Josh Kopelman

1. Snapshot and testable edge thesis

This is a falsifiable process hypothesis, not a finding of Kopelman-specific alpha. Acquisition values, IPO holdings, and private marks do not establish First Round proceeds, much less Josh's personal return. Pareto boundary [researcher inference · strong]

2. Core philosophy

Kopelman's surviving record repeatedly makes uncertainty the object of management. In Crystal Balls vs. Calculators, he separates seed judgment from the false precision available after product and revenue evidence. In The Watney Rule, the First Round partnership argues that a startup should not depend on a financing event outside its control. In recent interviews he describes venture decisions as probabilistic, asks partners to form views before group discussion, and treats founder adaptability as more durable than a fixed plan. Writings synthesis · Decision Education, 32:58–46:19 [investor-stated · strong]

The evidence-bounded synthesis is: find companies before social proof, choose people who can learn faster than the environment changes, price the check and ownership for fund contribution, design each round to answer a named uncertainty, and build an institution that stores learning rather than relying on one partner's intuition. [researcher inference · moderate] The last clause is supported as organizational intent, not as measured causal alpha; public records do not expose the rubric history, matched passes, or net cash results. Research boundaries [researcher inference · strong]

3. Sourcing

Reusable sourcing ledger: origin, introducer, prior relationship, access cost, unique evidence, conflict, owner, decision date, pass/invest reason, ownership offered, and eventual outcome. [researcher inference · strong] The claimed non-referral advantage remains unproved until matched referral and non-referral cohorts are available.

4. Picking

Founder and people evidence

Adaptability is the clearest repeated founder criterion. The relevant test is not whether a founder can tell a confident story, but whether they identify a wrong premise, run a discriminating test, recruit around a gap, and change direction before runway disappears. Decision Education · Upstart pivot lab [retrospective witness · strong] Persistence counts only when the next iteration produces new evidence; Notion is a visible survivor, not proof that repeated rebuilding is generally optimal. Notion lab [researcher inference · moderate]

Product, market, and evidence

At seed, Kopelman's public doctrine prefers the present problem and learning rate over a fully calculable forecast. A useful IC packet therefore names the user, current pain, why-now mechanism, cheapest falsifying test, financing dependency, and alternate explanation. Writing synthesis [investor-stated · moderate] For marketplaces and networks, analysts should map the atomic transaction, cold-start path, liquidity, trust, and platform dependence; for regulated products, legal and data gates precede growth. AppNexus lab · Clover lab · Project Fair Bid lab [researcher inference · strong]

Decision process and red flags

First Round's reported practice of independent pre-discussion judgments aims to preserve information before status and consensus contaminate it. Decision Education [investor-stated · strong] A decision-ready version records each partner's probability, confidence, premise, ownership requirement, kill criterion, and next-check milestone before discussion. Red flags are referral prestige without unique evidence, unfalsifiable vision, weak customer pull, a round that requires the next round to rescue it, regulatory ambiguity treated as a launch detail, and partner enthusiasm without named responsibility. [researcher inference · strong] Formal veto rules, historical votes, and a complete pass ledger remain [researcher inference · insufficient public record].

5. Deal and portfolio mechanics

First Round's 2011 SEC letter described a firm with more than 155 investments, usual initial checks of $250,000–$750,000, follow-ons, assistance, and conflict procedures. Those are firm-wide period facts, not fixed current policy or company-level economics. SEC letter, pp. 1–4 [contemporaneous record · strong]

No public evidence supports Kopelman-specific DPI, TVPI, IRR, MOIC, hit rate, or ex-outlier return. Portfolio evidence boundary [researcher inference · strong]

6. Vehicle and incentive fit

The strategy evolved inside numbered, LP-backed pooled vehicles. The SEC record identifies the adviser/fund complex, Upstart attributes its FRC III stake to entities controlled by a three-person GP committee, Clover identifies FRC V, and a 2026 Form D names Kopelman in FRC X-F. Vehicle map [contemporaneous record · strong]

• Capital form and duration: pooled, illiquid venture capital can fund many small entries and selective escalation; exact fund terms and extensions are unavailable. [researcher inference · moderate]
• Fundraising pressure and scale: larger successor vehicles can expand ownership and reserves while raising the outcome size required to matter. Fundraising dates and offered amounts do not establish deployed capital or returns. [researcher inference · strong]
• Compensation, carry, and GP commitment: not publicly reconstructed; do not infer alignment from title, tenure, or an issuer holding. Vehicle evidence boundary [researcher inference · strong boundary]
• Governance and decision rights: partner sponsorship, GP legal control, board participation, and company control are separate. Upstart's committee evidence rejects lone-partner ownership of the pooled stake. [researcher inference · strong]
• Liquidity and conflicts: founder liquidity, pass-through opportunities, pro rata rights, and allocation can create conflicts; the SEC letter shows rules existed, not their use in a named case. [researcher inference · moderate]
• Geography, regulation, succession, and key person: the firm spans Philadelphia, New York, and San Francisco and still names Kopelman formally, but geographic authority, succession, key-person clauses, and carry continuity are [researcher inference · insufficient public record].

The vehicle is compatible with the edge thesis only if small first checks genuinely buy information and later capital follows precommitted evidence rather than reputation or sunk cost. That causal chain is a prediction for private reserve records, not a proven result. [researcher inference · strong]

7. Team attribution and work with founders

The correct attribution ledger has separate fields for source, sponsor, diligence, approval, negotiation, board, recruiting, intervention, follow-on, exit, narrator, legal holder, and economic beneficiary. [researcher inference · strong]

8. Views on the craft

• Power law: extreme outcomes make ownership and fund size consequential, but the public record does not reveal a cash contribution curve. Crystal Balls analysis [investor-stated · moderate]
• Access versus selection: broad access supplies more independent shots; only matched opportunity and outcome records can show whether it improves judgment. [researcher inference · strong]
• Stage specialization: First Round's formation thesis treated seed as a distinct craft when company formation costs fell and incumbent funds wrote larger checks. Wharton interview [retrospective witness · strong]
• Governance: question-led boards should improve founder reasoning without seizing operator credit. 20VC, 14:12–15:25 [investor-stated · strong]
• Learning: forecasts, marks, passes, and failures should be stored as dated hypotheses, not rewritten from outcome. [researcher inference · strong]
• Great investor: the public doctrine prizes relevance, adaptability, truth-seeking, and institutional learning; whether those qualities caused superior net returns remains untested. Vintage, 04:33–08:13 [investor-stated · moderate]

9. Signature bets and decision process

The six no-hindsight decision labs freeze the evidence before revealing outcomes:

1. Notion tests whether repeated rebuilding is evidence-generating persistence or reserve escalation.
2. Flatiron Health tests data rights, workflow integration, regulation, and clinical-value evidence.
3. Upstart tests whether a board question can expose absent demand before runway vanishes.
4. AppNexus tests infrastructure latency, marketplace liquidity, platform dependence, and privacy risk.
5. Clover Health tests actuarial, compliance, related-party, and public-disclosure controls.
6. Project Fair Bid / BigDeal tests payments, consumer-loss, disclosure, jurisdiction, and legal classification before launch.

