observer or board rights, and financing milestones; exact target depends on valuation and custody liability.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: secure custody history, repeat funded accounts, merchant or transfer use, stable fiat rails, compliance leadership, and resilient unit economics across price cycles.

[contemporaneous record | strong] Actual decision and outcome: USV invested in the Series A; Wilson later joined the board. Coinbase's S-1 shows USV entities with 13.9 million Class B shares, an 8.2% position, shared partner power, and disclosed prelisting secondary transactions (Coinbase S-1, certain transactions, directors, and principal stockholders). It does not by itself establish listing completion or allocate outcome to Wilson.

Airbnb — winter 2009 pass

• [contemporaneous record | strong] Known then: Airbnb was in Y Combinator; Graham's consented emails record USV partner debate and a founder meeting. Wilson's later account records the cereal-box hustle and air-mattress/hosted-sleeping framing (Graham, “Subject: Airbnb,” emails dated 2009-01-23–02-21; Airbnb, ¶¶1–7).
• [retrospective witness | strong] Unknown / believed / disputed: Wilson says USV liked the founders but did not see the broader accommodation market and did not press for an investment. Price, allocation, partner vote, user cohorts, supply growth, and trust/safety data are absent. It is therefore supported that the opportunity was not pursued, not that a single thesis error fully caused the pass.
• [researcher inference | moderate] Alternatives and context: invest a small seed check; wait for repeat booking evidence; introduce another investor; pass because behavior looked niche. YC and founder resourcefulness were positive signals, while payments, host trust, regulation, and marketplace liquidity were major unknowns.
• [researcher inference | moderate] Risks / kill criteria: no repeat stays outside events, insufficient host supply, safety incidents without controls, adverse local regulation, or unit economics requiring subsidies without network improvement.
• [researcher inference | insufficient public record] Terms / milestones / follow-ons / governance: no proposed terms are public. Sensible milestones would have been city-level liquidity, repeat bookings, host retention, incident rates, and contribution margin before geographic scale.

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: a small option check is warranted because the founders repeatedly reframed and financed the company, but only if marketplace liquidity can be measured beyond conferences.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: demand limited to exceptional events, hosts unwilling to repeat, unmanageable safety risk, or regulation that makes the supply structurally unavailable.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: seed-sized minority ownership with pro-rata rights and explicit information on liquidity/safety; no public offered valuation permits a numerical ownership conclusion.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: repeat bookings, organic host acquisition, improving match rates in a second/third city, low incident rates, and positive unit economics before broad expansion.

[contemporaneous record | strong] Actual decision and outcome: USV passed. Wilson's 2011 post documents his later conclusion that the air-mattress framing was too narrow. Airbnb's 2020 S-1/A documents a global accommodation platform and public offering route (Airbnb; Airbnb S-1/A). The counterfactual allocation, dilution, and return cannot be computed.

Kozmo — expansion financing, approximately 1999–2000

• [retrospective witness | moderate] Known then: Wilson later says the service worked and was profitable in New York; substantial capital was available during the internet boom; management proposed rapid replication across many cities (Worthless Stock Certificates, Kozmo section and Wilson comment).
• [researcher inference | strong] Unknown / believed / disputed: contemporaneous unit economics, order density, customer retention, delivery cost, expansion budget, liabilities, and board materials are missing. Wilson later believed the model should have been proved in a second and third city; the record does not show whether this was proposed then.
• [researcher inference | moderate] Alternatives and context: stay in New York; add one city; license logistics; fund a staged rollout; or reject follow-on capital. Easy boom-era financing made broad expansion possible while masking whether local density was portable.
• [researcher inference | moderate] Risks / kill criteria: deteriorating contribution margin outside New York, insufficient order density, fixed obligations that make retreat impossible, rising cash burn, or inability to replicate service quality.
• [retrospective witness | moderate] Terms / milestones / follow-ons / governance: Wilson says hundreds of millions were raised and roughly 18–20 cities launched, but checks, preferences, board composition, budgets, and votes are absent. The central missing milestone was second/third-city proof before simultaneous expansion.

Decision exercise — answer before reading the outcome

1. [researcher inference | strong] Would you invest then? Analyst-created answer: retain the initial position only if a single new city can be funded as a falsifiable replication test; reject a many-city plan.
2. [researcher inference | strong] What would kill the deal? Analyst-created answer: no path to dense repeat demand and positive contribution margin in the second city, or liabilities that prevent shutting a failed market.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: tranched follow-on capital, board-approved city gates, burn covenants/information rights, and no funding of simultaneous expansion before replication.
4. [researcher inference | strong] What evidence unlocks the next check? Analyst-created answer: second-city cohort retention, route density, delivery cost per order, local payback, and evidence that service quality holds.

[investor-stated | strong] Actual decision and outcome: investors financed broad expansion—Wilson says hundreds of millions and roughly 18–20 cities—and he describes a total loss for Flatiron and other investors (Worthless Stock Certificates, ¶3 and Wilson comment). The amounts and city count are approximate; no bankruptcy/cap-table record in the corpus allocates loss by vehicle.

StarMedia — leadership decision, date not reconstructed

• [retrospective witness | strong] Known then: the Reboot account says the company had reached a stage where the founder was no longer the leader the company needed; the board nevertheless remained loyal too long (Reboot, PDF pp. 5–7).
• [researcher inference | insufficient public record] Unknown / believed / disputed: original investment thesis, user/revenue metrics, cash, competitive position, founder performance evidence, financing terms, and board minutes are not in the processed record. Wilson's and Colonna's 2015 recollection is not a contemporaneous adjudication of the founder.
• [researcher inference | moderate] Alternatives and context: retain the founder with a stronger operating team; set a timed improvement plan; change roles; recruit a CEO; sell or wind down. The relevant context is stage transition, not a generic anti-founder rule.
• [researcher inference | moderate] Risks / kill criteria: missed operating commitments, inability to recruit complementary executives, unreliable reporting, board paralysis, or financing dependence that removes decision time.
• [researcher inference | insufficient public record] Terms / milestones / follow-ons / governance: board duties and leadership fit are discussed, but director composition, voting thresholds, removal rights, reserves, and milestones are absent. A fair process would require documented role requirements, evidence, alternatives, and stakeholder impacts.

Decision exercise — answer before reading the outcome

1. [researcher inference | provisional] Would you invest then? Analyst-created answer: not decidable from the surviving record; the original opportunity cannot be reconstructed.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: refusal or inability to repair a documented leadership gap before cash runway closes.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: normal minority venture terms cannot substitute for a capable, balanced board; governance rights should permit an orderly leadership process without promising investor control.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: agreed operating goals, executive-team references, credible succession candidates, board alignment, and financing runway through transition.

[retrospective witness | strong] Actual decision and outcome: Wilson says the board delayed the leadership change and failed its duty to the company and stakeholders (Reboot, PDF pp. 5–7). The record does not support a quantified loss, a sole-Wilson decision, or an adverse finding against the founder.

Turntable — early social-music investment, approximately 2011–2012

• [retrospective witness | strong] Known then: aggregate usage grew rapidly and the social, synchronous music experience attracted attention. Cohort analysis later showed many users leaving after six to eight weeks (Business Insider, Turntable exchange).
• [retrospective witness | strong] Unknown / believed / disputed: Wilson says he led the investment and believed the product could work; the decision-time cohort data, licensing economics, acquisition channels, check, price, and board materials are absent. He later believed a passive mode might have retained users, but that is a counterfactual product thesis.
• [researcher inference | moderate] Alternatives and context: fund the live social experience; require an asynchronous/passive test first; negotiate rights economics; or wait for cohort durability. Music licensing and high engagement could create either defensibility or structural cost.
• [researcher inference | strong] Risks / kill criteria: six-to-eight-week churn persisting across cohorts; usage dependent on continuous active participation; licensing cost outrunning monetization; no credible passive use case.
• [retrospective witness | moderate] Terms / milestones / follow-ons / governance: Wilson says the board/company reacted too slowly. A milestone plan should have separated acquisition, week-eight retention, creator/DJ supply, session frequency, and licensing contribution before follow-on funding.

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: only a small evidence-gathering check until cohort durability and rights economics are visible.
2. [researcher inference | strong] What would kill the deal? Analyst-created answer: repeated cohort collapse by week eight with no successful passive/asynchronous experiment.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: staged minority position with pro-rata rights, monthly cohort reporting, and follow-on explicitly conditional on retention rather than gross users.
4. [researcher inference | strong] What evidence unlocks the next check? Analyst-created answer: stable week-eight/month-six cohorts, repeat listening without high effort, improving licensed-content economics, and organic reactivation.

[retrospective witness | strong] Actual decision and outcome: Wilson says he led the investment, the company and board diagnosed/reacted too slowly, and the original form did not endure (Business Insider, Turntable exchange). No decision memo, cohort export, liquidation record, or return was recovered.

Boxee — relationship and software-pivot case, 2006–2008

• [retrospective witness | moderate] Known then: Wilson says he first declined a hardware-oriented pitch, stayed in contact for roughly two years, then used the open-source software product before USV invested; the current firm directory dates the Series A to 2008 (MeetInnovators, PDF pp. 8–10; USV directory, Boxee entry).
• [researcher inference | strong] Unknown / believed / disputed: decision-time users, retention, hardware dependency, content rights, valuation, check, ownership, and competitive set are absent. The supported change is hardware pitch to software product plus Wilson's use; it does not prove a large addressable market or attractive terms.
• [researcher inference | moderate] Alternatives and context: remain in relationship without investing; fund the software layer; require distribution/content partnerships; or pass on living-room platform dependence. Open-source software reduced hardware capital needs while leaving incumbent-platform and content risks.
• [researcher inference | moderate] Risks / kill criteria: weak repeat living-room use, dependence on unavailable content, hardware/platform gatekeepers blocking access, or a return to capital-intensive hardware before product-market evidence.
• [researcher inference | insufficient public record] Terms / milestones / follow-ons / governance: original financing documents and board roles are absent. Sensible milestones would test repeat viewing, active installations, content availability, distribution reach, and burn before follow-on.

Decision exercise — answer before reading the outcome

1. [researcher inference | moderate] Would you invest then? Analyst-created answer: yes, at modest early-stage scale after the software pivot and product-use evidence, while treating platform/content control as the central risk.
2. [researcher inference | moderate] What would kill the deal? Analyst-created answer: content unavailable on acceptable terms, low repeat use, or required hardware/subsidy capital incompatible with the fund.
3. [researcher inference | moderate] What ownership/terms fit the risk? Analyst-created answer: minority Series A ownership with pro-rata/information rights and follow-on capacity tied to software adoption, not an assumption that hardware distribution will emerge.
4. [researcher inference | moderate] What evidence unlocks the next check? Analyst-created answer: retained active installations, content breadth, organic adoption, partner distribution, and controlled burn.

[retrospective witness | strong] Actual decision and outcome: Wilson later includes Boxee among a sold/wound-down group on which he was lead investor; the directory records acquisition by Samsung. The public sources do not disclose price, proceeds, or whether the outcome returned invested capital (Business Insider, lead-investor outcome passage; USV directory, Boxee entry). This is an ordinary/indeterminate acquisition case, not a proven win.

Portfolio-level forensics

Construction, concentration, and selection

[researcher inference | strong] Vintage pattern: Flatiron concentrated on a local internet ecosystem with flexible corporate capital; early USV intentionally constrained fund size and concentrated on application-layer networks; opportunity vehicles later separated scale-up/later-stage capital; later funds broadened people and themes. This is institutional evolution, not proof that every change improved net performance (Wired 1996; USV 2016; USV 2022).

[researcher inference | strong] Power-law and attribution test: large filing-date stakes in Twitter, Etsy, MongoDB, Zynga, and Coinbase make them plausible drivers, but only LACERA's one-commitment fund row supplies a public net-performance signal. Without company cash flows, the report cannot identify the dominant deal, calculate loss ratio by capital, or test results excluding Coinbase/another outlier.

[investor-stated | moderate] Personal 2007 distribution: Wilson grouped 26 realized/effectively realized direct investments into five failures, ten at reported: 1–5x (Why Early Stage Venture Investments Fail, outcome table, 2007), and eleven at reported: >5x (same outcome table, 2007), with six still unrealized. [researcher inference | strong] The table is unaudited, unnamed, count-based, and status-dependent; it cannot be added to later deal outcomes or used as a lifetime hit rate.

Modern comparison — what transfers as of 2026-08-21

Scrutiny and litigation boundary — excluded from economic/failure counts

Failure, miss, and anti-portfolio postmortems

[researcher inference | strong] Anti-portfolio rule: passes belong in a separate denominator from investments. Airbnb, Huffington Post, and Gilt are useful for selection-error analysis, but adding their later values to “lost returns” would ignore allocation, follow-ons, dilution, liquidity, portfolio capacity, and the possibility that terms/access were unavailable.

[researcher inference | strong] Failure-rate rule: named failures cannot be divided by the current firm directory or 230+ aggregate. The 2007 32-direct snapshot is the only explicit personal count table and should remain frozen at its publication date; Kozmo, StarMedia, Turntable, Canvas, Kik, and other later cases may overlap different vehicles and outcome definitions.

Uncertainty and retrieval ledger

Open questions after retrieval

1. [researcher inference | strong] Which named investments are actually among Wilson's anonymous 32, and how does the 2007 outcome table change when each is followed to final cash realization?
2. [researcher inference | strong] Which one or two companies drove each Flatiron and USV fund after costs, and does the record remain attractive when the largest outcome is excluded?
3. [researcher inference | strong] Which partner sourced, led diligence, negotiated, joined the board, approved reserves, and managed exit for every material company?
4. [researcher inference | strong] Did USV's small-fund/ownership/reserve design improve realized net results, or merely coincide with strong network vintages?
5. [researcher inference | strong] How often did Wilson/USV act on retention, replication, leadership, or regulatory kill criteria before additional capital was committed?
6. [researcher inference | strong] What share of passes arose from category priors, price, allocation, partner disagreement, conflicts, or access—and what changed after Airbnb, Huffington Post, and Gilt?
7. [researcher inference | strong] How do the 26 current ADV legal-entity rows map to the firm's stated 14 invested funds, and which parallel/feeder entities share company positions?
8. [researcher inference | strong] What do operative Patel and Grabski pleadings allege against Wilson specifically, and what later orders change his status without converting allegations into findings?

[researcher inference | strong] The retrieval order is deliberate: personal and vehicle denominators come before another winner story. Until the top four records are available, company rankings remain directional, public-filing stakes remain snapshots, and coverage fraction: not computable remains the only defensible lifetime statement.

Coverage and evidence base

Literal priority-source reconciliation

The source map routes 34 priority URLs representing 33 underlying works to WRITINGS. The difference is the two-post April 2009 venture-math sequence, analyzed as one work while preserving both URLs. The accessible priority corpus is fully accounted for: 32 full works and one partial MBA Mondays series were deeply processed. “Partial” describes series-level coverage—the live archive and seven named posts were read, but no canonical manifest permits a whole-series claim.

The dossier contains 36 source notes and 36 source-by-source subsections. Thirty-three notes correspond to the priority works; exactly three are supplements: “What a CEO Does”, “I've Moved Onchain”, and the Coin Center prepared text, retained as a written-text and excluded-talk control.

Evidence and attribution rules

The corpus is unusually rich in direct prose but still uneven as evidence. An essay can establish what Wilson or a named partner said at a date; it cannot by itself establish consistent practice, sole decision credit, ownership, realized return, or causal skill. Firm announcements are attributed to their bylined author and USV; SEC filings are used only as claim-local controls for vehicle ownership, board tenure, and transaction facts. Wilson's public narration is never silently converted into personal capital or sole authorship.

The record is split into three layers throughout: text-derived fact (what a source actually says), investor-stated belief (the prescriptive rule or explanation advanced by Wilson or a partner), and researcher synthesis (a cross-source inference with a stated support grade). Quotes are short and decorative; the analysis rests on paraphrase, locators, and source boundaries. All sources were accessed on 2026-08-21.

Source-by-source analysis

1. About Fred Wilson — the blog as a public reasoning system

Context and central claim. The living About page, three body paragraphs says Wilson has been a venture capitalist since 1986 and began AVC in September 2003 to think out loud, receive feedback, and meet people. It frames the archive as evolving personal thought rather than USV doctrine. source note

Reasoning and evidence boundary. [investor-stated | strong] The page establishes Wilson's declared use of writing as an open feedback loop and relationship channel. [researcher inference | moderate] That design can improve idea discovery and correction, but readership, comment volume, and deal flow are not evidence that feedback improved investment decisions. The About page's 1986 marker aligns with Wilson's detailed internship chronology, while the current USV biography uses 1987; the defensible sequence is internship in 1986 and a 1987 full-time career marker. About, biography paragraph · “I Got Lucky,” Euclid paragraphs

Decision implication. [researcher inference | strong] Treat AVC as a dated decision-and-revision record, not a house view. When a post names a USV belief, any attribution must identify byline, partner contribution, vehicle, and contemporaneous action before assigning credit.

2. I Got Lucky — biography, domain focus, and the limits of outcome stories

Context and argument. In “I Got Lucky,” ¶¶2–14, Wilson reconstructs a summer-1986 Euclid internship obtained through a direct ask, continued work during Wharton, roughly ten years of broad venture apprenticeship, an internet specialization in 1994–1995, and the early-1996 formation of Flatiron with Jerry Colonna. He credits Joanne Wilson, Bliss McCrum Jr., Colonna, Chase, and SoftBank rather than presenting a solitary origin story. source note

Reasoning and evidence. [retrospective witness | moderate] The account supports a sequence in which access created an apprenticeship, accumulated transaction craft met a new domain, and a complementary partner supplied operating/media knowledge and networks. Wilson reports that a $150 million pledge fund became $750 million in three years (I Got Lucky, Flatiron paragraphs), but gives no cash-flow, valuation, ownership, fee, or realization basis. [researcher inference | strong] The number is therefore a self-reported outcome marker, not audited performance or proof that the method caused the result.

Decision implication and contradiction. [investor-stated | strong] Wilson's prescription is to earn deep operating knowledge and relationships in a field before entering venture as a partner—even though he says his own path relied heavily on timing. His “no real operating position” phrasing coexists with two years writing software; the narrower, supportable distinction is technical employment versus responsibility for building and running a startup business. The essay is strongest as an anti-copying warning: biography explains conditions, not a portable recipe.

3. VC Cliche of the Week — style drift as a portfolio-level failure mode

Context and argument. “VC Cliche of the Week,” Flatiron postmortem paragraphs uses Flatiron to explain “style drift”: changes in sector, stage, team, and capital base weakened the coherence of the later portfolio. Wilson does not isolate a single error; he describes a bundle of deviations from the conditions under which the first strategy worked. source note

Reasoning and boundary. [retrospective witness | moderate] The essay is decision-relevant because it connects fund structure to opportunity selection: more capital and altered mandate can force a manager to accept deals that do not fit the original edge. [researcher inference | moderate] It does not prove that drift rather than the dot-com crash, valuation, vintage, or execution caused weak outcomes. Nor does it define a durable test separating disciplined thesis evolution from drift.

Decision implication. [researcher inference | strong] Before changing stage, sector, geography, check size, or partner set, write the hypothesized advantage and the evidence that it survives the change. A strategy change is adaptation only if the edge, decision rights, and loss budget are explicit; otherwise “larger opportunity set” can conceal mandate dilution.

4. Connectors — complementary networks, not generic popularity

Context and argument. In “Connectors,” ¶¶1–14, Wilson ranks useful introductions highly and maps complementary contributions: he brought deal craft, Jerry Colonna brought operating/media expertise and an internet network, Brad Burnham brought a largely non-overlapping AT&T network, and AVC expanded Wilson's relationship graph. By 2006 he estimated connecting consumed half his time. source note

Reasoning and evidence. [investor-stated | strong] The claimed quality test is bilateral value, not introduction count. [retrospective witness | moderate] Complementary networks plausibly reduced search and access costs at Flatiron and USV, but the essay supplies no referral denominator or outcome comparison. Fast pre-2000 gains can reflect market beta and valuation expansion as well as network skill.

Decision implication. [researcher inference | strong] Underwrite a network as a conversion system: which trusted person supplies missing evidence, talent, customers, capital, or governance help, and what changed afterward? Network centrality is not an edge unless the investor diagnoses mutual value and converts access into useful action without relaxing diligence.

5. Union Square Ventures — formation principles before signature outcomes

Context and argument. Wilson's 2005 formation announcement, ¶¶2–13 dates the partnership with Brad Burnham to October 2003, records the September 2005 final close of Union Square Ventures 2004, L.P., identifies Tacoda and Instant Information as pre-close investments, and states an early-stage technology-enabled-services remit. AVC is both discovery channel and fit filter, but Wilson explicitly says his posts do not represent partners, colleagues, or investors. source note

Reasoning and evidence. [contemporaneous record | strong] The announcement is rare ex-ante evidence for a small, partner-led organization, prior-founder sourcing, and direct founder work before the best-known USV outcomes. [researcher inference | moderate] It establishes aspiration, not later delivery or returns. The post says October 2003, while Coinbase's S-1, director biography dates Wilson's USV service to June 2003; preparation versus public formation is plausible but unproven.

Decision implication. [researcher inference | strong] Preserve the formation memo as a precommitment: later portfolio analysis should test stage, sectors, founder-service behavior, team size, and repeat-relationship sourcing against it. Do not backfill the original thesis from later winners.

6. Our Customer Is the Entrepreneur — service doctrine under governance pressure

Context and argument. “Our Customer Is the Entrepreneur,” ¶¶1–16 distinguishes entrepreneurs as the firm's customers and LPs as shareholders. Wilson's chain is that entrepreneurs create enterprise value, money is comparatively substitutable, service earns trust and future access, and that access can sustain returns. He also reports roughly 30 relevant inbound opportunities per week, about 1,500 per year, and around four investments at the time. source note

Reasoning and contradiction. [investor-stated | strong] Founder service is both an ethical posture and a sourcing/reputation mechanism. Yet after investment the “customer” expands to company, employees, and end customers; service can mean refusing capital, demanding change, coaching, or replacing a founder. [researcher inference | strong] Compassion governs how hard oversight is conducted, not whether control intervention occurs.

Decision implication and boundary. [researcher inference | moderate] A modern diligence process should measure response quality and clarity across accepted and rejected founders, while board records should distinguish founder preference from company interest. The 2005 funnel is a dated, unaudited snapshot—not a lifetime denominator or proof that founders experienced the stated service.

7. Early Stage Investing — define stage by formation remaining

Context and argument. “Early Stage Investing,” ¶¶1–11 defines stage functionally: USV must be able to participate materially in forming a thesis-aligned business, regardless of revenue, employee count, or round label. Delicious lacked a company, team, business model, and partnerships; Oddcast had revenue but still needed to form a new social/mobile opportunity. Wilson credits earlier thesis work to Brad Burnham. source note

Reasoning and evidence. [investor-stated | strong] The method accepts team, market-development, business-model, and partnership risk when the missing formation work is both legible and influenceable. Valuation and prior capital matter because the firm's money and counsel must remain significant. [researcher inference | moderate] This turns “early” into a joint test of uncertainty, ownership, and potential contribution, but the two cases disclose no checks, decision dissent, rights, or returns.

Decision implication. [researcher inference | strong] Ask what remains unformed, why this investor can change it, and what milestone will show that formation is occurring. The kill test is not the financing label: if capital and counsel cannot materially alter the path, the opportunity is later than the strategy's preferred stage.

8. Deal Size — buy learning without surrendering meaningful ownership

Context and argument. “Deal Size,” ¶¶1–13 records four of nine USV initial checks at $1 million or less, only one above $3 million, target ownership of 15–20%, and a practice of financing near-term milestones. Wilson says lower initial dollars let USV tolerate more team, adoption, market, and business-model uncertainty while preserving capacity to build a $5–8 million position and, exceptionally, as much as $20 million. source note

Reasoning and evidence. [contemporaneous record | strong] This is a dated firm-level construction snapshot and credits both Wilson and Burnham. [investor-stated | strong] Small initial checks are options on learning, not a preference for small ownership. Syndication and later capital are necessary parts of the model. The unnamed nine-company set, entry prices, reserves, rights, and outcomes prevent any performance claim.

Decision implication and counterweight. [researcher inference | strong] Size the first check to reach evidence that can reprice or kill the thesis, and pre-plan the ownership and reserve path. Too little capital can create fragile experiments; acquiring the same ownership later at higher prices can erase the apparent early-check advantage.

9. Reserves — concentrate after evidence, not after sunk cost

Context and argument. In “Reserves,” ¶¶1–15, Wilson starts with a $25 million early-career fund whose company cap sometimes blocked pro-rata investment. He argues that long-financed companies can still produce attractive later rounds, estimates perhaps 20% will require unexpectedly large capital, and describes a reserve model using invested capital, future-round timing and probability, and pro-rata share. Andrew Parker is credited with building the spreadsheet; Brad Burnham is also named. source note

Reasoning and evidence. [investor-stated | strong] The economic claim is asymmetric: preserve capacity to add exposure when evidence improves. The eleven-company USV snapshot averaged $2 million invested and ranged from $275,000 to $6 million within a $125 million fund. Wilson gives reported: 5–10x (“Reserves,” named examples ¶¶3–6, 2007) for Multex, Bigfoot Interactive, and comScore; the figures are self-reported and lack dates, cash flows, and audit support. [researcher inference | moderate] Expected value makes assumptions visible, but the added buffer admits model error without sizing it.

Decision implication and contradiction. [researcher inference | strong] Predefine positive evidence for follow-ons, compare each insider round with the best new opportunity, and distinguish pro-rata defense from rescue financing. “Pounce on the Z rounds” can mean disciplined concentration or escalation after sunk cost; a losing-follow-on denominator is required to tell which.

10. Why Early Stage Venture Investments Fail — runway is an option on search

Context and argument. “Why Early Stage Venture Investments Fail,” ¶1 and numbered categories gives Wilson's strongest public personal denominator: 32 investments he says he sourced, led, and managed over 17 years, with five failures, six then-unrealized USV positions, and three sold companies. He expected one or two additional failures and an eventual rate near 20%. source note

Reasoning and evidence. [investor-stated | strong] Wilson separates a bad idea killed early from a plausible but directionally wrong idea overfunded until burn removes the ability to adapt. Four of five failures fit the second pattern in his account, and he assigns responsibility to investors and boards, including himself. [researcher inference | strong] The operational asset is runway for learning: capital policy determines how many dead ends a company can explore before evidence arrives.

Boundary and implication. [retrospective witness | moderate] The 32-name set is self-selected, unnamed, unweighted, and mixes failed, sold, unrealized, and partner-boundary categories; it is not a lifetime hit rate. A low company-count failure rate can reflect rescue capital, early sales, or classification. [researcher inference | strong] Boards should predefine burn limits, learning milestones, and stop conditions while explicitly separating Wilson-led from partner-led decisions.

11. Twitter — underwrite emergent behavior before a settled business model

Context and argument. “Twitter,” investment rationale and team framing explains the firm's investment through product behavior: public, asynchronous messaging; optional reply; simple creation and consumption; and an API that let others extend the service. Wilson acknowledges that the business model was unknown and treats the user behavior and platform possibility as the early evidence. source note

Reasoning and evidence. [contemporaneous record | strong] This is decision-near thesis evidence, not a retrospective claim built from the eventual outcome. [investor-stated | strong] The central wager is that simple communication primitives and third-party extensibility can generate uses the company did not design. [researcher inference | moderate] Unknown monetization is tolerable only if behavior, retention, and ecosystem formation provide measurable option value.

