Brad Feld

1. Snapshot and testable edge thesis

Feld's proposed edge is technical-product judgment compounded by public teaching and networks, converted through a concentrated collective partnership. Product use and themes focus learning; transparent writing/Techstars/Give First increase trusted access; independent partner decisions resist social proof; meaningful ownership, reserves, board service and duration preserve upside. [researcher inference · moderate] product · themes · network

The thesis is most plausible in early software, Internet infrastructure, developer tools, networks/marketplaces and technically legible products from the 1990s through cloud-era venture. It is weaker for biological/regulatory science, hardware efficacy, late-stage price and public-market timing. [researcher inference · moderate]

Predictions: earlier warm/geographically distributed leads; direct product-use records; explicit pre-deal theme fit; independent partner votes; high ownership; long service; and network-sourced deals that outperform after stage/theme/price controls. Falsifiers: no advantage in a complete funnel, top-three dependence, post-hoc themes, consistently different lead partners, network underperformance, or weak cash returns after fees/reserves. [researcher inference · strong] profile edge

2. Core philosophy

1. Use the product: observe workflow and founder response, while checking whether the investor represents the actual user. [investor-stated · strong] product essay
2. Use themes, not narrow sectors: accumulate causal category knowledge without pretending to forecast one product. [investor-stated · strong] themes
3. Decide independently and collectively: record each partner's view; do not outsource judgment to syndicate prestige. [investor-stated · strong] decisions
4. Treat seed as commitment: reserve and serve, rather than buy cheap options. [investor-stated · strong] seed
5. Own enough for a win to matter: fund/partner capacity, dilution and follow-on are portfolio decisions. [investor-stated · moderate] Next 2018
6. Give first with boundaries: contribute before an immediate return is known, while preserving capacity/conflict clarity. [investor-stated · strong] definition
7. Support, do not shadow-manage: tell the truth, recruit, connect and challenge while the CEO decides. [investor-stated · strong] EFI 04:39
8. Stop when the system fails its design: FG Angels and Crucible show that walking away can dominate forcing activity. [documented behavior · strong] FG Angels

3. Sourcing

Feld's public sourcing system has six channels: technical/product curiosity; prepared themes; founder/portfolio referrals; Techstars and mentorship; Boulder/regional community; and partner funds. The channel evolved from personal angel/operator relationships into institutional network-driven investing. [researcher inference · strong] Techstars history · network

For every lead record: introducer, prior relationship, theme date, product-use date, economic interests, other vehicles eligible, partner sponsor, cold channels searched, populations missing, and independent references. A warm lead saves search cost but is not quality evidence. [researcher inference · strong]

Techstars, personal angels, Mobius, Foundry direct funds, partner funds and personal LP positions must never be pooled. Mentorship can be genuine and still create commercial access; the control is disclosure and allocation, not denial of dual benefit. [researcher inference · strong] vehicle table

4. Picking

Product. Use it or observe real users; map pain, frequency, time-to-value, retention, switching, reliability and buyer/user mismatch. Investor delight is not sufficient. [researcher inference · strong]

Founder/team. Look for obsession translated into action, learning after feedback, truth under bad news, recruiting and complementary capability. A founder need not agree with the VC; response quality matters. [investor-stated · strong] Stanford

Theme/market. State the causal shift, budget/user, boundary, adoption evidence and disconfirmers. Avoid themes broad enough to retrofit any company. [researcher inference · strong]

Decision card. Require independent partner votes, dissent, price/ownership, initial/follow-on capacity, next milestone, kill condition, governance owner, conflicts and evidence that would reopen a pass. Fitbit shows re-entry; iRobot shows category error. [researcher inference · strong] decision post

5. Deal and portfolio mechanics

Feld's 2010 seed policy stated checks from $250,000 to $10 million, all-partner engagement and willingness to follow on. The 2015 capacity post described fixed $225 million funds, roughly 30 companies/fund and about ten new investments/year. The 2018 Next description said ownership could exceed 30% across the lifecycle. These are dated policy statements, not universal company facts. [investor-stated · strong] seed · passes · Next 2018

A reusable portfolio model needs opportunity count, target ownership, dilution, initial check, reserves, partner/board capacity, correlation, duration, and concentration. Select requires a fresh opportunity-cost comparison for each “winner”; Next requires separate direct and partner-fund sleeves and fee look-through. [researcher inference · strong]

No public source supplies full reserve ratios, recycling, pro-rata use, loss ratio, write-off timing, sale discipline or ex-top-three returns. Acquisition consideration and IPO ownership remain proxies until cash flows are known. [researcher inference · insufficient public record]

6. Vehicle and incentive fit

Personal angel capital offers flexible duration but concentrated personal risk. Mobius/Foundry closed-end funds owe duties to LPs and face fees, carry, reserves, key-person and distribution constraints. Select concentrates follow-on capital in marked winners; Next adds growth and partner-fund exposure; FG Angels distributes economics/rights across syndicate backers; Techstars has separate institutional interests; Crucible introduced public-shareholder/redemption/deadline incentives. [researcher inference · strong] vehicle map

Ask for each deal: which legal vehicle; who approves; who benefits; duration; fee/carry layers; target ownership; reserve burden; liquidity horizon; geography/regulation; board/information rights; allocation conflicts; and succession/key-person coverage. [researcher inference · strong]

The 2022 final fund changes the operating equation: no future franchise fundraising can be assumed to retain partners/staff, while portfolio duties may last a decade. [researcher inference · strong] final fund

7. Team attribution and work with founders

Foundry's original four-equal-partner structure used collective investment decisions; later documents describe six or seven equal partners. Attribution must separate origin, mentorship, thesis sponsor, vote, capital, board, recruiting, follow-on, exit and narration. [researcher inference · strong]

Zynga is attributed to Brad in the firm's theme-distribution post; SendGrid is Ryan McIntyre-led/mentored; Fitbit establishes Feld board service but not sole sourcing; all three filings show fund/entity securities under shared control. [contemporaneous record · strong] theme attribution · SendGrid · filing map

Founder work should include direct truth, recruiting, references, financing strategy, crisis response and fiduciary challenge. Efficacy needs founder/board evidence; public doctrine alone cannot prove value-add. [researcher inference · moderate]

8. Views on the craft

Feld treats venture as a power-law craft constrained by fund mathematics and human attention: a small number of wins must matter, while partner service and reserves cannot scale without limit. Greatness is independent judgment, sufficient ownership, honest LP communication, direct founder support, learning from failure, and stewardship through liquidity—not deal celebrity. [researcher inference · strong] quick-pass arithmetic · LP candor

Access and selection interact: Techstars, writing and partner funds can improve access; themes, product use and collective votes attempt selection. A network is not selection proof, and public teaching is not behavior proof. [researcher inference · strong]

Risk is not one score. Separate product, market, founder/team, financing, ownership, governance, platform, regulation, hardware/science, security, reputation, concentration, key-person and vehicle risks. Name the evidence that reduces each and the person authorized to stop. [researcher inference · strong]

Interliant requires acquisition/integration, cash, reporting and conflict gates. Molekule requires independent efficacy and recapitalization gates. FG Angels requires partner-time gates. Crucible requires deadline/incentive/walk-away gates. iRobot requires a thesis-adjacency gate. [researcher inference · strong] labs

Failure review must preserve allegations versus findings, company versus investor causality, and unknown alternatives. A candid retrospective is learning evidence, not an acquittal or proof of foresight. [researcher inference · strong]

9. Signature bets and decision process

The signature cases are deliberately heterogeneous: FeedBurner tests product-use/protocol judgment; Zynga tests distribution and Feld/team attribution; Fitbit tests dissent and re-entry; SendGrid tests developer infrastructure and Ryan McIntyre credit; Moz tests culture against an ordinary outcome; Interliant and Molekule test failure; iRobot tests a pass; FG Angels and Crucible test stopping. [researcher inference · strong] nine no-hindsight labs

Readers should answer each lab’s invest/pass, kill, ownership/terms/capacity and next-check prompts before the reveal. A signature decision remains collective: sponsor, vote, capital, board, follow-on and outcome are separate. [researcher inference · strong] people map

Feld publicly prefers net cash-on-cash over short/intermediate IRR and, in 2025, again emphasizes cash-on-cash. That disciplines mark worship but is not a complete sell rule. [investor-stated · strong] Next 2018 · Alpha transcript

A vehicle-specific exit memo should separate company trajectory, security, cost, realized proceeds, remaining shares, dilution, reserves, taxes/fees/carry, LP liquidity, concentration and follow-on opportunity. Holding a winner can preserve convexity; it can also concentrate a fund and defer distributions. [researcher inference · strong]

No public record reveals Foundry's tender/IPO lockup sales, distribution-in-kind rules, recycling or decision owners for the main cases. Therefore the dossier does not infer a universal “hold forever” rule. [researcher inference · insufficient public record]

10. Mental models and maxims

The supported index is: use the product; themes as falsifiable filters; independent decisions and evidence-based reopening; seed as service/reserve commitment; ownership with concentration re-underwriting; Give First/network boundaries; CEO-support governance; and bounded experiments/walk-away rules. [researcher inference · strong]

Their dated origins run from Feld’s 2006 product essay and Foundry’s 2008 themes through 2010–12 seed/decision doctrine, Techstars/Give First, 2018 lifecycle ownership, 2024 CEO support, and the FG Angels/Crucible stops. They are reconstructions, not eight verbatim maxims.

1. Freeze the theme and why-now before the pitch.
2. Use/observe the product and log evidence, not taste.
3. Collect independent partner views before social proof.
4. Record the dissent reason and re-entry evidence.
5. Treat seed as full service and reserve commitment.
6. Set ownership/check/capacity before enthusiasm.
7. Pass quickly and explain “not us” honestly.
8. Separate economics and control in terms.
9. Mentor with directness, consent and conflicts.
10. Give first within a time/capacity budget.
11. Re-underwrite follow-ons against external alternatives.
12. Stop experiments when their operating model fails.
13. Report company, security, vehicle and LP return separately.

[researcher inference · strong] Each step traces to the 24-source writing analysis; it is a synthesized checklist, not a verbatim Feld algorithm.

11. Evolution and contradictions

The system evolves from operator/product learning (1980s–2006), to themes and a four-partner early-stage model (2006–13), to Select/Next/partner funds (2013–22), and finally a six-partner finite-fund stewardship phase (2023–26). Durable beliefs are product evidence, founder learning, independent judgment, long relationships and candor. [researcher inference · strong]

Unresolved tensions remain: Stanford's 1985 Feld Technologies start versus Foundry's 1987 account; combined “over 40 years” entrepreneur/investor shorthand versus distinct investing from 1994; $225m/$232m early funds; four versus six/seven partners; themes versus network-driven sourcing; founder control versus high ownership/veto/boards; seed commitment versus FG Angels; and cash-on-cash preference versus public mark/IRR evidence. [researcher inference · strong] chronology · network evolution

[researcher inference · strong] The pattern is coherent but not uniformly proven; each attractive rule has a failure mode in the same record.

12. Failures and limits

Interliant demands acquisition/integration, cash, reporting and conflict controls; Mobius VI controls vintage economics; Molekule demands independent efficacy and recapitalization gates; Crucible tests deadline/walk-away; FG Angels tests capacity; iRobot tests adjacency; Moz and Mattermark test ordinary outcomes. [researcher inference · strong] failure ledger

Limits include missing IC/funnel/cash-flow/partner-role records, winner publicity, team/vehicle attribution, changing fund architecture, era effects and no matched base rates. Failure review must separate allegation from finding and company from investor causality. [researcher inference · insufficient public record]

[researcher inference · strong]

13. Historical operating environment

Feld's formative operator/angel period benefited from expanding PC/Internet markets; Mobius confronted the dot-com crash; Foundry's first decade captured social/mobile/cloud with relatively small funds; Select/Next emerged during abundant private capital; Crucible/Molekule crossed the zero-rate/SPAC reset; the final fund now faces slower liquidity and long private duration. [researcher inference · strong] historical context

Rules must be formation-date specific. Product use was especially informative for early Internet applications; high ownership in a $225m early fund differs from a $750m multi-sleeve fund; an interim IRR under zero-rate marks differs from cash distributions after reset. [researcher inference · strong]

No quantified matched base-rate dataset was located for stage/theme/geography/vintage, so “outperformance” remains unproven. [researcher inference · insufficient public record]

14. Modern VC translation (as of 2026-08-22)

What survives: firsthand workflow evidence, prepared themes, independent votes, founder learning, meaningful ownership, reserves, direct truth, network/community and capacity stops. [researcher inference · strong]

What changed: AI products can demo before they are reliable; model/data rights, evaluation drift, security, safety, inference economics and distribution can dominate. Open-source/cloud diffusion weakens superficial moats; regulation is fragmented; later-stage private marks can delay loss recognition; remote/global networks increase access and conflicts. [researcher inference · strong]

Apply by requiring an evaluation suite and failure log, lawful data/model rights, buyer/user workflow evidence, security/red-team controls, unit economics at scale, rollback, independent references, and vehicle-specific cash/ownership. Misuse includes calling a demo “product love,” labeling every AI company a theme fit, using Give First for undisclosed access, or treating a famous network as diligence. [researcher inference · strong]

15. Comparison with completed peers

This is a process comparison, not a ranking. Wilson and Breyer are completed dossiers; Winblad provides the completed enterprise-software operator comparison. [researcher inference · strong]

16. The strongest case against greatness

Feld's public brand and prolific self-documentation may make a team/fund record look personal and systematic. The complete denominator is absent; Zynga/Fitbit/SendGrid could dominate; SendGrid is Ryan-led; company transaction values are not returns; public themes may be retrospective umbrellas; network access may be homophily; Give First may create commercial reciprocity; and large multi-vehicle funds contradict original simplicity. [researcher inference · strong]

Hard adverse records exist: Mobius VI at 0.6x/-5.80% for one LP/date, Interliant bankruptcy, Molekule common cancellation, Crucible liquidation, FG Angels stop, iRobot pass, and ordinary/insufficient Moz/Mattermark results. [contemporaneous record · strong] failure ledger

What would overturn the skeptical case: full cash flows with strong net DPI across vintages; partner-role evidence showing repeat decisive contribution; a complete pass/write-off denominator; source-channel performance; and independent founder evidence of causal board value. [researcher inference · insufficient public record]

17. Analyst study guide, glossary, and go deeper

Recommended order: (1) profile chronology; (2) five best talks; (3) writing playbook; (4) answer the labs before reveals; (5) apply mental models; (6) read review last.