These are analyst exercises, not recovered First Round memos. Each asks the reader to decide invest/pass, kill criteria, ownership/terms, and evidence for the next check before consulting the sourced actual outcome. [researcher inference · strong]

10. Mental models and maxims

The model inventory merges overlapping expressions and separates direct doctrine from reconstructed practice:

1. Adaptive founders with falsification gates — learning velocity plus a dated evidence threshold.
2. Start with a present problem, not a forecasted theme — provisional; two cases plus an exhausted-route ledger.
3. Independent judgment and the three clocks — private view, group debate, and later scoring.
4. Fund math before company enthusiasm — separate company quality from security and vehicle contribution.
5. Financing dependency and reserve gates — each check buys survival to a named proof.
6. Stage specialization and vehicle fit — match craft, check, ownership, duration, and LP incentives.
7. Designed access with channel measurement — broaden sourcing, then test whether the channel adds decision value.
8. Portfolio as trusted network and retrieval system — community and First Search as measurable infrastructure.
9. Legal, regulatory, data, and enforcement rights as product risk — gate regulated mechanics before growth.
10. Map roles before assigning credit — separate founder, partner, vehicle, board, and economic claims.
11. Questions for board decisions, escalation when duty requires — provisional; two publicly connected cases plus an exhausted-route ledger.

The field guide's inventory reconciles 40 distinct cited examples. Candidate slogans without enough application evidence stay excluded instead of being inflated into models. [researcher inference · strong]

11. Evolution and contradictions

Three contradictions remain productive rather than resolved: seed intuition versus measurable process; founder patience versus financing independence; and community collaboration versus independent judgment. A credible operating system needs both sides—judgment recorded before measurement, patience governed by milestones, and community information evaluated without social proof. [researcher inference · moderate]

12. Failures and limits

The adverse record prevents an all-winner narrative. Knewton's distressed sale, AltSchool's contraction, Path's shutdown/asset transfer, Boxed's bankruptcy, Project Fair Bid litigation, and the priced-out Twitter/Dropbox misses expose different failure modes. Failure and adverse evidence [researcher inference · strong]

For each, the public record is materially incomplete: original thesis, First Round check and vehicle, ownership, dilution, board interventions, reserve votes, stop decisions, loss, and any dated process change. Lessons such as “gate reserves on adoption” or “price the offered ownership, not the later winner” are therefore analyst reconstructions, not recovered Kopelman rules. [researcher inference · strong] The record is also selected toward narrated successes and notable failures; ordinary quiet outcomes and passes are missing. Denominator [researcher inference · strong]

13. Historical operating environment

Historical comparisons must preserve the financing alternatives and evidence available at the decision date. Later outcomes are outcome evidence, never an ex-ante premise. [researcher inference · strong]

14. Modern VC translation (as of 2026-08-24)

A current analyst should preserve a versioned decision record: source, private probability, decisive premise, alternatives, legal/data gates, ownership and vehicle fit, milestone, reserve rule, board owner, and scheduled postmortem. [researcher inference · strong] The method is transferable; the claimed historical edge is not assumed to transfer.

15. Comparison with completed peers

The closest defensible completed comparators are Fred Wilson, Ron Conway, and Aileen Lee. Wilson supplies a thesis/community and reserve contrast; Conway supplies high-volume founder-network access; Lee supplies a small-vehicle seed, learning, and governance comparison. No return ranking is possible because the packages do not expose dimension-matched audited cash flows. [researcher inference · strong]

The comparison suggests that First Round's distinctive public claim is not seed selection alone but the attempted institutionalization of learning and service at seed. Whether that is causal alpha, reproducible operating hygiene, or brand-backed access remains unresolved. [researcher inference · moderate]

16. The strongest case against greatness

The skeptical thesis is stronger than “venture is lucky.” First Round entered institutional seed when software formation costs fell and incumbent fund checks rose; Kopelman's founder exits, Yale anchor, early network, and firm brand could have created privileged access. Partners, LPs, platform teams, founders, co-investors, and pooled vehicles divide both causal and economic credit. Famous firm investments assigned to other partners cannot be imported into Josh's record. People/rights map [researcher inference · strong]

The visible record is radically selected: 62 searched candidates produce 36 retained associations, but only seven sufficiently reconstructed cases; the official “100+” is not an enumerated universe; 29 retained rows are unresolved; passes and most losses are absent. Notion may dominate paper narratives, while acquisition prices, enterprise values, and S-1 stakes are not cash returns. Denominator and Pareto [researcher inference · strong] Public essays and interviews may describe sensible operational hygiene without showing that the rules preceded decisions, differed from peers, or improved net returns.

The favorable thesis would gain support from dated memos showing the same rubric before wins and losses; matched referral/channel cohorts; recorded probabilities and calibration; reserve decisions tied to precommitted milestones; and audited vehicle cash flows that persist across partners, vintages, and after removing the top outcome. It would weaken if rules appear only after success, channel effects vanish, partners/era/outliers explain results, or later reserves perform as well as first checks. [researcher inference · strong]

17. Analyst study guide, glossary, and go deeper

Recommended path

1. Start with the chronological profile and its people/influence boundaries.
2. Read the written corpus and spoken corpus to separate dated doctrine from later synthesis.
3. Work the six no-hindsight decision labs before reading outcomes; include the adverse Project Fair Bid case and the Upstart pivot.
4. Use the portfolio ledger to test how much is known outside famous examples.
5. Apply the mental-model field guide, then read the adversarial review.

Glossary

Reusable checklists

Sourcing: source and prior tie; unique evidence; referral conflict; channel cost; comparable conversion and outcome; named owner. Diligence: present problem; user/payer/controller; decisive premise; alternative explanation; founder update record; legal/data rights; financing dependency. IC: private probability/confidence; dissent; company case; security/ownership case; vehicle contribution; kill criteria; next-check milestone; decision owner. Governance: founder-question versus board action; evidence dashboard; runway; hiring gap; compliance escalation; reserve/stop vote; dated postmortem. Attribution: source; sponsor; diligence; vote; negotiation; board; recruiting; follow-on; exit; narrator; holder; beneficiary. [researcher inference · strong]

Go deeper

1. Decision Education interview — clearest current direct account of probabilistic process and independent judgments.
2. 2011 SEC comment letter — strongest contemporaneous record of firm mechanics, initial checks, follow-ons, assistance, and conflicts.
3. Colossus interview — seed history, attribution, and portfolio community.
4. Crystal Balls vs. Calculators and The Watney Rule — concise primary doctrine with different attribution boundaries.
5. Upstart decision lab and Project Fair Bid lab — a survivor pivot and adverse legal-control case.
6. Complete source map — all processed, inaccessible, excluded, archive, filing, and adverse routes.