Attribution and boundary. Wilson is the public thesis narrator, not the personal owner of the investment; the public record recovered here does not establish internal sponsorship. Twitter's amended S-1, principal-stockholder footnote 16 attributes the position to USV vehicles and shared partner power. [researcher inference | strong] Later return or governance analysis must map vehicle ownership, partner credit, board decisions, follow-ons, and distributions separately.

12. I Want a New Platform — a Wenger thesis, not Wilson's MongoDB proof

Context and argument. Albert Wenger's “I Want a New Platform,” full post argues that application developers were repeatedly assembling servers, deployment, scaling, and operations from fragmented components. He asks for an integrated hosted platform with open interfaces that abstracts infrastructure without recreating proprietary lock-in. source note

Reasoning and evidence. [contemporaneous record | strong] The post supplies a partner-specific thesis based on observed developer pain and an architectural preference for integration plus openness. It does not mention 10gen or MongoDB, say it caused an investment, disclose a decision, or establish an outcome. Wilson separately said “Albert did the MongoDB deal” in The Business Insider Interview, partner-attribution passage; that retrospective credit must not be merged into the 2007 text.

Decision implication. [researcher inference | moderate] Look for repeated user-built integration as evidence that an infrastructure layer wants to exist, then test whether abstraction preserves portability and bargaining power. [researcher inference | strong] This source is also an attribution control: a USV thesis written by Wenger is not Wilson doctrine merely because it sits in the firm archive.

13. Zynga Game Network — update conviction when distribution changes

Context and argument. In “Zynga Game Network,” investment announcement and rationale, Wilson says USV initially lacked conviction about games but changed its view over roughly six months as social networks became distribution platforms and Zynga showed it could build across networks. The post presents an investment in the company and founder Mark Pincus through a firm-level “we.” source note

Reasoning and evidence. [contemporaneous record | strong] The essay records a belief update near the decision: a category became investable when distribution and cross-platform execution changed, not because the partners simply abandoned a prior view. [researcher inference | moderate] This supports an evidence-responsive thesis process, but the post omits entry price, ownership, retention, unit economics, internal dissent, and follow-on criteria.

Attribution and implication. Zynga's final S-1/A, principal-stockholders table and footnote shows 30,738,892 shares held through USV vehicles with shared power among Brad Burnham, Fred Wilson, Albert Wenger, and John Buttrick. [researcher inference | strong] Record the observation that changed the prior, the company evidence that satisfies it, and the shared vehicle/decision team; narration is not sole ownership.

14. Looking Back on 1999 — correct signature-deal shorthand

Context and claim. “Looking Back on 1999,” GeoCities correction paragraphs revisits Wilson's earlier phrase that GeoCities was the best deal he ever did. He states that Jerry Colonna brought the investment to Flatiron and led it; Wilson's role was to negotiate and structure the transaction. source note

Reasoning and evidence. [retrospective witness | strong] The correction is first-person role evidence and materially narrows Wilson's own shorthand. It also distinguishes a firm's successful position from the people who originated, negotiated, governed, financed, or narrated it. The post remains retrospective and does not allocate every later board, financing, or exit action.

Decision implication. [researcher inference | strong] Signature-deal histories should use a role map, not a hero label: origination, underwriting, term negotiation, board work, follow-ons, exit decision, vehicle capital, and later narration. This correction governs every later reference to GeoCities unless stronger decision-date evidence displaces it.

15. GeoCities — case method with team credit and incomplete economics

Context and argument. “GeoCities,” opening and ¶¶8–31 supplies a detailed Flatiron case: Jerry Colonna sourced and led; Wilson helped structure founder options and syndication; Jason Calacanis contributed analysis; Harry Lambert's Innocal team helped stabilize financial controls; Bob Greene represented Chase in the syndicate; David Bohnett founded the company; and Tom Evans later led the IPO and sale. Wilson retrospectively says the financing grew to $9 million and that the proposed structure put about 30% with Flatiron and its partners for a $6 million allocation; the realization basis and final allocation are not independently audited. He says Flatiron sold one-third of its position at reported: almost 10x (“GeoCities,” partial-sale ¶¶27–31, 2009). source note

Reasoning and evidence. [retrospective witness | moderate] Wilson presents partial liquidity as regret-minimizing risk control: reported: 3x (“GeoCities,” partial-sale ¶¶27–31, 2009) on the deal while preserving two-thirds of the position for further upside. [researcher inference | strong] The post demonstrates that venture outcomes are co-produced across sourcing, company building, finance, syndication, and exit. Its dollars and multiples omit dilution, timing, fees, allocations, and complete cash flows.

External control and implication. Yahoo's Form 8-K/A, Exhibit 99.2, MD&A overview and Note 11 corroborates transaction timing and consideration, not Flatiron ownership or return. [researcher inference | moderate] A defensible partial-sale rule needs the remaining exposure, fund concentration, liquidity needs, and scenario distribution—not hindsight comparison with the eventual price.

16. The Venture Capital Math Problem — expose assumptions and revise publicly

Context and argument. The original post, ¶¶1–22 and continuation, ¶¶1–14 form one work. Wilson models asset-class capacity from annual fundraising, required gross-to-net multiple, aggregate VC ownership, exit count, and exit-value distribution. The first pass uses $25 billion raised, 3x gross for 2.5x net, 50% VC ownership, 1,000 exits, and a $5 billion maximum exit. source note

Reasoning and revision. [investor-stated | strong] Wilson exposes uncertain assumptions and asks readers to correct them. VentureXpert data then reduce reported average exits to 365 annually; allowing for under-reporting, he revises the plausible range to 400–500. [documented behavior | strong] The next-day correction is direct evidence of public model revision. The exit-value curve remains unfinished, so the sequence is not a validated capacity theorem.

Decision implication and boundary. [researcher inference | strong] Fund-size claims should reveal fundraising, ownership, exit count, tail shape, duration, fees, recycling, and dilution assumptions, then show sensitivity rather than one output. Removing 1999–2000 may reduce anomaly but also remove tail behavior central to venture returns; a five-year steady state compresses a lumpy asset class.

17. What a CEO Does — three board-level duties, not a complete operating manual

Context and argument. “What a CEO Does,” ¶¶1–10 attributes a rule to an unnamed veteran investor during an early-career CEO search: set and communicate vision and strategy, recruit and retain the best people, and keep enough cash in the bank. Wilson says other work should be delegated and failure at any one duty defeats greatness. source note

Reasoning and evidence. [retrospective witness | moderate] The three duties make a concise board diagnostic across strategy, organization, and solvency. The remembered advice predates AVC and USV and belongs initially to the unnamed mentor, not Wilson. No company, candidate, board record, or outcome tests whether the rule predicted performance.

Decision implication and counterargument. [researcher inference | moderate] Boards can use the three duties as a first-pass scorecard, then require observable measures and owner assignments. Product judgment, execution systems, culture, customers, compliance, governance, and capital allocation beyond cash remain omitted; delegation itself depends on the talent duty already being satisfied.

18. MBA Mondays — practical teaching with an explicit aging-data warning

Corpus and sampling protocol. The MBA Mondays category archive has no stable deduplicated manifest, so this analysis has partial series access and processes the live category page plus seven disclosed posts: employee equity, litigation, convertible debt, retaining a team, turning a team, profitless prosperity, and the separately analyzed CEO essay. source note

Argument and evidence. [investor-stated | strong] Across the bounded sample, Wilson translates finance and management into explicit trade-offs: future versus current cash flow, convertible convenience versus unpriced equity risk, retention versus necessary team change, settlement cost versus existential litigation, and benchmarks versus company-specific value. In 2018 he warns that his 2010 employee-equity numbers were obsolete while defending the method and proposing new portfolio data. [documented behavior | strong] The archive therefore records teaching and revision, not a timeless rulebook.

Decision implication and boundary. [researcher inference | strong] Reuse the decision structure only after refreshing numeric inputs, legal assumptions, and market context. The category offers no exact work count, version history, case denominator, or decision-right map; the reported experience with 150-plus startups and roughly one-third reaching scale remains self-reported rather than audited.

19. Airbnb — a public miss that exposes frame lock

Context and argument. In “Airbnb,” email thread and reflection, Wilson publishes a decision-adjacent exchange and explains that USV focused on the then-visible air-mattress lodging wedge instead of the founders' ability to build a broader marketplace. He does not claim the company lacked promise; he admits the team and market were framed too narrowly. source note

Reasoning and evidence. [retrospective witness | strong] The email record makes this more useful than a generic “we passed” anecdote: it identifies the category frame Wilson says constrained judgment. [researcher inference | moderate] It still does not establish that framing was the only cause; price, allocation, access, partnership priorities, or portfolio conflict could also matter, and the corpus lacks an eligible-opportunity denominator.

Decision implication and bias control. [researcher inference | strong] For a strange current wedge, separately score present product evidence, founder learning velocity, market-expansion path, and the investor's reason for disbelief. The cereal-box reminder is an organizational memory device, not evidence that later behavior changed; only a named subsequent decision could test learning.

20. Investment Thesis @ USV — collective doctrine with exact authorship

Context and argument. Andy Weissman's “Investment Thesis @ USV,” full post presents a firm thesis around large networks of engaged users, differentiated network experiences, and the infrastructure enabling such networks. Weissman credits Brad Burnham with the core phrasing and frames the thesis as a team artifact rather than a Wilson-authored manifesto. source note

Reasoning and evidence. [contemporaneous record | strong] The post formalizes a transition from sector labels toward a system-level model: users, network effects, and enabling infrastructure. [researcher inference | moderate] It is broad enough to connect consumer and technical platforms, but without falsification criteria it can become a retrospective umbrella over dissimilar companies.

Decision implication and boundary. [researcher inference | strong] Translate “network” into measurable behavior: participant roles, engagement, retention, contribution incentives, interaction density, multihoming, and value created at each scale. Preserve exact byline and credited contribution; Wilson can be analyzed as a partner operating within the thesis, not its sole author.

21. Coinbase — protocol infrastructure as a network investment

Context and argument. “Coinbase,” complete investment announcement frames Bitcoin as an open financial protocol and Coinbase as an accessible service layer for holding and transacting in it. Wilson names the founders and co-investors, placing the company inside USV's network and infrastructure thesis rather than treating it as a speculative token-price wager. source note

Reasoning and evidence. [contemporaneous record | strong] The source captures decision-near belief in protocol adoption, usability, and a gateway business. [researcher inference | moderate] It does not disclose regulatory analysis, custody risk, security controls, economics, entry terms, or downside cases; those omissions matter in a financial infrastructure investment.

Attribution and implication. Coinbase's 2021 S-1, director biography and principal-stockholder section verifies Wilson's board tenure and shared USV vehicle power, not sole origination or personal ownership. [researcher inference | strong] Underwrite the protocol, interface, custody/control surface, regulatory regime, and vehicle economics separately; a correct protocol thesis does not establish that every gateway is defensible or compliant.

22. Yahoo Acquires Tumblr — partner narration and interested exit advocacy

Context and argument. Brad Burnham's “Yahoo Acquires Tumblr,” full post praises David Karp, Tumblr's community, and Yahoo's prospective stewardship. It is a partner-authored acquisition statement from an interested shareholder, not a Wilson-authored deal history. source note

Reasoning and evidence. [investor-stated | strong] Burnham emphasizes founder vision and network/community value while acknowledging that monetization was not the source of Tumblr's user devotion. [researcher inference | moderate] The post's celebratory transaction context weakens it as independent evidence that Yahoo could preserve the product or that the price captured durable value.

Transaction control and implication. Yahoo's Q2 2013 10-Q, acquisition note records a June 19 close for about $990 million, primarily cash; it does not disclose USV's realized return. [researcher inference | strong] Use the byline as a team-credit control and separate product thesis, transaction consideration, vehicle ownership, distributions, and post-acquisition outcome.

23. Coin Center Keynote — belief, resilience, and regulatory advocacy

Context and argument. Wilson's prepared Coin Center keynote, full text links internet resilience, open protocols, and blockchain networks, and argues that entrepreneurs should be free to build within rules that protect consumers without foreclosing innovation. He describes USV investing in the category since 2011 while acknowledging the absence, at that date, of a blockbuster outcome. source note

Reasoning and evidence. [investor-stated | strong] The speech makes persistence a product of a belief system: decentralization and open innovation are expected to survive cycles of skepticism. [investor-stated | moderate] The regulatory balance is advocacy by an economically interested investor, not legal analysis or neutral evidence. The admission that results were not yet conclusive is a useful decision-date boundary.

Decision implication and counterargument. [researcher inference | moderate] A long-duration thesis needs explicit empirical milestones—developer activity, useful transactions, security, decentralization, user protection, and sustainable economics—so “resilience” cannot absorb every adverse result. Belief supports patience only while dated falsifiers and conflict disclosures remain visible.

24. Valuing a Venture Capital Portfolio — marks are governance estimates, not returns

Context and argument. “Valuing a Venture Capital Portfolio,” opening and three rules describes a quarterly USV process in which every partner participates. Wilson advocates conservative marks, rapid downside adjustment using public comparables, and a backtest that M&A proceeds to USV should exceed the prior carrying value. source note

Reasoning and evidence. [investor-stated | strong] The doctrine separates information-rich company partners from finance-only marking and makes downside recognition faster than private financing cadence. [researcher inference | moderate] “Below market” can mean prudent, stale, or strategically managed; public comparables may mismatch stage, rights, growth, and liquidity. Wilson's claim that M&A proceeds have always exceeded the prior mark is unaudited and may exclude write-offs or other resolutions.

Decision implication and boundary. [researcher inference | strong] Record mark methodology, evidence date, rights/liquidity adjustments, dissent, and later realization for every material position. Interim NAV is an estimate used for reporting and governance; DPI, cash-flow timing, fees, and final net return answer different questions.

25. New Fund — Same Focus — adaptation with an explicit team byline

Context and argument. Brad Burnham's 2008 fund announcement, ¶¶1–9 records a $156 million vehicle, adds Albert Wenger as partner, retains Andrew Parker, and says a weakening economy and crowded web-services market require sharper differentiation, sustainable user growth, and selective later-stage activity. “Same focus” therefore means continuity of underlying strategy, not identical execution. source note

Evidence and implication. [contemporaneous record | strong] The byline, vehicle size, team change, and stated selection adjustments are decision-date firm evidence. They are not Wilson-authored or personal capital. [researcher inference | strong] A renewal memo should state invariants, adaptations, named decision owners, and vehicle differences, then test them against actual stage, sector, check, ownership, and concentration data. Full first-fund LP participation is an interested fundraising claim, not disclosed performance.

26. Worthless Stock Certificates — Kozmo and premature geographic scaling

Context and argument. In “Worthless Stock Certificates,” body and archived Wilson reply, Wilson names Kozmo as a total Flatiron loss. His postmortem is not that the initial New York experiment was necessarily wrong, but that investors financed expansion to roughly 18–20 cities before carefully proving repeatability in a second and third market. source note

Evidence and implication. [retrospective witness | moderate] The named failure and Wilson's admission are material, but “New York worked,” the capital raised, and the city count remain approximate and unaudited. [researcher inference | strong] Geographic expansion should use sequential cohorts for demand, contribution margin, service quality, and management capacity, while preserving board records and contrary evidence. The post does not isolate Wilson's role, the vehicle, or alternative causes such as logistics, competition, leadership, and capital-market closure.

27. The Opportunity Fund — stage extension with allocation risk

Context and argument. “The Opportunity Fund,” ¶¶1–12 describes a complement to core funds: follow established portfolio companies, enter later networks initially financed elsewhere, handle special situations, and respond to opportunities. Wilson acknowledges the risk of later-stage drift, says there was no promise to deploy fully, describes fees on invested capital, and adds John Buttrick. source note

Evidence and implication. [contemporaneous record | strong] The mandate, conflict awareness, and broad $250,000-to-$25 million capability are policy parameters, not invested distributions or returns (The Opportunity Fund, range paragraph). [researcher inference | strong] Stage-extension vehicles require precommitted eligibility, cross-fund allocation, fee, valuation, and conflict rules. The post does not disclose the LPA, final commitment, deployment, cash flows, or ceiling that would empirically keep the core strategy dominant.

28. USV 2016, LP — succession through vehicle documents and workload

Context and argument. John Buttrick's “USV 2016, LP,” ¶¶1–7 records a $175 million fifth early-stage fund, five investing partners, 20–23 planned companies, more than 60 active companies, and a key-person shift from Wilson and Burnham to Albert Wenger and Andy Weissman. Wilson, Burnham, and Buttrick would remain involved while cutting back. source note

Evidence and implication. [contemporaneous record | strong] This is concrete succession evidence because it ties leadership to a fund provision, partner set, workload, and multi-decade vehicle duration. [researcher inference | strong] Title changes are weaker than key-person clauses, lead assignments, actual board work, and legacy obligations. “Active in all decisions” does not disclose veto, voting, attendance, carry, or workload; executed LP documents and deal attribution remain required.

29. Kin — ex-ante token thesis and issuer-plan boundary

Context and argument. In “Kin,” whitepaper excerpts and closing paragraph, Wilson presents Kik's proposed general-purpose token, rewards engine, and planned move toward decentralized governance, then calls cryptocurrency an important business-model innovation. The post predates the public sale and SEC complaint, making it decision-near advocacy—but much of the substance is issuer forecast rather than observed result. source note

Counterevidence and implication. [documented behavior | strong] The SEC Kik case page, ¶¶1–3 records an unregistered-offering judgment and $5 million civil penalty paid by Kik. That is an issuer outcome, not a finding against Wilson or USV. [researcher inference | strong] A token thesis must separately precommit product use, decentralization, governance, distribution, purchaser-protection, and securities-law cases; whitepaper claims are forecasts, not adoption evidence.

30. DefendCrypto.org — advocacy under disclosed economic and governance interests

Context and argument. “DefendCrypto.org,” ¶¶1–12 criticizes regulation by enforcement, says Wilson's cooperation strategy had failed, discloses that Kik was a USV portfolio company and that he sat on its board, and promotes a crowdfunded legal challenge seeded by Kin and Wilson's personal token contribution. source note

Counterevidence and implication. [investor-stated | strong] The post establishes Wilson's policy view and interests, not governing law or a board vote. [documented behavior | strong] The SEC outcome attaches to Kik, and the case page contains no finding that Wilson or USV violated law. [researcher inference | strong] Portfolio enforcement advocacy should reproduce the authority's strongest factual and legal case, disclose interests, preserve board advice and recusals, and separate issuer defense from sector policy. The adverse judgment cuts against the proposed outcome for Kik's offering without deciding every token or later transaction.

31. An Appraisal — a firm admission that requires a measurable follow-through

Context and argument. The firm-bylined “An Appraisal,” ¶¶1–3 says USV was “part of the problem,” reports that its first sixteen years and more than 100 investments included very few self-identified Black founders, and commits to hire more Black investors, fund more Black founders, and share data. source note

Evidence and implication. [contemporaneous record | strong] The statement is an institutional admission and forward commitment, not a Wilson-only admission or a complete demographic denominator. [researcher inference | strong] Accountability requires cohort definitions and time series for sourced, screened, diligenced, funded, followed-on, hired, promoted, retained, ownership, and board roles. The source gives no baseline counts or linked follow-up dataset; a promise to disclose is not evidence that selection or allocation changed.

32. Investing at the Edge — a twenty-year umbrella that needs prospective falsifiers

Context and argument. Nick Grossman's 2024 firm retrospective, body and thesis-era bullets says USV had raised and invested 14 funds and made more than 230 investments, then traces internet applications, networks, infrastructure, decentralization, access/trust, climate, and AI into an umbrella of investing at the edge of large markets under transformative pressure. source note

Evidence and implication. [contemporaneous record | strong] The firm-level counts and authored thesis chronology are useful denominators, not Wilson's deal count or performance. [researcher inference | moderate] “Edge” can unify changing opportunities, but its breadth can also absorb almost any winner after the fact. Each use should predefine the pressure, new behavior, gatekeeper constraint, rejected comparison, disconfirming evidence, partner, and vehicle before investment.

33. Board Meeting Suggestions — shift live time from reporting to decisions

Context and argument. “Board Meeting Suggestions,” pre-read list and meeting design reproduces advice Wilson sent a founder: circulate operating, product, finance, cash, and people material two or three days ahead; spend 30–60 minutes discussing it; devote the balance to one or two strategic topics; and open and close with CEO/director executive sessions. source note

Evidence and implication. [contemporaneous record | strong] This is decision-near personal governance advice, not proof of improved board outcomes or universal USV practice. [researcher inference | strong] A modern version should confirm the pre-read was read, log decisions, dissent, owners, deadlines, attendance, and access controls, and reserve reporting time when controls are weak. CEO-only sessions can increase candor or suppress operating information; repeated outcomes must decide.

34. Reserves, Recycling, and Returns — mature doctrine with hypothetical arithmetic

Context and argument. In “Reserves, Recycling, and Returns,” ¶¶3–11, Wilson analogizes the initial check to a poker ante, says USV models company funding needs and round probabilities, describes recycling early proceeds within the same fund, and concedes that portfolio mechanics cannot create a top-decile fund without breakout companies. The $200 million/$250 million/3x/3.75x example is hypothetical arithmetic. source note

Evidence and implication. [investor-stated | strong] The 2024 essay shows continuity with the 2007 reserve model and makes legal-vehicle recycling explicit. [researcher inference | strong] Report invested capital, recycled proceeds, commitments, NAV, DPI, timing, fees, and carry separately; precommit increase, maintain, and stop criteria. Reserving “for every company” cannot mean unconditional support in a finite fund, and a higher multiple on commitment may still produce a weak net IRR.

35. Startup Mortality Rates — power-law heuristic without a defined denominator

Context and argument. “Startup Mortality Rates,” ¶¶1–8 and return bands offers Wilson's mature thirds heuristic—good investments, regretful partial outcomes, and zeros—and compares it with figures attributed to Marc Andreessen through Nate Silver. Wilson's durable claim is that a small percentage of investments produces the returns and that financing conditions affect mortality. source note

Evidence and implication. [retrospective witness | moderate] The heuristic spans roughly four decades but supplies no company list, vintage, valuation date, fund, or cash-flow table. The secondhand Andreessen bands do not literally equal thirds without regrouping. [researcher inference | strong] Define whether the unit is company, financing, or vehicle position; separate survival, loss, realized return, and current mark; and publish company-count and value-weighted distributions before invoking the power law to justify portfolio size.

36. I've Moved Onchain — corpus history and interested platform advocacy

Context and argument. “I've Moved Onchain,” ¶¶3–20 traces Typepad, avc.com, Mirror, and Paragraph/avc.xyz; says Wilson intended Paragraph to become the primary home; and advocates onchain storage, composability, and wallet-linked identity. Wilson discloses that USV invested in Mirror and Paragraph and expected to be a major shareholder after their merger. source note

Evidence and implication. [investor-stated | strong] The post is strong evidence of Wilson's stated migration and conflicts, not proof of permanence, portability, or complete historical transfer. [researcher inference | strong] Corpus research needs a canonical title-date-author index, content hashes, original/mirror/archive URLs, edits, comments, and independent retrieval tests. Hosted databases can be portable and onchain records can still depend on gateways, interfaces, keys, and indexes; platform ownership makes durability claims hypotheses, not neutral facts.

Cross-source synthesis

Public writing is part of the process, not merely publicity

[researcher inference | strong] Wilson's most distinctive documented practice is to externalize reasoning: AVC is declared a feedback channel; the venture-math sequence publishes assumptions and corrects an exit-count estimate the next day; MBA Mondays flags obsolete equity data; the Zynga post records a changed category view; and Airbnb preserves a miss. Together these are observable acts of publication and revision, not proof that every important decision used the same discipline. About, publication paragraph · venture-math original and continuation · MBA employee-equity update · Zynga · Airbnb

The useful mechanism is a dated loop: state a model, expose assumptions, receive contrary information, publish the update, and preserve the prior view. [researcher inference | moderate] The risks are performance theater, selective disclosure, and hindsight curation; internal dissent, lost memos, and decisions never blogged remain invisible. A public archive is a research asset only when paired with the complete eligible-opportunity and decision record.

The early USV system links firm design, stage, ownership, and reserves

[researcher inference | strong] The 2005–2007 essays form a coherent construction system. A small partner team seeks direct founder work and prior-relationship access; “early” means significant formation remains; smaller initial checks buy learning while targeting meaningful ownership; reserves preserve the option to concentrate after evidence; burn discipline protects the company's ability to search. formation announcement · early-stage essay · deal-size essay · reserves essay · failure essay

This system contains a real trade-off. Lower initial capital reduces loss severity and permits more uncertainty, but target ownership and meaningful influence may require higher-priced follow-ons. Reserves can exploit improved evidence or subsidize insider optimism. [researcher inference | strong] The missing control is a company-by-company ledger of initial check, milestone, follow-on choice, ownership, runway, and cash return; doctrine alone cannot distinguish option value from underfunding or sunk-cost escalation.

Networks operate at two levels: sourcing and product structure

[investor-stated | strong] In “Connectors,” Wilson treats relationships as investor infrastructure for sourcing, recruiting, introductions, and financing. [contemporaneous record | strong] In Twitter, the USV thesis, Coinbase, and Tumblr, the product itself is a network or service layer whose value emerges from participants and complementary builders. Connectors · Twitter · Investment Thesis @ USV · Coinbase · Tumblr

[researcher inference | moderate] The combined edge is not the word “network.” At the investor level, test referral conversion and useful action; at the product level, test roles, interaction frequency, retention, multihoming, governance, and marginal value with scale. Social centrality can bias sourcing, while a large user graph can lack monetization or participant power. Each level needs its own falsifiers.

Wilson's narration sits inside a team and vehicle system

The corpus names and constrains credit. Colonna sourced and led GeoCities; Wilson negotiated and structured. Burnham contributes network, early-stage, thesis, and Tumblr records. Parker built the reserve model. Wenger authored the platform thesis and is separately credited with MongoDB. Weissman authored the 2012 investment-thesis post and credits Burnham. Twitter, Zynga, and Coinbase filings show vehicle holdings and shared power rather than personal ownership. Looking Back · GeoCities · Reserves · Wenger platform thesis · Weissman investment thesis · Twitter S-1/A · Zynga S-1/A · Coinbase S-1

[researcher inference | strong] Later work should maintain separate columns for public narrator, thesis author, source, diligence lead, negotiator, board representative, decision group, legal vehicle, and economic beneficiary. “Wilson investment,” “USV investment,” and “Wilson wrote about the investment” are not synonyms.