Glossary:

• Theme: dated causal search hypothesis broader than a sector.
• Give First: contribute without knowing immediate return; not boundless altruism.
• Partner fund: external venture manager backed through a Foundry vehicle; not a direct company.
• Select: follow-on-only vehicle for existing Foundry companies.
• Next: vehicle with growth-company and partner-fund sleeves.
• Beneficial ownership: filing concept of voting/disposition power; not automatically personal economics.
• DPI / TVPI / MOIC / IRR: distributions, total value, multiple, and time-weighted return; not interchangeable.
• No-hindsight lab: formation evidence and decision prompts shown before outcome.

Go deeper: request dated IC/pass/follow-on memos; partner/board role logs; legal vehicle agreements; company security/cash-flow ledgers; audited LP reports; Techstars allocation/referral data; Interliant/Molekule/Crucible governance records; and independent founders across wins, ordinary outcomes and failures. [researcher inference · strong]

Study warning: Do not copy personality, network breadth, large ownership or final-fund structure. Copy only the process that fits your vehicle, competence, evidence, capacity and fiduciary duties—and retain the adverse ledger beside the doctrine.

Pareto 80/20 — what drove the record

Public evidence supports only directional ranking, not a career return table. Company value, affiliated beneficial ownership, financing amount, IPO stake, acquisition consideration, and LP-reported fund return are different quantities. No complete security-level cash-flow record lets us decide whether Zynga, Fitbit, or SendGrid generated the largest realized Foundry gain, let alone Feld's personal gain. [researcher inference · insufficient public record]

Evidence boundary and searched denominator

Foundry’s official investments route serves or redirects to a live collection that explicitly is not a complete lifetime ledger and mixes direct companies, exits, and partner funds. A one-session HTML/link inventory on 2026-08-22 found 136 distinct linked detail-card URLs after removing navigation and exact URL duplicates; because types and omissions are not stable in one export, this is a search frame, not a company denominator. The firm separately reports 200+ companies and nearly 50 venture-fund investments over its life. [contemporaneous record · strong] final-fund disclosure

There are nine labs total, but only seven are company lineages; FG Angels is an experiment and Crucible a special-purpose vehicle. No reconstruction fraction is reported against 136 or 200+ because the objects and omission rules differ. Career-wide coverage and hit rate are not computable. [researcher inference · strong]

Evidence hierarchy: issuer/regulator/LP records control securities, legal outcome and disclosed return; decision-date posts control stated thesis/role; later speech controls retrospective witness; acquirer releases control company transactions; researcher arithmetic is labeled. A company valuation or transaction never becomes fund or personal return. [researcher inference · strong]

Historical market, regulatory, and base-rate context

No formation-date numerical base-rate series matches Feld’s stages, themes, geographies and vehicles. LP snapshots use different vintages/dates and cannot be pooled. Context changes diligence questions; it does not excuse losses or manufacture a benchmark. [researcher inference · insufficient public record]

Vehicle economics and incentives

Across vehicles, duration and liquidity differ: personal angels can hold without LP deadlines; closed-end funds face reserves, key-person and distribution obligations; Select concentrates into marked winners; Next adds growth and look-through manager fees; a syndicate distributes rights across backers; and a SPAC has redemption/deadline mechanics. Public sources do not disclose complete fees, carry, recycling, side letters, pro-rata allocation, reserves or GP commitments, so incentive claims remain structural rather than quantified. [researcher inference · insufficient public record]

Governance evolved from four collective partners to six in 2023 and seven in the 2018 system description. Vehicles span US/Canada direct investing, broader partner-fund relationships, securities/privacy/advertising rules, hardware/science validation and public-market regulation. Stakeholders include founders, employees, customers, LPs, syndicate backers, partner managers and public shareholders; the final fund makes succession and key-person coverage especially material. [researcher inference · strong] Next 2018 · Foundry 2022

People, capital, and influence map

Team-credit rule: discovery, introduction, mentorship, thesis sponsorship, partnership vote, fund capital, board service, recruiting, follow-on, exit decision, later narration, and company building are separate columns. “Foundry invested” never means “Brad sourced,” and “Brad was a director” never means “Brad owned the fund's shares personally.” [researcher inference · strong]

Comprehensive deal ledger

Crucible and FG Angels are analyzed as vehicle/experiment cases, not forced into the 32-company ledger. Techstars is an institution and exposure channel, not one portfolio-company line. [researcher inference · strong]

Ex-ante forensic case studies and decision labs

Each lab freezes public formation-era evidence before revealing the later outcome. Answer four prompts first: invest/pass; primary kill condition; ownership/terms/capacity requirement; evidence required for the next check.

Lab 1 — FeedBurner (Mobius, win candidate)

Formation packet. A young feed-management product sits inside an emerging protocol/distribution shift. Feld can use it directly and observe founder response to product feedback. Market structure, durable pricing, Google platform risk, ownership, and reserves are unknown. decision-era post [contemporaneous record · strong]

Decide before reveal. Invest/pass? Kill if publishers do not retain or monetization depends on one platform. Require product cohort evidence, protocol dependency map, founder update, price/ownership and reserve budget.

Reveal and audit. Google announced the acquisition in 2007. That validates company importance, not investor return or product-use causality. The lesson is to combine lived product evidence with platform-dependency underwriting. Google [researcher inference · strong]

Lab 2 — Fitbit (Foundry/Select, win candidate after pass)

Formation packet. A connected-device opportunity initially fails Foundry's collective decision; later evidence changes the vote. Hardware, inventory, consumer retention, accuracy, privacy, margin, and platform competition are material. decision post [retrospective witness · moderate]

Decide before reveal. Pass initially unless the dissent reason can be named. Reopen only on decision-changing evidence. Kill on weak repeat use/unit economics. Set hardware working-capital and follow-on ceilings; require cohort, returns/warranty, margin, accuracy and channel evidence.

Reveal and audit. Feld joined the board in August 2010; Foundry/Select later held disclosed securities and Select invested roughly $10m in Series D. The filing validates governance/capital, not why the first pass changed or net return. A 2018 derivative opinion let certain allegations survive pleading but left proof unresolved; neither Feld nor Foundry was a defendant. Fitbit S-1/A · Delaware opinion [contemporaneous record · strong]

Lab 3 — SendGrid (Foundry, win; team-credit control)

Formation packet. A Techstars team offers email-delivery infrastructure to developers. Ryan McIntyre had mentored the company and writes the investment announcement; deliverability, abuse, customer concentration, platform dependence and reliability are key. Foundry post [contemporaneous record · strong]

Decide before reveal. Invest if usage/retention, delivery quality and developer pull beat incumbents. Kill on abuse/reputation or undifferentiated price. Require ownership, reliability/security milestones, reserve, and a named board lead.

Reveal and audit. Foundry disclosed 29.64% pre-IPO at the filing basis; Ryan and Feld were former directors; Twilio later completed an approximately $3b acquisition. Credit Ryan's lead/mentorship, shared Foundry capital/control, company execution, and separate governance. S-1 [contemporaneous record · strong]

Lab 4 — SEOmoz/Moz (ordinary/disappointing outcome)

Formation packet. A transparent founder, explicit TAGFEE culture, SEO software/community, and a stated $18m Foundry financing. Risks: channel dependence on search platforms, founder/CEO fit, recurring economics, culture scaling, and follow-on concentration. Feld post [investor-stated · strong]

Decide before reveal. Invest only if recurring retention/expansion and platform-independent value justify price. Kill if culture substitutes for economics. Require CEO/leadership plan, cohort metrics, platform scenarios, ownership and reserve cap.

Reveal and audit. Moz was acquired by iContact in 2021; consideration and Foundry proceeds are not public in the retrieved primary record. The case is ordinary/insufficient—not a proven loss—and shows cultural fit is not an outlier-return metric. company post [researcher inference · moderate]

Lab 5 — Interliant (failure and adverse governance)

Formation packet. Hosting consolidation in the late-1990s capital cycle, with acquisition integration, reporting, cash burn, leverage/financing, related-party and board-control risks. Contemporaneous IC/board packets are unavailable; do not pretend the later lesson was known. [researcher inference · insufficient public record]

Decide before reveal. Pass without stand-alone unit economics, integration owner, audited reporting, cash/covenant downside, conflict map, and stop authority. Kill on delinquent reporting or acquisitions exceeding integration capacity.

Reveal and audit. Interliant filed Chapter 11 in August 2002 after a reported 2001 $48.9m net loss and later delinquent reports. A committee alleged transfers/fiduciary breaches involving Feld-linked parties; the accessible decision addresses procedure and does not adjudicate Feld's personal liability. SEC order · decision [contemporaneous record · strong]

Lab 6 — Molekule (failure after substantial capital)

Formation packet. A hardware/air-quality company requires independently validated efficacy, regulatory/advertising discipline, manufacturing, warranty, channel, margin and financing evidence. In 2020 NARB found support for some PECO claims but recommended discontinuing or modifying several pollution-elimination, superiority, quantified-performance, testing and health-relief claims; Molekule agreed. This is a balanced warning, not proof the product was wholly ineffective. Foundry participated across 2018–22 through multiple entities; exact thesis memo is unavailable. NARB announcement [contemporaneous record · moderate]

Decide before reveal. Pass absent blinded efficacy and viable unit economics. Kill on claims exceeding evidence or repeated recapitalization without risk reduction. Cap cumulative ownership/capital and define independent science, gross margin, returns and cash milestones.

Reveal and audit. A filing aggregates about $58.44m of Foundry-entity legacy common-equivalent capital and 23.7% post-merger ownership; it names Feld among five shared managing members and as the required Foundry board nominee. The 2023 Chapter 11 led to common cancellation for no consideration. That supports severe equity impairment, not an exact $58.44m fund or personal loss because securities, transfers, recoveries and entity allocations differ. 13D · 8-K · 13D/A [contemporaneous record · strong]

Lab 7 — iRobot (miss / anti-portfolio)

Formation packet. A robotics company with a consumer vacuum product appears outside Mobius's software focus. Hardware/manufacturing/channel risk is high; embedded software, data, brand, and autonomy may create a broader system. Feld post [retrospective witness · moderate]

Decide before reveal. Pass under a strict software mandate unless system/software economics satisfy an adjacency exception. Kill on gross margin, reliability, inventory or weak repeat platform. Require round/access, ownership, working capital, and counterfactual portfolio fit.

Reveal and audit. Feld later loved the product and named the pass. iRobot's later success makes it a miss, but no public proposed terms prove Mobius could have invested or that relaxing the rule improves expected fund return. [researcher inference · strong]

Lab 8 — FG Angels (experiment stopped)

Formation packet. AngelList syndicates might extend Foundry's network and seed access at lower check size. Risks: partner time, adverse selection, signaling, founder service, conflicts, reserves, and dispersed backer obligations. launch-to-postmortem route [documented behavior · strong]

Decide before reveal. Run only with a time-box, responsible partner, success/learning metric, capital/attention budget, conflict rule, and support plan. Kill if per-company effort exceeds the model despite company quality.

Reveal and audit. Foundry stopped new investments after two years because the program consumed more capacity than expected. Existing obligations continued. This is a credible stop rule; absent cohort economics, it does not prove financial failure or success. [documented behavior · strong]

Lab 9 — Crucible Acquisition Corp. (special-vehicle failure)

Formation packet. A $225m SPAC sponsored by Foundry Next, chaired by Feld, has a deadline to find and complete a combination. Risks include sponsor/public-shareholder incentive divergence, diligence under time pressure, warrant dilution, market regime change, and opportunity cost. 424B4 [contemporaneous record · strong]

Decide before reveal. Accept only with a strict target thesis, conflict/valuation process, walk-away authority, deadline plan, sponsor-capital loss limit, and liquidation preference over a poor merger.

Reveal and audit. Crucible did not complete a combination and announced liquidation/return of trust funds. This is a failed search but may evidence discipline versus forcing a transaction; sponsor/warrant economics remain unreconstructed. 8-K [contemporaneous record · strong]

Portfolio-level forensics

The visible economic record is plausibly concentrated, but missing cash flows prevent ranking even the top three. A valid test needs cost, follow-on, dilution, distributions, remaining value, fees/carry and allocation plus ex-top-one/ex-top-three sensitivity. [researcher inference · strong]

Failure, miss, and anti-portfolio postmortems

[researcher inference · strong] The table does not imply Feld caused each outcome. It shows what evidence a student needs to test doctrine against adverse and ordinary cases.