Pareto 80/20 — what drove the record

The public record cannot support a literal Josh Kopelman return attribution. It exposes selected company outcomes, a few issuer ownership snapshots, First Round vehicles, and retrospective narratives, but not a complete chain of check size, entry ownership, dilution, secondary sales, distributions, fees, carry, and time. Acquisition consideration and private or public valuation are therefore not treated as investor proceeds. The three rankings below deliberately separate observable economic candidates, historical importance, and decision-learning value. First Round Josh profile, biography and selected investments · Upstart S-1, Principal Stockholders [researcher inference · strong]

Economic contribution — directional candidates, not a return table

Dominant-outlier test. No defensible Josh-level dominant outlier can be identified from public cash flows. If Notion is removed, the remaining candidates still cannot be ordered on a comparable realized basis. At firm level, Uber, Square, and Roblox may be economically important, but First Round's own portfolio credits Rob Hayes, Rob Hayes, and Chris Fralic respectively; importing them into a Josh ranking would be team-credit leakage. First Round companies, partner filters [researcher inference · strong]

Historical importance

Decision-learning value

Evidence boundary and searched denominator

The scoped unit is a normalized company family in which a public route specifically associates Josh Kopelman with investing, sponsorship, or governance. The search covered Kopelman, First Round Capital and numbered vehicles, personal/angel references, current and archived firm/company pages, financing press, S-1/8-K/Form D filings, litigation and criticism routes, deal/outcome reporting, 20VC metadata, and compiled databases used only for discovery. The temporal range runs from pre-First Round angel/operating-era leads through the current profile and a 2026 fund filing. It does not automatically include every First Round company, Kopelman's three operating companies, LP commitments, philanthropy, funds he advised or backed, or a successor name without a separate financing route. Aliases are collapsed: Project Fair Bid/BigDeal; Massdrop/Drop; BankSimple/Simple; Swipely/Upserve. Follow-ons remain within the company family rather than inflating the deal count. First Round Josh profile, current biography · SEC 2011 comment letter, pp. 1–4 [contemporaneous record · strong]

Reproducible reconciliation

Arithmetic check: 62 searched = 36 retained + 26 excluded; 36 retained = 7 sufficiently reconstructed + 29 unresolved. Coverage fraction: not computable because the official “100+” is a lower-bound statement rather than an enumerated denominator, while the 62 candidates are the auditable searched set rather than a complete career universe. The visible roster is subject to survivorship, publicity, current-site, outcome, and source-access bias. [researcher inference · strong]

Historical market, regulatory, and base-rate context

No comparable, contemporaneous numeric seed-success base rate survived the bounded record. That absence is itself a constraint: the 155-plus firm count establishes a broad portfolio, but neither loss frequency nor net-return distribution. Later famous wins cannot substitute for the missing cohort. [researcher inference · strong]

Vehicle economics and incentives

The vehicle plausibly rewards many early options and selective escalation, but that mechanism is a researcher inference, not a published Josh rule or audited source of excess return. Fundraising pressure, ownership targets, reserve ratios, write-down policy, and partner compensation are missing. [researcher inference · moderate]

People, capital, and decision-rights map

Comprehensive 36-company retained ledger

The ledger is comprehensive for the 36 retained names in this audit, not for the official “100+” career lower bound. “Sufficiently reconstructed” means multiple material decision/outcome links survived; it does not mean returns are known. “Unresolved” preserves the lead while blocking economic or causal claims. “Unknown” means not publicly established after the recorded routes; round totals are not First Round checks.

Twenty-six searched exclusions

Exclusion means “not admitted to this Josh-specific retained ledger under the stated rule,” not proof that no economic or professional relationship ever existed.

The exclusion arithmetic is 3 + 3 + 7 + 13 = 26. Company aliases and follow-on rounds are deduped before this count. [researcher inference · strong]

Six no-hindsight ex-ante decision labs

These labs freeze the record at the decision point. Later outcomes appear only after the prospective questions so that a successful exit cannot cure a weak decision and a failure cannot retroactively make every initial premise irrational.

Lab 1 — Notion

Freeze frame (2013 seed). Known then: the company raised a $2 million seed and First Round says Kopelman wrote its largest check; later accounts describe Ivan Zhao's pitch as paper-based and centered on an open digital workspace rather than a conventional screenshot demo. Unknown then: durable user pull, initial product usability, repeatable distribution, pricing, enterprise security, team-building, exact First Round ownership, reserve budget, and board rights. Believed: a flexible “paper for software” layer might create a horizontal category. Disputed/fragile: the pitch's abstraction could signal category insight or lack of a testable wedge. First Round, Notion company card [retrospective witness · moderate]

• Alternatives and dissent then: fund a narrower collaboration tool, wait for a working prototype, stage a smaller check, or pass on the horizontal vision. No First Round memo, partner vote, founder response, or dissent record was located. [researcher inference · strong that public record is absent]

• Would you invest then? Conditional yes only if founder references, prototype learning speed, and a narrow initial workflow can make the broad vision falsifiable. [researcher inference · moderate]

• What would kill the deal? No frequent user job, product complexity preventing activation, founder rigidity, or a financing plan dependent on immediate mass-market adoption. [researcher inference · moderate]

• What ownership/terms fit the risk? Seed board/information rights, pro rata, milestone-based reserve planning, and enough runway for a constrained build; no historic price, security, ownership, or Josh check amount should be invented from the $2m round. [researcher inference · strong]

• What evidence unlocks the next check? Weekly cohort retention, collaborative documents/workspaces, activation time, repeated user-created templates, product reliability, and evidence the broad model wins one specific job. [researcher inference · moderate]

• Actual decision and outcome. First Round backed the seed. First Round later said the product was rebuilt four times; reporting describes layoffs and a move to Japan before traction, with a $150,000 family loan during the lean period. Notion later reached high reported private valuations, but ownership, sales, and distributions remain unknown. First Round manifesto, Notion rebuild passage · Forbes Midas Touch, Notion funding-history paragraphs [retrospective witness · moderate]

• Learning verdict. The public case supports patience paired with repeated product change, not “never stop.” Without board minutes, reserve decisions, and counterfactual projects, Kopelman's causal contribution is unmeasured. [researcher inference · strong]