Five tensions should remain visible

1. Founder as customer versus company governance. [investor-stated | strong] Wilson advocates founder service yet explicitly allows withholding capital, behavior change, coaching, or replacement. [researcher inference | strong] The conflict is managed through stakeholder clarity and process, not erased by customer language. customer essay
2. Focus versus adaptation. [retrospective witness | moderate] Flatiron's changes are called style drift; [contemporaneous record | strong] USV's Zynga decision records an evidence-based category update. The distinction is whether the edge and falsifier are preserved, not whether the thesis wording changes. style-drift essay · Zynga
3. Small checks versus ownership and reserves. [investor-stated | strong] Small initial exposure buys learning, while meaningful ownership and follow-ons can require much more capital. [researcher inference | moderate] Portfolio liquidity and opportunity cost decide whether the pair is coherent. deal size · reserves
4. Conviction versus falsifiability. [investor-stated | strong] The Coin Center keynote prizes belief and resilience; the math and Airbnb posts prize correction. [researcher inference | strong] Conviction is disciplined only when milestones and disconfirming evidence are written before the outcome. keynote · math continuation · Airbnb
5. Conservative marks versus realized return. [investor-stated | strong] USV describes conservative quarterly valuation and a favorable M&A backtest. [researcher inference | strong] A mark is still an estimate; only complete cash flows, fees, timing, residual value, and write-offs establish net performance. valuation essay

Later writings convert doctrine into audit tests

The 2020–2024 corpus is more institutional and self-critical. USV admits a founder-representation failure and promises data; its succession and thesis retrospectives expose firm scale and authorship; Wilson updates reserves, board process, mortality, and publishing infrastructure. [researcher inference | strong] These later works make earlier doctrine testable rather than automatically validated: service can be compared with access; focus with actual portfolio drift; reserves with cash flows; board templates with decision follow-through; open publishing with independent archive retrieval. An Appraisal · Investing at the Edge · reserves/recycling · board suggestions · migration

Kik/Kin supplies the sharpest adverse test. Wilson published a bullish token thesis, later disclosed a board seat and personal advocacy contribution, and argued against the SEC's approach; Kik subsequently paid a penalty after judgment on its unregistered offering. [researcher inference | strong] The sequence neither proves Wilson wrongdoing nor invalidates every crypto thesis. It does show why governance role, economic interest, product forecast, legal analysis, issuer outcome, and sector policy must remain separate. Kin · DefendCrypto · SEC Kik case hub

What the written record establishes by decision function

Intellectual chronology, evolution, and contradictions

The main evolution is from a Wilson-centered public voice to an increasingly explicit institution. The early blog often says “I” while describing “we”; later firm posts expose distinct authors, vehicles, key-person clauses, partner additions, and a 230-plus-investment scale. [researcher inference | strong] The doctrine grows more portable, but personal attribution becomes less safe. About · formation announcement · USV 2016 · 2024 institutional retrospective

The main unresolved contradiction is adaptiveness versus unfalsifiability. Wilson's best writings show correction—math, Zynga, Airbnb, stale MBA data—while the broad “edge” and crypto-resilience frames can absorb many outcomes. [researcher inference | strong] The solution is not a static thesis; it is dated selection criteria, explicit disconfirming evidence, rejected-deal comparisons, and outcome ledgers. math continuation · Zynga · Airbnb · MBA equity revision · edge thesis · Kin · DefendCrypto

Decision playbook and modern translation

This playbook is a researcher reconstruction, not a Wilson-authored checklist. Each rule distinguishes the historical text from the analyst's current implementation and preserves a failure test.

Rule 1 — Define stage by formation remaining and ability to matter

[researcher inference | strong] Historical rule: an investment is early when important company elements remain unformed and the investor's capital, network, and counsel can materially affect them. early-stage essay, ¶¶1–11

Diagnostics: What is missing—team, product, distribution, market, business model, or governance? What specific investor capability changes the path? What ownership and rights make that contribution rational? Kill test: if the missing work is not influenceable or the company is already priced as formed, the stage does not fit.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: formation remaining and ability to contribute. Does not transfer: 2006 check sizes, ownership norms, and financing labels. Signals: named missing functions, investor-specific help, and dated 6-/12-/18-month evidence. Likely misuse: calling any small round “early.” Analyst-created hypothetical: decline a seed-priced company whose critical regulatory approval cannot be accelerated by the investor, despite its low revenue.

Rule 2 — Finance the next learning milestone, not an undifferentiated plan

[researcher inference | strong] Historical rule: smaller initial checks buy an option on uncertain teams and markets; burn discipline preserves attempts to find the viable path. deal-size essay, ¶¶7–13 · failure essay, categories 1–2

Diagnostics: Which uncertainty will the tranche retire? Is the round large enough to produce credible evidence? What burn and runway remain if the first route fails? Kill test: do not finance activity that cannot change the investment probability.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: milestone finance and runway for search. Does not transfer: historical software-development costs or a fixed round-size ceiling. Signals: predeclared cohort, technical, regulatory, or commercial thresholds and runway after a failed route. Likely misuse: underfunding a fixed-cost proof and mistaking starvation for falsification. Analyst-created hypothetical: finance two production-scale model evaluations plus a fallback architecture, not a calendar year of undifferentiated AI spend.

Rule 3 — Reserve for evidence, not familiarity

[researcher inference | strong] Historical rule: preserve pro-rata capacity because the best later round may arrive after years of learning; model future rounds and uncertainty explicitly. 2007 reserves · 2024 reserves/recycling

Diagnostics: What changed since the prior check? How does the follow-on compare with a new investment? What are dilution, downside rights, remaining vehicle capacity, and cross-fund conflicts? Kill test: stop when the case depends on sunk cost, insider coordination, or preserving a mark rather than improved company evidence.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: evidence-weighted follow-ons and vehicle-level reserve planning. Does not transfer: Wilson's historical probability inputs, fund sizes, or any assumption that recycling is contractually permitted. Signals: forecast error, pro-rata economics, new evidence, and comparison with the best new deal. Likely misuse: defending ownership or NAV in a weak insider round. Analyst-created hypothetical: fund a down round only when retention and gross-margin evidence improve enough to beat a new-investment alternative after dilution.

Rule 4 — Treat founders as service recipients without confusing that with control entitlement

[researcher inference | strong] Historical rule: provide fast, candid, compassionate service, while recognizing that company, employees, customers, and investors can require hard intervention. customer essay, ¶¶3–16

Diagnostics: Which stakeholder interest governs the decision? What was promised before investment? Are expectations, conflicts, coaching steps, and escalation rights documented? Kill test: founder friendliness cannot justify concealing risk; fiduciary language cannot justify arbitrary treatment.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: candid service and explicit stakeholder boundaries. Does not transfer: the metaphor that an entrepreneur is literally a consumer with unilateral control; legal duties and governance vary by entity and jurisdiction. Signals: response quality, expectation records, coaching milestones, conflicts, and decision process. Likely misuse: invoking “founder friendly” to avoid oversight or “fiduciary duty” to excuse surprise. Analyst-created hypothetical: document cash milestones and CEO-support steps before a board considers leadership change.

Rule 5 — Underwrite networks at the behavior level

[researcher inference | strong] Historical rule: relationship networks create access and portfolio help; product networks create value through participant interaction and complementary builders. Connectors · Twitter · USV thesis

Diagnostics: Who are the participant roles? What repeated action creates value? Does retention, interaction density, or ecosystem contribution improve with scale? Can participants multihome or leave? Kill test: a large audience without durable interaction or participant value is not a network effect.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: behavior-level network underwriting and complementary ecosystem analysis. Does not transfer: early social-web growth proxies or an assumption that API openness alone produces defensibility. Signals: cohort retention, interaction density, contribution, multihoming, switching, and participant economics. Likely misuse: relabeling an audience, marketplace inventory, or token holders as a network effect. Analyst-created hypothetical: reject a creator platform whose user count grows while creator retention and cross-user interaction fall.

Rule 6 — Make thesis changes explicit and reversible

[researcher inference | strong] Historical rule: focus protects a proven edge, but category beliefs can change when distribution or behavior changes. style-drift post · Zynga · new-fund notice

Diagnostics: What invariant edge remains? Which observation changed? Does fund size, team, stage, or geography force the shift? Kill test: if the thesis can explain both selecting and rejecting every company, it is not a selection rule.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: explicit invariants, belief updates, and rollback conditions. Does not transfer: Flatiron's sectors, vintage, capital structure, or any inference that all adaptation is drift. Signals: versioned thesis language, pre-scored investments and rejections, stage/check mix, and exception frequency. Likely misuse: writing a broad umbrella after winners emerge. Analyst-created hypothetical: pilot three climate-infrastructure deals under a dated thesis and stop if customer adoption and fund-ownership targets both miss.

Rule 7 — Separate public narrator, partner labor, vehicle, and economics

[researcher inference | strong] Historical rule: partnership outcomes arise from complementary sourcing, negotiation, operating help, boards, and capital. Looking Back · GeoCities · USV 2016

Diagnostics: Who sourced, sponsored, diligenced, negotiated, joined the board, approved follow-ons, and managed exit? Which legal vehicle owned the position? Kill test: never use byline, board seat, or shared voting power alone as sole-deal credit.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: role decomposition and legal-vehicle attribution. Does not transfer: informal Flatiron/early-USV job divisions or an assumption that the most public partner led. Signals: dated role ledgers, investment memos, board records, filings, carry, and cash flows. Likely misuse: converting narration or a board seat into personal ownership and sole skill. Analyst-created hypothetical: credit one partner with sourcing, another with diligence, the full committee with approval, and the fund—not any individual—with the position.

Rule 8 — Turn conviction into falsifiable patience

[researcher inference | strong] Historical rule: durable technology change can require long belief, but Wilson's own best process publishes corrections and misses. Coin Center prepared text · venture math · Airbnb

Diagnostics: What must be true by what date? What product, legal, adoption, or economic observation would lower conviction? Is the investor economically or institutionally conflicted? Kill test: resilience that treats every adverse result as proof of being early is unfalsifiable.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: patient conviction paired with published disconfirming evidence. Does not transfer: 2016 crypto regulatory, custody, market-structure, or adoption assumptions. Signals: dated product use, decentralization, security, unit economics, legal analysis, and independent countercases. Likely misuse: treating enforcement, weak use, or governance concentration as proof the market is merely early. Analyst-created hypothetical: pause token deployment when promised control dispersion and non-speculative use miss two consecutive dated gates.

Rule 9 — Design board time around decisions and traceable follow-through

[researcher inference | strong] Historical rule: boards should absorb reporting in advance and use live time for consequential discussion; CEO review still centers on strategy, talent, and cash. board suggestions · CEO essay

Diagnostics: Was the pre-read complete and read? What decisions, dissent, owner, evidence, and deadline emerged? Who was excluded and what information might be lost? Kill test: strategy discussion is not superior when controls, cash, compliance, or basic reporting are unreliable.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: pre-read discipline, strategic live time, and explicit CEO/board space. Does not transfer: one meeting cadence, attendance pattern, or assumption that reporting is low value across stages and regulated contexts. Signals: read confirmation, decision latency, dissent, owner/deadline completion, risk detection, and management candor. Likely misuse: excluding operators and calling unstructured discussion strategy. Analyst-created hypothetical: in a cash crisis, devote the live meeting first to forecast integrity and financing decisions, then strategy.

Rule 10 — Keep marks, multiples, mortality, and cash returns separate

[researcher inference | strong] Historical rule: mark conservatively, recognize downside quickly, and preserve breakout exposure; portfolio mechanics cannot manufacture a winner. valuation · mortality · reserves/recycling

Diagnostics: Is the figure company count, gross multiple, NAV, DPI, TVPI, IRR, or net LP cash flow? What is realized, residual, recycled, fee-bearing, and time-weighted? Kill test: never infer performance from a survival rate, current mark, hypothetical recycling example, or one named outcome.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: conservative marks, quick downside recognition, and metric separation. Does not transfer: historical comparables, one firm's valuation policy, or Wilson's unaudited M&A backtest. Signals: company-count and value-weighted outcomes, mark-to-realization error, DPI, TVPI, IRR, recycling, fees, and timing. Likely misuse: presenting a hypothetical gross multiple or current NAV as LP return. Analyst-created hypothetical: show the same vehicle's gross MOIC, net DPI, residual NAV, and dated net IRR side by side.

Rule 11 — Convert institutional commitments into comparable data

[researcher inference | strong] Historical rule: a firm should own the hiring and investment decisions it controls and expose its progress. An Appraisal, ¶¶1–3

Diagnostics: Are cohort definitions stable? Are sourcing, diligence, investment, ownership, reserves, governance, hiring, promotion, and retention all measured? Kill test: a public commitment without comparable follow-up cannot support an improvement claim.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: responsibility for decisions within the firm's control and comparable follow-through data. Does not transfer: undefined 2020 categories, a promise as outcome evidence, or disclosure that compromises voluntary identity privacy. Signals: stable counts and rates across the full sourcing-to-governance and hiring-to-promotion funnels. Likely misuse: reporting only initial checks or hires while ignoring ownership, reserves, retention, and authority. Analyst-created hypothetical: publish annual cohort rates with suppressed small cells and unchanged definitions.

Rule 12 — Preserve the corpus independently of the platform thesis

[researcher inference | strong] Historical rule: open publication and portable identity can make ideas composable, but migration and ownership create archival conflicts. About · I've Moved Onchain

Diagnostics: Can every artifact be exported, hashed, independently retrieved, dated, and attributed? Are comments, revisions, images, and mirrors preserved? Kill test: “permanent” content that depends on one funded interface or inaccessible gateway is not independently durable.

Modern VC translation (as of 2026-08-21). [researcher inference | strong] Transfers: open feedback, exportability, and independently verifiable archives. Does not transfer: any platform's permanence claim, wallet identity as a universal moderation solution, or USV-backed interface incentives. Signals: canonical manifests, hashes, independent gateways, timestamp parity, edit history, comments, and successful restore tests. Likely misuse: counting mirrors as new works or treating onchain location as accessible preservation. Analyst-created hypothetical: rebuild a sampled year from two independent exports and compare every hash, date, author, image, and comment count.

Best writings and unresolved record

Best five

1. “Why Early Stage Venture Investments Fail”. The best combination of a bounded personal sample, causal mechanism, self-implication, and operational rule. [investor-stated | strong] Burn can remove the option to search; [retrospective witness | moderate] the 32-investment denominator remains unnamed and unaudited. source note
2. “GeoCities”, read with “Looking Back on 1999”. The richest case method: roles, syndication, partial liquidity, reported economics, and Wilson's own correction that Jerry Colonna sourced and led. [retrospective witness | moderate] It teaches attribution and risk control while leaving full cash flows unresolved. GeoCities note · credit-correction note
3. “Twitter”. The strongest compact decision-near investment rationale: emergent public/asynchronous behavior, API extensibility, and an explicitly unknown business model. [contemporaneous record | strong] It captures belief at investment time; the issuer filing prevents sole or personal credit. source note
4. The Venture Capital Math Problem sequence. The model is incomplete, but the practice is exemplary: expose assumptions, invite correction, and update with better data the next day. [documented behavior | strong] The continuation corrects exit count while preserving unresolved tail-shape questions. source note
5. “Our Customer Is the Entrepreneur”. The clearest founder-service doctrine and its sharpest internal limit: service can require refusing capital or replacing a founder when broader company interests govern. [investor-stated | strong] It is a belief and reputation strategy, not founder-experience evidence. source note

The ranking favors decision proximity, explicit mechanisms, corrections, role boundaries, and falsifiable implications—not fame or eventual company value. “An Appraisal,” Airbnb, Kozmo, Kin/DefendCrypto, and the valuation essay are indispensable counterevidence but serve the audit better than the canonical teaching set.

Unresolved record

• Complete decision denominator. [researcher inference | insufficient public record] Recover the 32 Wilson-sourced/led/managed investments named in 2007, every subsequent Wilson-attributed decision, passes, follow-ons, outcomes, and measurement dates. failure essay
• Partner and vehicle attribution. [researcher inference | insufficient public record] Map source, sponsor, diligence, negotiation, board, reserve, exit, voting, carry, and legal vehicle for each signature position; public narration and shared filing power are insufficient.
• Fund cash flows and portfolio construction. [researcher inference | insufficient public record] Reconcile initial checks, reserves, recycling, dilution, ownership, gross and net cash flows, marks, DPI, TVPI, IRR, fees, and residuals by vehicle. The public essays supply policies and hypotheticals, not a complete ledger.
• Founder-service and board outcomes. [researcher inference | insufficient public record] Compare accepted and rejected founder experience, CEO interventions, board formats, decision latency, dissent, risk detection, and implementation against the published doctrine.
• Kik/Kin governance chronology. [researcher inference | insufficient public record] Obtain board minutes, legal opinions, offering materials, Wilson votes or recusals, purchaser communications, decentralization milestones, and DefendCrypto asset disposition; preserve the issuer-only enforcement boundary. Kin · DefendCrypto · SEC case hub
• Diversity commitment follow-through. [researcher inference | insufficient public record] Locate comparable post-2020 sourcing, investment, ownership, reserve, hiring, promotion, retention, and board data, including definitions and privacy protections. An Appraisal
• MBA Mondays denominator. [researcher inference | insufficient public record] Build a deduplicated manifest of originals, revisions, reruns, dates, and authors; this analysis remains partial at the series level and names every sampled post. MBA archive
• AVC archive integrity. [researcher inference | insufficient public record] Hash and independently retrieve Typepad, WordPress, Mirror, Paragraph, and Arweave versions, with images, edits, comments, timestamps, cross-posts, and author records. migration essay
• Prospective thesis falsifiability. [researcher inference | insufficient public record] Recover dated investment and rejection criteria for the successive network, infrastructure, crypto, climate, AI, and “edge” frames, then compare selected and rejected companies. 2024 thesis retrospective

Coverage and evidence base

The priority spoken corpus contains seven source routes and seven underlying appearances: priority coverage is 7/7 appearances (100%). All seven have source notes: five full-access artifacts and two partial-access artifacts. The full set is the 2009 MeetInnovators transcript, 2010 Mixergy transcript, 2014 Business Insider/AVC transcript, 2015 Reboot interview (transcript rereleased in 2020), 2016 Coin Center prepared keynote, 2018 a16z transcript, and 2026 USV oral history. Reboot's 2020 PDF is a rerelease of the 2015 appearance, not an eighth work. Coin Center is one TALKS priority appearance; its separate w-coin-center-keynote.md treatment as a WRITINGS supplement reuses the same underlying prepared text and does not create a second work.

“Full” means the mapped text artifact was read end to end; it does not mean the underlying recording was independently voice-verified. Mixergy is partial because the page documents failed video, transmission loss, and tape cutouts. The 2026 USV episode is partial because the official chapter map and metadata were accessible but the recording/transcript could not be voice-verified. The 2016 Coin Center artifact gives the complete prepared text but not proof that delivery matched it word for word. The 2018 a16z page exposes a full transcript, but speaker-label anomalies prevent fine-grained Wilson attribution without audio review.

One additional note, Risky Business with Fred Wilson, is a supplement, not part of the priority denominator: the edited transcript has no date, recording, or publication history. It is used only where it sharpens a retrieval question or corroborates a recurring retrospective story.

Source-by-source analysis

2009-11-19 — MeetInnovators interview

[contemporaneous record | strong] MeetInnovators' 13-page transcript captures Wilson before the largest later public outcomes could dominate his account. He defines USV's target as actively used web services, explains AVC as open-ended thinking and reputation building, describes founder relationships before financing, maps partner skills, and names Huffington Post and Gilt Groupe as misses (PDF pp. 1–5, 8–13).

• [investor-stated | strong] Framework: public thinking and product use are not ancillary marketing; they help the investor form a view, attract founders, and build the trust required to win an allocation (PDF pp. 1–5).
• [investor-stated | strong] Decision evidence: Boxee moved from an initially unattractive hardware pitch to an open-source software product and later financing after roughly two years of contact (PDF pp. 8–10). The account shows changed evidence, not a rule that persistence alone merits investment.
• [retrospective witness | strong] Miss evidence: Wilson attributes the Huffington Post pass to misreading a political-blog wedge and the Gilt Groupe pass to e-commerce bias (PDF p. 13). These are later diagnoses, not a complete decision-time record.
• [researcher inference | strong] Attribution boundary: Wilson assigns himself and Albert Wenger financial skill and Brad Burnham market/strategy strength (PDF pp. 11–12); portfolio work should therefore be decomposed by role rather than credited to the public spokesman.

2010-04-12 — Mixergy, The Biography of Union Square Ventures

[researcher inference | strong] The Mixergy interview is the most detailed early spoken institutional history. Wilson connects Flatiron's failure to overfunding, price discipline, and organizational growth; explains choosing Burnham and later Wenger; reconstructs the first USV thesis and fund; and treats blogging, founder meetings, and long relationship cultivation as the firm's distribution system (opening Flatiron exchange through the sourcing/Crowley sections).

• [retrospective witness | moderate] Flatiron lesson: scaling from two people to 25 separated experienced investors from new decisions and encouraged an extrapolation from early success that Wilson says did not hold (Mixergy, opening “what did you…screw up?” exchange). His reported $150 million/$750 million illustration in the same exchange has no defined realization or LP-return basis.
• [retrospective witness | strong] Team credit: Wenger had worked with Burnham, served as interim crisis CTO at Tacoda, found Etsy, and took the first USV Etsy board seat; Burnham and Wilson had worked together for more than a decade before founding USV (“Albert, how did you decide to partner up with him?” and preceding Burnham exchange).
• [investor-stated | strong] Institutional design: every partner sourced and supported companies inside a deliberately small organization, while a rigorous internal thesis was simplified into a six-slide LP explanation (partner-responsibility and thesis exchanges before/around 25:00).
• [researcher inference | strong] Conflict: the transcript says the first fund was $100 million; Wilson's first-party 2007 reserves essay says $125 million. This could be rounding, target size, or transcription error and should not be silently harmonized.
• [researcher inference | strong] Access boundary: missing recording segments around the anxiety discussion make the transcript partial even though the hosted text spans the interview (around 39:34 and labeled tape cutouts).

2014-04-14 — The Business Insider Interview

[researcher inference | strong] The full lightly edited Q&A repost is the strongest spoken team-attribution and midcareer reset source. Wilson says Wenger did MongoDB and found Etsy, Burnham deserved equal USV credit, and newer partners carried more new investment activity while incumbents held legacy boards (“tell me about Albert Wenger” through the Andy Weissman/fund-count exchange).

• [retrospective witness | strong] Attribution: “Albert did the MongoDB deal” is direct internal testimony. MongoDB's S-1 separately establishes USV vehicle ownership and shared power—not personal Wilson ownership or sole authority.
• [investor-stated | strong] Thesis evolution: Wilson describes horizontal social networks in 2004, vertical networks in 2008, and transactional marketplaces in 2012, while admitting that “been-there-done-that” preference can pull investors toward novelty (“what is USV's investment thesis?”).
• [retrospective witness | strong] Failure diagnosis: Turntable's aggregate growth obscured rapid cohort churn, and the company failed to build a less demanding passive mode in time (“investing in…Turntable has to be hard”). The lesson is data reaction speed, not merely that consumer products are fickle.
• [investor-stated | strong] Portfolio obligation: Wilson says a financing promise includes work through shutdown or sale, even when the company will not be a large outcome (“how do you deal with feeling uninspired?”).
• [retrospective witness | moderate] Crypto rationale: Wilson and Wenger studied Bitcoin for about a year, and Wilson pursued Coinbase after the earlier seed round was too small for USV (“discovering Bitcoin…why?”). The Coinbase S-1 verifies Wilson's board tenure and shared USV vehicle power, not personal economics.
• [researcher inference | strong] Editorial boundary: Wilson says the interview began as background and that he requested one unspecified fairness edit before publication (introductory paragraphs). It is full-access direct voice, not an untouched recording transcript.

2015-01-13 — Reboot, Investing in Trust; transcript rereleased 2020-02-14

[researcher inference | strong] The 22-page Reboot transcript places Wilson alongside former Flatiron partner Jerry Colonna. It is the deepest account of founder removal, board duty, trust, and Wilson's own fear at Flatiron's founding. The 2020 file is a transcript rerelease; the underlying conversation remains the original 2015 episode.

• [retrospective witness | moderate] Career loop: Wilson presents Euclid, Flatiron, and USV as successive corrections—learning what failed, trying a new structure, then codifying what survived (PDF pp. 4–5). This is interpretive memory, not a contemporaneous firm plan.
• [retrospective witness | moderate] Governance failure: in the StarMedia case he says the board stayed loyal to the founder too long and failed its company/stakeholder duty (PDF pp. 5–7). No board minutes or capitalization history are available in the processed corpus.
• [investor-stated | strong] Trust rule: an investor earns trust through useful action and listening, not reputation or promises; Wilson uses hands-on work, long board relationships, leadership changes, and regulatory help as examples (PDF pp. 8–16).
• [investor-stated | strong] Board rule: leadership fit changes by stage; boards should identify current needs, complement a CEO's weaknesses, and balance founder, investor, and independent directors rather than default to investor control (PDF pp. 11–13, 17–19).
• [retrospective witness | strong] GeoCities credit: the conversation's financing-role map is consistent with Wilson's written Jerry Colonna correction: Colonna brought/led the Flatiron opportunity, while Wilson and other syndicate participants had distinct roles. Repetition across the same interested participants supports consistency, not independence.

2016-05-04 — Coin Center keynote

[contemporaneous record | strong] Wilson says the posted keynote is the prepared text he delivered at Coin Center's New York gala. It states an internet-to-blockchain belief system—openness, decentralization, resilience, and absence of single-entity control—while admitting that no blockbuster blockchain company or clearly transformative application was then visible (speech ¶¶2–10).

• [investor-stated | strong] Thesis: protocol resilience and permissionless innovation can create durable application opportunities, but Wilson explicitly concedes uncertain timing and weak application proof as of 2016 (speech ¶¶8–10, 13–27).
• [investor-stated | strong] Team credit: he credits Rikki Tahta with introducing Bitcoin and says the entire USV investment team worked on the sector (speech ¶¶6–7). “We” cannot be collapsed into Wilson-only sourcing.
• [researcher inference | strong] Advocacy boundary: the event audience and belief-language make this a high-value conviction artifact but not a neutral empirical study, legal opinion, or company/fund return record.
• [researcher inference | moderate] Prepared-versus-spoken boundary: the complete intended text is accessible; no recording establishes exact delivery, deviations, or audience questions.

2018-09-23 — a16z, Technological Trends, Financial Capital, and the Dynamics of Disruption

[contemporaneous record | moderate] The a16z transcript and show notes broaden the frame from portfolio craft to technological cycles: innovation versus disruption, software plasticity, incumbent data advantage, cloud/SaaS economics, and the infrastructure prerequisites for crypto applications (show-note sections 01:20–49:30).

• [contemporaneous record | moderate] Framework: abundant innovation does not guarantee incumbent displacement; data, distribution, capital structure, and business-model changes determine whether a new technology becomes disruptive (transcript sections after 01:20 and 17:31).
• [contemporaneous record | moderate] Timing test: the crypto discussion names trust, developer experience, decentralization, security, and throughput as prerequisites rather than treating price alone as adoption (section after 49:30).
• [researcher inference | strong] Attribution boundary: the displayed transcript conflicts with the page summary on speaker assignment in a central exchange. Until the audio is voice-checked, the ideas are conversation-level evidence and unsafe as Wilson-only quotations.
• [researcher inference | strong] Counterargument: software's flexibility can enlarge markets, but it can also strengthen incumbents through data, compute, distribution, and switching costs; “innovation” must therefore be tested against a concrete displacement mechanism.

2026-04-29 — USV, Fred Wilson on 40 Years in Venture—and Why USV Is Automating Itself

[contemporaneous record | strong] The official Apple episode page verifies a 65-minute USV conversation between Wilson and Michael Mignano and provides a detailed chapter map. It routes career history, Flatiron, AVC, Twitter, Brad Burnham and the original thesis, Coinbase, open-source/decentralized AI, firm automation, risk, and price regrets (06:01–1:00:48).