Fund-performance controls and arithmetic

The reproducible WSIB distribution ratio is $49,282,966 ÷ $82,354,694 = 0.5984x, consistent with the table's rounded 0.6x. No other arithmetic invents undisclosed cost, fee, carry, tax, security conversion, or allocation. [researcher inference · strong] WSIB p.6

Uncertainty and retrieval ledger

Until these records emerge, the defensible conclusion is that Feld helped design and narrate an influential venture system with several major visible company outcomes and meaningful adverse controls. The public record supports process study more strongly than personal-alpha or career-return ranking. [researcher inference · strong]

Coverage and evidence base

The priority written inventory contains 24 ★★★/★★ rows in the source map: 22 full works and two mixed-access routes with full author-origin posts but only partial book access (Venture Deals and Do More Faster). No priority writing is inaccessible or metadata-only. All 24 were read, deduplicated, and given their own exact-schema note and analysis below: 24 mapped − 24 notes − 24 analyses = 0 missing. Feld's separate books page remains a ★ metadata route and does not enter this priority denominator. [researcher inference · strong]

The corpus spans three different voices: Feld-signed personal posts, coauthored books/doctrine, and Foundry firm posts. “Feld published” does not mean “Feld alone decided”; Foundry deal, fund, and team claims retain the named partner, collective vehicle, and beneficiary. [researcher inference · strong]

Source-by-source analysis

1. Product-Focused Venture Capital (2006)

• Context and question: Before Foundry's public theme framework, Feld asks what a software-oriented VC can learn by actually using a product. primary · note
• Argument: Product use exposes friction, founder responsiveness, and an investor's ability to contribute; FeedBurner and NewsGator make the rule concrete. [investor-stated · strong]
• Evidence and example: The signed contemporaneous essay is strong for the stated process, but it supplies no same-basis comparison of used-versus-not-used products or investment returns.
• Hidden assumptions: Investor taste represents target users; founders should react to one power user's feedback; product legibility arrives early enough for diligence.
• Decision implication: Require a dated product-use log, map persona mismatch, and score the founder's response to evidence rather than agreement.
• Contradiction/transfer: Enterprise infrastructure, biotech, and safety-critical AI may not be directly usable; “I used it” can create familiarity bias. The 2026 translation is instrumented workflow observation, not consumer intuition. [researcher inference · strong]

2. How I Think About Seed Investing as a VC (2010)

• Context and question: Seed programs were proliferating, so Feld distinguished institutional commitment from cheap optionality. primary · note
• Argument: A Foundry seed should fit a theme, involve all partners, support follow-on, and stand independent of syndicate signaling; stated initial checks ranged $250,000–$10 million. [investor-stated · strong]
• Evidence and example: Seven named seed examples form a prospective test set; no complete outcome/reserve ledger follows them.
• Hidden assumptions: A small equal partnership can spend full attention on small checks; fund reserves and ownership can absorb follow-ons.
• Decision implication: Write the seed's next-round commitment, target ownership, reserve ceiling, partner owner, milestone, and stop condition at entry.
• Contradiction/transfer: FG Angels later used lighter-touch syndication and stopped for capacity. The failure refines rather than erases the commitment rule. [documented behavior · strong]

3. Make Your Own Decisions (2012)

• Context and question: Feld responds to the tendency to infer quality from other VCs. primary · note
• Argument: Foundry invested in only roughly 5–10% of actively evaluated companies, required all four partners, and treated one partner's sustained resistance as a pass. [investor-stated · strong]
• Evidence and example: Fitbit and SEOmoz are valuable false-negative/re-entry cases showing that a first “no” need not be permanent.
• Hidden assumptions: Partners form sufficiently independent views and dissent is informational rather than political.
• Decision implication: Record pre-meeting priors, independent votes, reasons for dissent, and the evidence required to reopen a pass.
• Contradiction/transfer: Social proof can still enter through founder/network access. A unanimous system also lets one risk-averse partner suppress unusual bets. [researcher inference · strong]

4. Why We Pass Quickly on Things (2015)

• Context and question: A four-partner firm with a fixed fund cannot pursue every credible company. primary · note
• Argument: Approximately 30 companies per fund and ten new investments per year made stage, geography, theme, check size, and partner energy capacity controls—not judgments of founder worth. [investor-stated · strong]
• Evidence and example: The post provides funnel arithmetic but not a logged pass denominator or false-negative rate.
• Hidden assumptions: The fund size and partner model are themselves optimal; quick rejection cost is lower than delayed discovery value.
• Decision implication: Publish filter failures promptly, distinguish “not us” from “bad,” and audit passes that later become large outcomes.
• Contradiction/transfer: Later Next/partner-fund vehicles widened the opportunity set, and the team expanded; capacity filters must be recalibrated by vehicle. [documented behavior · strong]

5. VCs: Don't Bullshit (2014)

• Context and question: Feld uses Mobius and Foundry fundraising to attack performative certainty. primary · note
• Argument: State problems plainly, accept rejection, and build LP trust through truth; the post recalls roughly 90 meetings and 20 immediate rejections for Foundry's first fund. [retrospective witness · moderate]
• Evidence and example: Named LP relationships and Mobius adversity make this richer than generic advice, but no subscription ledger or contemporaneous meeting notes are public.
• Hidden assumptions: Candor is recognized and rewarded; gatekeepers do not punish adverse disclosure.
• Decision implication: Keep a dated risk register shared with LPs/founders and separate unresolved problems from confident forecasts.
• Contradiction/transfer: Transparency cannot cure poor economics. The principle should be tested against adverse board/fund records, not likes or brand warmth. [researcher inference · strong]

6. Term Sheet Series / Venture Deals (2005–2011)

• Context and question: Feld and Jason Mendelson sought to reduce information asymmetry in venture documents. primary · note
• Argument: Explain economics and control separately so founders understand negotiation rather than defer to lawyers/VCs. [investor-stated · strong]
• Evidence and example: The accessible 20-post series is decision-relevant; it is less than 20% of the later book and remains coauthored.
• Hidden assumptions: Disclosure creates bargaining competence despite unequal capital, time, and information.
• Decision implication: Model dilution, liquidation preference, participation, option-pool, anti-dilution, board and protective provisions in scenarios.
• Contradiction/transfer: Teaching terms may also standardize investor-favorable norms. Modern translation adds SAFEs, secondaries, pay-to-play, AI/data rights, and financing-control failure modes. [researcher inference · strong]

7. The Techstars Mentor Manifesto (2011)

• Context and question: An accelerator needed behavioral rules for mentors. primary · note
• Argument: Be direct, responsive, Socratic, network-helpful, relationship-oriented, and founder-centered without controlling. [investor-stated · strong]
• Evidence and example: The attribution chain matters: Jon Bradford prompted the discussion, David Cohen authored the manifesto, Feld amplified it.
• Hidden assumptions: Mentors know when advice is useful and can avoid portfolio/conflict incentives.
• Decision implication: Set mentor disclosure, response-time, conflict, confidentiality, and escalation rules; collect founder-side feedback.
• Contradiction/transfer: A dense mentor network can create consensus, status pressure, and extraction. “Mentor whiplash” is a predictable boundary. [researcher inference · strong]

8. Definition of Give First (2020)

• Context and question: A slogan had diffused enough to require a precise definition. primary · note
• Argument: Contribute without knowing the immediate return, but not with no boundaries or expectation that reciprocity never matters. [investor-stated · strong]
• Evidence and example: Feld dates the phrase from “Give Before You Get” in 2012 through the hashtag and 2020 refinement.
• Hidden assumptions: Contributions and future benefits can be separated; privileged actors are not merely branding normal networking as generosity.
• Decision implication: Record beneficiary, cost, conflict, consent, expected public/private benefit, and a capacity ceiling.
• Contradiction/transfer: FG Angels shows that contribution without a bounded operating model can overload a small partnership. [documented behavior · strong]

9. Foundry Group Next 2018

• Context and question: Foundry explains why a $750 million seventh vehicle fits rather than abandons its model. primary · note
• Argument: Combine direct early/growth ownership, partner funds, long holding, and a seven-equal-partner network; prefer net cash-on-cash over short/interim IRR. [investor-stated · strong]
• Evidence and example: The post discloses prior vehicle sizes, US/Canada focus, thematic continuity, and “often over 30%” ownership; it does not disclose cash flows.
• Hidden assumptions: Network breadth and larger capital do not dilute partner attention or increase adverse selection.
• Decision implication: Model each sleeve's fee layer, decision owner, duration, overlap, reserve burden, and look-through concentration.
• Contradiction/transfer: Earlier Foundry celebrated four partners/no associates and fixed $225m funds; the 2018 system is materially different. [researcher inference · strong]

10. Status of Our FG Angels Experiment (2016)

• Context and question: Foundry tested AngelList syndicates as a lightweight extension of its network. primary · note
• Argument: Stop new investments when a supposedly low-effort experiment consumes scarce partner capacity, even if individual companies remain promising. [documented behavior · strong]
• Evidence and example: Two years of operation and an explicit stop make this a real control against innovation theater; cohort economics remain absent.
• Hidden assumptions: Partner time was the binding constraint and discontinuing was better than staffing/delegating.
• Decision implication: Pre-register time budget, owner, learning goal, economic threshold, continuation date, and obligations to existing companies.
• Contradiction/transfer: The experiment's syndicate/light-touch form conflicts with the 2010 claim that seed is full institutional commitment. [researcher inference · strong]

11. Who Are We? (2008)

• Context and question: A newly formed firm defines its architecture. primary · note
• Argument: Small, experienced, equal partners make collective decisions and remain directly involved; Boulder is a network node, not a geographic constraint. [investor-stated · strong]
• Evidence and example: Founding-era timing is valuable, while organizational claims need later behavior and deal records.
• Hidden assumptions: Equality produces challenge and coverage rather than ambiguity or veto gridlock.
• Decision implication: Map vote, sponsor, board, reserves, conflicts, key-person, and succession responsibility for every company.
• Contradiction/transfer: Six equal partners in 2023 and a decade-long final-fund wind-down test the original architecture's durability. [researcher inference · strong]

12. What Does Foundry Group Invest In? (2008)

• Context and question: Foundry needs a repeatable search map without becoming sector-bound. primary · note
• Argument: Themes capture protocols, distribution shifts, and interaction changes broad enough for discovery but narrow enough for accumulated expertise. [investor-stated · strong]
• Evidence and example: The dated list constrains hindsight; later Gnip, SendGrid, Fitbit, Mapbox and others can be tested for actual entry fit.
• Hidden assumptions: Themes are stable, mutually informative, and not so broad that every software company fits.
• Decision implication: Freeze the theme definition, causal mechanism, disconfirmers, entry evidence, and adjacency budget before meeting a company.
• Contradiction/transfer: The system later became “network-driven,” which may improve access or replace thesis-led discovery with referrals. [researcher inference · strong]

13. Foundry Group Select (2013)

• Context and question: Successful portfolio companies needed more capital than the early-stage vehicle could prudently supply. primary · note
• Argument: A separate $225 million fund invests up to $25 million only in existing winners. [investor-stated · strong]
• Evidence and example: Fitbit's S-1/A confirms a Select Series D investment around $10 million, independently demonstrating vehicle use. Fitbit filing
• Hidden assumptions: Follow-on selection beats external opportunities and added concentration fits LP risk.
• Decision implication: Re-underwrite each follow-on against a fresh external opportunity, ownership value, reserves, and total look-through exposure.
• Contradiction/transfer: Calling companies “winners” before liquidity risks mark-driven selection; no audited Select DPI/TVPI was found. [researcher inference · insufficient public record]

14. Our Newest Fund — Next (2016)

• Context and question: Foundry expands beyond Select into partner funds and growth companies. primary · note
• Argument: A $500 million dual-sleeve vehicle treats network managers and later-stage direct investments as reinforcing. [investor-stated · strong]
• Evidence and example: Official mandate is clear; sleeve allocations, fee look-through, and outcomes are private.
• Hidden assumptions: Managers share high-quality deals and information without allocation conflict; growth checks remain inside competence.
• Decision implication: Separate direct and fund-return denominators; record referral, overlap, fees, information rights, and allocation.
• Contradiction/transfer: Network benefit is causal self-report, not a channel-matched outcome study. [researcher inference · insufficient public record]

15. Network-Driven Investing (2022)

• Context and question: Foundry retrospectively names the mechanism connecting its partner-fund and direct-company strategies. primary · note
• Argument: Dense trusted relationships improve discovery, expertise, recruiting, and collaboration. [investor-stated · strong]
• Evidence and example: The post gives institutional history and early-fund size of $232m each; it does not identify every network-sourced deal or counterfactual.
• Hidden assumptions: Warm-network quality exceeds homophily, circular validation, and adverse selection.
• Decision implication: Tag lead channel, conflicts, independent references, excluded populations, and outcome by source cohort.
• Contradiction/transfer: The $232m description conflicts with $225m shorthand; preserve possible commitment-versus-investable/rounding scope rather than choose silently. [researcher inference · strong]

16. The Final Foundry Fund (2024)

• Context and question: A partnership designed not to be generational states its terminal capital-raising decision. primary · note
• Argument: Foundry 2022 is final, but active investing and portfolio stewardship continue for at least a decade. [investor-stated · strong]
• Evidence and example: The firm reports 200+ direct companies and nearly 50 fund investments; current 2025 direct voice gives roughly 80 active companies/50 fund positions.
• Hidden assumptions: A long closed-end wind-down preserves recruiting, incentives, key-person coverage, and follow-on discipline.
• Decision implication: Treat “final fund” as a lifecycle state: map remaining new-deal authority, reserves, boards, succession, staffing, and LP communication.
• Contradiction/transfer: Current promotional summaries may compress counts; neither disclosure becomes a Feld-only denominator. [researcher inference · strong]

17. Your First Big Success — What's Next? (2014)

• Context and question: Feld narrates how early liquidity, bubble adversity, and health shaped later institutions. primary · note
• Argument: Success creates options, not immunity; deliberate reinvention and personal systems matter after a win. [retrospective witness · moderate]
• Evidence and example: Named periods—Feld Technologies, angel deals, Mobius, Techstars, Foundry, 2013 depression—support chronology, not causal alpha.
• Hidden assumptions: Later meaning accurately reconstructs earlier decisions.
• Decision implication: Pair career retrospectives with date-fixed records and include health/capacity in key-person risk.
• Contradiction/transfer: Public candor is evidence of reflective practice, not proof that contemporaneous governance anticipated the risks. [researcher inference · strong]

18. Every Generation Learns the Same Lessons (2018)

• Context and question: Interliant becomes a cycle/capital/governance failure case. primary · note
• Argument: Fast growth, acquisition integration, and abundant capital can repeat familiar errors when pattern memory is ignored. [retrospective witness · moderate]
• Evidence and example: SEC records Chapter 11 and delinquent reporting; the bankruptcy decision records committee allegations. SEC order · decision mirror
• Hidden assumptions: The lesson selected after failure was the binding cause, rather than product, market, financing, control, or conflict.
• Decision implication: Reconstruct acquisition cadence, cash, covenant, reporting, board escalation, alternatives, related-party conflicts, and stop authority.
• Contradiction/transfer: Allegations are not adjudicated liability; the record must not imply Feld wrongdoing. [contemporaneous record · strong]

19. Roomba / iRobot Pass (2005)

• Context and question: A product Feld loved had failed Mobius's software-category filter. primary · note
• Argument: Category discipline protects focus but can misclassify hardware with software/network economics. [retrospective witness · moderate]
• Evidence and example: The near-period post names the pass and product use; round, price, allocation, and vote remain unknown.
• Hidden assumptions: The counterfactual investment was available and would have fit vehicle timing/ownership.
• Decision implication: For a thesis mismatch, write the adjacent category that would make it eligible and evidence that could reopen the pass.
• Contradiction/transfer: A celebrated miss is not proof that relaxing the rule raises expected returns; test all drift deals, not only iRobot. [researcher inference · strong]