Lab 2 — Flatiron Health

Freeze frame (2012 formation / January 2013 financing). Known then: Nat Turner and Zach Weinberg formed Flatiron after encountering a clinical follow-up-data gap through Foundation Medicine; a January 2013 $8 million round included Google Ventures, First Round, and angels. Unknown then: access to oncology records, interoperability, HIPAA/security execution, hospital sales cycles, clinician trust, data quality, regulatory evolution, and per-investor checks. Believed: fragmented cancer-care data could be normalized into a valuable longitudinal evidence layer. Disputed/fragile: prior founder success and a large problem did not establish data rights or clinical usefulness. Forbes founder profile, formation/data-gap account · Fortune, prior-round paragraph [retrospective witness · strong]

• Alternatives and dissent then: build a services business, license data infrastructure, partner with a strategic, wait for practice access, stage capital, or pass. No First Round memo, partner vote, founder response, or dissent record was located. [researcher inference · strong that public record is absent]

• Would you invest then? Conditional yes if representative oncology practices would provide data access and use the resulting workflow, not merely endorse the mission. [researcher inference · moderate]

• What would kill the deal? Inability to obtain longitudinal data lawfully, poor normalization accuracy, workflow burden on clinicians, a services-only model, or dependence on one strategic buyer. [researcher inference · moderate]

• What ownership/terms fit the risk? Board/information rights, security and regulatory reporting, pro rata, and milestone reserves; the $8m round is not First Round's check. [researcher inference · strong]

• What evidence unlocks the next check? Signed practice access, normalized-record accuracy, time saved, retention, expansion, de-identification/security audit, and diversified customer/payer/research demand. [researcher inference · moderate]

• Actual decision and outcome. First Round identifies Kopelman as Seed partner; Roche completed a $1.9 billion fully diluted acquisition in April 2018. No public route supplies First Round's cost, stake at sale, distributions, or Josh economics. First Round, Flatiron company card · Roche, 2018-04-06 closing release [contemporaneous record · strong]

• Learning verdict. A mission-aligned strategic outcome is compatible with good underwriting, but price alone does not prove the seed thesis, governance contribution, or return. [researcher inference · strong]

Lab 3 — Upstart

Freeze frame (2012–2013 seed and pre-pivot board period). Known then: former Google executive Dave Girouard founded Upstart around income-share agreements and a belief that data science could improve access to credit; the founder later recalled weak demand and roughly six months of cash before the pivot. Unknown then: borrower willingness, investor funding appetite, consumer-law treatment, adverse selection, model fairness, bank partnership, loss behavior, and whether a loan product would be permitted. Believed: conventional credit scoring left useful variables unused. Disputed/fragile: a compelling model thesis did not make the income-share product desirable. Upstart S-1, Founder Letter · First Round Review, founder board/pivot account [retrospective witness · strong]

• Alternatives and dissent then: persist with income-share agreements, pivot to conventional loans, sell underwriting tools to lenders, stage capital, or wind down. No First Round memo, partner vote, formal alternative set, or dissent record was located. [researcher inference · strong that public record is absent]

• Would you invest then? Only as an experiment with explicit product-demand, regulatory, model-risk, and runway gates; founder pedigree cannot substitute for pull. [researcher inference · moderate]

• What would kill the deal? No customers “at the gate,” unclear legal enforceability, inability to fund contracts, biased or unstable underwriting, or refusal to abandon the founding product. [researcher inference · moderate]

• What ownership/terms fit the risk? Seed information/board rights, pro rata, model and compliance reporting, and a staged reserve plan. The later 5.2% FRC III pre-IPO position must not be reverse-engineered into seed ownership. [researcher inference · strong]

• What evidence unlocks the next check? Application conversion, lender funding, repayment/vintage loss data, approval uplift at controlled risk, bank/legal sign-off, cohort fairness, and complaint metrics. [researcher inference · moderate]

• Actual decision and outcome. The founder says Kopelman asked at a seed-board meeting whether customers were massing at the gate; after loans produced immediate demand, the team killed income-share agreements within days. Upstart later went public. Its S-1 showed concentrated bank-partner dependence—Cross River Bank originated 89% and 72% of loans in cited periods—and “true lender,” usury, model, and regulatory risks. First Round Review, board-question and pivot sections · Upstart S-1, Risk Factors and Principal Stockholders [retrospective witness · strong]

• Learning verdict. The observable contribution is a sharp demand question in a board context, not proof Kopelman invented the pivot or controlled the company. The case rewards falsification speed while preserving regulated-credit downside. [researcher inference · strong]

Lab 4 — AppNexus

Freeze frame (2007–2008 seed). Known then: a decision-era Venrock memo described a $2.5 million seed at a $10.5 million post-money valuation from First Round, Marc Andreessen, Ben Horowitz, and Vinod Khosla; AppNexus was seeking a larger round to build advertising infrastructure. Unknown then: real-time marketplace liquidity, publisher/advertiser concentration, latency at scale, fraud, privacy regulation, platform gatekeepers, burn, and First Round's exact allocation. Believed: a technically superior neutral platform could intermediate rapidly automating ad markets. Disputed/fragile: category growth could accrue to dominant platforms rather than an independent intermediary. Venrock memo, Company and Financing sections [contemporaneous record · strong]

• Alternatives and dissent then: back one side of the market, partner with an incumbent, wait for liquidity evidence, stage financing, or pass. Venrock records its view; no First Round memo, partner vote, founder response, or dissent record was located. [researcher inference · strong that public record is absent]

• Would you invest then? Conditional yes if the technical team can prove low-latency reliability and balanced buyer/seller pull before scaling spend. [researcher inference · moderate]

• What would kill the deal? Subsidized volume without gross profit, a single dominant customer/platform, unbounded fraud/privacy exposure, or infrastructure costs that erase marketplace economics. [researcher inference · moderate]

• What ownership/terms fit the risk? Board/information rights, pro rata, security/privacy reporting, customer-concentration covenants where feasible, and a reserve plan; the round post-money is not proof of First Round ownership. [researcher inference · strong]

• What evidence unlocks the next check? Bid volume and take rate, buyer/seller concentration, latency/uptime, fraud loss, contribution margin, retention, privacy/security posture, and cash burn. [researcher inference · moderate]

• Actual decision and outcome. First Round joined the seed syndicate; later metadata says Kopelman served on the board. AT&T acquired AppNexus in 2018, with credible reporting around $1.6 billion while official terms were undisclosed. 20VC show notes, board-role metadata · AT&T 2018 annual report, AppNexus acquisition disclosure · Axios, 2018-06-25 reported consideration [retrospective witness · moderate]