• [contemporaneous record | strong] Recent-role evidence: USV presents Wilson as an active general partner speaking about four decades in venture (episode description and Information fields).
• [contemporaneous record | strong] Retrieval map: the official chapters identify exact audio targets for biography (06:01–21:46), partner/thesis credit (27:27–29:19), Coinbase and AI (31:29–36:55), automation (42:37–47:00), and price/career advice (58:10–1:00:48).
• [researcher inference | strong] Access boundary: these are producer-written labels, not voice-verified quotations. They establish subject matter, not Wilson's exact claim, causal reasoning, or supporting evidence.
• [researcher inference | strong] Automation boundary: a chapter label about agents replacing partners cannot establish legal authority, fiduciary delegation, job elimination, or implemented workflows without the recording and operating evidence.

Recurring frameworks

1. Invest in participatory networks, then test the mechanism

• [investor-stated | strong] In 2009, Wilson describes actively used web services as the target; in 2014, he names an evolution from horizontal social networks to vertical networks and transactional marketplaces (MeetInnovators, PDF p. 1; Business Insider, thesis exchange).
• [researcher inference | strong] The reusable diagnostic is not the word “network.” Identify participants, repeated action, contribution incentives, retention, cross-side liquidity, marginal acquisition cost, governance, and the mechanism by which scale improves utility.
• [researcher inference | moderate] The failure boundary is category abstraction: a weak product can be described as a network, and mature platforms can be mistaken for exhausted categories. Wilson's own missed Huffington Post and Gilt Groupe accounts show how labels can hide wedges.

2. Relationships precede transactions

• [investor-stated | strong] Boxee, Foursquare, Disqus, blogging, and repeated founder meetings all support a multi-year relationship model rather than pitch-day selection (MeetInnovators, PDF pp. 4–5, 8–10; Mixergy, sourcing and Crowley exchanges).
• [researcher inference | moderate] Relationship duration is an information channel, not a quality score. Analysts should record what changed—product, usage, team, market, or terms—and compare relationship-sourced decisions with cold inbound/outbound decisions.

3. Venture is a service and fiduciary business

• [investor-stated | strong] Across 2009–2015, Wilson describes entrepreneurs as primary value creators, venture as support through success and failure, and directors as owing duties to the company rather than to founder loyalty (MeetInnovators, PDF pp. 4–5; Business Insider, failure-work section; Reboot, PDF pp. 5–8).
• [researcher inference | strong] The test is behavioral: response during a financing gap, leadership conflict, shutdown, or regulatory crisis. Public founder-friendly language without dated actions is not evidence.

4. Small teams and fund size protect decision quality

• [retrospective witness | moderate] Flatiron's growth separated senior judgment from new deployment; USV responded with a small partner model, bounded fund size, and partner renewal rather than a large hierarchy (Mixergy, Flatiron and partner-responsibility exchanges).
• [investor-stated | moderate] The 2014 succession account adds a capacity mechanism: new partners can lead more decisions while incumbents carry legacy boards (Business Insider, Weissman/fund-count exchange).
• [researcher inference | strong] The diagnostic is workload and authority, not headcount alone: boards per partner, new decisions per partner, reserve decisions, response time, and whether experienced decision-makers still own selection.

5. Treat failure as evidence, not an exception

• [retrospective witness | strong] Wilson names misclassification (Huffington Post/Gilt), delayed reaction to cohort churn (Turntable), overexpansion (Flatiron), and delayed leadership change (StarMedia) as distinct failure modes (MeetInnovators, PDF p. 13; Business Insider, Turntable exchange; Reboot, PDF pp. 5–7).
• [researcher inference | strong] A useful postmortem names the decision-date signal, ignored evidence, delayed action, and changed process. A broad “mistakes were made” story cannot be tested.

6. Separate protocol conviction from application value capture

• [investor-stated | strong] The 2016 keynote presents open, decentralized infrastructure as resilient while conceding the lack of a blockbuster application and uncertain timing (Coin Center keynote, speech ¶¶8–10, 13–27).
• [contemporaneous record | moderate] The 2018 conversation adds infrastructure prerequisites such as security, developer experience, throughput, and trust (a16z, section after 49:30).
• [researcher inference | strong] Protocol persistence, asset price, application retention, company value capture, and venture cash return are different variables. No one substitutes for the others.

Evolution and biographical turns

• 2009 — public thinker and relationship investor. [investor-stated | strong] Wilson presents AVC, product use, founder contact, and network reputation as an integrated sourcing and learning system; he is already willing to name major passes (MeetInnovators, PDF pp. 1–5, 8–13).
• 2010 — institutional design from Flatiron's failure. [retrospective witness | moderate] He links fund/team architecture to investment quality, identifies Burnham and Wenger's distinct roles, and frames USV as a deliberate reaction to Flatiron (Mixergy, opening through partner/thesis exchanges).
• 2014 — midcareer capacity reset. [investor-stated | strong] Wilson discloses a poor 2012, names failed companies, describes partner renewal, and transfers public credit to Burnham and Wenger while identifying Bitcoin as renewed intellectual energy (Business Insider, partner, 2012, and Bitcoin exchanges).
• 2015 — governance becomes the central moral language. [retrospective witness | strong] The lesson moves from “serve entrepreneurs” to the harder conflict between founder trust and company/stakeholder duty, using StarMedia and leadership transitions (Reboot, PDF pp. 5–13).
• 2016–2018 — portfolio craft expands into system belief. [investor-stated | strong] Wilson argues for decentralized infrastructure while conceding missing applications; the later cross-firm conversation introduces disruption and enabling-condition tests (Coin Center keynote; a16z transcript).
• 2026 — automation, AI, and price regret become the advertised frontier. [contemporaneous record | moderate] The chapter map shows those themes, but exact evolution cannot be claimed until the audio is recovered and voice-checked (USV episode, 33:18–47:00 and 58:10–1:00:48).

Deal rationales and attribution ledger

Recollection versus contemporaneous record

• [researcher inference | strong] Flatiron scale and economics: Mixergy's recollected $100 million USV first fund conflicts with the contemporaneous first-party $125 million figure in Reserves. The Flatiron $150 million/$750 million scaling example is undefined and cannot be treated as net return. Later memory must not replace vehicle documents.
• [researcher inference | strong] GeoCities: Reboot's team story and Wilson's written 2009 correction align on Colonna's causal role, but both are retrospective and connected to the same principals. Yahoo's GeoCities acquisition filing controls transaction timing/consideration, not Flatiron role or return.
• [researcher inference | strong] MongoDB: the 2014 transcript is the clearest role statement—Wenger did the deal. The issuer S-1 establishes a USV vehicle position and shared power. Role testimony and ownership evidence answer different questions and are complementary, not interchangeable.
• [researcher inference | strong] Twitter, Coinbase, and Tumblr: Wilson's spoken rationales establish public sponsorship and later interpretation. The relevant issuer filings establish vehicles, governance, and transaction facts; Burnham's Tumblr post preserves partner authorship. None permits a Wilson-personal return or sole-decision claim.
• [researcher inference | strong] StarMedia and Turntable: these are unusually candid failure stories but remain participant recollection. StarMedia needs board/capitalization records; Turntable needs cohort data and decision chronology before the causal claims can be tested.
• [researcher inference | strong] 2026 oral history: official chapter labels are useful retrieval coordinates only. Exact claims about automated partners, AI, risk, or price regret remain unprocessed audio, not Wilson quotation evidence.

Listen to these five first

1. [researcher inference | strong] Reboot — Investing in Trust (2015; transcript rereleased 2020). Best for board duty, founder removal, trust earned through action, and Wilson/Colonna's candid Flatiron-era self-critique; start with PDF pp. 4–19.
2. [researcher inference | strong] Business Insider/AVC — full Q&A (2014). Best for partner attribution, succession mechanics, a named bad year, Turntable's cohort failure, and the Wilson/Wenger Coinbase path; read the Wenger, Weissman, 2012, Turntable, Tumblr, and Bitcoin question blocks.
3. [researcher inference | strong] Mixergy — The Biography of Union Square Ventures (2010). Best for Flatiron-to-USV institutional design, thesis construction, partner selection, sourcing, and anxiety; use the opening through 39:34 with the transcript-cutout warning and $100 million/$125 million conflict.
4. [researcher inference | strong] MeetInnovators (2009). Best early snapshot before later public outcomes: blogging as cognition/distribution, relationship-before-money, partner skills, and candid Huffington Post/Gilt misses; read PDF pp. 1–5 and 8–13.
5. [researcher inference | strong] Coin Center keynote (2016). Best statement of Wilson's protocol belief system because it also records adverse evidence and timing uncertainty (speech ¶¶2–10, 13–27); treat it as prepared text and advocacy, not a voice-verified event transcript or proof of returns.

[researcher inference | strong] The 2018 a16z appearance ranks just outside the five because speaker-label conflicts block person-level attribution, although its innovation/disruption and crypto-prerequisite discussion is analytically valuable. The 2026 USV episode may ultimately rank highly, but only its editorial chapter map—not the spoken content—was verifiable from the accessible page.

Unresolved spoken record

• Recover and voice-check the 2018 a16z audio, especially the turns where transcript and page summary conflict.
• Obtain an accessible recording or transcript for the 2026 USV oral history and transcribe the chapters at 27:27–36:55, 42:37–47:00, and 58:10–1:00:48 before quoting or asserting evolution.
• Locate any surviving full Mixergy recording/video or alternate transcript to restore the documented cutouts and resolve obvious name errors.
• Locate a recording of the Coin Center keynote to compare delivered remarks and questions with the prepared text.
• Date and recover the recording/editorial history for the supplementary Startups.com transcript before using its deal multiples or chronology.

Model 1: Build domain expertise before committing capital

• Rule: [researcher inference | moderate] Enter a field only after the investor or a clearly identified partner can explain its technical or market structure, name the participants and failure modes, and show a repeatable information advantage. Domain knowledge may come from operating, transactions, product use, or disciplined study; a celebrated outcome is not proof of expertise.
• Evidence status: [investor-stated | moderate] Wilson prescribes deep operating knowledge and a network in one field, but his own path relied on transaction apprenticeship, timing, and a newly forming internet domain. [retrospective witness | strong] His accounts consistently credit complementary partners and later joint study. [researcher inference | insufficient public record] No matched set compares Wilson/USV decisions inside and outside a defined circle of competence.
• Provenance and origin: [retrospective witness | strong] Wilson dates the model to Euclid, where a decade of sourcing, structuring, board, management, and exit work preceded his 1994–1995 internet focus; he later warns that his timing-dependent route should not be universalized. I Got Lucky, paragraphs 2–14 His concise origin claim—“Timing is everything in life”—is a humility boundary, not the model itself. I Got Lucky, paragraph 9
• Mechanism and reasoning: [researcher inference | moderate] Expertise improves the investor's ability to notice a discontinuity, ask specific diligence questions, judge whether a founder's evidence is unusual, and recruit the right complementary specialist. The causal chain assumes the field has learnable structure, the knowledge is current, and access is not mistaken for judgment. Wilson's later account of a year studying Bitcoin with Albert Wenger illustrates deliberate preparation, while his credit to Wenger for MongoDB shows expertise can reside with another partner. Business Insider interview, Wenger and Bitcoin exchanges
• Historical and vehicle context: [retrospective witness | moderate] Euclid's broad life-sciences and IT remit gave Wilson craft but no obvious sector edge; the early commercial internet reduced incumbent expertise and created a New York opportunity that Wilson and Jerry Colonna approached with complementary transaction, operating, media, and network skills. I Got Lucky, paragraphs 4–9 [researcher inference | strong] A partner-led fund can distribute expertise across people, but only if decision credit and workload are explicit; shared vehicle ownership does not make every partner equally expert in every deal.
• When to use: [researcher inference | strong] Use before opening a new sector, hiring a specialist, or treating a thematic insight as an investable company thesis. Do not use “domain expertise” as a credential screen that excludes new entrants, as a license to ignore price and terms, or where the field is changing faster than prior patterns remain predictive.
• How to apply it: [researcher inference | strong] (1) Write the system map, vocabulary, participants, value flows, incumbent constraints, and regulation. (2) Name which partner owns the knowledge and which gaps require external diligence. (3) Record decision-date predictions and disconfirming observations. (4) Compare at least one company, one pass, and the do-nothing alternative. (5) Ask: what can this team see that a competent generalist cannot; what evidence would prove the expertise stale; who has authority if experts disagree? Kill criteria: no differentiated questions, no independent references, no identified expert owner, or a thesis that rests only on reputation.
• Examples:
  • Documented formation — internet focus. [retrospective witness | moderate] Wilson says the internet let him establish a focus after a mediocre first decade, while Colonna supplied complementary operating/media knowledge and relationships. The favorable 1996–2000 regime and undefined economics prevent attributing results to expertise alone. I Got Lucky, paragraphs 7–10
  • Documented study — Coinbase. [retrospective witness | strong] Wilson says he and Wenger studied Bitcoin for about a year, then Wilson pursued Coinbase after its seed round was too small for USV. The account supports preparation and role separation; the filing supports later board/vehicle facts, not causal skill. Business Insider interview, Bitcoin exchange Coinbase S-1, director biography and principal-stockholder footnotes
  • Named-partner application — MongoDB. [retrospective witness | strong] Wilson explicitly says Wenger did the deal and connects Wenger's computer-science background to the database founders. This supports partner-specific expertise and directly rejects a Wilson-only deal narrative. Business Insider interview, Wenger exchange
  • Boundary case — Wilson's own entry. [retrospective witness | strong] Wilson entered venture from software work and business school without a substantive operating leadership role; a decade of apprenticeship preceded sector focus. This contradicts any rule requiring founder experience while supporting accumulated craft. I Got Lucky, paragraphs 2–7 and 11–14
• Evidence: [researcher inference | moderate] The career essay establishes the prescription and biographical counterexample; the 2014 interview supplies independent deal-role and Bitcoin-study testimony; Coinbase and MongoDB filings constrain ownership and governance inference. I Got Lucky, paragraphs 2–14 Business Insider interview, Wenger and Bitcoin exchanges MongoDB S-1, principal-stockholder footnote The record supports a reusable preparation rule, not its hit-rate or return contribution.
• Limits, failure modes, and exception ledger: [researcher inference | strong] Expertise can become rigidity, prestige can masquerade as knowledge, and a sector network can narrow sourcing. New domains sometimes reward first-principles newcomers; Wilson's internet timing is such a boundary. Falsifiers: expert-led investments do not outperform comparable generalist decisions; decision memos contain no differentiated evidence; or prior patterns repeatedly miss platform/regulatory shifts. Record exception: no Euclid roster, contemporaneous 1994–1995 thesis memo, complete sector opportunity set, or partner-by-partner return ledger was recovered; the SCOUT archive, four long-form interviews, firm writings, issuer filings, and named-deal routes were searched without producing those controls.
• Evolution and contradictions: [researcher inference | moderate] The doctrine moves from Wilson's retrospective “learn one field” prescription to an institutional form in which Burnham, Wenger, Weissman, and later partners carry distinct expertise. Wilson's high public profile conflicts with that distributed reality; his explicit MongoDB/Etsy credit corrections partly resolve it. Business Insider interview, Wenger and Burnham exchanges USV 2016, LP, key-person and active-decision paragraphs
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: require a named knowledge owner, system map, dated predictions, and independent references. Does not transfer: a fixed operating-background requirement or 1990s internet patterns. Changed assumptions (analyst-created scenario): Assume AI, crypto, climate, health, and regulated infrastructure combine technical, policy, and capital expertise that rarely resides in one partner. Observable signals: differentiated diligence, correctable predictions, expert disagreement logged, and decisions that change with new evidence. Likely misuse: calling personal brand or deal access expertise. Analyst-created hypothetical: an AI-infrastructure IC assigns model architecture, buyer economics, security, and power constraints to separate owners, then declines if no one can reconcile them.

Model 2: Use relationships as longitudinal evidence, not as a quality score

• Rule: [researcher inference | strong] Build relationships before a financing need, but invest only when the longitudinal record shows decision-relevant change in product, team, market, or behavior. Familiarity creates an observation channel; it does not substitute for comparison, references, price, or a falsifiable thesis.
• Evidence status: [investor-stated | strong] Wilson repeatedly describes useful introductions, product use, blogging, and multi-year founder contact as sourcing infrastructure. [retrospective witness | moderate] Boxee and Foursquare connect long observation to eventual investment. [researcher inference | moderate] The combined “what changed?” procedure is reconstructed; no relationship-sourced versus cold-deal denominator exists.
• Provenance and origin: [investor-stated | strong] The 2005 formation post joins trusted prior relationships with an open public channel, and the 2006 “Connectors” essay says selective introductions must create mutual value—“You have to think about the win/win.” USV formation, sourcing and operating-design paragraphs Connectors, paragraphs 7–13
• Mechanism and reasoning: [researcher inference | strong] Repeated contact exposes how founders process feedback, recruit, pivot, ship, and earn trust. Product use gives the investor direct behavioral evidence; mutually useful introductions test judgment and deepen the information network. The assumptions are that observations are recorded, alternative opportunities remain visible, and social affinity is not mistaken for improvement. MeetInnovators transcript, PDF pages 4–5 and 8–10
• Historical and vehicle context: [retrospective witness | moderate] Early USV operated with few partners and a public blog during the rise of New York web services, lowering the cost of repeated founder contact. Wilson later described blogging as both cognition and relationship distribution, while a small team retained direct portfolio work. Mixergy transcript, blogging, Crowley, and partner-role exchanges [researcher inference | strong] This channel depends on partner time and reputation; it becomes harder to interpret when legacy boards, platform followers, and larger firm funnels expand.
• When to use: [researcher inference | strong] Use when product or founder evolution can be observed before a round, especially seed, developer, creator, marketplace, and community products. Do not use when regulatory, technical, or financing gates require independent expert proof; when the relationship creates a conflict; or when access to a round is being mistaken for investment quality.
• How to apply it: [researcher inference | strong] (1) Date the first contact and each meaningful observation. (2) Record what changed in product, cohort behavior, team, distribution, or terms. (3) Compare the founder with cold-sourced and partner-sourced alternatives. (4) Seek disconfirming references outside the relationship. (5) Ask: did the founder update faster than the market; did the investor contribute specific value; would the same evidence persuade us without familiarity? Kill criteria: unchanged evidence, social obligation, reference homogeneity, or a decision justified mainly by years known.
• Examples:
  • Documented application — Boxee. [retrospective witness | moderate] Wilson says he passed on a hardware pitch, stayed in contact for about two years, used the software product, and then invested after the pivot. The later Samsung acquisition has no disclosed fund economics. MeetInnovators transcript, PDF pages 8–10 USV directory, Boxee entry
  • Documented observation with attribution boundary — Foursquare. [retrospective witness | moderate] Wilson describes observing Dennis Crowley from Dodgeball onward, but later assigns the Foursquare deal to Wenger. Relationship evidence and deal leadership are different facts. Mixergy transcript, Crowley exchange Business Insider interview, Wenger portfolio exchange
  • Researcher-applied miss — Airbnb. [researcher inference | strong] The consented 2009 emails show debate, a founder meeting, market-size skepticism, and continued data gathering; Wilson later says USV liked the team but failed to chase. The failure was not lack of contact, but underweighting founder expansion and the broader market. Paul Graham, “Subject: Airbnb,” emails dated 2009-01-23–02-21 Airbnb, retrospective paragraphs 1–11
  • Researcher-applied update — Zynga. [researcher inference | moderate] Six months of Mark Pincus demonstrating a business across a new social platform changed Wilson's prior. The public post supports belief updating, not the full relationship, diligence, or sole decision. Zynga Game Network, paragraphs 1–8
• Evidence: [researcher inference | strong] “Connectors” supplies the mutual-value mechanism; the formation post and two interviews document public and multi-year sourcing channels; the Airbnb emails supply rare decision-time evidence; Boxee, Foursquare, and Zynga provide distinct outcomes and boundaries. Connectors, paragraphs 7–13 USV formation MeetInnovators transcript, PDF pages 4–10 Paul Graham, Airbnb email thread
• Limits, failure modes, and exception ledger: [researcher inference | strong] Familiarity bias, homophily, reciprocity, public-personality selection, and insider access can corrupt the signal. Airbnb shows that contact does not guarantee correct framing; Foursquare shows that the relationship narrator may not be the deal lead. Falsifiers: relationship length has no relation to observed improvement; cold alternatives repeatedly show better evidence; or decisions remain unchanged after adverse references. Record exception: no full sourcing funnel, referral conversion, pass log, relationship-duration cohort, or matched return data was recovered after the AVC/USV corpus, interviews, directory, and issuer routes were processed.
• Evolution and contradictions: [researcher inference | moderate] The early doctrine treats a connector network as an edge; later evidence makes it a fallible input. Wilson's misses at Airbnb, Huffington Post, and Gilt show that a network can deliver the opportunity while category framing still fails. MeetInnovators transcript, PDF page 13 The institutional evolution toward more partners also separates Wilson's relationships from the firm's actual investment ownership.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: maintain a dated observation log and use relationships for richer evidence. Does not transfer: blog engagement, years known, or warm access as quality proxies. Changed assumptions (analyst-created scenario): Assume online communities, scout networks, founder databases, and AI-assisted sourcing multiply weak ties and selection bias. Observable signals: rate and quality of founder updates, retained product use, independent references, and relative funnel conversion. Likely misuse: rewarding social proximity. Analyst-created hypothetical: a partner who has known a founder for five years must present the strongest cold alternative and the specific evidence that changed since the prior pass.

Model 3: Define stage by formation remaining and the investor’s ability to matter

• Rule: [researcher inference | strong] Classify a company as early only when important elements remain unformed and the investor can name how capital, counsel, relationships, or governance will materially improve that formation. Round label, revenue, and age are secondary; ownership and price must still make the work rational.
• Evidence status: [investor-stated | strong] Wilson defines stage by business formation remaining, thesis fit, and USV's ability to participate materially. [contemporaneous record | moderate] Delicious and Oddcast were contemporaneous examples. [researcher inference | insufficient public record] The record lacks the complete stage-classification scorecard and outcomes needed to test efficacy.
• Provenance and origin: [investor-stated | strong] The earliest explicit articulation is the 2006 “Early Stage Investing” essay: “Most of our investment approach is built on top of our foundation as early stage investors.” Wilson credits earlier Burnham thesis work, limiting sole authorship. Early Stage Investing, paragraphs 1–3 and 9–11
• Mechanism and reasoning: [researcher inference | strong] Formation remaining creates learning and value-creation opportunities; the ability to matter links stage to investor capability rather than company immaturity. The model assumes missing functions are influenceable, the founder wants the help, the investor can secure enough ownership/rights, and formation risk is not already fully priced. Early Stage Investing, Delicious and Oddcast examples, paragraphs 4–10
• Historical and vehicle context: [contemporaneous record | strong] In 2004–2007, cheaper web infrastructure let USV make modest initial checks while seeking meaningful ownership and partner involvement; those dollar ranges and ownership norms are historical. Deal Size, paragraphs 1–13 New Fund — Same Focus, opening and market-context paragraphs [researcher inference | strong] Larger later vehicles, climate, health, tokens, and opportunity investing weaken any simple equivalence between USV and this original functional stage.
• When to use: [researcher inference | strong] Use at initial screening and when a nominal seed, Series A, or later round contains formation risk in team, product, distribution, business model, market, or governance. Do not use when the critical gate is outside the investor's influence, when the company is priced as fully formed, or when “help” is an excuse for operational overreach.
• How to apply it: [researcher inference | strong] (1) List what remains unformed. (2) Separate founder work from investor-controllable help. (3) Name the relevant partner capability and time budget. (4) Define the evidence expected in 6, 12, and 18 months. (5) Model ownership, dilution, rights, and reserves. (6) Ask: if our help disappeared, would the thesis change; can the company absorb it; is this formation or repair? Kill criteria: no material missing element, no credible contribution, misaligned governance, or price/ownership that makes the promised work uneconomic.
• Examples:
  • Documented application — Delicious. [investor-stated | strong] Wilson describes passionate users and a founder but no company, team, office, partnerships, or business model; USV's role was to help form those missing pieces. The sale interrupted the independent-company test. Early Stage Investing, paragraphs 4–5
  • Documented boundary — Oddcast. [investor-stated | moderate] A revenue-producing company could still be early for the new social/mobile avatar opportunity being financed. The source establishes the classification logic, not the outcome or return. Early Stage Investing, paragraphs 6–8
  • Researcher-applied illustration — Twitter. [researcher inference | strong] Decision-time evidence included emergent user behavior, an open API, unknown monetization, and reliability needs; it fits formation remaining without implying Wilson alone made the decision. Twitter, investment-rationale and business-model paragraphs Twitter S-1/A, principal-stockholder footnote 16
  • Boundary failure — Airbnb pass. [researcher inference | moderate] USV saw a small current wedge and an incomplete market, but failed to treat founder adaptability and adjacent accommodation as formation potential. Price, access, and partner factors remain unknown. Airbnb, paragraphs 1–11 Paul Graham, “Subject: Airbnb,” 2009 email thread
• Evidence: [researcher inference | strong] The 2006 essay states the rule and two contemporaneous applications; “Deal Size” connects it to ownership and capital; Twitter and Airbnb stress-test it with a win and miss from separate records. Early Stage Investing, paragraphs 1–11 Deal Size, paragraphs 5–13 Twitter Paul Graham, Airbnb emails
• Limits, failure modes, and exception ledger: [researcher inference | strong] Investors can overstate their ability to help, “formation” can conceal a weak market, and early control rights can harm founders. A later-stage or capital-intensive company may be attractive even when the investor cannot shape it. Falsifiers: named support produces no measurable change; the company succeeds without the claimed capability; or early-classified deals do not show more influence/learning than later ones. Record exception: no complete initial-stage classifications, time-allocation records, ownership targets by deal, board interventions, or matched outcomes were recovered from the 21-company first-fund claim or current directory.
• Evolution and contradictions: [researcher inference | moderate] The functional definition remains useful, but USV later created opportunity vehicles and expanded into markets with different formation costs. The Opportunity Fund, mandate paragraphs The tension is resolvable only by naming the legal vehicle and the specific formation work; “early” cannot be inferred from firm brand.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: formation remaining, ability to matter, and ownership/time fit. Does not transfer: 2006 round labels, check ranges, or web-service costs. Changed assumptions (analyst-created scenario): Assume AI compute, biotech, climate hardware, security, and regulation can make early proof capital-intensive or externally gated. Observable signals: named missing functions, founder consent, partner time, milestone evidence, and rational ownership. Likely misuse: calling every small round early. Analyst-created hypothetical: decline a nominal seed deal whose only missing element is regulatory approval the investor cannot accelerate and whose price assumes success.