20. Small Company / Big Company (2015)

• Context and question: Portfolio startups repeatedly depend on slow incumbents. primary · note
• Argument: Named ownership, executive sponsorship, economic alignment, and process clarity determine whether a partnership helps rather than kills a startup. [investor-stated · strong]
• Evidence and example: The numbered rules reflect experience but anonymize cases and omit denominator/outcomes.
• Hidden assumptions: The startup has negotiating alternatives and can survive the incumbent's cycle.
• Decision implication: Underwrite concentration, termination, data/IP, integration, payment, champion departure, and fallback distribution.
• Contradiction/transfer: Partnership advice can normalize dependency. Kill a deal when survival rests on one uncontracted champion. [researcher inference · strong]

21. SEOmoz, TAGFEE, and Me (2012)

• Context and question: Feld publicly explains Foundry's relationship with Rand Fishkin and a culture of transparency. primary · note
• Argument: Explicit values and radical communication can be investable operating systems, not decoration. [investor-stated · moderate]
• Evidence and example: A stated $18m financing and Feld board role anchor involvement; later leadership/outcome does not reveal return.
• Hidden assumptions: Public transparency predicts private truth and execution; values scale through leadership changes.
• Decision implication: Test values with hiring, firing, bad-news, customer, and financing decisions rather than slogans.
• Contradiction/transfer: The case belongs in the ordinary/disappointing ledger precisely because cultural fit did not guarantee outlier economics. [researcher inference · strong]

22. Foundry Group Next (2015)

• Context and question: Feld previews a vehicle and relates it to LP history/succession. primary · note
• Argument: Growth and partner funds can deepen a network without building a permanent institution; partner careers, not a brand dynasty, define the horizon. [investor-stated · strong]
• Evidence and example: The UTIMCO relationship and anti-legacy position predate the 2024 final-fund statement.
• Hidden assumptions: LP duration aligns with finite partner careers and long portfolio obligations.
• Decision implication: Put succession, key-person, post-investment staffing, and terminal fundraising policy into vehicle design.
• Contradiction/transfer: A final firm can still require institutional continuity longer than any partner's active career. [researcher inference · strong]

23. Why Do We Care About Techstars? (2009)

• Context and question: Feld explains a young accelerator's purpose beyond investment return. primary · note
• Argument: Mentorship and dense community increase entrepreneurial learning and ecosystem capacity. [investor-stated · strong]
• Evidence and example: Techstars' official history independently fixes founders and dates; no channel-adjusted return data proves sourcing advantage.
• Hidden assumptions: Mentor incentives align; access is inclusive; advice quality exceeds confusion; commercial and public-good goals coexist transparently.
• Decision implication: Separate Techstars, personal, Foundry, and later partner-fund capital; disclose allocation and conflicts.
• Contradiction/transfer: Ecosystem-building can become privileged access dressed as public service. Measure participant outcomes and excluded populations. [researcher inference · strong]

24. Do More Faster (2010)

• Context and question: Feld and David Cohen convert accelerator lessons into a founder playbook. primary · note
• Argument: Bias toward action, learning, focus, and founder resilience; the short-chapter form reflects mentor plurality. [investor-stated · moderate]
• Evidence and example: The accessible author post controls purpose/coauthorship but not the full book's evidence or rule ownership.
• Hidden assumptions: Speed raises learning faster than it compounds legal, safety, product, or organizational debt.
• Decision implication: Define the decision a faster experiment changes, its guardrails, rollback, owner, and evidence threshold.
• Contradiction/transfer: “Faster” is dangerous in regulated AI, health, security, and finance without right-to-operate gates. [researcher inference · strong]

Cross-source synthesis

The writings evolve in four layers: technical product observation (2005–06), small-partnership filters and full commitment (2008–15), community/teaching as network infrastructure (2009–20), and multi-vehicle network/lifecycle design (2013–24). The later network doctrine does not replace the earlier theme/product method; it changes sourcing, scale, attribution, and conflicts around it. [researcher inference · strong]

The strongest repeatable playbook is: use the product; define a dated theme; decide independently; treat seed as commitment; reserve enough for winners; reject quickly when the vehicle cannot serve; support the CEO without control; teach openly; and stop experiments that exceed capacity. The strongest counterevidence is Interliant, iRobot, SEOmoz, FG Angels, Molekule, Mobius VI, and the early-fund/partner-count inconsistencies. [researcher inference · strong]

Intellectual chronology, evolution, and contradictions

The intellectual chronology begins with product use and category misses (2005–06), formalizes collective themes and seed commitment (2008–12), expands through culture, quick passes, LP candor, Select/Next and explicit experiment stops (2013–18), then defines Give First, network-driven investing and a final-fund lifecycle (2020–24). [researcher inference · strong]

Durable ideas are firsthand evidence, independent judgment, full commitment, direct truth and long relationships. Contradictions are productive controls: iRobot versus software focus; Fitbit/Moz re-entry versus quick passes; FG Angels versus full seed service; four partners versus six/seven; themes versus referral networks; cash-on-cash preference versus mark/IRR-heavy public records; and finite partnership versus decade-long stewardship. [researcher inference · strong] iRobot · network

Decision playbook and modern VC translation as of 2026-08-22

A reusable decision sequence is: date the theme and why-now; use or observe the product; collect independent partner votes; state dissent/reopen evidence; name eligible vehicle and conflicts; size price/ownership/reserves; define next milestone and kill; assign founder-service/governance; and pre-commit follow-on/exit metrics. [researcher inference · strong] seed doctrine · decision doctrine

For AI/cloud/developer-tool investing, translate product use into traceable workflow evaluation: permissions, model/data provenance, evaluation drift, security, latency, inference cost, switching, buyer/user separation, and failure recovery. Translate themes into falsifiable causal maps rather than labels; translate Give First into disclosed, bounded contributions; translate networks into channel-tagged sourcing audited for homophily and conflict; and translate cash-on-cash preference into a vehicle-specific cash-flow ledger rather than an excuse to ignore interim risk. [researcher inference · strong]

Recommended order: read product focus → themes → seed commitment → independent decisions → quick passes → Mentor Manifesto/Give First → Select/Next/network → final fund → Interliant/iRobot/FG Angels. Then use the decision labs and models to test whether the attractive rules survive adverse cases.

Best five writings and unresolved record

1. Product-Focused Venture Capital (2006) — clearest origin of firsthand product diligence.
2. How I Think About Seed Investing (2010) — most complete stage, reserve and commitment policy.
3. Make Your Own Decisions (2012) — rare funnel, veto, false-negative and re-entry evidence.
4. Foundry Group Next 2018 — fullest vehicle, ownership, partner and return-metric statement.
5. Status of FG Angels (2016) — strongest operational experiment postmortem.

Unresolved are the complete nine-book texts, a deduplicated full Feld/Foundry archive denominator, every priority-quality deal announcement/byline, deleted or migrated posts, full contemporaneous Mobius records, and source-by-source outcome testing for the seven named 2010 seeds. These gaps are logged as partial/★ routes rather than silently represented as read. [researcher inference · insufficient public record] source coverage

Coverage and evidence base

The priority spoken inventory contains seven ★★★/★★ appearances in the source map. Six provide full official or publisher transcript routes; MeetInnovators provides the complete audio plus an edited transcript. None is inaccessible or metadata-only. All seven have exact-schema notes and analyses below: 7 mapped − 7 notes − 7 analyses = 0 missing. Publisher transcripts are strong for Feld's words but not for host framing; no inferred audio quote is used where complete text was unavailable. [researcher inference · strong]

Source-by-source analysis

1. MeetInnovators — Brad Feld / Foundry Group (2010)

• Publication and historical context: A long interview near Foundry's first-fund era and before its later Select/Next expansion. canonical interview · note
• Reasoning chain: Technical founder experience → small company sale → roughly 40 personal angels → institutional venture → small-partnership redesign. [retrospective witness · moderate]
• Evidence/examples: NetGenesis, Critical Path, and Harmonix anchor the personal-angel period; the recollection of roughly 40 old Mobius companies in 2010 bounds ongoing work, not performance.
• Assumptions and access boundary: Edited transcript/audio cannot validate exact cash flows, decision dates, or interview omissions.
• Decision/biography implication: Separate operator liquidity, personal angel capital, SoftBank/Mobius fund roles, and Foundry capital before discussing “Feld's investments.”
• Evolution/contradiction: The early personal sale is described as a few million/two million across retellings; preserve approximate self-report. [retrospective witness · moderate]

2. Stanford eCorner — Great Entrepreneurs Go Out and Do (2011)

• Publication and historical context: Official full transcript of a public Stanford talk after Techstars/Foundry formation. official transcript PDF · note
• Reasoning chain: Build, learn, sell, angel-invest, join institutional venture, fail, then construct a preferred partnership model; founder/investor selection is mutual. [investor-stated · strong]
• Evidence/examples: Feld describes starting Feld Technologies with $10, the 1993 sale, later Mobius/Foundry transitions, and founder obsession as a key signal.
• Assumptions and access boundary: The official transcript controls speech, but career dates remain retrospective; its 1985 start/Dave Jilk joining in 1987 conflicts with Foundry's 2008 history saying Feld created the company with a partner in 1987.
• Decision/biography implication: Use founder action and obsession as a reference hypothesis, then test ethics, learning, team-building, and market evidence rather than glorify intensity.
• Evolution/contradiction: “Great entrepreneurs do” anticipates Do More Faster, but speed without safety/legal gates is not universal. [researcher inference · strong]

3. SaaStr — Feld Thoughts Live (2017)

• Publication and historical context: A software-investing stage interview after Next's formation. publisher transcript · note
• Reasoning chain: Mobius Fund I was “extremely well,” Fund II a “complete disaster,” and Fund III depended on one company; large organization complexity helped motivate Foundry's smaller equal-partner system. [retrospective witness · moderate]
• Evidence/examples: The 70-person/ten-partner organizational memory and fund contrast are rare adverse direct voice; WSIB later independently shows poor Mobius VI LP performance on one basis. WSIB report
• Assumptions and access boundary: Fund numbering/scope in speech may not map directly to legal vehicles; adjectives are not audited multiples.
• Decision/biography implication: Treat firm architecture and fund size as investment variables, not backstage administration.
• Evolution/contradiction: Foundry itself later expanded partners and vehicle complexity; the original anti-scale lesson needs re-testing. [researcher inference · strong]

4. Tim Ferriss — Brad Feld (2020)

• Publication and historical context: Long-form pandemic-era interview revisiting dot-com collapse, relationships, mental health, and work systems. publisher transcript · note
• Reasoning chain: Crisis and overextension impaired judgment/capacity; off-grid breaks, therapy, direct communication, and partnership systems became operating controls. [retrospective witness · moderate]
• Evidence/examples: The direct account adds human/key-person risk to the Mobius transition and explains later public mental-health advocacy.
• Assumptions and access boundary: Therapeutic meaning and causality are personal testimony; do not medicalize or infer portfolio outcomes.
• Decision/biography implication: A venture firm should monitor partner load, dependency, communication, leave coverage, and health-related key-person continuity without invading privacy.
• Evolution/contradiction: Extreme endurance and “do more faster” rhetoric require counterweights: recovery, boundaries, and explicit capacity stops. [researcher inference · strong]

5. Growth Everywhere (2017)

• Publication and historical context: A founder-marketing interview focused on career, network, and firm construction. timestamped page · note
• Reasoning chain: Small operator exit funded angel experimentation; repeated company exposure generated patterns and a network that later supported funds and Techstars. [retrospective witness · moderate]
• Evidence/examples: Feld self-reports an approximately $2m company sale and about 40 angel investments; no transaction documents or angel cash-flow ledger were located.
• Assumptions and access boundary: Promotional editing and rounded memory limit precision; named wins are vulnerable to survivorship bias.
• Decision/biography implication: Operator capital, relationship capital, and institutional capital are separate inputs with separate beneficiaries.
• Evolution/contradiction: “Investor since 1987” official shorthand conflicts with a first personal angel period after the 1993 sale and “accidental VC” around 1996. [researcher inference · strong]

6. Entrepreneurs for Impact (2024)

• Publication and historical context: Climate-founder interview during Foundry's final-fund lifecycle. interview · note
• Reasoning chain: The board investor supports rather than commands the CEO; mentorship and direct truth increase the founder's decision quality. [investor-stated · strong]
• Evidence/examples: At 04:39 Feld says that while he supports the CEO he works for her; the interview also discusses mentoring emerging fund managers.
• Assumptions and access boundary: A stated governance ideal is not founder-side evidence that every board relationship operated this way.
• Decision/biography implication: Define board service as truth, recruitment, pattern access, crisis support, and fiduciary challenge—not shadow management.
• Evolution/contradiction: All-partner veto and high ownership create real power even when rhetoric emphasizes founder control. [researcher inference · strong]

7. Alpha Partners — Driving Alpha (2025)

• Publication and historical context: Current direct voice after the 2024 final-fund announcement. publisher transcript · note
• Reasoning chain: A finite partnership can stop fundraising while working through roughly 80 active companies and about 50 fund positions; realized cash-on-cash is more meaningful than interim IRR. [investor-stated · moderate]
• Evidence/examples: Feld says the 2022 fund closed in 2023, only a handful of new investments remained, and mentorship shapes his “third third.”
• Assumptions and access boundary: Counts are rounded self-report and time-specific; sponsor/host deal lists are promotional and not personal attribution.
• Decision/biography implication: Evaluate late-career performance through remaining deployment, reserves, boards, exits, team continuity, and cash distributions—not new-fund cadence.
• Evolution/contradiction: TMRS interim IRRs are useful LP snapshots but, by Feld's own metric preference, cannot settle outcome. TMRS report [contemporaneous record · strong]

Cross-talk synthesis

Across fifteen years, the durable direct-voice structure is: technical operator learning, deliberate transition through angel/Mobius adversity, small-partnership architecture, mutual founder selection, community compounding, CEO-support governance, and a finite institutional lifecycle. The interviews add material absent from polished firm announcements—fund disasters, organizational overload, mental health, approximate angel counts, and the difference between interim IRR and cash-on-cash. [researcher inference · strong]

The most important contradictions remain live rather than “resolved”: Stanford's 1985 Feld Technologies start versus Foundry's 1987 team history; the current About page's combined “over 40 years” shorthand versus a distinct 1994 angel/around-1996 institutional-investor chronology; Mobius fund labels versus legal-vintage records; four-partner simplicity versus later six/seven equal partners; and founder-control rhetoric versus board/veto/ownership power. [researcher inference · strong] current bio · 1994 account

Decision playbook and modern translation as of 2026-08-22

Use a direct-voice interview as doctrine evidence, then require behavior and outcome controls. For a live deal: identify vehicle; freeze theme and product evidence; collect independent partner votes; map founder/investor power; state ownership/reserve/capacity; log bad-news channels; and pre-commit what would trigger follow-on, pass, leadership intervention, or stop. In AI companies, add data/model rights, evaluation drift, inference cost, security, safety, regulatory authority, and rollback. [researcher inference · strong]

Best five talks and unresolved record

1. Stanford (2011): best complete career/process transcript.
2. SaaStr (2017): strongest fund/organization self-critique.
3. Tim Ferriss (2020): best capacity, health, and crisis account.
4. MeetInnovators (2010): earliest rich angel/Mobius/Foundry boundary.
5. Alpha Partners (2025): current final-fund and metric preference.