• Learning verdict. A reported strategic sale validates neither privacy/governance quality nor investor cash return. Subsequent Xandr ownership and sale belong to AT&T/Microsoft history, not the original First Round result. [researcher inference · strong]

Lab 5 — Clover Health

Freeze frame (initial Series A). Known then: First Round identifies Kopelman as the Series A partner in a company aiming to improve Medicare care through data and software. Unknown then: risk adjustment and coding practices, CMS/DOJ scrutiny, network adequacy, unit economics by cohort, claims maturity, related-party exposure, sales practices, data advantages, investor rights, and reserve plan. Believed: software and longitudinal data might improve care coordination and insurance economics. Disputed/fragile: regulated-insurance outcomes may be driven by selection, reimbursement, and coding rather than product quality. First Round, Clover company card [contemporaneous record · strong attribution / provisional thesis]

• Alternatives and dissent then: back provider software rather than an insurer, wait for mature cohorts/regulatory validation, stage a smaller check, or pass. No First Round memo, compliance dissent, partner vote, or founder response was located. [researcher inference · strong that public record is absent]

• Would you invest then? Only with independent actuarial, compliance, clinical, related-party, and data-quality diligence plus explicit board reporting. [researcher inference · moderate]

• What would kill the deal? Economics dependent on aggressive coding, undisclosed regulator contact, related-party conflicts, weak medical-cost improvement, unreliable cohorts, or growth incentives outrunning compliance. [researcher inference · moderate]

• What ownership/terms fit the risk? Board/audit and information rights, compliance escalation, pro rata, milestone reserves, conflict approvals, and public-transition disclosure gates; historical terms remain unknown. [researcher inference · strong]

• What evidence unlocks the next check? Mature medical-loss ratios, risk-adjustment audit, CMS star/complaint measures, broker conduct, provider retention, cohort contribution margin, related-party register, and regulator correspondence. [researcher inference · moderate]

• Actual decision and outcome. First Round invested and Kopelman later signed the SPAC support agreement as partner. The transaction announcement stated $3.7 billion enterprise value. In 2021 Hindenburg alleged undisclosed government scrutiny and other problems; those were allegations, not findings. A Stanford securities-litigation index records the related shareholder action and a 2023 settlement, but it is a secondary docket index and not a finding that every allegation was true. Clover disclosed in 2024 that SEC staff did not intend to recommend enforcement after an investigation begun in 2021; that does not adjudicate every Hindenburg claim, settle every shareholder allegation on the merits, or resolve any separate DOJ issue. SEC transaction announcement, Transaction Overview · Hindenburg, allegations · Stanford case tracker, settlement metadata · Clover 8-K, Item 8.01 [contemporaneous record · strong fence]

• Learning verdict. The case warns against equating SPAC enterprise value with realized success and against allowing product enthusiasm to compress regulated-market diligence. Josh-specific board votes and proceeds remain unknown. [researcher inference · strong]

Lab 6 — Project Fair Bid / BigDeal

Freeze frame (July 2009 financing). Known then: the stealth auction platform announced a $4.5 million round, and TechCrunch reported Kopelman would join the board. Unknown then: final consumer mechanics, loss distribution, marketing representations, payments/chargebacks, state-by-state auction/gambling treatment, venue/terms, customer support, unit economics, board compliance process, and reserve budget. Believed: a new auction mechanism could create engaging commerce. Disputed/fragile: “entertainment shopping” could be characterized as a game of skill, auction, or unlawful gambling depending on facts and law. TechCrunch, 2009-07-23 financing/board report [contemporaneous record · strong]

• Alternatives and dissent then: invest only in a conventional auction, restrict launch to counsel-cleared jurisdictions, stage capital behind legal tests, or pass. No First Round memo, legal opinion, partner vote, founder response, or dissent record was located. [researcher inference · strong that public record is absent]

• Would you invest then? No until independent gaming/auction counsel maps every target jurisdiction, product simulations quantify consumer loss, and disclosures/controls pass a plain-language fairness review. Conditional investment would require those gates, not a generic legal opinion. [researcher inference · strong]

• What would kill the deal? Revenue depending on confused consumers, adverse jurisdictional advice, mechanics legally indistinguishable from chance, misleading odds/fees, uncontrollable chargebacks, or inability to geo-restrict. [researcher inference · moderate]

• What ownership/terms fit the risk? Board/audit rights, compliance-officer authority, jurisdiction-launch covenants, complaint/chargeback reporting, insurance, indemnification, and tranche milestones. Historic terms are unknown. [researcher inference · moderate]

• What evidence unlocks the next check? Written jurisdiction matrix, product/odds simulations, disclosure-comprehension tests, complaint rates, refunds, payment-processor approval, and board minutes for launch decisions. [researcher inference · moderate]

• Actual decision and outcome. A later complaint named Project Fair Bid, First Round Capital II, Mayfield, Foundation, and others and asserted RICO/illegal-gambling theories. These are plaintiff allegations only. In Doe v. Project Fair Bid, W.D. Wash. 2:11-cv-00809, the plaintiff voluntarily dismissed First Round and other investor defendants without prejudice amid personal-jurisdiction issues; the remaining matter was dismissed for service/venue defects. A refiled N.D. Cal. 4:13-cv-00759 case ended after a settlement notice with all claims dismissed without prejudice. There was no merits finding; public terms do not disclose the settlement. Complaint exhibit, pleaded allegations · Courthouse News, complaint summary · W.D. Wash. order, 2011-08-11 dismissal analysis · N.D. Cal. order, 2013-12-23 disposition [contemporaneous record · strong fence]

• Learning verdict. This is a governance failure lab because product-law classification and consumer fairness belonged in ex-ante board diligence. The record cannot establish liability, causation, Josh's vote, total loss, shutdown reason, or the content of any process change. [researcher inference · strong]

Portfolio-level forensics

Pattern tests

• First-check specialization is supported at firm level. The 2011 SEC letter and current profile emphasize the earliest company-building period. Whether the pattern produced superior net returns is not public. SEC 2011 comment letter, pp. 1–3 [contemporaneous record · strong]
• Adaptation appears repeatedly but is not automatically virtuous. Notion rebuilt, Upstart pivoted, and Knewton/AltSchool pivoted or restructured with much weaker observable outcomes. The discriminating variable should be evidence per dollar and governance stop rules, not persistence alone. [researcher inference · strong]
• Regulation is portfolio-level, not an edge case. Upstart, Clover, Flatiron, AppNexus, and Project Fair Bid expose credit, insurance, clinical-data, privacy, consumer-protection, and gaming/auction risks. Product diligence without a legal/data-governance lane would be incomplete. [researcher inference · strong]
• Partner attribution must be deal-specific. Official pages assign some companies to Josh, some jointly, and Uber/Square/Roblox to other partners. Legal GP control and firm economic exposure should not be rewritten as sole sourcing or decision ownership. First Round companies, partner filters [researcher inference · strong]

Failure, miss, anti-portfolio, and adverse evidence

Uncertainty and retrieval ledger

These gaps qualify the assessment; they are not invitations to replace missing evidence with a valuation-derived return or a lone-investor story.