Model 4: Finance the next falsifiable learning or replication milestone

• Rule: [researcher inference | strong] Size each tranche to retire a named uncertainty while preserving enough runway to try a credible alternative. Fund expansion only after the product, cohort, unit, geographic, technical, or regulatory mechanism replicates; neither starvation nor abundant capital is evidence.
• Evidence status: [investor-stated | strong] Wilson explicitly advocates near-term milestone finance and argues that high burn removes time to correct strategy. [retrospective witness | strong] Kozmo and Turntable supply adverse replication and cohort cases. [researcher inference | moderate] The integrated “learning per unit of burn” procedure is reconstructed; company-level budgets and stop decisions are mostly absent.
• Provenance and origin: [investor-stated | strong] “Deal Size” gives the 2006 rule—“raise only enough money to achieve a set of near term milestones”—and the 2007 failure essay ties runway to the ability to change strategy. Deal Size, paragraphs 7–13 Why Early Stage Venture Investments Fail, failure categories and burn paragraphs
• Mechanism and reasoning: [researcher inference | strong] A smaller initial exposure buys an option on uncertain teams and markets; a milestone changes the posterior probability and valuation before more capital is committed. Runway protects search. The model assumes the milestone is measurable, the tranche can reach it, failure is interpretable, and fixed costs are not being ignored. Deal Size, paragraphs 5–13
• Historical and vehicle context: [contemporaneous record | strong] Early USV web services could often test product and distribution with modest capital while targeting 15–20% ownership; the figures do not transfer to capital-intensive sectors. Deal Size, paragraphs 1–13 [retrospective witness | moderate] Flatiron's boom-era access to large corporate capital made simultaneous Kozmo expansion possible and reduced the discipline imposed by scarcity. Worthless Stock Certificates, Kozmo section and Wilson comment
• When to use: [researcher inference | strong] Use for initial rounds, follow-ons, geographic rollout, product expansion, hiring plans, regulated launches, and capital-intensive proofs when uncertainty can be staged. Do not use rigid tranching when a genuine network land-grab, safety requirement, or indivisible infrastructure build demands scale; instead define the larger proof and downside case explicitly.
• How to apply it: [researcher inference | strong] (1) Name the uncertainty. (2) Choose the smallest credible experiment, not the smallest check. (3) Define metric, cohort, date, data owner, and alternative explanation. (4) Fund runway through one failed route and the next decision. (5) Precommit increase, maintain, resize, and stop gates. (6) Ask: what probability changes if this succeeds; what do we learn if it fails; can the next market reproduce economics; what is the no-new-capital case? Kill criteria: activity without probability change, a second geography/product before first replication, or burn that removes the ability to adapt.
• Examples:
  • Documented doctrine — early USV checks. [contemporaneous record | moderate] Wilson reports four of nine initial checks at $1 million or less while still seeking meaningful ownership and later capacity. The unnamed sample shows policy, not outcome efficacy. Deal Size, paragraphs 1–13
  • Failure — Kozmo. [retrospective witness | strong] Wilson says New York worked, but investors financed roughly 18–20 cities before proving the second and third; he describes a total loss. The account is retrospective and the amounts approximate. Worthless Stock Certificates, Kozmo section and Wilson comment
  • Failure — Turntable. [retrospective witness | strong] Six-to-eight-week churn was visible while gross acquisition hid it; Wilson says the board/company reacted too slowly and the service wound down. The missing decision-time cohort export prevents a causal audit. Business Insider interview, Turntable exchange
  • Researcher-applied application — Twitter. [researcher inference | moderate] Reliability, repeated usage, and ecosystem activity were more informative early gates than forcing a complete business model. The source does not disclose a formal milestone schedule or check. Twitter, rationale and business-model paragraphs
• Evidence: [researcher inference | strong] “Deal Size” and the failure essay supply independent contemporaneous doctrine; Kozmo and Turntable supply two distinct adverse cases; Twitter supplies a positive researcher-applied boundary. Deal Size, paragraphs 7–13 Why Early Stage Venture Investments Fail, categories 1–2 Worthless Stock Certificates Business Insider interview, Turntable exchange
• Limits, failure modes, and exception ledger: [researcher inference | strong] Small tranches can underfund a valid proof, milestones can be gamed, and a sequence of local optimizations can miss a brief winner-take-most window. Kozmo and Turntable are selected failures, not a denominator. Falsifiers: milestone success does not predict replication; added capital does not improve evidence; or strict staging repeatedly loses viable companies. Record exception: no original budgets, board packs, city cohorts, Turntable dashboards, reserve votes, rejected follow-ons, or matched staged-versus-unstaged outcomes were recovered after direct posts, interview transcripts, company/filing routes, and the deal-retrieval sweep.
• Evolution and contradictions: [researcher inference | moderate] The doctrine matures from small checks and runway into explicit replication/retention lessons and 2024 reserve mechanics. It conflicts with Wilson's willingness to support uncertain networks before monetization and with markets requiring rapid scale; the reconciliation is to stage the decisive mechanism, not demand the same metric everywhere. Reserves, Recycling, and Returns, reserve/recycling sections
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: risk-retirement per dollar, replication gates, fallback runway, and precommitted stops. Does not transfer: fixed 2006 check sizes or universal low burn. Changed assumptions (analyst-created scenario): Assume AI compute, hardware, clinical work, and regulatory submissions can make evidence expensive and lumpy. Observable signals: cohort retention, cost per accepted output, second-market unit economics, technical reliability, regulatory progress, and remaining runway. Likely misuse: equating underfunding with discipline. Analyst-created hypothetical: finance two production-scale AI evaluations plus a fallback architecture, not twelve months of undifferentiated hiring.

Model 5: Reserve for changed evidence, not familiarity or sunk cost

• Rule: [researcher inference | strong] Treat initial checks and reserves as one allocation system, and release follow-on capital only when new evidence improves the risk-adjusted case—not because the company is familiar, marked up, or costly to abandon.
• Evidence status: [investor-stated | strong] Wilson explicitly describes reserves as both protection for winners and an opportunity cost. [contemporaneous record | strong] USV created a separate opportunity vehicle for later rounds. [researcher inference | moderate] No public vintage-level ledger tests whether evidence-gated follow-ons outperformed alternatives.
• Provenance and origin: [investor-stated | strong] Wilson's 2007 essay says reserves are necessary because future needs of the best companies are unknowable, while every reserved dollar competes with a new investment. Reserves, full essay His 2024 treatment adds recycling and distribution mechanics but uses hypothetical, not realized, return arithmetic. Reserves, Recycling, and Returns, full essay
• Mechanism and reasoning: [researcher inference | strong] A reserve is a real option on improving information. Retention, unit economics, network density, technical proof, or team learning can justify exercising it; prior exposure and social proximity cannot. The comparison set is follow-on, a new deal, distribution, or cash, subject to the vehicle's legal and tax constraints.
• Historical and vehicle context: [contemporaneous record | strong] Early USV paired modest initial checks with follow-on capacity; the 2011 Opportunity Fund then separated some later-stage investing from the core early-stage mandate. Deal Size, paragraphs 1–13 The Opportunity Fund, announcement and mandate paragraphs These structures do not disclose Wilson's personal economics or a universal reserve ratio.
• When to use: [researcher inference | strong] Use at initial portfolio construction, every pro-rata or rescue decision, and whenever a round may belong in another vehicle. Do not use the rule mechanically where participation preserves governance rights, prevents a destructive financing, or new evidence will arrive after the option expires; make that exception explicit.
• How to apply it: [researcher inference | strong] (1) Record the initial reserve thesis and ownership target. (2) Rewrite the company case from zero at each round. (3) Compare follow-on with the best new deal and distribution. (4) Test fund, recycling, concentration, and conflict constraints. (5) Ask what changed, what remains falsifiable, and whether the same facts would justify a first check today. Kill criteria: sunk-cost language, valuation-only evidence, no updated operating data, or vehicle ineligibility.
• Examples:
  • Documented doctrine — initial checks and later capacity. [investor-stated | moderate] Wilson describes small initial checks alongside meaningful ownership and an ability to invest more; the unnamed nine-deal sample shows policy, not efficacy. Deal Size, paragraphs 1–13
  • Documented vehicle control — Opportunity Fund. [contemporaneous record | strong] USV created a separate fund for later-stage investments in existing portfolio companies and selected network opportunities, preventing silent mandate stretch. The notice supplies no return outcome. The Opportunity Fund, announcement and mandate paragraphs
  • Researcher-applied adverse case — Kozmo. [researcher inference | strong] Repeated capital and rapid multi-city expansion did not repair unproven replication and cash intensity; Wilson describes the investment as a total loss. Worthless Stock Certificates, Kozmo section and Wilson comment
  • Researcher-applied boundary — Twitter. [researcher inference | moderate] Continued support before a settled business model can be rational when usage, ecosystem, and reliability evidence improves; the public record does not reveal each reserve vote or alternative considered. Twitter, rationale and business-model paragraphs
• Evidence: [researcher inference | strong] The 2007 reserve essay and 2024 mechanics are Wilson-authored doctrine; the independent opportunity-fund notice is a vehicle control; Kozmo and Twitter provide adverse and positive boundaries. Reserves, full essay Reserves, Recycling, and Returns The Opportunity Fund Worthless Stock Certificates
• Limits, failure modes, and exception ledger: [researcher inference | strong] Reserve choices are path-dependent; declining can damage a viable company, while insiders can possess both better information and stronger bias. Selected anecdotes cannot establish portfolio optimization. Falsifiers: contemporaneous memos show familiarity overrode adverse evidence; follow-ons systematically trail comparable new investments; or vehicle separation adds cost without preserving mandate. Record exception: no complete reserve schedule, rejected follow-on log, fund cash-flow ledger, or matched alternative set was recovered from public fund notices, writings, filings, and deal histories.
• Evolution and contradictions: [researcher inference | moderate] The doctrine evolves from simple follow-on capacity in 2007, through a dedicated 2011 vehicle, to explicit recycling mechanics in 2024. Opportunity investing can conflict with early-stage focus; the defensible reconciliation is transparent vehicle separation and a fresh evidence test, not assuming every later round belongs in the core fund.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: a written reserve thesis, zero-based re-underwriting, comparable alternatives, and precommitted stop conditions. Does not transfer: fixed historical check sizes or hypothetical return arithmetic. Changed assumptions (analyst-created scenario): Assume AI compute commitments, structured rounds, secondary liquidity, and continuation vehicles complicate concentration. Observable signals: cohort retention, gross-margin trajectory, technical progress, ownership, and runway. Likely misuse: treating a markup or strategic logo as changed evidence. Analyst-created hypothetical: release an AI follow-on only after paid-workload retention and inference economics meet the decision-date threshold.

Model 6: Match fund, vehicle, team, and succession to the opportunity set

• Rule: [researcher inference | strong] Design the fund's size, mandate, partner capabilities, workload, and succession as one system; reject opportunities that require a different system unless a disclosed vehicle and accountable team can support them.
• Evidence status: [retrospective witness | strong] Wilson links Flatiron's deterioration to capital and style drift, while USV's founding and later notices document deliberate size, focus, opportunity vehicles, and generational change. [researcher inference | moderate] Public records do not reveal partner economics, key-person deliberations, or comparative fund returns.
• Provenance and origin: [investor-stated | strong] Wilson's Flatiron postmortem describes how abundant capital and a larger institution changed investment behavior. VC Cliche of the Week, Flatiron and Chase sections USV's founding response was a small partnership with direct partner work and explicit focus. Union Square Ventures, formation and operating-design paragraphs
• Mechanism and reasoning: [researcher inference | strong] Fund size changes the minimum useful check, ownership needs, follow-on pressure, and number of outcomes required; team size changes information flow and board capacity. A mandate constrains selection, while succession preserves or deliberately revises it. Misalignment turns an attractive company into a bad vehicle investment or converts scale into style drift.
• Historical and vehicle context: [contemporaneous record | strong] USV reiterated its focus in 2008, added an Opportunity Fund in 2011, and in 2016 disclosed Burnham and Wilson's reduced key-person role while preserving active decisions and LP duties. New Fund — Same Focus, full notice USV 2016, LP, key-person and active-decision paragraphs Later firm history spans multiple funds and partners; it is not a Wilson-only record. Investing at the Edge, article body
• When to use: [researcher inference | strong] Use during fund formation, vehicle allocation, partner hiring, board-capacity planning, and succession. Do not use mandate discipline to suppress genuine thesis learning; revise the institution explicitly when the opportunity set changes.
• How to apply it: [researcher inference | strong] (1) Define the opportunity set and evidence unit. (2) Model check size, ownership, reserves, and outcome concentration. (3) Map each partner's expertise, decision rights, and board load. (4) state which vehicle owns each deal. (5) stress-test key-person loss and transition. (6) Ask whether capital or staffing requires weaker selection. Kill criteria: no credible ownership path, unnamed capability owner, hidden cross-vehicle discretion, or deployment pressure replacing thesis fit.
• Examples:
  • Failure — Flatiron style drift. [retrospective witness | strong] Wilson says a larger capital base pulled Flatiron away from its original practice; underlying LP records are missing, so causal and return claims remain bounded. VC Cliche of the Week, Flatiron postmortem
  • Documented design — early USV. [contemporaneous record | strong] The founding post describes a small team, focused thesis, and partners doing the work directly. It is a design statement, not performance proof. Union Square Ventures, operating-design paragraphs
  • Documented vehicle separation — Opportunity Fund. [contemporaneous record | strong] A later-stage vehicle expanded capacity without redefining every core investment as later stage. The Opportunity Fund, mandate paragraphs
  • Documented succession — USV 2016. [contemporaneous record | strong] The firm disclosed the senior partners' changed key-person role and continued decision involvement, showing that succession is both governance and LP disclosure. USV 2016, LP, key-person paragraphs
• Evidence: [researcher inference | strong] A Wilson-authored failure account, three official fund notices, and a current institutional history triangulate the system. VC Cliche Union Square Ventures USV 2016, LP Investing at the Edge None supplies a controlled comparison between alternative fund designs.
• Limits, failure modes, and exception ledger: [researcher inference | strong] Small teams can concentrate blind spots; a narrow mandate can miss adjacent markets; separate vehicles can create conflicts; succession disclosures can conceal informal influence. Falsifiers: larger funds preserve selection and returns without drift; board coverage is unaffected by team load; or disclosed succession does not change actual authority. Record exception: no partnership agreements, fund models, allocation policies, partner compensation, key-person negotiations, or complete vintage cash flows were recovered.
• Evolution and contradictions: [researcher inference | strong] Wilson moves from a personal Flatiron lesson to USV's institutional design, then to opportunity vehicles and explicit succession. The tension between “same focus” and later expanded edge markets is real; it is disciplined adaptation only when the changed opportunity set, vehicle, authorship, and team are named.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: model the fund and team from the work required, disclose allocation, and plan succession before it is urgent. Does not transfer: a universal small-fund ideal. Changed assumptions (analyst-created scenario): Assume solo GPs, rolling funds, opportunity SPVs, platform teams, and AI-assisted sourcing separate brand from decision capacity. Observable signals: partner time, ownership, reserve concentration, vehicle conflicts, and key-person coverage. Likely misuse: raising more because demand exists. Analyst-created hypothetical: an AI seed fund declines a capital-intensive data-center build unless a separate vehicle and infrastructure partner own the underwriting.

Model 7: Underwrite networks at the participant-behavior level

• Rule: [researcher inference | strong] Underwrite a network from repeated participant behavior, differentiated value, and reinforcing contribution—not from registered users, category fashion, or a large addressable-market slide.
• Evidence status: [investor-stated | strong] USV's dated thesis makes engaged networks, differentiated user value, defensibility, and capital efficiency explicit. [contemporaneous record | moderate] Twitter and Zynga investment posts apply related logic. [researcher inference | moderate] Public evidence lacks comparable decision-date cohort data across wins, failures, and passes.
• Provenance and origin: [contemporaneous record | strong] USV's 2012 institutional thesis describes “large networks of engaged users” as a core criterion and links those networks to value and defensibility. Investment Thesis @ USV, thesis criteria and byline The byline and firm framing prevent sole attribution to Wilson.
• Mechanism and reasoning: [researcher inference | strong] A durable network grows useful interactions faster than coordination, moderation, or acquisition costs. Participants return because other participants create value; contribution improves the product; switching sacrifices relationships, reputation, data, or liquidity. The model assumes engagement is authentic, not subsidized, and that the company can capture enough value without destroying participation.
• Historical and vehicle context: [contemporaneous record | strong] Twitter's open API and emergent communication behavior and Zynga's social-game distribution were early web-network applications; Etsy combined marketplace liquidity and community. Twitter, investment rationale Zynga Game Network, investment rationale SEC filings establish vehicle ownership and governance, not Wilson-only sourcing or network causality.
• When to use: [researcher inference | strong] Use for social, marketplace, protocol, creator, developer, and data networks. Do not use where the product is primarily workflow software, content inventory, or one-way distribution; network language must identify participants, contribution, and feedback loops.
• How to apply it: [researcher inference | strong] (1) Name each participant and job. (2) map the atomic interaction and contribution loop. (3) measure retained cohorts, contribution concentration, match quality, and cross-side effects. (4) test whether value improves with relevant density. (5) model capture, governance, abuse, and multi-homing. Kill criteria: growth without retained interaction, one subsidized side, no density effect, deteriorating trust, or value captured mainly by an external platform.
• Examples:
  • Documented application — Twitter. [contemporaneous record | strong] Wilson's decision-near post emphasizes user-created communication and an open ecosystem while admitting the business model was unknown. Twitter, rationale and business-model paragraphs The filing places the stake in USV vehicles and identifies shared power. Twitter S-1/A, principal-stockholder footnote 16
  • Documented application with platform risk — Zynga. [contemporaneous record | strong] The thesis was social gaming across an emerging distribution platform; dependence on another platform limits defensibility. Zynga Game Network, rationale The filing shows shared USV power, not personal Wilson ownership. Zynga S-1/A, principal-stockholder footnote 6
  • Governance-rich application — Etsy. [researcher inference | strong] Marketplace liquidity, seller identity, and buyer trust fit the participant model, while the filing supports Wilson's board role and USV ownership rather than sole deal credit. Etsy S-1, principal stockholders and director biography
  • Failure — Turntable. [retrospective witness | strong] Gross acquisition hid six-to-eight-week churn; novelty did not become durable participant behavior. Business Insider interview, Turntable exchange
• Evidence: [researcher inference | strong] The firm thesis states the mechanism; Twitter and Zynga are decision-near applications; SEC filings constrain role; Turntable supplies the adverse retention case. Investment Thesis @ USV Twitter Zynga Game Network Business Insider interview
• Limits, failure modes, and exception ledger: [researcher inference | strong] Networks can produce congestion, abuse, regulatory exposure, creator dependence, or value captured by an upstream platform. Selected successes create survivorship bias. Falsifiers: density does not improve participant outcomes; cohorts decay after subsidies stop; contribution concentrates without governance; or a simpler non-network product wins. Record exception: no comparable decision-date retention, graph, liquidity, moderation, or acquisition-cost exports were recovered across the portfolio.
• Evolution and contradictions: [researcher inference | moderate] The thesis moves from consumer web networks toward protocols and broader market edges. Twitter shows that monetization can follow behavior; Turntable shows behavior must endure; Zynga shows a network can still depend on a more powerful network. “Network effects” therefore remains a mechanism to test, not a category label.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: participant maps, retained interaction, contribution loops, and governance. Does not transfer: registered-user growth as network proof. Changed assumptions (analyst-created scenario): Assume AI agents can create synthetic activity, distribution can be platform-rented, and protocol incentives can purchase temporary participation. Observable signals: retained human or economic work, density-adjusted match quality, contribution diversity, and post-subsidy use. Likely misuse: calling every AI data flywheel a network. Analyst-created hypothetical: decline an agent marketplace whose transaction count is entirely generated by one subsidized model provider.

Model 8: Map enabling disruption and incumbent pressure before calling a market an “edge”

• Rule: [researcher inference | strong] Invest at a market edge only after naming the enabling technical change, the incumbent constraint it exploits, the new participant behavior it permits, and the evidence that the wedge can become an independent value chain.
• Evidence status: [contemporaneous record | moderate] Flatiron's New York internet positioning and USV's 2024 institutional history document repeated edge framing. [retrospective witness | moderate] Wilson and Chris Dixon discuss technology-capital cycles. [researcher inference | moderate] “Edge” is a firm-level synthesis, and speaker attribution in the 2018 transcript is unreliable at individual-line resolution.
• Provenance and origin: [contemporaneous record | strong] A 1996 Wired profile records Flatiron's thesis around New York's media, advertising, and finance base as commercial internet infrastructure emerged. Wired, fund strategy and first-deal sections USV's later history describes successive theses at the edge of large markets under transformative pressure. Investing at the Edge, article body
• Mechanism and reasoning: [researcher inference | strong] A platform or cost discontinuity weakens an incumbent bottleneck; a startup enters through a use case the incumbent cannot serve without cannibalization, coordination failure, or architectural change. Capital matters only if the wedge compounds into distribution, data, community, or infrastructure rather than remaining dependent on the disrupted platform.
• Historical and vehicle context: [retrospective witness | moderate] Flatiron pursued commercial internet services; early USV focused on application networks; later USV expanded into crypto and broader systemic pressure. The 2018 a16z conversation is useful for cycle logic but should be cited as a conversation-level record because transcript speaker labels are inconsistent. a16z transcript, full conversation and timestamps
• When to use: [researcher inference | strong] Use in thesis formation and category diligence where technology, regulation, cost, or distribution is shifting. Do not use “disruption” to excuse weak unit economics, illegal distribution, dependency on a single incumbent, or a market too small beyond the wedge.
• How to apply it: [researcher inference | strong] (1) Name the old stack and profit pool. (2) identify the technical, regulatory, or cost discontinuity. (3) specify the constrained incumbent response. (4) map the wedge, adoption path, and independent control point. (5) model what happens if the incumbent copies, blocks, acquires, or litigates. Kill criteria: no observable pressure, a wedge without expansion, permanent upstream dependence, or legality essential to the thesis but untested.
• Examples:
  • Documented application — New York internet. [contemporaneous record | moderate] Flatiron linked local industry density to emerging internet services; the article reports strategy and capital, not audited outcome attribution. Wired, positioning and portfolio sections
  • Researcher-applied application — Twitter. [researcher inference | strong] Mobile, web, and open-API behavior opened a communication edge; the investment post acknowledges that capture remained unresolved. Twitter, rationale and business-model paragraphs
  • Researcher-applied application — Coinbase. [researcher inference | strong] Bitcoin enabled a new asset and payment interface while custody, regulation, and exchange infrastructure became the company-level control points. Wilson's study account and the issuer filing support preparation, board tenure, and vehicle ownership—not sole causality. Business Insider interview, Bitcoin exchange Coinbase S-1, director and principal-stockholder sections
  • Adverse boundary — Kik/Kin. [documented behavior | strong] A token thesis confronted securities law; Kik paid a civil penalty after an unregistered offering. The regulator's record establishes issuer outcome, not Wilson or USV wrongdoing. SEC, Kik case and distribution page
• Evidence: [researcher inference | strong] Contemporaneous reporting, official firm history, a cross-firm transcript, decision-near company writing, and a regulator outcome triangulate the model. Wired, Flatiron profile Investing at the Edge a16z transcript SEC, Kik
• Limits, failure modes, and exception ledger: [researcher inference | strong] “Disruption” is easy to narrate after success; incumbents can adapt, regulation can close the wedge, and enabling technology can commoditize startups. Falsifiers: adoption occurs without the claimed discontinuity; incumbent response is unconstrained; startup capture does not improve; or legal dependence fails. Record exception: no full Flatiron or USV opportunity-set memos, rejected edge theses, comparable incumbent-response dataset, or investment-level return ledger was recovered.
• Evolution and contradictions: [researcher inference | moderate] The vocabulary evolves from a geographic internet cluster to application networks, crypto, and broad market pressure. Kik shows that technical possibility does not settle legal viability; Twitter shows participant value can precede capture. A changing thesis is defensible only when its enabling mechanism and prior errors are dated.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: discontinuity, incumbent constraint, wedge, control point, and adverse-response map. Does not transfer: labeling AI, crypto, or climate inherently disruptive. Changed assumptions (analyst-created scenario): Assume foundation models and cloud platforms can absorb application features quickly, while regulation changes distribution. Observable signals: declining input costs, new behavior, incumbent cannibalization, independent retention, and legal pathway. Likely misuse: mistaking a model wrapper for a durable edge. Analyst-created hypothetical: pass on an AI workflow tool if the only edge disappears when the platform adds the same feature and the company owns no data, distribution, or trust layer.

Model 9: Separate protocol resilience from application capture, token structure, and law

• Rule: [researcher inference | strong] Underwrite a decentralized protocol, the company built around it, the token or security, and the legal path as four separate assets; strength in one cannot cure failure in another.
• Evidence status: [investor-stated | strong] Wilson argues that open protocols can be resilient and investable. [documented behavior | strong] Coinbase supplies a company outcome; Kik/Kin supplies an adverse token-and-law outcome. [researcher inference | moderate] No complete crypto opportunity set or token-level cash-flow record is public.
• Provenance and origin: [investor-stated | strong] In his prepared Coin Center keynote, Wilson distinguishes internet protocols from the companies built on them and admits, “I cannot point to a blockbuster company” from an earlier protocol thesis. Coin Center Keynote, protocol and investing sections That contemporaneous limitation is part of the model.
• Mechanism and reasoning: [researcher inference | strong] Protocol resilience comes from distributed validation and exit; application capture comes from trust, custody, interfaces, compliance, or distribution; token value depends on rights and demand; legality depends on offering facts and jurisdiction. Conflating them converts ideological conviction into unpriced company and enforcement risk.
• Historical and vehicle context: [retrospective witness | strong] Wilson says he and Wenger studied Bitcoin before USV invested in Coinbase; the issuer filing later verifies Wilson's board tenure and USV vehicle ownership. Business Insider interview, Bitcoin exchange Coinbase S-1, director biography and principal stockholders
• When to use: [researcher inference | strong] Use for protocols, exchanges, wallets, token issuers, decentralized infrastructure, and regulated financial networks. Do not use token demand, decentralization rhetoric, or protocol uptime as a proxy for company revenue, security compliance, or investor return.
• How to apply it: [researcher inference | strong] (1) Map protocol governance and failure modes. (2) identify the company's control point and replaceability. (3) specify token rights, supply, distribution, and value flow. (4) obtain jurisdiction-specific legal analysis. (5) model enforcement and protocol-without-company outcomes. Kill criteria: undisclosed token allocation, capture dependent on regulatory violation, no company moat beyond the protocol, or governance controlled by an unaccountable minority.
• Examples:
  • Company-layer application — Coinbase. [documented behavior | strong] Custody, exchange, compliance, and interface capabilities created company-level value around Bitcoin; the filing supports governance and ownership, not a fund multiple. Coinbase S-1, business, risk, director, and principal-stockholder sections
  • Protocol doctrine — Bitcoin. [investor-stated | strong] Wilson's keynote argues resilience and openness while acknowledging that protocol enthusiasm had not yet produced a cited blockbuster company. Coin Center Keynote, protocol-resilience and company sections
  • Adverse token case — Kik/Kin. [documented behavior | strong] Wilson advocated for Kin and disclosed his Kik board seat; Kik later paid a civil penalty for an unregistered offering. DefendCrypto.org, board and advocacy paragraphs SEC, Kik case and distribution page
• Evidence: [researcher inference | strong] The keynote supplies direct doctrine, the Coinbase filing supplies independent company facts, and Wilson's advocacy plus the regulator record supplies an adverse sequence. Coin Center Keynote Coinbase S-1 Kin SEC, Kik
• Limits, failure modes, and exception ledger: [researcher inference | strong] Protocol and application boundaries can change; legal opinions are jurisdiction- and fact-specific; enforcement after investment is not proof the original decision was irrational or that an investor committed wrongdoing. Falsifiers: company value persists without the claimed control point; token demand is unrelated to protocol use; or legal structure fails under disclosed facts. Record exception: no Kik board minutes, token allocation ledger, USV investment memo, outside legal advice, or crypto-portfolio return denominator was recovered.
• Evolution and contradictions: [researcher inference | strong] Wilson's protocol optimism predates the Kin enforcement outcome. The later regulator record does not refute protocol resilience, but it does refute treating that resilience as a complete security, offering, or company thesis. Advocacy and conflicted economic interest must be labeled alongside the claim.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: four separate underwritings, explicit governance, legal downside, and value-capture mapping. Does not transfer: 2016 regulatory assumptions. Changed assumptions (analyst-created scenario): Assume stablecoins, tokenized assets, sanctions, and jurisdictional fragmentation add issuer and intermediary risk. Observable signals: validator concentration, application retention, fee capture, token ownership, and licensed operations. Likely misuse: calling decentralization a moat. Analyst-created hypothetical: pass on a protocol company if token appreciation is the only return path and counsel cannot defend the distribution facts.