Unresolved: complete audio-verifiable transcripts for several later podcasts; contemporaneous 1985/1987 Feld Technologies records; exact Mobius legal-vehicle mapping to spoken Fund I/II/III; independent founder accounts across wins and failures; current final-fund authority/capacity; and any direct-voice discussion of Molekule or Crucible. [researcher inference · insufficient public record]

Model 1 — use the product, then correct for whose experience it is

• Rule — Use or closely observe the product before investing; record friction and founder response, then test whether your experience represents the target user. [researcher inference · strong] product essay
• Evidence status — Strong that Feld stated and repeatedly illustrated the practice; moderate that it improved selection; insufficient public record for a matched used/not-used outcome cohort. [researcher inference · insufficient public record]
• Provenance and origin — Feld's operator background and 2006 FeedBurner/NewsGator essay provide the clearest dated origin; iRobot later shows use after a pass. [investor-stated · strong] iRobot
• Mechanism and reasoning — Direct use reveals time-to-value, workflow, reliability and founder learning that decks/references can hide. Persona mismatch, familiarity and novelty can contaminate the signal. [researcher inference · strong]
• Historical and vehicle context — Consumer/SMB Internet software was directly usable by a technical partner; enterprise, hardware and regulated science require customer observation and independent validation. Partner time remains a closed-end-fund cost. [researcher inference · strong]
• When to use — Software, developer tools, consumer products, marketplaces, connected devices and workflows where repeated behavior can be observed before the round. [researcher inference · moderate]
• How to apply it — Log persona, task, baseline, activation, repeated use, failure, switching, telemetry, buyer/user difference, and founder response; require two target-user observations and one disconfirming reference. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented application: FeedBurner usage and founder dialogue. decision-era source
  • Documented application: NewsGator as another lived-product case in the same essay. note
  • Exception/miss: Feld loved Roomba after Mobius passed iRobot, showing use and category fit can disagree. iRobot lab
• Evidence — A dated signed essay controls doctrine, a separate near-period pass post supplies an adverse case, and Google controls FeedBurner's later acquisition. No public decision/outcome cohort establishes causality. [contemporaneous record · strong] Google
• Limits, failure modes, and exception ledger — Investor taste can overfit affluent technical users; a good demo can hide retention, unit economics, safety or buyer friction; founder agreement can be mistaken for coachability. [researcher inference · strong]
• Evolution and contradictions — The product habit expands into themes/APIs and connected devices. iRobot contradicts a narrow software lens; Molekule shows firsthand product experience cannot replace independent efficacy. [researcher inference · strong] Molekule filing
• Modern VC translation (as of 2026-08-22) — For AI, require an evaluation suite, traceable data/model rights, workflow adoption, hallucination/failure logs, security, latency, inference cost and rollback; do not call prompt delight diligence. [researcher inference · strong]

Model 2 — themes are falsifiable filters, not forecasts or labels

• Rule — Define a durable causal shift before a company appears; use it to focus search and accumulated learning, then reject post-hoc theme assignment. [researcher inference · strong] Foundry themes
• Evidence status — Strong for the founding-era stated method and repeated theme language; moderate for consistent execution; insufficient for theme-level return/outperformance. [researcher inference · insufficient public record]
• Provenance and origin — Foundry's 2008 theme essay distinguishes themes from sectors; the 2011 distribution post assigns deals/partners and makes part of the map testable. [investor-stated · strong] distribution
• Mechanism and reasoning — A causal frame improves search, question quality and pattern accumulation. Breadth preserves serendipity; explicit boundaries and disconfirmers prevent rationalization. [researcher inference · strong]
• Historical and vehicle context — $225m/$232m early funds needed capacity filters; Select/Next and partner funds widened stage/object exposure, making one theme denominator less coherent. [researcher inference · strong] network essay
• When to use — Protocol transitions, developer platforms, distribution shifts, human-computer interaction, networks/marketplaces and new technical architecture with multiple product expressions. [researcher inference · moderate]
• How to apply it — Date the causal change, beneficiary, evidence, stage/geography, non-fit boundary, adjacency budget, expected lead channels, disconfirmers, and examples that would falsify it. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented fit: Zynga mapped to distribution, with Brad deal credit and filing-controlled shared vehicle. theme post · S-1
  • Researcher-applied fit: SendGrid as developer/email infrastructure, while Ryan retains lead credit. Foundry post
  • Exception: iRobot fell outside the software category although the product later mattered. pass note
• Evidence — Founding and decision-date firm posts are independent of later issuer filings; the filing controls securities/role, not causal theme. No complete deal/pass map tests false positives/negatives. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Themes can be too broad, mutually overlapping, consensus-driven, or indifferent to price; they can hide platform, regulatory and hardware risks. [researcher inference · strong]
• Evolution and contradictions — The original thematic model evolves into network-driven sourcing. Network information can strengthen a prepared mind or invert it by letting referrals define the theme. [researcher inference · strong] network note
• Modern VC translation (as of 2026-08-22) — An AI theme must specify workflow, buyer, data/model advantage, distribution, evaluation, cost curve, rights and disconfirmers—not merely “AI-native.” Audit outcome by theme date and lead channel. [researcher inference · strong]

Model 3 — decide independently, preserve dissent, and reopen on evidence

• Rule — Form partner views independently, avoid syndicate social proof, treat sustained dissent as information, and reopen a pass only when named evidence changes. [researcher inference · strong] decision post
• Evidence status — Strong for Feld's public all-four-partner rule and two re-entry examples; insufficient for full vote logs or comparative decision accuracy. [researcher inference · insufficient public record]
• Provenance and origin — The 2012 essay reports a 5–10% active-evaluation conversion, all-partner decision process, Fitbit/SEOmoz reversals and social-proof resistance. [investor-stated · strong]
• Mechanism and reasoning — Independent priors reduce informational cascades; dissent finds hidden risk; an explicit reopen threshold preserves learning without letting charisma erase a no. [researcher inference · strong]
• Historical and vehicle context — Four equal partners could deliberate every early deal. Six/seven partners, growth checks, partner funds and a final-fund tail change information load and veto cost. [researcher inference · strong] Foundry 2022
• When to use — High-uncertainty early-stage deals, crowded rounds, famous founders, inside rounds, partner-fund referrals, re-entry after a pass and follow-ons in marked winners. [researcher inference · moderate]
• How to apply it — Collect written pre-discussion priors, confidence, kill reason, sponsor, conflicts and evidence-to-reopen; reveal votes simultaneously; log decision and later outcome without rewriting reasons. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented re-entry: Fitbit moved from no to investment; filing later records Feld's board role. Feld · S-1/A
  • Documented re-entry: SEOmoz/Moz also overcame initial resistance; later ordinary outcome keeps the rule honest. Feld
  • Exception: A quick-pass system can discard iRobot-like adjacency before decisive evidence arrives. iRobot
• Evidence — Signed posts establish the rule and examples; issuer filing independently establishes Fitbit role/securities; later company acquisition establishes outcome but not decision causality. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Unanimity can create lowest-common-denominator selection, political veto, delay and sponsor gaming; dissent is not automatically insight; repeated reopening can waste capacity. [researcher inference · strong]
• Evolution and contradictions — Foundry's partnership expanded while still describing equal partners. The unchanged doctrine must operate differently at six/seven people and across fund managers. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — Use blinded IC memos before celebrity metrics or co-investor lists; preserve model/data/security dissent; require a dated delta memo for any reopened AI deal. [researcher inference · strong]

Model 4 — seed is a service-and-reserve commitment, not a cheap option

• Rule — Make a seed only if the vehicle will treat it as a real investment: named partner attention, follow-on willingness, target ownership, and explicit stopping conditions. [researcher inference · strong] seed essay
• Evidence status — Strong for stated 2010 policy and later lifecycle vehicles; moderate for observed large ownership; insufficient for reserve ratios and seed-cohort returns. [researcher inference · insufficient public record]
• Provenance and origin — Feld separates two personal angel periods from Mobius/Foundry and rejects option-style seed behavior; checks spanned $250k–$10m in that dated policy. [investor-stated · strong]
• Mechanism and reasoning — Full commitment improves founder trust, information and follow-on continuity; reserves protect ownership. It consumes partner/fund capacity and can escalate sunk costs. [researcher inference · strong]
• Historical and vehicle context — Roughly $225m/$232m early funds and ~30 companies made reserves/attention scarce; Select and Next later financed winners and growth; FG Angels tested a different lightweight form. [researcher inference · strong] Select
• When to use — Pre-seed/seed companies inside core competence where a partner can serve, ownership can matter and follow-on milestones fit the vehicle's duration/capital. [researcher inference · moderate]
• How to apply it — Record initial/total check ceiling, ownership/dilution, reserve, partner time, next risk, follow-on threshold, bridge rule, pro-rata owner, governance and kill condition. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented application: Seven named 2010 seed examples create a prospective set, though outcomes remain incomplete. seed note
  • Documented lifecycle: Fitbit received Foundry and later Select capital; filing records about $10m Select Series D. Fitbit S-1/A
  • Exception: FG Angels' supposedly lighter model exceeded partner capacity and stopped new investments. postmortem
• Evidence — Direct doctrine, official fund announcements and an issuer filing triangulate policy/behavior; missing cash-flow and reserve records prevent performance inference. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Commitment can become escalation, signaling harm on no-follow, over-reserving, or weak service spread across too many seeds; bridge capital can postpone failure. [researcher inference · strong]
• Evolution and contradictions — A core all-partner seed model coexisted with FG Angels syndicates and later partner-fund exposure. “Commitment” must be vehicle-specific rather than a brand-wide claim. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — For AI seed, reserve against proof milestones—workflow retention, evaluation reliability, data rights, security and unit economics—not model-demo progress; state compute and future capital needs. [researcher inference · strong]

Model 5 — own enough for wins to matter, but re-underwrite concentration

• Rule — Size ownership and reserves so a successful company can return meaningful fund capital, then re-underwrite price, dilution, concentration and liquidity at every follow-on. [researcher inference · strong] Next 2018
• Evidence status — Strong that Foundry publicly emphasized lifecycle ownership, sometimes above 30%; strong for disclosed stakes in three issuers; insufficient for cost-adjusted net return or optimality. [researcher inference · insufficient public record]
• Provenance and origin — Small fixed early funds/30-company capacity precede Select's winner concentration and Next's growth support. Feld prefers net cash-on-cash over interim IRR. [investor-stated · strong] quick passes
• Mechanism and reasoning — Venture power laws reward ownership in winners; reserves defend it. The same mechanism amplifies wrong theses, late marks, financing needs and single-company risk. [researcher inference · strong]
• Historical and vehicle context — Zynga/Fitbit/SendGrid stakes sit across early funds/Select and shared entities; partner funds add look-through concentration; final-fund liquidity extends duration. [researcher inference · strong]
• When to use — High-conviction companies with proven risk reduction, sufficient market, governance truth and vehicle capacity; never solely because an existing mark rose. [researcher inference · moderate]
• How to apply it — Model base/downside/upside ownership, dilution, total capital, reserve, exit value, time, liquidity, correlation and ex-top-one/ex-top-three fund result; compare each follow-on with a fresh external deal. [researcher inference · strong]
• Examples — Three bounded cases. [contemporaneous record · strong]
  • Documented application: SendGrid's 29.64% pre-IPO Foundry position and later ~$3b company sale. S-1 · Twilio
  • Documented application: Fitbit's 52.7m Class B shares across Foundry-linked entities and Select follow-on. S-1/A
  • Exception: Molekule received about $58.44m affiliated legacy common-equivalent capital yet common was cancelled for no consideration. 13D · 13D/A
• Evidence — Issuer filings control securities/capital and acquirer releases control transactions; none reveals all cost, sale, distribution, fee or allocation cash flows. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Sunk-cost escalation, marked “winner” bias, fund concentration, governance entrenchment, financing dependence, opportunity cost and delayed distributions can overwhelm ownership benefit. [researcher inference · strong]
• Evolution and contradictions — Early fund simplicity expands through Select/Next. Cash-on-cash rhetoric conflicts with a public evidence set dominated by interim IRRs, marks and securities. [researcher inference · strong] TMRS
• Modern VC translation (as of 2026-08-22) — Price AI follow-ons on normalized inference economics, retention, model dependence, security and future compute/capital; show realized/unrealized and direct/look-through exposure separately. [researcher inference · strong]