Reasonable public routes for DEALS were attempted: official firm/company pages and archives; SEC EDGAR S-1, 8-K, agreement, comment-letter and Form D records; company/acquirer releases; founder and Josh direct-voice material; decision-era and outcome press; litigation/criticism/SEC queries; complaint and court dispositions; aliases; compiled lists used only for discovery; and named company/partner searches. Dynamic pages provide no stable row locators, some press may be paywalled or change, and private LP/board/cap-table records were not publicly available. Further research should begin with the exact missing packets above rather than another famous-win search. [researcher inference · strong]

Coverage reconciliation

The writing universe is 27 items = 17 priority + 10 lower priority. The 17 priority items divide into 10 accessible sources with exact-schema deep notes and 7 inaccessible Redeye routes. This is an item denominator, not a claim that Kopelman wrote only 27 things: the LinkedIn index exposes eight articles, while Redeye's complete post count cannot be recovered. LinkedIn profile, Articles by Josh · Redeye canonical route [researcher inference · strong]

Exact priority set

Ten lower-priority items

Inaccessible Redeye route ledger

No body below is represented as read. Canonical URLs were attempted directly; the LinkedIn profile's blog/RSS routes, FeedBurner, title/slug variants, citation trails, and third-party archive indexes were swept. The surviving material supports metadata or short-fragment discovery only. LinkedIn profile, Websites · Redeye parser archive [researcher inference · strong on access boundary]

Deep source-by-source analysis

1. 15 Rejections That Changed My Life

Access boundary. Full LinkedIn article and explicit 2026-02-11 date; no contemporaneous rejection letters or fundraising records. Article, title/date and ¶5–11 [retrospective witness · strong on testimony]

Biography and intellectual development. The article connects Infonautics' college formation to rejected employment options and later financing. It helps explain attention to persistence under sparse validation, while the clean 1992 start date on the firm biography remains a date variant. Article, ¶5–11 · First Round profile, biography [researcher inference · strong sequence, moderate date]

Actual-decision comparison and evolution. Unlike the later Watney rule, which responds to changed operating evidence, this story is about preserving a path before a decisive external signal. The distinction matters: persistence is supportable only when new evidence changes. Watney, embedded letter operating guidance [researcher inference · moderate]

Modern transfer test — 2026-08-24. Ask: what independent signal improved, which assumption changed, and what dated milestone would stop funding? A labeled hypothetical AI company with repeated fundraising rejection but rising paid retention could continue; flat retention with unchanged assumptions is the failure boundary. The article's counterfactual—an accepted job would have ended the company—remains untestable.

2. First Round investor role post

Access boundary. Full post, but the activity ID/current crawl does not expose a stable publication date; “2026” is provisional rather than activity-ID certainty. Canonical post [investor-stated · strong text, provisional date]

Biography and intellectual development. The role institutionalizes sourcing work once associated with an individual investor: cold outreach, repeated follow-up, events, market/customer/technical curiosity, simplification, self-direction, and apprenticeship. Post, role and candidate paragraphs [investor-stated · strong]

Actual-decision comparison and evolution. Finding 9 in the 2015 portfolio analysis motivated open/direct inbound; the role adds active outbound and relationship creation. It is an evolution from channel openness to a repeatable access function, not proof of better outcomes. 10 Years, Finding 9, lines 81–83 [researcher inference · moderate]

Modern transfer test — 2026-08-24. Record source channel, relationship age, independent partner score, founder references, first falsifiable customer/technical claim, and conversion by channel. The current post is a cited implementation example; outcome evidence is not yet available. Failure boundary: “builders enjoy them” becomes affinity bias unless audited against channel, geography, and demographic mix.

3. Yale anchor post

Access boundary. Full canonical post; publication date stays provisional because LinkedIn exposes activity metadata/relative age rather than a stable date. LP documents are not public. Canonical post [retrospective witness · strong text, provisional date]

Biography and intellectual development. Kopelman says Nick Shalek's July 2007 cold email led Yale to anchor First Round's first institutional venture fund and credits David Swensen, Matt Mendelsohn, Tim Sullivan, and the Yale team with humility/alignment norms. Post, opening through influence passage [investor-stated · strong]

Actual-decision comparison and evolution. The post places the model's institutionalization after a $7 million first fund, but it does not expose vehicle terms or which strategy changes Yale caused. Origins also names Chris Douvos in the fundraising story, so “first” requires legal-document resolution. Origins show notes [contemporaneous record · moderate]

Modern transfer test — 2026-08-24. Before adding an LP/vehicle, ask whether fee/carry, ownership targets, reserves, succession, and reporting fit the claimed stage edge. Changed assumption: institutional scale can alter decision incentives. Failure boundary: prestige-based deference suppresses independent model review. A current example is Fund X-F's formal scale/related-person record, which still reveals no terms. Fund X-F Form D [contemporaneous record · strong role, insufficient economics]

4. 10 Years of Investing, 10 Lessons Learned

Access boundary. Full article and disclosed method; no raw company-level dataset, code, or 2026 replication. Article, lines 34–41 and 92–96 [researcher inference · moderate]

Biography and intellectual development. The essay marks a transition from intuition-only seed judgment to inspecting roughly 300 companies/nearly 600 founders. It removes Uber, mixes exit/fair-value marks, and explicitly disclaims representativeness/statistical significance. Article, lines 36–41 and 92–96 [researcher inference · strong]

Actual-decision comparison and evolution. The clearest stated action was the Open App after direct/non-referral sourcing appeared strong; the article says other findings would not dictate investment selection. No commenter assertion is used. Article, Finding 9 and closing, lines 81–90 [investor-stated · strong]

Modern transfer test — 2026-08-24. Preregister outcome and vintage, include write-offs/stale marks, separate sourcing from selection, and rerun after removing the top three outcomes. Diagnostic questions: does a result survive later write-downs and channel-cost adjustment? Failure boundary: turning one portfolio's multiple-tested correlations into founder stereotypes.