Model 10: Version the thesis; distinguish disciplined adaptation from style drift

• Rule: [researcher inference | strong] Date every thesis, state what evidence changed, and preserve the prior prediction; expand only when the new mechanism is explicit and the fund can still do the work.
• Evidence status: [retrospective witness | strong] Wilson calls Flatiron's expansion style drift. [contemporaneous record | strong] USV repeatedly published focus and thesis revisions. [researcher inference | moderate] Public posts do not reveal all internal disagreements, passes, or silent exceptions.
• Provenance and origin: [investor-stated | strong] Wilson's Flatiron postmortem is the negative origin: capital and success encouraged departure from the original practice. VC Cliche of the Week, Flatiron postmortem USV's 2008 notice answers with “same focus” while updating market conditions. New Fund — Same Focus, full notice
• Mechanism and reasoning: [researcher inference | strong] A dated thesis makes learning distinguishable from hindsight. New evidence can change market scope, mechanism, or vehicle; unchanged evidence plus deployment pressure is drift. Version history also exposes whether a new category is genuinely adjacent or merely fashionable.
• Historical and vehicle context: [contemporaneous record | strong] The 2012 network thesis and 2024 edge history are firm-level artifacts with multiple authors, not a single Wilson doctrine. Investment Thesis @ USV, criteria and byline Investing at the Edge, article body and archive byline control
• When to use: [researcher inference | strong] Use at annual strategy reviews, new-fund formation, category expansion, and every IC debate that invokes “what has changed.” Do not freeze a thesis when falsifying evidence accumulates; the purpose is accountable learning, not permanence.
• How to apply it: [researcher inference | strong] (1) Preserve the prior thesis verbatim and date it. (2) list predictions and outcomes. (3) name the new enabling mechanism and disconfirming evidence. (4) map changes to vehicle and team. (5) compare one in-thesis deal, one exception, and one pass. Kill criteria: no dated prior, change justified by prices or competitors alone, unnamed author, or expansion that destroys ownership/workload fit.
• Examples:
  • Failure — Flatiron. [retrospective witness | strong] Wilson says institutional capital altered style; absent fund records, the account is testimony, not a measured drift coefficient. VC Cliche, Flatiron and Chase sections
  • Documented update — Zynga. [investor-stated | strong] Wilson asks what changed and identifies a new social platform and founder evidence; the post is decision-near but not the complete IC record. Zynga Game Network, paragraphs 1–8
  • Institutional continuity — 2008 fund. [contemporaneous record | strong] USV explicitly paired a new fund with the same focus, allowing readers to compare claim and later behavior. New Fund — Same Focus
  • Institutional expansion — 2024 edge. [contemporaneous record | strong] The later history broadens markets and team while explaining the pressure-based continuity; it remains a self-assessment, not performance proof. Investing at the Edge
• Evidence: [researcher inference | strong] The Flatiron failure account, decision-near Zynga post, dated 2008 notice, 2012 thesis, and 2024 firm history create a longitudinal record. VC Cliche Zynga Game Network Investment Thesis @ USV Investing at the Edge
• Limits, failure modes, and exception ledger: [researcher inference | strong] Published theses can be marketing documents; categories overlap; a correct change can look inconsistent before outcomes emerge. Falsifiers: portfolio choices routinely contradict published versions without explanation; updates follow fundraising fashion rather than evidence; or preserved predictions show no learning. Record exception: no full internal thesis decks, exception logs, rejected deals, partner votes, or time-stamped portfolio classifications were recovered.
• Evolution and contradictions: [researcher inference | strong] Wilson moves from warning about drift to participating in a firm that deliberately broadens its market-edge vocabulary. The contradiction cannot be erased; it is resolved only deal by deal through dated mechanisms, named authors, appropriate vehicles, and a record of what was abandoned.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: version control, prediction logs, named authors, and explicit change evidence. Does not transfer: preserving old category labels after mechanisms change. Changed assumptions (analyst-created scenario): Assume rapid AI capability shifts shorten thesis half-lives and amplify herd behavior. Observable signals: dated memos, falsified claims, allocation changes, and partner capability. Likely misuse: rebranding momentum as learning. Analyst-created hypothetical: an AI fund adds robotics only after documenting the cost and reliability discontinuity, specialist owner, and vehicle impact.

Model 11: Turn conviction into falsifiable patience

• Rule: [researcher inference | strong] Hold an uncertain thesis only while named evidence can still validate it; patience without a dated falsifier is identity, not discipline.
• Evidence status: [investor-stated | strong] Wilson publicly tolerated unresolved monetization and long protocol development. [retrospective witness | strong] Airbnb shows conviction can be too weak; Turntable and Kik show it can persist past adverse evidence. [researcher inference | moderate] Decision-time stop rules are generally unavailable.
• Provenance and origin: [investor-stated | strong] Twitter's investment post separates observed product behavior from an unknown business model. Twitter, rationale and business-model paragraphs The Coin Center keynote similarly frames long protocol development while acknowledging missing company proof. Coin Center Keynote, protocol and company sections
• Mechanism and reasoning: [researcher inference | strong] Conviction protects a non-consensus mechanism from premature abandonment; falsifiers prevent belief from absorbing every outcome. Patience is rational when runway remains, evidence cadence is specified, and adverse observations update probability.
• Historical and vehicle context: [researcher inference | strong] Early networks often formed before monetization, while venture funds still imposed reserve, concentration, and life constraints. Advocacy, board duty, and vehicle exposure can intensify commitment and must be disclosed separately.
• When to use: [researcher inference | strong] Use where product behavior precedes capture, technical progress is lumpy, or regulation is unresolved. Do not use when survival depends on illegal conduct, repeated capital without learning, hidden cohort decay, or a thesis that cannot be disproved.
• How to apply it: [researcher inference | strong] (1) Write the mechanism and strongest contrary case. (2) define dated validation and falsification signals. (3) budget runway and reserves. (4) assign an independent challenger. (5) update after every material event. Kill criteria: falsifier triggered, evidence cadence missed, runway below the next test, or advocacy replacing analysis.
• Examples:
  • Positive boundary — Twitter. [researcher inference | strong] Continued conviction rested on observed communication behavior and ecosystem development despite unknown monetization; the public record lacks the actual reserve votes. Twitter
  • Miss — Airbnb. [retrospective witness | strong] USV liked the founders but did not extend the market thesis; the email thread preserves skepticism that later evidence could have challenged. Paul Graham, 2009 email thread Airbnb, retrospective
  • Failure — Turntable. [retrospective witness | strong] Persistent gross growth obscured rapid churn and the response came too slowly. Business Insider interview, Turntable exchange
  • Adverse boundary — Kik/Kin. [documented behavior | strong] Wilson's advocacy continued into enforcement; the SEC outcome shows why legal falsifiers must be independent of protocol conviction. DefendCrypto.org SEC, Kik
• Evidence: [researcher inference | strong] Twitter and the keynote state patient theses; Airbnb supplies a miss; Turntable and the SEC record supply distinct adverse outcomes. Twitter Coin Center Keynote Paul Graham, Airbnb SEC, Kik
• Limits, failure modes, and exception ledger: [researcher inference | strong] Outcomes can arrive after fund constraints force action; premature exits can be rational; a failed company does not automatically falsify a market. Falsifiers: stated stop rules do not change decisions; adverse evidence is repeatedly reframed as confirmation; or patient positions underperform matched alternatives. Record exception: no complete decision logs, reserve votes, valuation histories, or rejected sale records were recovered.
• Evolution and contradictions: [researcher inference | strong] Wilson's public record contains both too little pursuit at Airbnb and extended advocacy at Kik. The reconciliation is not “more conviction”; it is conviction proportional to observable evidence, exposure, legal risk, and the cost of another test.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: dated falsifiers, independent challenge, runway, and update cadence. Does not transfer: patience as a personality trait. Changed assumptions (analyst-created scenario): Assume AI benchmarks move quickly and can manufacture apparent progress. Observable signals: customer retention, technical reliability, legal milestones, and cash runway. Likely misuse: defending a visible thesis after contrary data. Analyst-created hypothetical: hold an AI-agent investment through two weak quarters only if repeat task completion and paid retention improve by the prewritten date.

Model 12: Serve founders while preserving duty to the company and stakeholders

• Rule: [researcher inference | strong] Treat the entrepreneur as the venture firm's customer during sourcing and support, but on the board distinguish service to the founder from fiduciary and governance duties to the company and its stakeholders.
• Evidence status: [investor-stated | strong] Wilson explicitly states founder-service doctrine and later discusses trust, CEO replacement, and board duty. [retrospective witness | moderate] Interviews supply selected founder interactions, not representative satisfaction data.
• Provenance and origin: [investor-stated | strong] The 2005 essay “Our Customer Is The Entrepreneur” defines the service orientation. Our Customer Is The Entrepreneur, full essay Wilson's Reboot account of StarMedia and founder removal supplies the adverse governance boundary. Reboot transcript, PDF pages 3–10
• Mechanism and reasoning: [researcher inference | strong] Service improves trust, information flow, recruiting, and decision quality; board independence protects the company when founder interests diverge. The model fails if “service” becomes deference, investor control, or marketing unsupported by founder experience.
• Historical and vehicle context: [contemporaneous record | strong] Early USV positioned partners as direct service providers rather than a large operating platform. Union Square Ventures, operating-design paragraphs Board seats and vehicle ownership create legal and economic roles that a blog relationship alone does not.
• When to use: [researcher inference | strong] Use when setting expectations before investment, providing portfolio help, evaluating CEO performance, or managing conflict. Do not promise founder primacy where law, insolvency, employee harm, or shareholder duty requires another course.
• How to apply it: [researcher inference | strong] (1) Define requested help and founder consent. (2) separate advice, board decision, and investor economics. (3) document conflicts and stakeholder effects. (4) give direct feedback with process and evidence. (5) preserve independent escalation. Kill criteria: hidden control, retaliation for disagreement, founder favoritism over company survival, or support promises the partner cannot staff.
• Examples:
  • Doctrine — entrepreneur as customer. [investor-stated | strong] The essay states the service promise but provides no founder-satisfaction denominator. Our Customer Is The Entrepreneur
  • Documented relationship — Boxee. [retrospective witness | moderate] Wilson describes product use, a long relationship, and investment after a pivot; this illustrates observation and help, not the founder's independent assessment. MeetInnovators transcript, PDF pages 8–10
  • Adverse governance case — StarMedia. [retrospective witness | strong] Wilson recounts founder removal and lasting emotional cost, showing that board duty can conflict with founder service. Reboot transcript, PDF pages 3–7
  • Failure boundary — Turntable. [retrospective witness | strong] Slow response to churn shows that supportive relationships do not excuse delayed operating or board action. Business Insider interview, Turntable exchange
• Evidence: [researcher inference | strong] Founder-service doctrine, a contemporaneous firm design, two long-form interviews, and distinct adverse cases support the tension. Our Customer Is The Entrepreneur Union Square Ventures Reboot transcript MeetInnovators transcript
• Limits, failure modes, and exception ledger: [researcher inference | strong] Founder accounts are underrepresented; service quality may vary by partner, company outcome, identity, and power. Falsifiers: founders report systematic coercion or neglect; board records show economics overriding company duty; or service claims do not correspond to partner time. Record exception: no representative founder survey, support ticket log, complete board minutes, or conflict register was recovered.
• Evolution and contradictions: [researcher inference | strong] The early slogan is simple; Wilson's later trust and failure accounts make it conditional. StarMedia shows service can end in a painful removal; Turntable shows loyalty can coexist with delayed correction. The durable rule is transparent role separation, not “founder-friendly” branding.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: consented help, direct feedback, conflict logs, and role separation. Does not transfer: assuming every founder wants the same platform services. Changed assumptions (analyst-created scenario): Assume distributed teams, founder liquidity, AI monitoring, and complex stakeholder claims expand conflicts. Observable signals: response time, promised versus delivered help, board escalation, and independent founder feedback. Likely misuse: using “founder-friendly” to avoid hard governance. Analyst-created hypothetical: support an AI founder through a model failure while the independent board separately decides disclosure and safety remediation.

Model 13: Design boards around decisions, CEO constraints, and traceable follow-through

• Rule: [researcher inference | strong] Make the board a decision system: center the CEO's hardest constraints, distribute materials early, separate reporting from debate, assign owners and dates, and preserve dissent and follow-through.
• Evidence status: [investor-stated | strong] Wilson's 2010 CEO essay and 2024 board template state operating doctrine. [retrospective witness | strong] StarMedia and Turntable supply failure boundaries. [documented behavior | strong] Etsy filings verify sustained governance roles but not meeting quality.
• Provenance and origin: [investor-stated | strong] “What a CEO Does” reduces the role to vision, team, and resources; the 2024 board essay turns recurring meetings toward prepared discussion and clear decisions. What a CEO Does, full essay Board Meeting Suggestions, full essay
• Mechanism and reasoning: [researcher inference | strong] Pre-reads preserve scarce meeting time for exceptions and decisions; explicit constraints prevent generic advice; owners and dates create accountability; recorded dissent improves later learning. The board adds value only when information is candid and authority remains clear.
• Historical and vehicle context: [retrospective witness | strong] Wilson's board experience spans Flatiron and USV, but governance credit is company- and director-specific. Etsy's filing verifies his director tenure and USV ownership; it does not reveal internal interventions. Etsy S-1, director biography and principal stockholders
• When to use: [researcher inference | strong] Use for every recurring board, crisis review, CEO evaluation, financing, and strategic pivot. Do not impose the same cadence on a two-person seed company and a public issuer; scale the controls to risk and legal duty.
• How to apply it: [researcher inference | strong] (1) Circulate metrics and narrative before the meeting. (2) name the two or three decisions and CEO constraints. (3) reserve reporting for exceptions. (4) record decision, owner, date, dissent, and dependency. (5) open the next meeting with follow-through. Kill criteria: surprise materials, no decision owner, unresolved conflicts, performative updates, or recurring adverse metric without a response threshold.
• Examples:
  • Documented doctrine — board template. [investor-stated | strong] Wilson's 2024 suggestions emphasize preparation and discussion; the source supplies no controlled evidence of better outcomes. Board Meeting Suggestions
  • Adverse case — StarMedia. [retrospective witness | strong] Founder removal and trust damage show that even a board's necessary decision can fail relationally or procedurally. Reboot transcript, PDF pages 3–7
  • Failure — Turntable. [retrospective witness | strong] A decisive churn signal existed, but Wilson says the board/company reacted too slowly. Business Insider interview, Turntable exchange
  • Governance boundary — Etsy. [documented behavior | strong] Filings verify long board service and evolving chair roles, not sole deal credit or meeting efficacy. Etsy 2026 proxy, board chronology
• Evidence: [researcher inference | strong] Two Wilson-authored operating essays, a long-form failure transcript, an interview, and issuer filings supply doctrine, adverse cases, and role controls. What a CEO Does Board Meeting Suggestions Reboot transcript Etsy proxy
• Limits, failure modes, and exception ledger: [researcher inference | strong] Process can create false certainty, minutes can chill debate, and boards can overreach into management. Falsifiers: decisions remain unimplemented despite logs; prepared boards detect crises no earlier; or CEO constraints are consistently better resolved outside the board. Record exception: no complete board packs, minutes, dissent logs, CEO evaluations, or matched governance outcomes were recovered.
• Evolution and contradictions: [researcher inference | strong] The concise 2010 CEO slogan is necessary but not sufficient; the 2024 template operationalizes board support. StarMedia and Turntable demonstrate that role clarity and meeting form do not guarantee timely or humane decisions.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: pre-reads, decision agendas, owner/date logs, dissent, and follow-up. Does not transfer: one fixed board cadence. Changed assumptions (analyst-created scenario): Assume distributed work and faster technical incidents compress decision windows. Observable signals: late materials, unresolved actions, metric exceptions, and escalation time. Likely misuse: adding dashboards instead of decisions. Analyst-created hypothetical: an AI safety incident board packet separates verified facts, model-risk hypotheses, disclosure duty, and a named remediation owner.

Model 14: Separate narrator, partner labor, decision rights, vehicle ownership, and economics

• Rule: [researcher inference | strong] Attribute every deal claim across five distinct fields—public narrator, sourcing or labor, decision rights, legal vehicle ownership, and realized economics—and leave a field unknown when evidence does not resolve it.
• Evidence status: [retrospective witness | strong] Wilson repeatedly corrects partner credit. [documented behavior | strong] SEC filings identify vehicles, boards, and shared power. [researcher inference | strong] Neither source type usually discloses sourcing votes, personal economics, or fund cash returns.
• Provenance and origin: [investor-stated | strong] Wilson's “Looking Back on 1999” explicitly corrects “best deal I ever did” shorthand by crediting Jerry Colonna with bringing and leading GeoCities. Looking Back on 1999, GeoCities credit paragraphs The correction makes attribution a decision-quality control rather than etiquette.
• Mechanism and reasoning: [researcher inference | strong] Accurate role fields reveal whose expertise produced a judgment, where governance authority sat, which vehicle bore risk, and whether an outcome reached investors. Collapsing them into a celebrity narrative prevents learning and manufactures track records.
• Historical and vehicle context: [retrospective witness | strong] Flatiron and USV are partnerships using fund vehicles; Wilson's blog made him a frequent narrator. In the Business Insider interview he assigns MongoDB, Foursquare, and Etsy to Albert Wenger while giving Brad Burnham equal credit for building the firm. Business Insider interview, partner-credit exchanges
• When to use: [researcher inference | strong] Use in investment histories, IC postmortems, board biographies, media claims, return analysis, and succession. Do not infer personal ownership from a board seat, sole credit from an announcement byline, or realized return from public-company value.
• How to apply it: [researcher inference | strong] (1) Create the five role fields. (2) enter only claim-local evidence and date. (3) distinguish individual, partner group, firm, and fund. (4) reconcile narrative with filings. (5) label conflicts and unknowns. Kill criteria: “his investment” without a vehicle, a multiple without cash-flow support, or a sole-deal claim contradicted by partner testimony or shared-power filings.
• Examples:
  • Credit correction — GeoCities. [retrospective witness | strong] Wilson names Colonna as source and lead, while Wilson's public retrospective remains the narrative route. Looking Back on 1999
  • Partner-specific credit — MongoDB. [retrospective witness | strong] Wilson identifies Wenger as deal lead; the issuer filing identifies a USV vehicle, not personal Wilson ownership. Business Insider interview MongoDB S-1, principal-stockholder footnote
  • Shared power — Zynga. [documented behavior | strong] The filing reports USV shares and shared voting/investment power among named partners, blocking sole or personal attribution. Zynga S-1/A, principal-stockholder footnote 6
  • Board-and-vehicle boundary — Coinbase. [documented behavior | strong] The filing verifies Wilson's board tenure and USV ownership while leaving personal proceeds and fund multiple unresolved. Coinbase S-1, director and principal-stockholder sections
• Evidence: [researcher inference | strong] Wilson's explicit correction and partner-credit interview are independent of issuer filings that control legal ownership and governance. Looking Back on 1999 Business Insider interview Zynga S-1/A Coinbase S-1
• Limits, failure modes, and exception ledger: [researcher inference | strong] Memories can be generous or selective; filings aggregate entities and do not reveal informal influence; team decisions resist single-lead labels. Falsifiers: contemporaneous memos contradict retrospective credit; vehicle records show different ownership; or cash-flow ledgers disprove attributed economics. Record exception: no complete investment memos, allocation schedules, partner votes, carry ledgers, or realized cash flows were recovered for signature deals.
• Evolution and contradictions: [researcher inference | strong] Wilson's visibility created Wilson-centric shorthand, but his later corrections expose a distributed institution. Succession makes the distinction more important: preserving firm history requires naming the person, action, vehicle, and evidence rather than transferring every outcome to the most visible partner.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: five-field attribution and filing reconciliation. Does not transfer: celebrity-led deal lists. Changed assumptions (analyst-created scenario): Assume syndicates, SPVs, scouts, rolling funds, and public online narration multiply role ambiguity. Observable signals: memo author, vote, board role, cap-table entity, and cash distribution. Likely misuse: treating a partner's post as ownership proof. Analyst-created hypothetical: an AI deal page credits the scout, diligence lead, IC, board director, fund vehicle, and realized proceeds separately.

Model 15: Externalize reasoning; preserve assumptions, corrections, and misses

• Rule: [researcher inference | strong] Write the thesis, assumptions, uncertainty, and authorship before outcome; publish or archive corrections without erasing the original so later readers can distinguish learning from hindsight.
• Evidence status: [investor-stated | strong] Wilson describes AVC as “thinking out loud,” built a long educational corpus, and publicly corrected math, deal credit, and misses. [researcher inference | moderate] The public corpus is selected and cannot substitute for complete internal records.
• Provenance and origin: [investor-stated | strong] Wilson's About page dates AVC to 2003 and frames it as public thinking rather than a finished doctrine. About Fred Wilson, AVC-purpose paragraphs The 2009 venture-math pair visibly revises assumptions after feedback. The Venture Capital Math Problem, both-post sequence
• Mechanism and reasoning: [researcher inference | strong] A dated record exposes base rates, attribution, and changing beliefs; outside critique finds errors; preserved versions deter hindsight editing; misses reveal model boundaries. Public writing adds selection and performance incentives, so private decision records remain necessary.
• Historical and vehicle context: [contemporaneous record | strong] AVC, USV writing, and MBA Mondays span personal, firm, and teaching contexts. MBA Mondays category archive, pagination A Wilson byline does not make a firm thesis, partner deal, or fund return personal.
• When to use: [researcher inference | strong] Use for IC memos, thesis changes, passes, valuation models, attribution disputes, and postmortems. Do not publish confidential founder, LP, employee, or legal information; preserve it in access-controlled records instead.
• How to apply it: [researcher inference | strong] (1) Date claims and evidence. (2) label author, role, and vehicle. (3) record priors, ranges, falsifiers, and alternatives. (4) append corrections with reasons. (5) revisit selected wins, failures, and passes. Kill criteria: post-outcome reconstruction without an original, silent edits, uncited numbers, or correction that omits what changed.
• Examples:
  • Visible correction — venture math. [documented behavior | strong] Wilson revised assumptions in a linked continuation; the pair is one evolving work, not two independent confirmations. Venture Capital Math Problem Continued, revised assumptions
  • Attribution correction — GeoCities. [investor-stated | strong] Wilson publicly credits Colonna with bringing and leading the deal. Looking Back on 1999
  • Miss record — Airbnb. [retrospective witness | strong] Wilson acknowledges the pass, while Paul Graham's email thread preserves decision-time reasoning. Airbnb Paul Graham, email thread
  • Educational corpus — MBA Mondays. [documented behavior | strong] The bounded archive externalizes operating and finance frameworks; it is one series and not evidence that readers or portfolio companies applied them. MBA Mondays archive
• Evidence: [researcher inference | strong] The self-description, visible math correction, credit correction, independent Airbnb emails, and bounded teaching archive provide distinct evidence types. About Venture Capital Math Problem Looking Back on 1999 Paul Graham, Airbnb
• Limits, failure modes, and exception ledger: [researcher inference | strong] Public writing selects publishable episodes, may alter behavior, and can privilege the most visible partner. Deletions and migrations can break provenance. Falsifiers: archived versions show silent narrative changes; corrections avoid material errors; or private memos systematically contradict public rationales. Record exception: no complete AVC edit history, internal memo archive, pass log, partner-comment record, or readership-to-decision dataset was recovered.
• Evolution and contradictions: [researcher inference | strong] The corpus moves from blog diary to firm archive, teaching series, interviews, and migrated onchain publishing. The expansion increases transparency but also duplicate routes and lost metadata. Public candor coexists with missing economics and selected failures; it is a learning input, not an audited track record.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: dated memos, version history, correction logs, and explicit attribution. Does not transfer: public posting as a universal requirement. Changed assumptions (analyst-created scenario): Assume searchable archives and generated summaries can amplify both errors and silent revisions. Observable signals: preserved originals, correction dates, linked evidence, and changed decisions. Likely misuse: content production without decision accountability. Analyst-created hypothetical: an IC stores the original AI-market forecast, model-generated summary, human edits, and the later falsification in one versioned record.

Model 16: Keep marks, mortality, multiples, and net cash returns separate

• Rule: [researcher inference | strong] State the unit, vehicle, valuation date, denominator, cash-flow status, fees, and source before comparing performance; never turn an interim mark, company outcome, or selected LP row into a personal or firm-wide realized return.
• Evidence status: [investor-stated | moderate] Wilson publishes valuation, mortality, and reserve heuristics. [documented behavior | strong] LACERA supplies one current LP schedule. [researcher inference | strong] These records are noncomparable without explicit unit and denominator controls.
• Provenance and origin: [investor-stated | strong] Wilson's valuation essay separates quarterly marks from exit proceeds and says, “Never Mark Higher Than Potential Sale Value.” Valuing a Venture Capital Portfolio, rule 3 His mortality essay is an approximate long-run heuristic, not an audited portfolio table. Startup Mortality Rates, paragraphs 3–4
• Mechanism and reasoning: [researcher inference | strong] Marks estimate current value; mortality counts outcome states; company multiples ignore fund construction; TVPI mixes distributions and residual value; DPI records distributions; net IRR depends on timing and fees. Converting among them without cash flows and a common denominator creates false precision.
• Historical and vehicle context: [documented behavior | strong] The LACERA schedule reports one pension commitment to USV 2012, not the whole fund, firm, or Wilson. Its exact locator is viewer page 40, internal printed page 40, “Top Performing Investments (since inception),” Union Square Ventures 2012 Fund, L.P. row. LACERA, PDF and printed p. 40
• When to use: [researcher inference | strong] Use whenever a return, hit rate, valuation, write-off, exit, or fund comparison appears. Do not average unlike vintages, mix gross and net, or count a directory as an investor-attributed portfolio.
• How to apply it: [researcher inference | strong] (1) Define company, investment, fund, LP commitment, or firm. (2) set the as-of date and denominator. (3) separate cost, mark, proceeds, TVPI, DPI, and net IRR. (4) identify reported versus audited evidence. (5) reconcile cash flows before computing a multiple. Kill criteria: missing vehicle, missing date, press valuation used as proceeds, residual value treated as cash, or selected winners presented without losses.
• Examples:
  • Interim-mark doctrine. [investor-stated | moderate] Wilson describes quarterly valuation and a conservative exit backtest but publishes no company schedule or auditor work. Valuing a Venture Capital Portfolio
  • Heuristic denominator — mortality. [investor-stated | moderate] Wilson's thirds are approximate and omit vintages, unit, and valuation basis; they must not be presented as audited USV loss rates. Startup Mortality Rates
  • Actual LP schedule — LACERA. [contemporaneous record | strong] The row distinguishes TVPI, DPI, and net IRR for one commitment as of 2026-03-31; it cannot establish Wilson-specific economics. LACERA, PDF and printed p. 40, named row
  • Company loss — Kozmo. [retrospective witness | strong] Wilson calls Kozmo a total loss, which documents one adverse investment but not Flatiron's fund-wide net return. Worthless Stock Certificates, Kozmo section
• Evidence: [researcher inference | strong] The valuation and mortality essays expose Wilson's definitions and gaps; LACERA is an independent institutional schedule; Kozmo is a selected adverse case. Valuing a Venture Capital Portfolio Startup Mortality Rates LACERA Worthless Stock Certificates
• Limits, failure modes, and exception ledger: [researcher inference | strong] Even audited LP schedules can differ by fee class, timing, and reporting policy; private marks remain estimates; selected commitments invite survivorship bias. Falsifiers: reconciled cash flows support the broad claims; full vintages show the heuristic is stable; or comparable LP schedules eliminate selection effects. Record exception: no complete USV or Flatiron capital-account statements, fee schedules, investment-level cash flows, audit workpapers, or Wilson carry ledger was recovered.
• Evolution and contradictions: [researcher inference | strong] Wilson's public language moves from venture-math models to mark discipline, mortality heuristics, and reserve simulations. These improve conceptual transparency but remain largely self-reported. LACERA adds an external LP observation without supplying a firm composite or causal attribution.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: unit, vehicle, date, denominator, cash status, and net/gross labels. Does not transfer: headline valuation as performance. Changed assumptions (analyst-created scenario): Assume secondary prices and continuation structures can create more partial-liquidity states. Observable signals: reconciled cash flows, valuation policy, DPI, residual concentration, and fee class. Likely misuse: ranking managers by selected TVPI. Analyst-created hypothetical: an AI company markup remains an interim mark until the fund records proceeds; the memo shows both values separately.