Model 6 — Give First and networks compound only with boundaries and audits

• Rule — Contribute before an immediate return is known, build durable communities, and treat resulting access as interested network evidence requiring capacity, conflict and inclusion controls. [researcher inference · strong] Give First
• Evidence status — Strong for stated doctrine and institutional behavior through Techstars/partner funds; moderate for access mechanism; insufficient for source-channel outperformance or net social benefit. [researcher inference · insufficient public record]
• Provenance and origin — Techstars began in 2006/07; Mentor Manifesto and “Give Before You Get” preceded the 2020 definition; Foundry later formalized partner-fund network investing. [contemporaneous record · strong] Techstars
• Mechanism and reasoning — Useful contribution creates trust, information flow, referrals, reputation and reciprocity over long horizons. Dense networks also reproduce status, homophily, hidden obligation and allocation conflict. [researcher inference · strong]
• Historical and vehicle context — Community activity spans personal time, accelerator entities, direct funds, fund-of-funds sleeves and LP mentorship—different beneficiaries and fiduciary rules. [researcher inference · strong] network essay
• When to use — Founder education, mentorship, ecosystem gaps, emerging-manager relationships, recruiting/reference networks and long-cycle communities where value can be given without coercion. [researcher inference · moderate]
• How to apply it — Record giver/beneficiary, resource, consent, conflicts, eligible vehicles, capacity, immediate/possible future benefits, inclusion gaps and no-reciprocity condition; audit lead/outcome channels. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented institution: Techstars' four-person founding and first class. official history
  • Documented doctrine: Mentor Manifesto preserves Jon Bradford/David Cohen/Feld attribution. manifesto
  • Exception: FG Angels became more work than intended and stopped new deals. FG Angels
• Evidence — Official history, signed doctrine and operational postmortem independently establish system/limits; no network graph or matched cold-source cohort proves alpha. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Mentorship can create pressure, advice overload, signaling, confidentiality/conflict, insider access, sponsor dependence and burnout; generosity branding can obscure extraction. [researcher inference · strong]
• Evolution and contradictions — Informal founder/community activity became Techstars, a manifesto, a named doctrine and partner-fund strategy. Commercial and public-good benefits increasingly overlap. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — Publish mentor/conflict/allocation rules; compensate hidden labor; measure participation and excluded groups; separate community KPIs from investment funnel/returns; cap partner time. [researcher inference · strong]

Model 7 — guide the CEO, tell the truth, and preserve decision ownership

• Rule — The investor's board role is to surface truth, recruit, connect, challenge and support; the CEO owns company decisions unless fiduciary/legal thresholds require formal action. [researcher inference · strong] EFI interview
• Evidence status — Strong for Feld's stated CEO-support philosophy and multiple board records; moderate for repeated behavior; insufficient for causal value-add or founder experience across outcomes. [researcher inference · insufficient public record]
• Provenance and origin — Feld's operator/board apprenticeship, Mentor Manifesto and 2024 “I work for her” statement supply a long arc. [investor-stated · strong] Stanford
• Mechanism and reasoning — Clear ownership speeds execution; direct truth reduces hidden risk; board network/recruiting closes gaps. Formal fiduciary challenge protects the company when support alone fails. [researcher inference · strong]
• Historical and vehicle context — Feld served on Zynga/Fitbit/SendGrid boards while Foundry vehicles held meaningful stakes and collective veto. Power is real even when rhetoric is founder-centered. [contemporaneous record · strong] Zynga S-1
• When to use — Board service, CEO coaching, hiring, financing, crisis, strategy disagreement, founder transition, misconduct, insolvency and conflicts. [researcher inference · moderate]
• How to apply it — Agree on CEO/board domains, start with bad news, state advice versus vote, document conflicts/recusal, build escalation/whistleblower channels, and evaluate board contribution with founder/management feedback. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented board role: Feld director at Zynga from Nov. 2007; shared vehicle stake remains separate. S-1
  • Documented board role: Feld at Fitbit from Aug. 2010–May 2015; filing controls dates. S-1/A
  • Exception/adverse: Interliant bankruptcy/legal record demonstrates that governance requires reporting, conflict and stop controls beyond supportive rhetoric. SEC order
• Evidence — Issuer/regulator filings independently control roles/outcomes; direct speech controls philosophy; no minutes or broad founder sample controls board behavior. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Founder deference can delay intervention; control can masquerade as advice; high ownership can distort judgment; board access can create conflicts; bad-news culture can still fail. [researcher inference · strong]
• Evolution and contradictions — Mentor language moves from advice norms to explicit CEO service. Collective veto and concentrated ownership remain a necessary counterweight to the no-control narrative. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — Add model/data/security/safety escalation to board dashboards; define who can stop deployment, disclose investor conflicts and preserve management ownership of implementation. [researcher inference · strong]

Model 8 — design experiments with time, capital, and walk-away rules

• Rule — Treat new investing formats as bounded experiments: pre-register learning, capacity and economic thresholds; support existing obligations; stop new activity when the model fails. [researcher inference · strong] FG Angels
• Evidence status — Strong for FG Angels' explicit stop and Crucible liquidation; moderate for a generalized Feld method; insufficient for counterfactual benefit/cost. [researcher inference · insufficient public record]
• Provenance and origin — Feld's startup bias toward action meets hard partnership-capacity lessons from Mobius and public experiments. [researcher inference · moderate] SaaStr
• Mechanism and reasoning — Time-boxes convert novelty into learning; kill rules prevent sunk-cost and deadline pressure; continuing duties preserve trust. Poor metrics can terminate promising systems or conceal economics. [researcher inference · strong]
• Historical and vehicle context — FG Angels was a syndicate overlay; Crucible a regulated public SPAC; Foundry Next and partner funds were larger experiments retained. Each has different stakeholders and stop costs. [researcher inference · strong]
• When to use — New sourcing channels, scout/syndicate programs, accelerators, geographic offices, fund sleeves, SPACs, platform services and AI diligence tools. [researcher inference · moderate]
• How to apply it — Specify hypothesis, baseline, owner, capital/time, stakeholders, conflicts, leading/lagging measures, review date, continuation/kill threshold, obligations and rollback. [researcher inference · strong]
• Examples — Three bounded cases. [researcher inference · strong]
  • Documented stop: FG Angels stopped new investments after unexpected work. postmortem
  • Documented walk-away: Crucible liquidated without forcing a business combination. 424B4 · 8-K
  • Failure boundary: Molekule follow-on capital did not prevent common cancellation, emphasizing company-level recap gates. 13D/A
• Evidence — Firm postmortem and SEC SPAC/restructuring filings independently show stops/outcomes; initial hypotheses and counterfactuals remain private. [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Easy-to-measure activity can crowd out long learning; sponsors can redefine goals; kill rules can abandon option value; existing companies still require service after program stop. [researcher inference · strong]
• Evolution and contradictions — Foundry stopped FG Angels/Crucible yet expanded through Select/Next/partner funds and later ended fundraising entirely. The pattern is selective experimentation, not anti-growth. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — For AI scouts/platform services, pre-register source quality, founder benefit, conflicts, human review, privacy/security, partner-time and cost; stop before tooling becomes uncontrolled selection infrastructure. [researcher inference · strong]

Evidence gaps and candidate models

[researcher inference · insufficient public record] Eight promoted models each contain twelve labeled components, at least three bounded examples/exceptions, four-plus citation placements, multiple independent underlying sources, an adverse boundary, and dated modern transfer. Candidate ideas remain hypotheses rather than hidden models.

What the evidence supports

The defensible claim is that Feld helped build an unusually explicit venture operating system: technical/product observation, themes, independent collective decisions, seed commitment, ownership/reserves, open teaching, mentorship/community, founder-support governance, partner-fund networks and bounded experiments. The record supports material roles in Foundry, Techstars, Zynga/Fitbit governance and decades of public doctrine. [researcher inference · strong] profile · philosophy

The record does not prove lone sourcing, a complete denominator, audited career outperformance, personal ownership of Foundry securities, or causal board value. SendGrid is Ryan McIntyre-led; Foundry stakes have shared control; major company transactions lack cost/distribution data; one early Foundry LP snapshot is strong while one Mobius vintage is poor. [researcher inference · strong] Pareto

What could falsify the thesis

The edge thesis is falsified by a complete funnel showing no advantage in product/theme/network sourcing after stage/price controls; returns disappearing outside the top three; themes assigned after decisions; partner-role records assigning repeated decisive work elsewhere; network leads underperforming cold leads; or founder evidence contradicting service doctrine. [researcher inference · strong]

It would strengthen with complete IC/pass/follow-on records, channel tags, partner roles, security cash flows, audited vintage DPI/TVPI/MOIC, ex-top-one/ex-top-three results, Techstars/partner-fund allocation records and independent founder/board accounts across wins, ordinary outcomes and failures. [researcher inference · insufficient public record]

Attribution and denominator audit

The reconstructed chain separates Feld Technologies, personal angels, Intensity/SoftBank, Mobius, each Foundry vehicle, FG Angels, Techstars entities, personal LP positions and Crucible. Discovery, mentorship, thesis sponsorship, vote, vehicle capital, board, follow-on, exit and narration remain separate. [researcher inference · strong] vehicle table · people map

The official collection search observed 136 distinct detail-card URLs in one session but the dynamic page is incomplete and mixes companies, exits and partner funds. Foundry reports 200+ companies/nearly 50 fund investments. The material research subset contains 32 company lineages plus FG Angels/Crucible; seven companies plus two noncompany cases support nine labs; 25 company lineages remain unresolved. No ratio against the mixed 136 or rounded 200+ scope is used as performance. [researcher inference · strong] funnel

Interliant allegations are described as allegations in a procedural bankruptcy decision, never as adjudicated Feld liability. Molekule's mixed NARB findings, shared governance, aggregate legacy capital and common cancellation are bounded rather than converted into total product failure or an exact fund loss. Techstars' later organizational criticism is not attributed personally to Feld. [researcher inference · strong] Molekule controls · Techstars controls

Transfer limits for modern VCs

The transferable processes are dated themes, firsthand workflow evidence, independent IC votes, explicit re-entry evidence, vehicle/ownership/reserve discipline, direct founder truth, bounded networks and pre-committed stop rules. They transfer as hypotheses, not as proof that a large ownership position or famous community will outperform. [researcher inference · strong] modern translation

Current AI and venture conditions add model/data rights, evaluation drift, security, safety, inference economics, regulation, slow liquidity and large-fund/private-mark risk. Do not copy a 2008 theme, 2010 check range, four-partner process, >30% lifecycle ownership, partner-fund network or final-fund structure without matching stage, competence, fund duration, fees, reserves and key-person coverage. [researcher inference · strong]

Unresolved evidence

The highest-value missing records are complete personal/Mobius/Foundry company and pass ledgers; dated IC and partner-role logs; security costs, sales and distributions; audited full-vintage DPI/TVPI/MOIC on one date; Techstars and partner-fund allocation/source-channel outcomes; founder/board evidence across wins and failures; Interliant merits/board/conflict records; and entity-level Molekule/Crucible economics. [researcher inference · insufficient public record] terminal ledger

Books and several audio/video routes remain partial, the live Foundry collection is incomplete/dynamic, and no matched base-rate dataset covers stage, theme, geography and vintage. These are terminal public-record gaps for this run, not permission to estimate. [researcher inference · insufficient public record]

Corrections made during review

• Replaced a stale About-page citation and removed the invented “formal founding” reconciliation; preserved Stanford's 1985 start/Dave Jilk joining in 1987 against Foundry's 1987 company-with-partner account. profile
• Replaced unsupported “investor since 1987” wording with the current combined “over 40 years” biography shorthand and direct 1994 angel/around-1996 institutional chronology. profile
• Preserved $225m versus $232m early-fund scope/rounding conflict. writings
• Preserved four-partner founding model versus six/seven later equal partners. philosophy
• Reframed thematic evolution into network-driven sourcing rather than treating the systems as static synonyms.
• Assigned SendGrid lead/mentorship to Ryan McIntyre; kept Feld's former-director role and shared vehicle capital separate. lab
• Added filing-controlled Rally and AvidXchange board, shared-control and stake boundaries; neither filing establishes personal ownership or proceeds.
• Added the pleading-stage Fitbit litigation boundary, balanced Molekule advertising warning, and source-matched Techstars institutional criticism.
• Restricted Zynga to Brad-attributed deal/board contribution plus shared vehicle securities; did not call every Foundry distribution deal Feld's.
• Did not convert Fitbit, Zynga or SendGrid shares/transaction value into personal or net fund return.
• Used UTIMCO, WSIB and TMRS only at their LP/date/metric scopes; did not pool vintages.
• Kept Interliant committee allegations distinct from merits/liability findings.
• Treated Molekule's mixed NARB findings and common cancellation as warning/severe impairment, not total product ineffectiveness or an exact $58.44m fund/personal loss.
• Treated Crucible as a liquidated SPAC search vehicle, not a portfolio-company write-off; returned trust capital remains separate from sponsor costs.
• Treated FG Angels as an operational stop with unknown cohort economics.
• Treated Mattermark consideration as unverified; did not repeat a paywalled “<$1m” figure.
• Removed performance fractions against the mixed/incomplete current Foundry collection.