5. Crystal Balls vs. Calculators

Access boundary. Full article embedding a May 5, 2015 institutional LP letter. Published scorecards are described but not exposed; audited return verification is unavailable. Article, lines 31–82 [contemporaneous record · strong]

Biography and intellectual development. The letter reduces return intuition to entry/exit price, rejects cycle timing, and argues for stable fund size/pace while acknowledging unrealized marks and future write-offs. Article, lines 47–77 [investor-stated · strong]

Actual-decision comparison and evolution. First Round says it did not materially enlarge its fund or pace and temporarily slowed to ten investments over two quarters. That is stated contemporaneous behavior, not evidence the choice later outperformed. Article, lines 62 and 68–71 [researcher inference · strong]

Modern transfer test — 2026-08-24. Compute required exit value from check, ownership, dilution, reserve, fund size, time, fees, and target return; stress 50% lower exit multiples and two extra rounds. Diagnostic question: is conviction actually higher or merely the entry price? Failure boundary: the two-number slogan hides ownership, loss rate, duration, and liquidity.

6. The Watney Rule for Startups

Access boundary. Full article with an anonymous board anecdote and a republished First Round partners' letter; the core letter is collective, not Josh-only. Article, opening and embedded letter [investor-stated · strong; anecdote provisional]

Biography and intellectual development. It extends 2015 entry-price discipline into portfolio operating plans: prior assumptions favored growth and future capital; changed conditions elevate runway, milestones, unit economics, syndicate capacity, and self-sufficiency. Embedded letter, operating guidance [investor-stated · strong]

Actual-decision comparison and evolution. The anonymous company prevents company/outcome verification. The letter says First Round kept investing across cycles while thinking more defensively about reserves; no stop-rule denominator is supplied. [investor-stated · moderate]

Modern transfer test — 2026-08-24. Model no-funding, flat-round, and delayed-funding cases. Ask which costs are reversible, which regulated duties cannot be deferred, which milestone earns the next dollar, and which syndicate member can actually fund. Changed assumption: capital availability. Failure boundary: reserves preserve a weak company and crowd out better options.

7. What It's Like to Lead a Startup According to 700+ Founders

Access boundary. Full wrapper article; the linked 2016 report is a legacy route without reliably accessible versioned data. Article, opening · report route [investor-stated · moderate]

Biography and intellectual development. The article describes a repeat survey with more than 700 founders inside/outside First Round and new lead-investor/diversity questions. It shows founder-side measurement becoming recurrent. Article, opening paragraphs [investor-stated · strong]

Actual-decision comparison and evolution. The new questions evolve the 2015 instrument; no public route ties a particular result to an investment, board, or platform decision.

Modern transfer test — 2026-08-24. Preserve question text, response rate, stage, portfolio status, geography, and nonresponse analysis. Ask whether experience predicts retention/referral without pretending it equals investment performance. Failure boundary: voluntary founder sentiment is marketed as causal proof.

8. What does it mean to be a startup entrepreneur in 2015?

Access boundary. Full short wrapper; underlying interactive report remains partial. Comments are excluded. Article body · report route [investor-stated · moderate]

Biography and intellectual development. The article says the survey reached more than 500 venture-backed founders and fewer than one quarter were in First Round's community, demonstrating an attempt to broaden feedback beyond the portfolio. Article body [investor-stated · strong]

Actual-decision comparison and evolution. It sits alongside the 2015 portfolio analysis: one studies investor records, the other founder perceptions. No specific behavior change is documented.

Modern transfer test — 2026-08-24. Report recruitment, response/nonresponse, stage, company status, and exact denominators; compare portfolio/non-portfolio under one instrument. Changed assumption: venture-backed respondents still omit unfunded and pre-financing failures. Failure boundary: cross-sectional perception is treated as a causal performance measure.

9. What the Seed Funding Boom Means

Access boundary. Partial legacy Inc/First Round Review page; exact original date, full text, edit history, and exhibits remain unresolved. Legacy page [investor-stated · moderate]

Biography and intellectual development. The surviving route attributes to Kopelman the view that falling startup costs created a financing gap and that seed work differs operationally from later-stage venture. The 2011 letter supplies harder contemporaneous check/stage evidence. SEC letter, pp. 1–3 [researcher inference · strong]

Actual-decision comparison and evolution. First Round's stated $250,000–$750,000 initial-check range is behaviorally consistent with specialization, but no company memo or comparison cohort proves the mechanism. SEC letter, pp. 2–3 [contemporaneous record · strong at firm level]

Modern transfer test — 2026-08-24. Compare check to total round, expected ownership, reserve need, milestone length, and partner service load. Ask whether a “seed” label conceals growth-style pricing or platform overhead. Changed assumption: today's seed rounds can be much larger. Failure boundary: stage branding without aligned vehicle economics. Uncapped is the cited current example for stage/fund-model arithmetic. Uncapped, 10:49–26:36 [investor-stated · strong]

10. The defunct startup that beat Best Buy in court

Access boundary. Full/archived narrative route, checked against court docket, Best Buy filing, contemporaneous reporting, discussion, and legal scholarship. Comments are not facts. Fortune · docket, entries 220–235 · Best Buy 10-K, Legal Proceedings [researcher inference · strong event]

Biography and intellectual development. It is rare public evidence of First Round supporting a post-operating-company asset-enforcement decision. The record does not expose the partnership vote, costs, vehicle, settlement alternatives, collection, or net distribution.

Actual-decision comparison and evolution. Unlike Watney's operating-runway decision, TechForward funded legal rights after operating failure. The docket corroborates verdict/order chronology; the narrative alone cannot establish every legal claim. Docket · TechCrunch, 2012-12-05 [contemporaneous record · strong]

Modern transfer test — 2026-08-24. Obtain independent merits counsel; probability-weight recovery, appeal, collection, delay, and fees; identify conflicts; set a stop-loss. Diagnostic question: is this enforceable residual value or emotional escalation? Boundary: jurisdiction-specific litigation-finance and professional-duty rules. Independent legal scholarship is a current research check, not controlling law. NYU JLB PDF, TechForward discussion [researcher inference · moderate]

Cross-source synthesis

Access becomes an operating system

The evidence progresses from legacy claims that opportunities can come from beyond conventional referrals, to Finding 9's direct/open channel, to First Search, and then to a 2026 role built around proactive relationship creation. 10 Years, Finding 9 · First Search thread · investor role post [researcher inference · moderate] The strongest support is behavioral channel design; the missing evidence is channel-adjusted hit rate and selection bias.

Judgment is paired with explicit arithmetic

Crystal Balls says uncertainty does not excuse ignoring entry/exit arithmetic; Watney translates changed capital conditions into company runway/milestones; the seed-boom essay and SEC letter constrain the mechanism to stage/vehicle fit. Crystal Balls, lines 31–71 · Watney, embedded letter · SEC letter, pp. 1–3 [researcher inference · strong stated consistency] This is not a published return record.