Model 17: Convert institutional commitments into comparable funnel and outcome data

• Rule: [researcher inference | strong] Translate every institutional value or reform promise into a defined population, dated funnel stages, accountable owner, intervention, outcome measure, and publication cadence.
• Evidence status: [contemporaneous record | strong] USV's “An Appraisal” publicly acknowledges a severe representation failure and commits to change; succession and governance writings provide adjacent accountability mechanisms. [researcher inference | moderate] Public records do not yet supply a complete comparable funnel or causal outcome series.
• Provenance and origin: [contemporaneous record | strong] In 2020 USV wrote that after 16 years and more than 100 investments it had backed very few self-identified Black founders and was “part of the problem.” An Appraisal, self-assessment and commitments This is firm-level evidence and must not be reassigned to Wilson alone.
• Mechanism and reasoning: [researcher inference | strong] A denominator exposes where access, screening, diligence, selection, ownership, or support diverges; named ownership and cadence make corrective action testable; outcome measures distinguish activity from changed allocation or experience. Without these controls, commitments remain narratives.
• Historical and vehicle context: [contemporaneous record | strong] USV's partnership and funds changed across the measurement period, including disclosed succession in 2016. USV 2016, LP, key-person and active-decision paragraphs Comparisons must therefore hold fund, stage, geography, check size, and team definition constant or disclose the change.
• When to use: [researcher inference | strong] Use for representation, sourcing access, follow-on allocation, founder service, hiring, board coverage, and any public institutional commitment. Do not infer causality from a rising count without the eligible population and selection stages.
• How to apply it: [researcher inference | strong] (1) Define the eligible population and self-identification rules. (2) record source, screen, diligence, IC, investment, ownership, follow-on, and outcome. (3) assign an owner and intervention at each leakage point. (4) publish comparable period and exclusions. (5) audit adverse effects and founder experience. Kill criteria: changing denominator, missing stage data, activity-only goals, privacy-unsafe collection, or no authority to act on the result.
• Examples:
  • Adverse self-assessment — representation. [contemporaneous record | strong] USV supplies a broad historical denominator and acknowledges failure, but not a complete founder-level funnel or later outcome table. An Appraisal
  • Governance accountability — succession. [contemporaneous record | strong] The 2016 LP notice names changed key-person roles and continued decision responsibilities, allowing future outcomes to be associated with an evolving team rather than a static founder narrative. USV 2016, LP
  • Decision accountability — board follow-through. [investor-stated | strong] Wilson's board template supplies owners, discussion, and follow-up at company level; it does not itself prove institutional reform. Board Meeting Suggestions
  • Measurement boundary — mortality. [investor-stated | moderate] The approximate thirds show why an undefined unit and vintage cannot support subgroup or policy comparisons. Startup Mortality Rates
• Evidence: [researcher inference | strong] The firm appraisal establishes the problem and commitments; a fund succession notice establishes changing accountable actors; board and mortality essays expose action and denominator requirements. An Appraisal USV 2016, LP Board Meeting Suggestions Startup Mortality Rates
• Limits, failure modes, and exception ledger: [researcher inference | strong] Small samples, privacy, self-identification, stage mix, and long outcome lags complicate comparison; targets can create gaming or tokenism. Falsifiers: a stable funnel shows no leakage where claimed; interventions do not change comparable selection or founder experience; or definitions shift without restatement. Record exception: no complete pre- and post-2020 sourcing funnel, self-identification protocol, ownership-weighted outcomes, founder-experience series, or external assurance was recovered.
• Evolution and contradictions: [researcher inference | strong] USV's early open-network and founder-service claims did not prevent the later representation failure. “An Appraisal” is an important correction, not evidence of completed remediation. The model therefore requires the same versioning and denominator discipline applied elsewhere in this document.
• Modern VC translation (as of 2026-08-21): [researcher inference | strong] Transfers: explicit population, stage funnel, owner, intervention, outcome, and cadence. Does not transfer: one demographic definition across jurisdictions without consent and legal review. Changed assumptions (analyst-created scenario): Assume remote sourcing and automated screening can change both access and hidden selection effects. Observable signals: stage conversion, ownership, follow-on, exits, founder experience, and restated definitions. Likely misuse: publishing activity counts as outcomes. Analyst-created hypothetical: an AI-assisted sourcing program is evaluated by comparable stage conversion and founder feedback, with model decisions auditable and privacy-protected.

Evidence gaps and candidate models

• Retained inventory: [researcher inference | strong] Seventeen models survive reconciliation across 36 writing notes, eight talk notes, writings.md, talks.md, and investments.md. Their titles preserve distinct decision mechanisms; close synonyms were merged into the relevant rule rather than split to reach a target count.
• No exhausted-record model waiver used: [researcher inference | strong] Every retained model has at least three grounded cases and at least four claim-supporting citation placements from at least two genuinely independent underlying sources, including primary material. Each model still carries a specific record-exception ledger because evidence breadth does not cure missing decision-time or economic records.
• Candidate not promoted — corpus portability/onchain migration: [researcher inference | strong] The AVC migration is useful provenance and archival infrastructure, but the recovered record does not establish a distinct investing decision mechanism beyond Model 15's versioned reasoning and source preservation.
• Candidate not promoted — portfolio-wide partial liquidity: [researcher inference | strong] GeoCities provides a historical transaction route, but one recoverable case cannot support a general Wilson rule. Partial liquidity remains a boundary inside Models 5 and 16 pending company- and vehicle-level cash-flow evidence.
• Candidate not promoted — generic CEO triad: [researcher inference | strong] “What a CEO Does” is retained as origin evidence inside Model 13; it does not warrant a separate model without outcomes showing that the three-part slogan predicts governance or operating quality.
• Candidate not promoted — AI automation and price-regret doctrine: [researcher inference | strong] The 2026 official episode supplies chapter metadata but no accessible transcript or audio in the recovered record; unsupported chapter-level inference is excluded until a citable artifact is retrieved.
• Candidate not promoted — syndication as a standalone edge: [researcher inference | moderate] Relationship and complementary-capability evidence is real, but a separate syndication rule would duplicate Models 2, 6, and 14 without a comparable co-investment denominator.
• Candidate not promoted — platform pain/open source and replication/scale: [researcher inference | strong] The supported mechanisms are already represented in Models 4, 7, 8, 9, and 10. Splitting labels by company or technology would create synonyms rather than reusable decision tools.
• Highest-value missing records: [researcher inference | strong] Decision-date memos and pass logs; partner votes and allocation records; board packs and minutes; fund agreements, capital accounts, fees, and cash flows; comparable company cohorts; founder-experience and post-2020 institutional-funnel data; and a verified transcript or accessible audio for the 2026 USV interview.

What the evidence supports

[researcher inference | moderate] The defensible favorable conclusion is narrower than “Fred Wilson was a great picker.” From roughly 2003 through the early 2010s, Wilson helped build a coherent learning-and-capital system for capital-efficient internet services: publish and cultivate relationships to observe emerging behavior; invest while meaningful company formation remains; make a small first commitment; demand an identifiable network mechanism; preserve ownership and follow-on capacity; and keep experienced partners close to companies. The architecture was published before Twitter, Zynga, and Coinbase became famous, which reduces—but does not eliminate—retrospective-fitting risk. USV formation, partner model and 2004 deployment · Early Stage Investing, stage and ownership · Reserves, follow-on rationale · profile edge test

[researcher inference | strong] The strongest decision evidence is behavioral, not financial. Twitter's 2007 announcement identifies optional-follow public communication, user-created conventions, an open API, and unknown monetization before the later filing; Zynga records a six-month change of mind about applications using social graphs; Coinbase frames Bitcoin as an open value-transfer protocol and Coinbase as application infrastructure. These sources establish dated beliefs and public narration. They do not prove sole sourcing, internal approval, price, personal ownership, or net return. Twitter, rationale and monetization paragraphs · Zynga, thesis-change paragraphs · Coinbase, investment announcement · deal-role matrix

[researcher inference | strong] The record also supports an unusual willingness to leave falsifying material in public: Wilson corrected the venture-math model after reader criticism, corrected his own GeoCities credit shorthand, published the Airbnb miss, described Kozmo and StarMedia failures, and separated portfolio marks from actual liquidity. Publication is documented behavior; the stronger proposition—that every correction changed the next comparable decision in time to improve returns—remains unproved. venture-math continuation, revised assumptions · Looking Back on 1999, Colonna correction · Airbnb · Valuing a Venture Capital Portfolio, rules and liquidity caveat

[researcher inference | strong] The biography's causal story survives only in bounded form. A 1986 Euclid internship, continued school-year work, and a 1987 full-time career marker are supportable; USV's small-team/fund design can plausibly be read as a response to Flatiron's overcapitalization and organizational drift. But the response mechanism rests substantially on Wilson's retrospective accounts, not a recovered 2003 design memo or matched fund test. I Got Lucky, Euclid and Flatiron paragraphs · Wired 1996, Flatiron capital and first deals · Reboot transcript, PDF pp. 4–7

[researcher inference | strong] Team and vehicle evidence materially changes the greatness claim. Jerry Colonna brought and led GeoCities; Albert Wenger found Etsy and is credited by Wilson with MongoDB and several other investments; Brad Burnham authored the Tumblr acquisition post; the Twitter, Zynga, MongoDB, and Coinbase filings place stakes in USV entities with shared power. Wilson's public narration and later board work are real contributions, but neither a byline nor deemed beneficial ownership converts a team/vehicle result into a personal result. GeoCities, ¶¶4–34 · Business Insider interview, Wenger/partner passages · Yahoo Acquires Tumblr · attribution audit below

[researcher inference | moderate] Public economics are compatible with a highly successful firm record but cannot establish Wilson-specific alpha. Coinbase disclosed affiliate secondary transactions and a remaining vehicle stake; issuer filings show meaningful pre-offering stakes in Twitter, Zynga, Etsy, and MongoDB filing and role cross-checks. For its one commitment to Union Square Ventures 2012 Fund, L.P., LACERA gives reported: 23.75x TVPI (LACERA, PDF/printed p. 40, “Top Performing Investments (since inception),” USV 2012 row, 2026), reported: 23.15x DPI (same table and row, 2026), and reported: 53.4% net IRR (same table and row, 2026), as of 2026-03-31. The row explicitly labels IRR net; TVPI includes residual value; DPI reflects distributions; all three are commitment-specific. It is not a firm composite, company attribution, gross result, or Wilson-only record. investment economics boundary

[researcher inference | moderate] Boxee is the most useful ordinary control recovered. Wilson describes a two-year relationship, an initial hardware pass, a software pivot, use of the product, and then investment; later evidence places it in a sold/wound-down group but supplies no price or return. It supports longitudinal observation and changed-evidence discipline without permitting a winner narrative. MeetInnovators transcript, PDF pp. 8–10 · Boxee no-hindsight lab

What could falsify the thesis

The package's thesis is a process hypothesis: public learning and relationships surfaced changing participant behavior; a small, partner-led vehicle could invest before monetization settled; ownership, board work, and reserves made learning consequential; and an application-network thesis focused the search. It is deliberately not a demonstrated personal-return claim. [researcher inference | moderate] profile, proposed edge and denominator boundary

The strongest skeptical explanation

[researcher inference | strong] A plausible skeptical account is that Wilson combined privileged access, a favorable post-crash internet vintage, a strong partnership, attractive small-fund economics, and several power-law outcomes, then used an unusually rich public corpus to make the resulting process legible. Under this account, the writing is evidence of intellectual honesty and brand-building, not proof of selection alpha; public narration exaggerates personal credit; missing losers, passes, prices, and cash flows make the record untestable; and the apparent doctrine is partly the common pattern of successful cases reconstructed after the fact. skill/team/vehicle/era tests · portfolio-level adversarial tests

[researcher inference | strong] The recovered record does not defeat that skeptical account. It does, however, prevent the strongest cynical version. USV's partner/fund architecture and several selection rules predate the famous outcomes; the corpus contains named failures and misses; Wilson corrects team credit; and issuer/LP evidence confirms meaningful vehicle stakes and one exceptional fund-level signal. The discriminating evidence is still private: dated memos, pass logs, company cash flows, partner-role records, reserve decisions, and board minutes. Until those appear, “repeatable process contribution” is supportable; “Wilson-specific economic alpha” is not. philosophy, strongest case against greatness · profile skill/team/vehicle tests

Attribution and denominator audit

Denominator reconciliation

Coverage fraction: not computable. [researcher inference | strong] The person, fund, firm, vehicle, and named-case universes use different units and overlap. The package correctly refuses winner-count or return ratios rather than dividing by 32, ~55, 21, or 230+. Evidence boundary and searched denominator

Deal and team attribution cross-checks

Source and corpus accounting

[researcher inference | strong] The live source map contains 80 rows = 19 ★★★ + 42 ★★ + 19 ★; 61 priority rows. Accepted SCOUT accounting was 19/41/20. The DEALS phase reported replacing an unused ★ press route with the ★★ Airbnb S-1/A; the present row and arithmetic are verifiable, while the historical swap cause is retained as phase-owner reporting rather than reconstructed fact. The swap did not alter the WRITINGS/TALKS direct-work denominators. source-map coverage assessment

• [researcher inference | strong] WRITINGS reconciles 34 priority URLs / 33 underlying works: 32 full works and one partial MBA Mondays series; 36 notes/subsections include exactly three supplements. The two venture-math URLs are one sequence. The MBA source row now lists all six sampled child URLs plus the separately mapped CEO essay without adding source rows. writings coverage
• [researcher inference | strong] TALKS reconciles 7/7 priority appearances (100%): five full-access text artifacts and two partial artifacts, plus one nonpriority supplement. Coin Center is one appearance reused as a WRITINGS supplement, not two works. The 2026 episode is correctly partial because only producer-written chapter navigation—not a voice-verified transcript—was accessible. talks coverage
• [researcher inference | strong] The 44 note files have metadata plus the exact six required H2s. The full/partial field is normalized; quotation metadata is mechanically reconcilable and every excerpt remains under 25 words. Source checkboxes mark exactly 69/80 processed/cited routes = all 61 priority routes + eight ★ context routes. Every checked row has explicit reliability, access, accessed-date, and locator metadata. The remaining 11 unchecked ★ routes are retrieval/context leads rather than implied evidence.

Transfer limits for modern VCs

[researcher inference | strong] The historical method transfers as a diagnostic discipline, not as a list of 2007 sectors or fund sizes. Every modern translation in the philosophy/models is dated 2026-08-21 and either rests on the historical record or is explicitly framed as an analyst-created scenario. No uncited claim about current AI, energy, defense, fund, platform, or regulatory conditions is treated as fact. modern translation matrix · 17-model field guide

[researcher inference | strong] Martin Casado and Trae Stephens are used only as completed modern controls for public doctrine, interested-speaker evidence, team attribution, governance visibility, and vehicle opacity. They are not presented as consumer-network, New York ecosystem, vintage, strategy, or performance peers. Therefore their comparison can test research method and disclosure boundaries, not rank investing skill. completed-peer comparison

Unresolved evidence

The gaps below are residual public-record limits, not permission to fill them with inference.

1. Personal and company-level economics. [researcher inference | insufficient public record] Recover complete vehicle cash flows, cost bases, follow-ons, dilution, distributions, fees, carry, and personal allocations. Without them, personal alpha, outlier contribution, and record-without-top-deal tests are not computable. vehicle/economic boundary
2. Decision files and pass ledger. [researcher inference | insufficient public record] Recover dated investment memoranda, dissent, price/allocation, rejected cases, reserve decisions, and stop rules. These are the decisive tests of whether the public doctrine preceded and governed decisions. ranked retrieval ledger
3. Partner labor and authority. [researcher inference | insufficient public record] Recover source, sponsor, diligence, negotiation, approval, board, recruiting, follow-on, and exit roles for every material case. Filings establish vehicle/control facts, not the complete work ledger. people and deal-role map
4. Flatiron records. [researcher inference | insufficient public record] The ~55-company roster, pledge/direct-investment agreements, Chase/SoftBank capital records, LP reports, and realized cash flows remain missing. Wilson's $150 million/$750 million recollection lacks a defined realization basis. Flatiron evidence boundary · source-map retrieval priorities
5. Portfolio denominator. [researcher inference | insufficient public record] The 32 personal, ~55 Flatiron, 21 first-fund, 230+ firm, and 41 named-case universes remain noncomparable. A deduplicated company/round/vehicle roster with aliases and exits is required. denominator audit
6. Board evidence. [researcher inference | insufficient public record] StarMedia, Turntable, Kik, Etsy, and other governance lessons lack complete board decks, minutes, dissent, delegations, and decision/action dates. failure/postmortem boundaries
7. Regulatory and legal posture. [researcher inference | strong] The SEC Kik page establishes issuer liability and a $5 million penalty, not Wilson/USV wrongdoing. Recovered Patel, Grabski, Risley, and SEC Coinbase materials establish only dated procedural postures. Operative pleadings, current dockets, and Wilson-specific status still require verification before any personal allegation is repeated. SEC Kik, complaint/judgment summary and 2024-11-13 update · scrutiny ledger
8. Diversity outcomes. [contemporaneous record | strong] USV's 2020 appraisal acknowledges 16 years/100+ investments and very few self-identified Black founders. Comparable pre/post funnel, ownership, outcome, founder-experience, definition, and assurance data were not recovered, so the admission is strong evidence of the problem—not proof of remediation. An Appraisal, ¶¶1–4
9. Partial/inaccessible voice and archives. [researcher inference | strong] The 2026 episode lacks a voice-verified transcript; Mixergy has recording/transmission limits; 25iq, books, paywalled press, and wildcard archives remain partial/search routes. Producer labels, snippets, and inaccessible text were not promoted into Wilson quotations. talks access accounting · source-map access boundaries
10. Current-role boundary. [researcher inference | insufficient public record] Public fund notices and filings show formal succession, collaborative decision-making, legal entities, and selected board roles; Wilson's current workload, informal influence, deal leadership, economics, and key-person authority remain private. profile succession boundary

Twenty-point acceptance evidence

Central-claim citation audit trail

The following 16 high-risk central claims were reopened on 2026-08-21. “Pass” means the source supports the bounded adjacent claim—not that a favorable economic conclusion is proved.

Ordinary-claim citation audit trail

The following 15 ordinary claims use a reproducible spread: at least one from every core document, plus multiple writings, talks, and deal facts. They were selected by document order rather than favorable outcome.

Mechanical QA evidence

• Root structure: eight core root documents exist; the next nonblank line after each H1 is a substantive H2. investments.md begins with the exact Pareto H2. review.md begins with ## What the evidence supports.
• Required REVIEW structure: the five reader-facing H2s appear first and in exact order; the acceptance table has exactly 20 numbered rows; central and ordinary trails contain 16 and 15 rows respectively; attribution cross-checks contain seven material cases.
• Source accounting: 80 rows = 19 ★★★ / 42 ★★ / 19 ★; 61 priority. Exactly 69 processed/cited rows are [x] (19 ★★★ + 42 ★★ + eight ★) and 11 retrieval/context ★ rows remain unchecked. WRITINGS is 34 priority URLs / 33 works and TALKS is 7/7 appearances. Post-SCOUT distribution change is disclosed without asserting an unrecovered edit history as direct fact.
• Notes and quotes: 44 note files = 36 w- + 8 t-; all have metadata and exactly the six required H2s. Exact whitespace-token quote-count audit reports zero mismatches; no excerpt exceeds 25 words.
• Deals: 41 ledger rows reconcile as 14 material + 27 additional and 36 invested/associated + three passes + two outside; eight decision labs each contain four IC prompts and a separate actual-decision/outcome block; ten required H2s appear in order.
• Models: 17 exact retained titles in philosophy/model order; 12 component labels per model; 67 example bullets; every model meets the stated evidence floor and discloses its record exception.
• Links: every local Markdown target and fragment across the eight root documents resolves. Every external URL used by a root document or source note is either literal in sources.md or is a declared child of the one MBA series route. No raw URL or stale Twitter footnote-18 reference remains.
• Boundaries: all multiplier claims retain reported: syntax; LACERA is one LP commitment; acquisition price is not investor proceeds; filing stake is not sourcing or personal ownership; issuer/legal procedures are not personal wrongdoing findings; coverage fraction: not computable remains unchanged.

Identity, chronology, and long-form direct voice

• ★★ | Union Square Ventures — Fred Wilson | https://www.usv.com/people/fred-wilson/ | live; accessed 2026-08-21 | official biography and author index | Controls current role, education, Flatiron link, and 1987 career start; locator: bio paragraph and dated recent/most-read lists. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — About Fred Wilson | https://avc.com/about/ | live; accessed 2026-08-21 | first-person biography | Establishes AVC's September 2003 start and “thinking out loud” purpose, while creating a 1986/1987 career-start discrepancy to resolve. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — I Got Lucky | https://avc.com/2008/05/i-got-lucky/ | 2008-05-30 | first-person career retrospective | Resolves 1986 Euclid internship/Wharton schedule, early-1996 departure, Flatiron formation/capital, operating-experience gap, and luck claims; retrospective and self-reported. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — VC Cliche of the Week | https://avc.com/2005/09/vc_cliche_of_th-5/ | 2005-09-07 | investor-authored Flatiron postmortem | Route for the later Chase commitment, style drift, overcapitalization, and fund-structure lesson; figures need underlying vehicle records. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Connectors | https://avc.com/2006/11/connectors/ | 2006-11-14 | first-person partnership history | Clarifies Flatiron team formation, new-investment stop, portfolio salvage, and USV transition while crediting Jerry Colonna and Brad Burnham. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Wired — Flatiron VC Fund Pins Hopes on Silicon Alley | https://www.wired.com/1996/12/flatiron-vc-fund-pins-hopes-on-silicon-alley/ | 1996-12-23 | contemporaneous reported profile | Independent control for initial $50 million plus $100 million options, first deals, and period positioning; not audited fund performance. (reliability: B; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Forbes — Flatiron Partners Called Back Home | https://www.forbes.com/2001/05/17/0517flatiron.html | 2001-05-17 | contemporaneous reported profile | Near-period adverse route on Flatiron after the crash and Chase relationship; access may be partial/paywalled. (reliability: B; access: partial; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Union Square Ventures | https://www.usv.com/writing/2005/09/union-square-ventures/ | 2005-09-27 | investor-authored firm announcement | Decision-date source for October 2003 formation, 2004 fund naming, September 2005 final close, partner model, initial remit, and explicit AVC/USV attribution boundary. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | MeetInnovators — Interview with Fred Wilson from Union Square Ventures | https://meetinnovators.com/2009/11/19/fred-wilson-from-union-square-ventures-docs/full-transcript.pdf | 2009-11-19 | full interview transcript PDF | Long direct-voice route on career, AVC, New York, portfolio work, and investment method; locator: PDF pp. 1–13. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Reboot — Investing in Trust with Fred Wilson | https://www.reboot.io/wp-content/uploads/2020/02/Reboot-119-InvestinginTrust.pdf | original 2015-01-13; rereleased 2020-02-14 | full podcast transcript PDF | Best candid Euclid/Flatiron/StarMedia, founder-removal, board, failure, trust, and partnership postmortem; locator: PDF pp. 3–10. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Union Square Ventures — Fred Wilson on 40 Years in Venture — and Why USV Is Automating Itself | https://podcasts.apple.com/nz/podcast/fred-wilson-on-40-years-in-venture-and-why-usv-is/id1895741985?i=1000764128219 | 2026-04-29 | official video-podcast episode and chapter map | The accessible official artifact supplies producer-written chapter navigation across the 65-minute retrospective, but no voice-verified transcript; locator: official chapter map, 03:22–1:04:17. (reliability: A; access: partial; accessed: 2026-08-21; locator: producer-written chapter labels and timestamps 03:22–1:04:17)

• ★ | Startups.com — Risky Business with Fred Wilson | https://www.startups.com/founder-interviews/fred-wilson | undated; accessed 2026-08-21 | edited interview transcript | Rich direct voice on luck, slow early career, failures, boards, and deal judgment; publication date and editing history require recovery before quotation. (reliability: B; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Mixergy — Fred Wilson on Building Union Square Ventures | https://mixergy.com/interviews/fred-wilson/ | 2010-04-12 | video interview and transcript | Long route on USV formation, earlier firms, service-business economics, sourcing, and founder relations. (reliability: B; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Tren Griffin / 25iq — A Dozen Things I've Learned from Fred Wilson | https://25iq.com/2014/07/05/a-dozen-things-ive-learned-from-fred-wilson/ | 2014-07-05 | secondary quotation synthesis | Required interpretive route and quotation lead only; current fetch is a cache miss, so every quote must be recovered from the underlying Wilson source. (reliability: B; access: partial; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Business Insider / AVC — The Business Insider Interview | https://avc.com/2014/04/the-business-insider-interview/ | 2014-04-14 | full interview transcript repost | Direct account of partner deal attribution (including “Albert did the MongoDB deal”), 2012 setbacks, board allocation, succession, Coinbase, and Sidecar; retrospective claims still require deal records. (reliability: A; access: full; accessed: 2026-08-21; locator: MongoDB attribution and partner-portfolio passages)