[researcher inference · strong]

Twenty-point acceptance matrix

The matrix has exactly 20 rows. A PASS means the repository requirement is satisfied by the cited public package, not that the underlying investment claim is proven beyond its evidence grade. [researcher inference · strong]

Central-claim citation audit trail

Ordinary-claim sample audit

Independent deal and team cross-checks

Seven checks exceed the five-case floor. [researcher inference · strong]

Arithmetic, parity, and denominator checks

Citation, label, link, and structure QA

Every root document begins H1 then a substantive H2; INVESTMENTS begins with the exact Pareto H2. The 31 notes each require one metadata line and six H2 sections. Material labels use one attribution—investor-stated, contemporaneous record, documented behavior, retrospective witness, or researcher inference—and one support—strong, moderate, provisional, or insufficient public record. [researcher inference · strong]

Canonical external links sit adjacent to high-risk claims; local note links aid navigation but do not substitute for external evidence. Filings/LP PDFs use named tables, form sections or pages; posts/transcripts use headings, timestamps, speaker turns or disclosed locator limits. Quotes remain short and confined mainly to notes. [researcher inference · strong]

Mechanical QA must verify H1/H2 openings, exact note schema, 24/7 parity, 17 philosophy sections, 8×12 model fields, exact 20 review rows, local links/anchors, allowed labels, no malformed Markdown, and bookkeeping consistency. External links remain access-time evidence and may later move. [researcher inference · strong]

Reader conclusion

Study Feld for a transparent, testable system that connects product judgment, thematic preparation, collective independence, ownership/reserves, community, founder service and institutional lifecycle. Study the same dossier to see why team attribution, vehicle incentives, cash-flow evidence, power-law concentration, capacity and adverse records determine what “great investor” can responsibly mean. [researcher inference · strong]

Identity, chronology, firms, and current state

• ★★★ | Brad Feld — About | https://feld.com/about/ | live; accessed 2026-08-22 | first-person/official biography | Current Foundry/Techstars roles, Feld Technologies, nine books, blog, MIT BS/MS, and combined “over 40 years” experience; the live page does not date company formation or the start of investing. (reliability: A; access: full; locator: biography paragraphs)
• ★★★ | Techstars — About and History | https://www.techstars.com/about | live; accessed 2026-08-22 | official institutional history | Names David Cohen, David Brown, Brad Feld, and Jared Polis; incorporation November 14, 2006 and first class 2007. (reliability: A; access: full; locator: “Our History” timeline)
• ★★★ | Foundry — The Final Foundry Fund | https://foundry.vc/blog/2024/01/the-final-foundry-fund/ | 2024-01 | official firm announcement | Final-fund status, no generational succession, 200+ companies, nearly 50 fund investments, and continuing portfolio work; note. (reliability: A; access: full; locator: opening and portfolio-work paragraphs)
• ★★ | Foundry — Announcing Foundry 2022 | https://foundry.vc/blog/2023/05/announcing-foundry-2022/ | 2023-05 | official fund announcement | Decision-date framing for the 2022 vehicle before it was publicly labeled final. (reliability: A; access: full; locator: fund-size, team, and strategy passages)
• ★★ | Foundry — Who Are We? | https://foundry.vc/blog/2008/04/who-are-we/ | 2008-04 | founding-era firm essay | Says Feld created Feld Technologies with a partner in 1987 and began angels while at AmeriData; also documents Foundry's original small equal-partner design; note. (reliability: A; access: full; locator: Brad and team-history paragraphs)
• ★★ | Brad Feld — About Foundry Group | https://feld.com/about-foundry-group/ | live; accessed 2026-08-22 | first-person firm biography | Cross-checks cofounding/partner role but remains self-authored and current-state compressed. (reliability: A; access: full; locator: role and firm-description paragraphs)
• ★★ | MIT — Brad Feld '87, SM '88 | https://alum.mit.edu/ | archive/search route; accessed 2026-08-22 | institutional identity lead | Controls degree-era leads; precise alumni profile route was not stable, so degree claims rely on Feld's official bio. (reliability: B; access: partial; locator: search result/profile metadata)
• ★★ | Alpha Partners — Brad Feld: From Techstars to AI Trends | https://alphapartners.com/podcast/brad-feld-techstars/ | 2025 | publisher transcript | Current self-report: last fund, roughly 80 active companies/50 fund investments, few new deals, cash-on-cash preference; note. (reliability: A for direct voice/B for host framing; access: full; locator: transcript speaker turns)

Priority direct writings and firm doctrine

• ★★★ | Feld — Product-Focused Venture Capital | https://feld.com/archives/2006/08/product-focused-venture-capital/ | 2006-08 | signed essay | Product-use diligence and founder-response test; note. (reliability: A; access: full; locator: FeedBurner/NewsGator examples)
• ★★★ | Feld — How I Think About Seed Investing as a VC | https://feld.com/archives/2010/08/how-i-think-about-seed-investing-as-a-vc/ | 2010-08-02 | signed essay | Seed as commitment, $250k–$10m, reserves, all-partner engagement, syndication agnosticism; note. (reliability: A; access: full; locator: paragraphs 8–19)
• ★★★ | Feld — Make Your Own Decisions | https://feld.com/archives/2012/06/make-your-own-decisions/ | 2012-06 | signed essay | 5–10% evaluation-to-invest range, all-partner veto, Fitbit/Moz re-entry; note. (reliability: A; access: full; locator: decision and counterexample passages)
• ★★★ | Feld — Why We Pass Quickly on Things | https://feld.com/archives/2015/04/why-we-pass-quickly-on-things/ | 2015-04 | signed essay | Fixed-fund funnel, roughly 30 companies/fund and ten new decisions/year; note. (reliability: A; access: full; locator: capacity arithmetic and filters)
• ★★★ | Feld — VCs: Don't Bullshit | https://feld.com/archives/2014/07/vcs-dont-bullshit/ | 2014-07 | signed LP/fundraising retrospective | Mobius and Foundry fundraising candor, 90 meetings, rejection and shared-LP claims; note. (reliability: A; access: full; locator: Mobius/Foundry fundraising sequence)
• ★★★ | Feld — Term Sheet Series Wrap-Up / Venture Deals | https://feld.com/archives/2005/08/term-sheet-series-wrap-up/ | 2005-08 and 2011 book | signed series/book route | Founder-negotiation transparency coauthored with Jason Mendelson; note. (reliability: A; access: full for post/partial for book; locator: wrap-up and book-origin passages)
• ★★★ | Feld — The Techstars Mentor Manifesto | https://feld.com/archives/2011/08/the-techstars-mentor-manifesto/ | 2011-08 | signed ecosystem essay | Preserves Jon Bradford prompt, David Cohen manifesto, and Feld amplification; note. (reliability: A; access: full; locator: attribution and 18-item list)
• ★★★ | Feld — Definition of #GiveFirst | https://feld.com/archives/2020/10/definition-of-givefirst/ | 2020-10 | signed doctrine essay | Separates Give First from altruism and records naming evolution; note. (reliability: A; access: full; locator: definition/evolution paragraphs)
• ★★★ | Foundry — Our New Fund: Foundry Group Next 2018 | https://feld.com/archives/2018/08/our-new-fund-foundry-group-next-2018/ | 2018-08 | signed fund essay | $750m seventh fund, seven equal partners, lifecycle, ownership, cash-on-cash preference; note. (reliability: A; access: full; locator: fund-history and strategy passages)
• ★★★ | Foundry — Status of Our FG Angels Experiment | https://foundry.vc/blog/2016/01/status-of-our-fg-angels-experiment/ | 2016-01 | official experiment postmortem | Stops new syndicate deals because effort/capacity exceeded design; note. (reliability: A; access: full; locator: results and stop-decision passages)
• ★★★ | Foundry — What Does Foundry Group Invest In? | https://foundry.vc/blog/2008/11/what-does-foundry-group-invest-in/ | 2008-11 | founding-era firm essay | Original theme map and theme-versus-sector definition; note. (reliability: A; access: full; locator: named themes)
• ★★★ | Foundry — Foundry Group Select | https://foundry.vc/blog/2013/10/foundry-group-announces-its-newest-fund-foundry-group-select/ | 2013-10 | official fund announcement | $225m winner-only follow-on vehicle, up to $25m/company; note. (reliability: A; access: full; locator: mandate paragraph)
• ★★★ | Foundry — Our Newest Fund: Foundry Group Next | https://foundry.vc/blog/2016/09/our-newest-fund-foundry-group-next/ | 2016-09 | official fund announcement | $500m growth-plus-partner-fund vehicle; note. (reliability: A; access: full; locator: mandate and rationale)
• ★★★ | Foundry — Network-Driven Investing | https://foundry.vc/blog/2022/06/network-driven-investing/ | 2022-06 | official strategy essay | Partner-fund mechanism and $232m early-vintage scope; note. (reliability: A; access: full; locator: network and fund-history sections)
• ★★ | Feld — Your First Big Success: What's Next? | https://feld.com/archives/2014/06/first-big-success-whats-next/ | 2014-06 | signed retrospective | Career, bubble, Mobius, Foundry/Techstars, and depression chronology; note. (reliability: A; access: full; locator: chronological body)
• ★★★ | Feld — Every Generation Learns the Same Lessons | https://feld.com/archives/2018/08/every-generation-learns-the-same-lessons/ | 2018-08 | signed failure retrospective | Interliant capital/governance failure framing; note. (reliability: A; access: full; locator: Interliant narrative)
• ★★★ | Feld — Roomba, July's Toy of the Month | https://feld.com/archives/2005/07/roomba-julys-toy-of-the-month/ | 2005-07 | signed product/pass essay | Named iRobot anti-portfolio case and category-error test; note. (reliability: A; access: full; locator: product-use/pass paragraphs)
• ★★ | Feld — How a Small Company Can Make a Big Company Successful | https://feld.com/archives/2015/01/small-company-make-big-company-successful-2/ | 2015-01 | signed operating essay | Bigco partnership controls and distribution-dependency lesson; note. (reliability: A; access: full; locator: numbered lessons)
• ★★★ | Feld — SEOmoz, TAGFEE, and Me | https://feld.com/archives/2012/05/seomoz-tagfee-and-me/ | 2012-05 | signed investment/board essay | $18m financing, Feld board role, culture thesis; note. (reliability: A; access: full; locator: financing/board/TAGFEE passages)
• ★★ | Feld — Foundry Group Next | https://feld.com/archives/2015/12/foundry-group-next/ | 2015-12 | signed fund-design essay | UTIMCO relationship, partner-fund/growth rationale, anti-legacy position; note. (reliability: A; access: full; locator: Next and succession passages)
• ★★ | Feld — Why Do We Care About Techstars? | https://feld.com/archives/2009/04/why-do-we-care-about-techstars/ | 2009-04 | signed ecosystem essay | Community/mentorship rationale and commercial-public-good tension; note. (reliability: A; access: full; locator: rationale/community paragraphs)
• ★★ | Feld — Do More Faster | https://feld.com/archives/2010/10/do-more-faster/ | 2010-10 | signed book/Techstars reflection | Coauthored operating corpus and teaching flywheel; note. (reliability: A; access: full for post/partial for book; locator: purpose/Techstars paragraphs)
• ★★★ | Foundry — The Final Foundry Fund | https://foundry.vc/blog/2024/01/the-final-foundry-fund/ | 2024-01 | firm strategy announcement | Final vehicle and disclosed lifetime scale; note. (reliability: A; access: full; locator: opening and closing)

Priority talks, interviews, and direct-voice records

• ★★★ | MeetInnovators — Brad Feld, Foundry Group | https://meetinnovators.com/2010/11/01/brad-feld-foundry-group/ | 2010-11-01 | audio + edited transcript | Direct career, angel, Mobius, Foundry, NetGenesis/Critical Path/Harmonix evidence; note. (reliability: A for voice/B for edit; access: full; locator: named transcript questions)
• ★★★ | Stanford eCorner — Great Entrepreneurs Go Out and Do | https://stvp.stanford.edu/wp-content/uploads/sites/3/2024/09/great-entrepreneurs-go-out-and-do-entire-talk-transcript.pdf | 2011 talk; 2024 transcript file | official full transcript PDF | Feld Technologies chronology, $10 start, 1993 sale, founder obsession and mutual selection; note. (reliability: A; access: full; locator: PDF pages/segment headings)
• ★★★ | SaaStr — Feld Thoughts Live | https://www.saastr.com/feld-thoughts-live-one-one-brad-feld-video-transcript/ | 2017 | publisher transcript | Mobius Fund I/II/III performance language and organizational scale; note. (reliability: A for voice/B for transcript; access: full; locator: transcript passages)
• ★★★ | Tim Ferriss — Brad Feld Transcript | https://tim.blog/2020/07/24/brad-feld-transcript/ | 2020-07-24 | publisher transcript | Dot-com collapse, off-grid reset, mental health, partnership and capacity; note. (reliability: A for voice/B for transcript; access: full; locator: speaker turns)
• ★★ | Growth Everywhere — Brad Feld Interview | https://www.levelingup.com/growth-everywhere-interview/brad-feld-foundry-group/ | 2017 | interview transcript/timestamps | Personal company-sale and angel-history self-report; note. (reliability: A for voice/B for edit; access: full; locator: publisher timestamps)
• ★★★ | Entrepreneurs for Impact — Brad Feld | https://www.entrepreneursforimpact.com/brad-feld-cofounder-of-foundry | 2024 | video + transcript | CEO-support doctrine and fund-manager mentoring; note. (reliability: A for voice/B for host framing; access: full; locator: 04:39 and named sections)
• ★★★ | Alpha Partners — Driving Alpha | https://alphapartners.com/podcast/brad-feld-techstars/ | 2025 | podcast transcript | Current final-fund/wind-down and return-metric direct voice; note. (reliability: A for voice/B for promotion; access: full; locator: speaker turns)