AVC and USV writings: doctrine, decisions, and longitudinal corpus

• ★★★ | Fred Wilson — Our Customer Is The Entrepreneur | https://www.usv.com/writing/2005/11/our-customer-is-the-entrepreneur/ | 2005-11-12 | investor-authored essay | Foundational founder-service doctrine; controls what Wilson claimed, not whether founders experienced it consistently. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Early Stage Investing | https://www.usv.com/writing/2006/09/early-stage-investing/ | 2006-09-19 | investor-authored essay | Core stage, ownership, capital-efficiency, and support thesis suitable for later decision testing. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Deal Size | https://www.usv.com/writing/2006/10/deal-size/ | 2006-10-24 | investor-authored essay | Gives initial-check and ownership logic; examples are historical policy statements, not proof of every fund's practice. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Reserves | https://www.usv.com/writing/2007/04/reserves/ | 2007-04-12 | investor-authored essay | Early fund-construction route on follow-ons and reserves; any return references remain self-reported. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Why Early Stage Venture Investments Fail | https://www.usv.com/writing/2007/11/why-early-stage-venture-investments-fail/ | 2007-11-30 | investor-authored track-record post | Candidate personal denominator: 32 direct investments and outcome table, plus burn-rate/failure lessons; unaudited, dated, unnamed, and not a lifetime hit rate. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Looking Back on 1999 | https://avc.com/2009/01/looking-back-on-1999/ | 2009-01-27 | first-person deal-credit correction | Corrects “best deal I ever did” shorthand by crediting Jerry Colonna for bringing and leading GeoCities; linked GeoCities post is the same attribution thread. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — The Venture Capital Math Problem | https://avc.com/2009/04/the-venture-capital-math-problem/ | 2009-04-29–2009-04-30 | two-post analytical sequence | Reveals asset-class capacity model and public revision process; treat the linked continuation as the same underlying work and benchmark assumptions to period data. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Fred Wilson — What a CEO Does | https://avc.com/2010/08/what-a-ceo-does/ | 2010-08-30 | investor-authored operating essay | Concise CEO-role doctrine that can be compared with portfolio governance and failure cases. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Airbnb | https://avc.com/2011/03/airbnb/ | 2011-03-16 | investor-authored pass post with email record | Rare decision-time-adjacent miss explaining USV's framing error; Paul Graham's canonical “Subject: Airbnb” email thread supplies the dated January–February 2009 exchange with one disclosed redaction. Neither route proves the absence of price, access, or partner factors. (reliability: A; access: full; accessed: 2026-08-21; locator: AVC post body; Graham email dates and redaction note)

• ★★ | Fred Wilson — Worthless Stock Certificates | https://avc.com/2011/05/worthless-stock-certificates/ | 2011-05-20 | investor-authored failure reflection | Names Kozmo.com as a total-loss reminder and supplies a concrete Flatiron-era setback without reconstructing fund-wide economics. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | AVC — MBA Mondays category archive | https://avc.com/category/mba-mondays/ | 2010–2019; accessed 2026-08-21 | first-party series archive | Canonical route for the MBA Mondays operating and finance corpus; WRITINGS bounded its partial series sample to the category page plus The Employee Equity Project, Litigation, From The Archives: Convertible Debt, From The Archives: Retaining Your Team, From The Archives: Turning Your Team, Profitless Prosperity, and the separately mapped What a CEO Does. These child links do not create six additional source-map routes or underlying works. (reliability: A; access: partial; accessed: 2026-08-21; locator: category pagination and the seven named sampled posts; no stable series-level paragraph locator)

• ★★★ | Fred Wilson — Coinbase | https://www.usv.com/writing/2013/05/coinbase/ | 2013-05-08 | investor-authored investment announcement | Decision-near route for bitcoin/payment-network thesis and co-investors, independently checkable against Coinbase filings. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Twitter | https://www.usv.com/writing/2007/07/twitter/ | 2007-07-25 | investor-authored investment announcement | Decision-near account of USV investing in Twitter; establishes Wilson as public narrator, while internal thesis sponsorship and decision roles remain unknown. It does not establish sole vehicle ownership or team credit. (reliability: A; access: full; accessed: 2026-08-21; locator: investment rationale and team framing)

• ★★ | Brad Burnham — Yahoo Acquires Tumblr | https://www.usv.com/writing/2013/05/yahoo-acquires-tumblr/ | 2013-05-20 | partner-authored acquisition post | Team-credit control: Brad Burnham is the named USV narrator; transaction economics require Yahoo's 10-Q rather than Wilson shorthand. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Zynga Game Network | https://www.usv.com/writing/2008/01/zynga-game-network/ | 2008-01-15 | investor-authored investment announcement | Decision-near USV route for Zynga and its social-gaming thesis; pair with the S-1/A to preserve shared vehicle control and avoid sole or personal attribution. (reliability: A; access: full; accessed: 2026-08-21; locator: investment announcement and rationale)

• ★★★ | Fred Wilson — Coin Center Keynote | https://avc.com/2016/05/coin-center-keynote/ | 2016-05-04 | prepared keynote text | Direct statement on internet resilience, blockchain, regulation, and investing since 2011; stronger than later paraphrases. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — DefendCrypto.org | https://avc.com/2019/05/defendcrypto-org/ | 2019-05-28 | investor-authored advocacy post | Discloses Wilson's Kik board seat and front-row role while arguing against SEC treatment of Kin; essential conflict/advocacy route to pair with the judgment. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Chris Dixon and Fred Wilson — Technological Trends, Financial Capital, and the Dynamics of Disruption | https://a16z.com/podcast/a16z-podcast-technological-trends-financial-capital-and-the-dynamics-of-disruption/ | 2018-09-23 | full podcast transcript | Cross-firm conversation on technology cycles, capital, and crypto with timestamps; retrospective and interested, but analytically rich. (reliability: B; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Valuing a Venture Capital Portfolio | https://avc.com/2022/07/valuing-a-venture-capital-portfolio/ | 2022-07-18 | investor-authored fund essay | Separates interim marks from liquidity and supports disciplined handling of reported performance. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Startup Mortality Rates | https://avc.xyz/startup-mortality-rates | 2024-09-11 | investor-authored essay | Recent forty-year failure-distribution heuristic and market/team comments; self-reported, approximate, and not an audited USV dataset. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Fred Wilson — Reserves, Recycling, and Returns | https://avc.xyz/reserves%2C-recycling%2C-and-returns | 2024-08-30 | investor-authored fund essay | Recent reserve simulation, recycling, and portfolio-construction doctrine; hypothetical multiples must not be presented as actual USV returns. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — Board Meeting Suggestions | https://avc.xyz/board-meeting-suggestions | 2024-07-09 | investor-authored governance essay | Current board-practice template suitable for comparison with earlier CEO/board doctrine and adverse cases. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Brad Burnham — New Fund — Same Focus | https://www.usv.com/writing/2008/03/new-fund-same-focus/ | 2008-03-05 | official partner-authored fund notice | Primary route for the $156 million 2008 fund and continuity of USV focus; firm-vehicle fact, not Wilson-only capital, decision rights, or performance. (reliability: A; access: full; accessed: 2026-08-21; locator: opening fund-size and strategy paragraphs)

• ★ | Fred Wilson — AIVC | https://avc.com/2023/05/aivc-2/ | 2023-05-04 | investor-authored corpus note | Wilson states a 9,059-post corpus and his reuse/attribution policy; self-reported count that needs archive/export validation before denominator use. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Fred Wilson — I've Moved Onchain | https://avc.com/2024/05/ive-moved-onchain/ | 2024-05-02 | investor-authored migration history | Maps Typepad (September 2003), avc.com/WordPress, Mirror (2021–23), Paragraph/avc.xyz (from 2023), Arweave, and the Mirror-Paragraph merger; also discloses USV interests. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | AVC — Archives | https://avc.com/archive/ | live; accessed 2026-08-21 | first-party subject/month archive | Corpus navigation from 2003 onward; WordPress counts, Typepad gaps, migrated duplicates, comments, and onchain copies require deduplication. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Union Square Ventures — USV Investment Thesis archive | https://www.usv.com/writing/topic/usv-investment-thesis/ | 2006–present; accessed 2026-08-21 | official firm topic archive | Tracks thesis versions and exact bylines, preventing Wilson from receiving sole credit for Burnham/Weissman/Kaden/Wenger work. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Union Square Ventures — An Appraisal | https://www.usv.com/writing/2020/06/an-appraisal/ | 2020-06-03 | firm self-assessment | Adverse primary route acknowledging that 16 years and 100+ investments included very few self-identified Black founders and that USV was “part of the problem.” (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Fred Wilson — GeoCities | https://avc.com/2009/04/geocities/ | 2009-04-24 | investor-authored deal retrospective | Wilson calls GeoCities Flatiron's best investment while explicitly naming Jerry Colonna as the partner who did the deal; controls team credit and later shorthand. (reliability: A; access: full; accessed: 2026-08-21; locator: opening deal-credit paragraphs)

• ★★ | Fred Wilson — The Venture Capital Math Problem (Continued) | https://avc.com/2009/04/the-venture-capital-math-problem-continued/ | 2009-04-30 | investor-authored analytical continuation | Completes and revises the April 29 model after reader feedback; count with the first post as one underlying work, not an independent denominator item. (reliability: A; access: full; accessed: 2026-08-21; locator: revised assumptions and correction discussion)

• ★★ | Fred Wilson — Kin | https://avc.com/2017/05/kin/ | 2017-05-25 | investor-authored portfolio advocacy | Records Wilson's favorable, economically interested view of Kik's token plan before enforcement; pair with DefendCrypto and the SEC judgment, without attributing issuer liability to Wilson or USV. (reliability: A; access: full; accessed: 2026-08-21; locator: token rationale and USV/Kik disclosure)

• ★★ | Albert Wenger — I Want a New Platform | https://www.usv.com/writing/2007/09/i-want-a-new-platform/ | 2007-09-19 | partner-authored thesis post | Decision-era platform request; the post itself does not claim it caused the 10gen/MongoDB investment, while the Business Insider transcript separately attributes that deal to Wenger. (reliability: A; access: full; accessed: 2026-08-21; locator: platform requirements and author byline)

• ★★ | Union Square Ventures — Investment Thesis @ USV | https://www.usv.com/writing/2012/05/investment-thesis-usv/ | 2012-05-27 | official firm thesis essay | Dated firm-level articulation of large networks, differentiated user value, defensibility, and capital efficiency; byline and collective ownership must be preserved. (reliability: A; access: full; accessed: 2026-08-21; locator: thesis criteria and byline)

Vehicles, portfolios, deals, boards, and performance controls

• ★★ | Union Square Ventures — Companies | https://www.usv.com/companies/ | live; accessed 2026-08-21 | official firm portfolio directory | Candidate firm-level universe only; filters, current/exited status, follow-ons, and missing historical companies require archived snapshots and cannot create a Wilson hit rate. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Yahoo — Form 8-K/A, Exhibit 99.2 (GeoCities acquisition) | https://www.sec.gov/Archives/edgar/data/1011006/000091205799005431/0000912057-99-005431-d2.html | filed 1999-11-12; acquisition completed 1999-05-28 | SEC issuer acquisition filing | Corroborates GeoCities transaction timing and consideration route only; it does not establish Flatiron ownership, Wilson attribution, or realized fund economics. Pair with Wilson’s Jerry Colonna credit correction. (reliability: A; access: full; accessed: 2026-08-21; locator: MD&A Overview; Note 11, “Acquisition of GeoCities”)

• ★★ | SEC — Union Square Ventures Form ADV | https://reports.adviserinfo.sec.gov/reports/ADV/162375/PDF/162375.pdf | 2026-03-27 | regulatory filing PDF | Schedule D reports 26 legal-entity rows as private funds; this is an entity/vehicle count, not 26 strategies, and the filing does not disclose partner carry, deal leadership, or audited returns. (reliability: A; access: full; accessed: 2026-08-21; locator: Schedule D §7.B(1), private-fund rows)

• ★ | SEC — Union Square Ventures 2012 Fund Form D filing detail | https://www.sec.gov/Archives/edgar/data/1532179/000153217911000001/0001532179-11-000001-index.html | 2011-10-14 | SEC filing index | Primary vehicle and offering route for the 2012 fund; filing amount is fundraising disclosure, not performance. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Union Square Ventures — USV 2016, LP | https://www.usv.com/writing/2016/06/usv-2016-lp/ | 2016-06-22 | official fund and succession notice | $175 million fifth early-stage fund plus key-man/authority transition to Albert Wenger and Andy Weissman and five-partner decision structure. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Union Square Ventures — USV 2019 Funds | https://www.usv.com/writing/2019/04/usv-2019-funds/ | 2019-04-30 | official fund announcement | $200 million core and $250 million opportunity vehicles; fundraising size is not investment performance. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★ | Union Square Ventures — USV 2022 Funds | https://www.usv.com/writing/2022/04/usv-2022-funds/ | 2022-04-28 | official fund announcement | $275 million core and $350 million opportunity vehicles plus team context; does not assign Wilson-specific decisions or economics. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Union Square Ventures — Investing at the Edge of Large Markets Under Transformative Pressure | https://blog.usv.com/investing-at-the-edge-of-large-markets-under-transformative-pressure | 2024-01-23 | official firm retrospective | Firm-level history states 14 funds and 230+ investments and describes team/thesis evolution; a candidate firm denominator, not Wilson's denominator. Nick Grossman's byline and the date are controlled by the USV January 2024 archive as part of the same underlying work. (reliability: A; access: full; accessed: 2026-08-21; locator: article body; January archive entry dated 2024-01-23)

• ★★★ | Twitter — Amendment No. 1 to Form S-1 | https://www.sec.gov/Archives/edgar/data/1418091/000119312513400028/d564001ds1a.htm | 2013-10-15 | SEC issuer filing | Establishes USV 2004 vehicle holdings and shared control by Wilson/Burnham; no sole sourcing, personal ownership, or realized-return inference. (reliability: A; access: full; accessed: 2026-08-21; locator: principal stockholders table and footnote 16)

• ★★ | Zynga — Final Form S-1/A | https://www.sec.gov/Archives/edgar/data/1439404/000119312511336851/d198836ds1a.htm | 2011-12-12 | SEC issuer filing | Shows 30,738,892 USV shares, 5.4% pre-offering, and shared power among Brad/Fred/Albert/John; explicitly blocks sole/personal attribution. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Etsy — 2026 Proxy Statement | https://www.sec.gov/Archives/edgar/data/1370637/000137063726000026/etsy-20260417.htm | 2026-04-17 | SEC issuer proxy | Current board chronology: director since 2007, chair May 2017–December 2025, lead independent director again from January 2026; not a decision-date memo. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | MongoDB — Form S-1 | https://www.sec.gov/Archives/edgar/data/1441816/000104746917006014/a2233365zs-1.htm | 2017-09-21 | SEC issuer filing | Establishes USV 2008 vehicle ownership and shared voting/investment power; firm-level holding, not personal deal credit. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | Coinbase — Form S-1 | https://www.sec.gov/Archives/edgar/data/1679788/000162828021003168/coinbaseglobalincs-1.htm | 2021-02-25 | SEC issuer filing | Verifies Wilson's board tenure, USV greater-than-five-percent position, shared power, and secondary-sale disclosures; no personal proceeds or fund multiple. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★ | Yahoo — Q2 2013 Form 10-Q | https://www.sec.gov/Archives/edgar/data/1011006/000119312513327068/d540713d10q.htm | 2013-08-06 | SEC issuer filing | Primary route showing the Tumblr acquisition completed June 19 for approximately $990 million, primarily cash; pair with Brad Burnham’s USV post, and do not infer Wilson or USV realized return. (reliability: A; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

• ★★★ | LACERA — Total Fund Performance, 2026 Q1 | https://www.lacera.gov/sites/default/files/assets/documents/financials/total_fund_performance-2026Q1.pdf | 2026-03-31 | public-pension performance schedule PDF | For the USV 2012 row, LACERA gives reported: 23.75x TVPI, reported: 23.15x DPI, and reported: 53.4% net IRR. TVPI includes residual value; DPI is distributed capital; LP cash-flow timing and fee/class terms may differ. This is one LACERA commitment as of 2026-03-31, not a current whole-fund mark, firm composite, gross return, or Fred-only result. (reliability: A; access: full; accessed: 2026-08-21; locator: PDF p. 40 / printed p. 40, “Top Performing Investments (since inception),” Union Square Ventures 2012 Fund, L.P. row)

• ★★ | Union Square Ventures — The Opportunity Fund | https://www.usv.com/writing/2011/01/the-opportunity-fund/ | 2011-01-15 | official firm fund notice | Primary route for USV first Opportunity Fund and its follow-on mandate; fundraising and remit do not establish returns or Wilson-specific economics. (reliability: A; access: full; accessed: 2026-08-21; locator: fund announcement and mandate paragraphs)

• ★ | Fred Wilson — Freedom and Innovation | https://avc.com/2014/01/freedom/ | 2014-01-29 | investor-authored regulatory testimony account | Wilson says he testified at the NYDFS virtual-currency hearing and states his innovation-versus-regulation position; this establishes participation and advocacy, not misconduct or regulatory endorsement. (reliability: A; access: full; accessed: 2026-08-21; locator: opening testimony paragraph and policy argument)

• ★★ | Etsy — Initial Form S-1 | https://www.sec.gov/Archives/edgar/data/1370637/000119312515077045/d806992ds1.htm | 2015-03-04 | SEC issuer filing | IPO-era route for USV vehicle ownership and Wilson board/governance history; it cannot establish sole sourcing, personal ownership, realized proceeds, or fund return. (reliability: A; access: full; accessed: 2026-08-21; locator: principal-stockholders table, director biography, and related-party sections)

Press, books, historical context, and independent witnesses

• ★ | Claire Cain Miller — A New Kind of Venture Capitalist Makes Small Bets on Young Firms | https://www.nytimes.com/2008/09/22/technology/22venture.html | 2008-09-21 | reported firm profile | Near-contemporary view of USV's small-fund model and reputation; paywall/robots limit current extraction and Wilson's own response warns about puff-piece risk. (reliability: B; access: partial; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)
• ★★ | Airbnb — Form S-1/A | https://www.sec.gov/Archives/edgar/data/1559720/000119312520311265/d81668ds1a.htm | 2020-12-07 | SEC issuer filing | Outcome control for the 2009 USV pass: establishes Airbnb's later public-company scale and offering route, not the terms USV was offered, foregone ownership, dilution, or a computable counterfactual return. (reliability: A; access: full; accessed: 2026-08-21; locator: business description, history, and offering sections)
• ★ | Wired — This Man Makes Founders Cry | https://www.wired.com/2017/02/this-man-makes-founders-cry/ | 2017-02-28 | reported Jerry Colonna profile | Independent route through Wilson's Flatiron cofounder and dot-com aftermath; centered on Colonna, so Wilson claims need exact-page checks. (reliability: B; access: full; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)
• ★ | Sebastian Mallaby — The Power Law | https://www.penguinrandomhouse.com/books/580120/the-power-law-by-sebastian-mallaby/ | 2022-02-01 | reported book and publisher record | Independent VC-industry history and power-law critique; obtain indexed Wilson/USV pages from an authorized copy before quoting. (reliability: B; access: partial; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)
• ★ | Nick Bilton — Hatching Twitter | https://www.hodder.co.uk/titles/nick-bilton/hatching-twitter/9781444761979/ | 2013-11-05; publisher page accessed 2026-08-21 | reported book and publisher record | Investigative Twitter governance counterweight; claims need page-level locators and comparison with filings/participant accounts. (reliability: B; access: partial; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

Scrutiny, enforcement, litigation searches, and archives

• ★★★ | SEC — Kik Interactive Inc. case and distribution page | https://www.sec.gov/enforcement-litigation/distributions-for-harmed-investors/kik | 2019-06-04–2020-10-21; page last reviewed or updated 2024-11-13 | regulator case hub | Primary complaint/judgment route: Kik paid a $5 million civil penalty after the unregistered Kin offering; proves issuer enforcement, not Wilson/USV wrongdoing or decision rights. (reliability: A; access: full; accessed: 2026-08-21; locator: page heading, Key Dates, and final judgment/civil-penalty paragraphs)
• ★★ | Kessler Topaz Meltzer & Check — Coinbase Securities Case Status | https://www.ktmc.com/featured-case/coinbase-global-inc/ | live status accessed 2026-08-21 | plaintiff-counsel case-status page | Reports that the third amended complaint was filed 2025-10-21 and motion-to-dismiss briefing is complete and pending; the current summary does not itself establish Wilson's status in the operative pleading. Use only for current posture, not merits or wrongdoing. (reliability: B; access: full; accessed: 2026-08-21; locator: Current Status section)
• ★★ | Justia — mirror of D.N.J. Patel v. Coinbase Global, Document 135 | https://law.justia.com/cases/federal/district-courts/new-jersey/njdce/2%3A2022cv04915/500015/135/ | 2025-09-30 | full-text federal securities-class-action order mirror | Wilson is a named director defendant; the court partly granted and partly denied Rule 12(c), denied interlocutory certification, and allowed a third amended complaint. Allegations and procedural survival are not wrongdoing findings. (reliability: B; access: full; accessed: 2026-08-21; locator: Document 135, Conclusion and Order ¶¶1–8)
• ★★ | Delaware Court of Chancery — Grabski ex rel. Coinbase v. Andreessen | https://courts.delaware.gov/Opinions/Download.aspx?id=391020 | 2026-01-30 | derivative-action letter opinion | In the Coinbase derivative action naming Wilson among director defendants, the court granted a motion to strike but denied the SLC motion to terminate. This is a procedural ruling, not a merits finding against Wilson. (reliability: A; access: full; accessed: 2026-08-21; locator: pp. 18–20 and concluding disposition)
• ★★ | Justia — mirror of S.D.N.Y. Risley v. Universal Navigation Inc., Document 126 | https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1%3A2022cv02780/577791/126/ | 2026-03-02 | full-text federal dismissal-order mirror | After the Second Circuit affirmed federal dismissal and remanded state claims, the district court dismissed the second amended complaint with prejudice and closed the case. USV was an earlier VC defendant; the operative remand defendants were Uniswap Labs and Adams. (reliability: B; access: full; accessed: 2026-08-21; locator: procedural-history section and final dismissal paragraph)
• ★★ | SEC — Dismissal of Civil Enforcement Action Against Coinbase | https://www.sec.gov/newsroom/press-releases/2025-47 | 2025-02-27 | regulator press release and stipulation route | The Commission and Coinbase filed a joint stipulation to dismiss for policy reasons; the Commission expressly said the decision was not an assessment of the merits. The linked press release does not itself establish the exact prejudice language of a court order. (reliability: A; access: full; accessed: 2026-08-21; locator: opening dismissal paragraph and final not-merits paragraph)
• ★ | Internet Archive — AVC URL history | https://web.archive.org/web/*/http://www.avc.com/* | captures from 2003 onward; accessed 2026-08-21 | web archive route | Recover original timestamps, deleted/changed text, images, and outbound links; wildcard listing is a retrieval route, not substantive evidence. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)
• ★ | Internet Archive — Flatiron Partners URL history | https://web.archive.org/web/*/http://www.flatironpartners.com/* | historical captures; accessed 2026-08-21 | web archive route | Candidate source for contemporaneous team, portfolio, and fund positioning; snapshot completeness and robots gaps must be logged. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)
• ★ | Internet Archive — Union Square Ventures URL history | https://web.archive.org/web/*/http://www.unionsquareventures.com/* | captures from 2003 onward; accessed 2026-08-21 | web archive route | Recover early firm pages, investment lists, thesis language, and team changes without treating current pages as historical snapshots. (reliability: C; access: search-route; accessed: 2026-08-21; locator: no stable locator; use named heading or paragraph in annotation)

Coverage assessment

SCOUT searched the name/date variants Fred Wilson, Frederick R. Wilson, Fred Wilson VC, Euclid Partners, Flatiron Partners, Union Square Ventures, Union Square Ventures 2004/2008/2012, and combinations with interview, transcript, podcast, portfolio, returns, IRR, failure, criticism, lawsuit, complaint, SEC, bankruptcy, and signature companies. It opened official USV/AVC pages, full transcripts, podcast chapter maps, SEC filings, regulator case pages, pension performance, publisher pages, and period reporting rather than relying on search snippets alone.

Direct voice is abundant in AVC/USV writings and the MeetInnovators, Reboot, Mixergy, a16z, and Business Insider appearances. The 2026 USV artifact contributes an official producer-written chapter map, not a voice-verified transcript. The bottleneck is not access to opinions; it is decision-time and economic evidence. AVC migrations across avc.com, usv.com, Paragraph, and avc.xyz create duplicate URLs, lost dates, and comment noise; title/date deduplication treats a post, mirror, transcript, and replay as one underlying work. The New York Times page is paywalled/robots-limited; 25iq, CB Insights, and some conference pages were blocked or partial; archive wildcard pages are retrieval leads; commercial PitchBook/Preqin/Crunchbase results are not included in the core denominator.

The live map contains 80 source routes: 19 ★★★, 42 ★★, and 19 ★. Accepted SCOUT accounting was 19 ★★★ / 41 ★★ / 20 ★; the DEALS phase reported that it replaced an unused low-priority press route with the ★★ Airbnb S-1/A, yielding the present distribution without changing the 80-route total. The current arithmetic and S-1/A row are directly verifiable here; the historical swap attribution is phase-owner reported because the accepted pre-swap artifact is not retained in this file. This did not alter the already reconciled WRITINGS denominator of 34 priority URLs / 33 underlying works or the TALKS denominator of seven priority appearances. The MBA Mondays corpus and 2009 math pair each count as one underlying work. The 61 priority routes span WRITINGS/TALKS, PROFILE chronology, DEALS/vehicles/performance, and REVIEW/scrutiny; stars rank decision relevance within each phase rather than defining one combined denominator. Partial, metadata-only, and search-route items remain outside an evidence denominator until their underlying artifact is retrieved.

Highest-value retrieval leads, in order: (1) Flatiron pledge-fund agreements, Chase/SoftBank capital-account and distribution records, and 1996–2003 LP reports; (2) USV investment memos, partner attribution, board minutes, and vehicle cash-flow schedules for Twitter, Etsy, MongoDB, Coinbase, Zynga, Tumblr, and Kik; (3) underlying LP reports for press-reported IRRs and the LACERA USV 2012 cash-flow/residual schedule; (4) current operative pleadings and dockets in Patel and Grabski plus archived Flatiron/early-USV pages; (5) authorized, page-located copies of The Power Law and Hatching Twitter, and recoverable 25iq quotation originals.

For deals, start with Wilson's 2007 self-reported 32 direct investments, explicit Wilson-authored investment announcements, verified board roles, and SEC ownership records. Build separate tables for Wilson-attributed, shared/collective, and vehicle-only positions. The live USV company directory is a candidate firm universe, not Wilson's denominator; the LACERA row is one LP commitment; reported press IRRs remain reported; public-company market value is not fund cash return. GeoCities, for example, should be credited to Jerry Colonna unless contrary decision-date evidence emerges, and Zynga's filing shows shared rather than personal power.

The criticism/enforcement sweep found USV's diversity appraisal, Kik/Kin SEC enforcement, adverse reporting around profile mythology, and active portfolio-company litigation leads. It did not establish a court or regulator finding of wrongdoing by Wilson personally. The operative pleadings/current dockets and Wilson’s status in each matter remain retrieval requirements; recovered orders establish only their stated procedural postures. Search-result allegations and procedural rulings remain labeled as such.

Historical controls include period NVCA data, investigative books, and independent founder/partner accounts. Completed repository peers are deliberately narrow: the Martin Casado profile helps with public-doctrine and partner/firm attribution, and the Trae Stephens profile helps with interested-speaker and board-role boundaries. Neither is a close consumer-network or New York ecosystem peer; no completed direct peer should be forced into the comparison.

Source notes