Deal, team, security, and outcome controls

• ★★★ | Zynga — Form S-1 | https://www.sec.gov/Archives/edgar/data/1439404/000119312511180285/ds1.htm | 2011-07-01 | SEC issuer filing | Feld director since November 2007; 34,560,060 Foundry-linked pre-IPO shares/6.1%; four-partner shared power and pecuniary-interest limits. (reliability: A; access: full; locator: directors and principal-stockholder table/footnote)
• ★★★ | Fitbit — Form S-1/A | https://www.sec.gov/Archives/edgar/data/1447599/000119312515223928/d875679ds1a.htm | 2015-06 | SEC issuer filing | Feld board Aug 2010–May 2015; Foundry/Select securities, $10m Select Series D transaction, shared control. (reliability: A; access: full; locator: director biography, related transactions, principal stockholders)
• ★★★ | SendGrid — Form S-1 | https://www.sec.gov/Archives/edgar/data/1477425/000104746917006437/a2233508zs-1.htm | 2017-10 | SEC issuer filing | 9,629,727 Foundry shares/29.64% pre-IPO at filing basis; four-person control; Ryan McIntyre and Feld former directors. (reliability: A; access: full; locator: principal stockholders and director biographies)
• ★★★ | Rally Software — Form 424B4 | https://www.sec.gov/Archives/edgar/data/1313911/000104746913004318/a2214381z424b4.htm | 2013-04 | SEC issuer filing | Mobius entities held 3,406,213 shares/20.3% pre-offering; Feld and Jason Mendelson shared control; Feld served as director September 2003–March 2013. (reliability: A; access: full; locator: directors, principal stockholders, footnote 2)
• ★★★ | AvidXchange — Form S-1/A | https://www.sec.gov/Archives/edgar/data/1858257/000119312521290302/d219478ds1a.htm | 2021-09 | SEC issuer filing | Foundry Select held 5,032,964 shares/2.97% pre-offering; Feld and other managing directors shared control; Feld was to resign from the board immediately before effectiveness. (reliability: A; access: full; locator: directors, principal stockholders, footnote 14)
• ★★★ | Foundry — Foundry Group Invests in SendGrid | https://foundry.vc/blog/2010/04/foundry-group-invests-in-sendgrid/ | 2010-04 | decision-date firm post | Identifies Ryan McIntyre as lead/mentor; prevents Feld-only sourcing attribution. (reliability: A; access: full; locator: author and investment-rationale passages)
• ★★★ | Twilio — Completes Acquisition of SendGrid | https://www.twilio.com/en-us/press/releases/twilio-completes-acquisition-sendgrid | 2019-02-01 | acquirer release | Roughly $3b company transaction consideration; not Foundry proceeds or return. (reliability: A; access: full; locator: opening transaction paragraph)
• ★★★ | Molekule/AeroClean — Schedule 13D | https://www.sec.gov/Archives/edgar/data/1872356/000110465923005825/tm234190d1_sc13d.htm | 2023-01 | SEC beneficial-ownership filing | Aggregates $58,437,457 of Foundry-entity legacy capital/common-equivalent securities; names Feld among five shared managing members and as the required Foundry board nominee. (reliability: A; access: full; locator: Items 3–5 and exhibits)
• ★★★ | Molekule/AeroClean — Form 8-K | https://www.sec.gov/Archives/edgar/data/1872356/000110465923106353/tm2327546d1_8k.htm | 2023-10 | SEC issuer filing | Chapter 11 commencement and restructuring record. (reliability: A; access: full; locator: Item 1.03)
• ★★★ | Molekule — Schedule 13D/A | https://www.sec.gov/Archives/edgar/data/1671899/000110465924029980/tm247755d1_sc13da.htm | 2024-03 | SEC ownership filing | Common cancellation/no consideration after bankruptcy; does not allocate exact fund loss. (reliability: A; access: full; locator: Item 5 and restructuring description)
• ★★★ | Take-Two — Completes Combination with Zynga | https://www.take2games.com/ir/news/take-two-interactive-software-inc-and-zynga-inc-complete-combination | 2022-05 | acquirer release | Later company outcome; public consideration is not a 2007 investor cash-flow reconstruction. (reliability: A; access: full; locator: opening transaction paragraph)
• ★★ | Google — Completes Acquisition of Fitbit | https://blog.google/products/devices-services/fitbit-acquisition/ | 2021-01 | acquirer announcement | Later Fitbit outcome and stewardship statements; not Foundry proceeds. (reliability: A; access: full; locator: announcement body)
• ★★ | Twitter — Acquires Gnip | https://blog.x.com/engineering/en_us/a/2014/twitter-welcomes-gnip | 2014-04 | acquirer announcement | Confirms acquisition and data-business rationale; terms/Foundry cash flows absent. (reliability: A; access: full; locator: announcement body)
• ★★ | Google — Acquires AdMeld | https://googleblog.blogspot.com/2011/12/google-acquires-admeld.html | 2011-12 | acquirer announcement | Confirms exit; consideration and Foundry proceeds require filings. (reliability: A; access: full; locator: opening paragraphs)
• ★★ | Stratasys — Completes MakerBot Combination | https://investors.stratasys.com/news-events/press-releases | 2013 archive | acquirer release route | Confirms stock transaction route; current archive requires date search and does not expose investor economics. (reliability: A; access: partial; locator: June/August 2013 release metadata)
• ★★ | Google — Acquires FeedBurner | https://googleblog.blogspot.com/2007/06/feedburner-hello-google.html | 2007-06 | acquirer announcement | Confirms acquisition and product rationale; terms/vehicle returns unavailable. (reliability: A; access: full; locator: announcement body)
• ★★ | Return Path — Validity Combination | https://www.validity.com/press-releases/validity-acquires-return-path/ | 2019 | acquirer release | Later outcome route for a Mobius-linked company; no investor proceeds. (reliability: B; access: full; locator: opening transaction paragraph)
• ★★ | Alteryx — Acquires Trifacta | https://investor.alteryx.com/news-and-events/news-releases/ | 2022 archive | acquirer route | Outcome lead for a Foundry company; terms and partner attribution require separate filings. (reliability: A; access: partial; locator: 2022 acquisition release metadata)
• ★★ | Moz — iContact Acquires Moz | https://moz.com/blog/moz-acquired-by-icontact | 2021 | company announcement | Company outcome and continuity framing; no disclosed Foundry proceeds. (reliability: B; access: full; locator: acquisition announcement)
• ★★ | Techstars — Portfolio | https://www.techstars.com/portfolio | live; accessed 2026-08-22 | official accelerator portfolio | Establishes a separate Techstars exposure universe; not Feld personal or Foundry ownership. (reliability: A; access: full; locator: portfolio filters/list)
• ★★ | In re Fitbit, Inc. Stockholder Derivative Litigation | https://law.justia.com/cases/delaware/court-of-chancery/2018/ca-2017-0402-jrs.html | 2018 | Delaware Chancery opinion mirror | Certain derivative allegations survived pleading, but merits/proof remained unresolved and neither Feld nor Foundry was a defendant. (reliability: B mirror of judicial record; access: full; locator: opinion posture, parties, conclusion)

Fund performance, vehicle, legal, and adverse primary records

• ★★★ | UTIMCO — Board Book, Nov. 29, 2017 | https://www.utimco.org/media/iqxdvo3z/board-book-november-29-2017.pdf | 2017-11-29 | institutional LP report | Foundry 2007 since-inception net IRR 43.8% at one LP/time snapshot; no DPI/TVPI or personal result. (reliability: A; access: full; locator: PDF p.61/table row)
• ★★★ | Washington State Investment Board — Private Markets IRR, June 30 2023 | https://www.sib.wa.gov/docs/reports/quarterly/ir063023.pdf | 2023-06-30 | institutional LP report | Mobius Technology Ventures VI: $82.35m paid-in, $49.28m distributed, 0.6x, -5.80% net IRR; Side Fund VI separately 1.5x/6.34%. (reliability: A; access: full; locator: PDF p.6 rows)
• ★★★ | TMRS — Private Investment Report, Dec. 31 2022 | https://www.tmrs.com/uploads/files/general/down/Private_Investement_Report_12_31_2022.pdf | 2022-12-31 | institutional LP report | One-LP interim Foundry 2022/Next positions, largely unrealized and valuation-sensitive. (reliability: A; access: full; locator: PDF table p.3)
• ★★★ | SEC — Interliant Administrative Order | https://www.sec.gov/files/litigation/admin/2008/34-58155.pdf | 2008-07-14 | regulator order | Delinquent filings, August 2002 Chapter 11, and 2001 loss; controls company outcome, not investor wrongdoing. (reliability: A; access: full; locator: findings ¶¶1–4)
• ★★★ | I Successor/Interliant, 321 B.R. 640 | https://www.casemine.com/judgement/us/5914756badd7b049343b0346 | 2005 | bankruptcy decision mirror | Records committee allegations involving Feld Group/Mobius/Bradley Feld and a procedural ruling; allegations are not adjudicated liability. (reliability: B mirror of judicial record; access: full; locator: factual background and claims)
• ★★ | NARB — Molekule advertising decision announcement | https://www.prnewswire.com/news-releases/narb-finds-supported-claims-by-molekule-that-the-peco-filter-of-its-mh1-air-purifier-can-address-bioaerosol-and-voc-pollution-recommends-discontinuance-or-modification-of-other-claims-including-discontinuance-of-superiority-clai-301030514.html | 2020-03-26 | official decision announcement distributed by PR Newswire | Found support for some PECO claims but recommended discontinuing or modifying pollution-elimination, superiority, quantified-performance, independent-testing, and allergy/asthma-relief claims; Molekule agreed. (reliability: A/B; access: full; locator: findings and advertiser statement)
• ★★ | TechCrunch — Techstars retools amid former-staff criticism | https://techcrunch.com/2024/02/23/as-techstars-retools-some-former-staffers-say-it-lost-focus-on-what-made-it-successful/ | 2024-02-23 | reported institutional criticism | Former staff criticized centralization and sponsor dependence; evidence concerns Techstars and later management, not personal misconduct by Feld. (reliability: B; access: full; locator: former-staff and company-response passages)
• ★★ | TechCrunch — Techstars layoffs and JPM program closures | https://techcrunch.com/2024/08/07/techstars-is-laying-off-17-percent-ending-jp-morgan-backed-programs/ | 2024-08-07 | reported institutional outcome | Reports 17% layoffs, JPM-backed program closures, and David Cohen's overbuilding/overhiring explanation; does not establish Feld's personal causality. (reliability: B; access: full; locator: opening and Cohen-statement passages)
• ★★★ | Foundry — Current Portfolio | https://foundry.vc/portfolio/ | live; accessed 2026-08-22 | official current/mixed list | Reproducible visible roster mixing active companies, exits, and partner funds; not a complete lifetime or Feld-only ledger. (reliability: A; access: full; locator: filterable company/fund cards)
• ★★ | Foundry — Team | https://foundry.vc/team/ | live; accessed 2026-08-22 | official people page | Current equal partners and staff; not historical deal-role proof. (reliability: A; access: full; locator: partner cards)
• ★★ | Foundry — 2022 Responsible Investment Policy | https://foundry.vc/ | archive/search route | Current governance/sustainability lead; stable document not recovered in this pass, so no substantive claim depends on it. (reliability: C; access: search-route; locator: site search)

Additional decision, failure, and corpus routes

• ★ | Feld — Protocol tag archive | https://feld.com/archives/tag/protocol/ | archive | Mapbox/SimpleGeo/deCarta decision chain and theme evolution; individual underlying posts control any claim. (reliability: C; access: full archive; locator: dated post list)
• ★ | Feld — The Foundry Group Themes | https://feld.com/archives/tag/foundry-group-themes/ | archive | Additional theme corpus route; not counted as a separate underlying work. (reliability: C; access: full archive; locator: dated post list)
• ★ | Feld — Venture Deals category | https://feld.com/archives/category/venture-deals/ | archive | Long term-sheet/financing corpus; sampled through the mapped series/book route. (reliability: C; access: full archive; locator: pagination)
• ★ | Feld — Techstars category | https://feld.com/archives/tag/techstars/ | archive | Large multi-author/history lead; priority sample is bounded to doctrine and founder-support notes. (reliability: C; access: full archive; locator: pagination)
• ★ | Feld — Depression category | https://feld.com/archives/tag/depression/ | archive | Mental-health writing route relevant to capacity and leadership, not an investing-return source. (reliability: A for first-person/C for archive completeness; access: full; locator: dated posts)
• ★ | Stanford eCorner — Ending and Starting Venture Capital Firms | https://stvp.stanford.edu/wp-content/uploads/sites/3/2024/09/ending-and-starting-venture-capital-firms-transcript.pdf | 2011 talk; 2024 transcript | Official direct-voice clip transcript on Mobius ending/Foundry formation; treated as a segment of the Stanford session, not a new priority work. (reliability: A; access: full; locator: PDF full clip)
• ★ | Techstars — Brad Feld Returns: The Power Behind Give First | https://www.techstars.com/blog/podcasts/brad-feld-returns-the-power-behind-give-first | 2025 | official podcast page | Current promotional route; no complete voice-verifiable transcript located, so no direct quotation used. (reliability: B; access: partial; locator: episode description)
• ★ | Feld — Books | https://feld.com/books/ | live; accessed 2026-08-22 | official bibliography | Nine-book metadata and coauthor boundaries; full texts not uniformly accessible. (reliability: A for metadata; access: partial; locator: book cards)
• ★ | Internet Archive — Feld.com captures | https://web.archive.org/web/*/feld.com | archive route | Recovery route for migrated/dead posts; capture completeness varies. (reliability: C; access: partial; locator: timestamped captures)
• ★ | SEC EDGAR — Foundry-linked issuer search | https://www.sec.gov/edgar/search/ | filing search route | Exhausted for named public controls; search results are navigation, cited filings are controlling artifacts. (reliability: A for filings/C for search completeness; access: full; locator: issuer/form/date queries)
• ★ | Feld — Small Company/Big Company corpus | https://feld.com/archives/tag/big-companies/ | archive route | Partnership-failure denominator lead; one priority essay sampled, no full case corpus reconstructed. (reliability: C; access: partial; locator: archive entries)
• ★ | Foundry — Blog archive | https://foundry.vc/blog/ | archive | Firm-authored strategy, deal announcements, partner attribution and postmortems; bylines preserved. (reliability: A for posts/C for archive denominator; access: full; locator: post cards/pagination)

Coverage assessment

The map contains 82 visible source rows and 80 unique underlying URLs: the final-fund artifact and Alpha transcript each appear once in current-state mapping and once in literal corpus reconciliation. Of the 82 rows, 70 are ★★★/★★ priority rows. The 31 priority direct writings/talks have 31 exact-schema notes and 31 source-specific analyses. The remaining priority rows are filings, regulator/LP reports, official histories, outcome controls, and adverse reporting processed directly in PROFILE, INVESTMENTS, and REVIEW; the methodology does not require note files for non-writing/non-talk controls. [researcher inference · strong]

The bounded company universe starts with Foundry's live mixed portfolio page: 136 visible linked cards observed on 2026-08-22, including direct companies, exits, and partner funds. Foundry separately reports 200+ direct-company investments and nearly 50 fund investments over its life. Because the live page is explicitly not a lifetime list and mixes object types, neither number is a Brad Feld personal denominator. INVESTMENTS lists a 32-name researched company subset, reconstructs seven company cases plus two noncompany cases sufficiently for nine labs, and leaves 25 materially researched company names unresolved; career-wide coverage and hit rate are not computable. [contemporaneous record · strong] final-fund disclosure · ledger

Primary-source strengths are unusual: long first-person archives, official full transcripts, issuer filings for Zynga/Fitbit/SendGrid/Rally/AvidXchange/Molekule, regulator and judicial records, and three LP disclosures. Independent adverse controls cover Techstars' institutional retrenchment and Molekule's advertising claims. Weaknesses remain decisive: no full Feld/Mobius/Foundry cash-flow ledger; no consistent vintage DPI/TVPI/MOIC; no complete pass/write-off denominator; no IC memos; no partner allocation records; and only selected founder/board perspectives. Public acquisition values and securities are never converted into personal return. [researcher inference · insufficient public record]

Access gaps are itemized rather than silently waived: books are sampled through author posts and metadata, the 2025 Techstars page lacks a full transcript, archives can omit migrated posts, the current Foundry roster is not historical, and the case-law route is a mirror. These sources remain ★ navigation/context routes and do not create unsupported priority parity obligations. [researcher inference · strong]

Source notes