Learning moves from stories to instrumentation

The rejection essay supplies a personal origin story; 10 Years mines portfolio records; State of Startups asks founders; current Decision Education describes more explicit decision process. 15 Rejections, ¶5–12 · 10 Years, lines 34–41 · Decision Education, 32:58–46:19 [researcher inference · moderate] Each layer has a different bias: memory, marks, voluntary respondents, and retrospective process narration.

Failure changes the asset and the decision

Watney treats runway as an operating option; TechForward treats enforceable rights as a residual asset; the same word “support” cannot obscure different expected-value models. Watney · TechForward docket [researcher inference · strong on distinction] Crown-jewel enforcement has only one documented application and should not become a full general model.

Authorship and attribution remain layered

Watney is a partner letter, State of Startups is a firm research product introduced by Josh, TechForward is an interested first-person narrative, and the Yale/role posts are personal accounts. Firm/company outcomes therefore cannot be assigned to Josh solely because he published or narrated them. Watney · 700+ Founders [researcher inference · strong]

Chronology of the written philosophy

Decision playbook and modern translation

Best five writing sources to study first

1. Crystal Balls vs. Calculators — clearest contemporaneous stage/price/pace/mark boundary.
2. 10 Years of Investing — rare method disclosure, findings, behavior change, and explicit statistical caveats.
3. The Watney Rule — regime-change translation, with collective-authorship and anonymous-case fences.
4. 15 Rejections — personal origin evidence and a clean counterfactual/survivor-bias lesson.
5. TechForward account plus docket — strongest example of why interested narrative requires procedural triangulation.

Unresolved evidence gaps

• Complete Redeye bodies and exact publication dates for seven priority routes.
• Raw 2015 portfolio data, definitions, marks, write-offs, dilution, and later-vintage replication.
• Static State of Startups instruments, response rates, subgroup denominators, and versioned exports.
• Josh-specific deal/rejection denominator, contemporaneous memos, votes, reserve choices, and price/ownership histories.
• Vehicle fees/carry, LP terms, Yale commitment documents, partner compensation, and realized cash flows.
• TechForward cost, vehicle, appeal/collection history, and net distribution.
• Evidence that the 2026 investor role improves access or outcomes after channel/time controls.
• Falsification records: where the stated heuristics failed, were overturned, or produced ordinary outcomes.

Coverage and evidence base

The spoken-source sweep returned 30 search results representing 25 distinct Josh Kopelman appearances after removing mirrors, reposts, duplicate podcast feeds, and the unrelated restaurateur Josh Kopel. Eleven appearances were processed into source notes: all eight distinct ★★★/★★ priority appearances in the source map, plus three contextual sources that add chronology, vehicle, or decision-process evidence. Eight have a full official transcript, edited transcript, or complete public caption route; three are partial because a first-party transcript is absent or gated. Repetition below is evidence of consistency, not proof that a rule improves returns.

The remaining fourteen distinct appearances are accounted for in sources.md or the retrieval ledger below. Their public routes were metadata-only, lower-yield panels, edited interviews, or archive items; none is silently treated as content-analyzed. The Colossus legacy URL and the Upfront article/video pair are version routes for two appearances, not extra denominator entries.

Source-by-source guide

2014 — Yale Entrepreneurship Speaker Series

Contemporaneous-to-early-fund-development; retrospective on named deals. The lecture is the cleanest early statement of seed as hypothesis financing: back founders who adapt, fund cheap tests, and prefer observable customer improvement. The Twitter and Uber stories are already success-conditioned by 2014, and Uber is a firm association rather than personal selection evidence. Use this for the early framework; do not use it as a realized-return record. Evidence note.

2015 — FundersClub / UPROUND fireside

Contemporaneous stage commitment; retrospective career and deal cases. Kopelman says First Round will stay at seed, explains Dorm Room Fund’s student-led selection with a retained veto, and describes roughly 40 pre-fund angel decisions as repetitions that built judgment. The later growth of reserves qualifies the purity of the stage claim. Evidence note.

2018 — Knowledge at Wharton

Contemporaneous mandate explanation; retrospective capital figures. The edited transcript explains the original capital-gap logic, the first-24-month focus, and why founder quality dominates sparse seed evidence. “Grit” is directionally useful but not operationally defined; the company-capital figures are recollections rather than financing-ledger evidence. Evidence note.

2019 — 20VC

Contemporaneous firm process; retrospective misses. The official audio and show notes connect price, first-check ownership, two-thirds partnership approval, question-led boards, and explicit Blue Apron handoff to Phin Barnes. The canonical page has no full first-party transcript, so the note paraphrases checked timestamp windows rather than manufacturing quotations. Evidence note.

2020 — Invest Like the Best / Colossus

Contemporaneous philosophy; retrospective portfolio examples. The episode supplies the navigator-versus-cartographer distinction, present-problem underwriting, single-user wedges, question-led boards, and First Round’s community/platform model. The full Colossus transcript now requires login; public chapters and a third-party ASR transcript support the locators, including Roblox at 16:27–17:01 and community at 43:29–47:28. Evidence note.

2020 — Upfront Summit with Mark Suster

Contemporaneous market diagnosis; retrospective firm metrics. Kopelman argues that the seed capital-gap arbitrage had inverted, making selection and firm product more important. He describes First Round as a company serving founders and reports reducing average decision time from roughly 90 days to nine. The metric has no visible distribution or quality control, and the video explicitly attributes Uber to Rob Hayes. Evidence note.

2020 — Origins

Retrospective biography. The episode is most useful for family influence, founder-to-angel-to-institution chronology, team expansion, and Kopelman’s description of First Round as a “forever job.” Apple offers official audio but not a transcript; chapter and clip text are third-party. Its platform biography contains an Infonautics chronology error and is not used to settle dates. Evidence note.

2022 — LDV Vision Summit

Contemporaneous philosophy; retrospective portfolio associations. Kopelman calls the product a “DVR” of founder decisions, warns operator-investors not to impose their recipe, and says First Round did not have a computer-vision thesis. The source is valuable for anti-theme and founder-autonomy logic, but product remains an imperfect founder-level proxy. Evidence note.

2024 — Decision Education with Annie Duke

Current process description; retrospective cases. This is the strongest process source: independent pre-votes, explicit rubrics, disagreement, a non-investing moderator, intermediate PMF states, and preserved funded/passed “game tape.” It also makes the most explicit distinction between decision quality and outcome. The official transcript is AI-assisted and the process efficacy is not independently evaluated. Evidence note.

2025 — Uncapped with Jack Altman

Current fund architecture; retrospective misses. Kopelman criticizes logo-based venture reputation, defines the first 24–36 months as First Round’s “power alley,” and says most capital is now reserved for follow-ons while doubting the firm’s follow-on skill. That candid tension is more useful than a clean seed-purity story. Evidence note.

2026 — Vintage CEO Summit with Dick Costolo

Current process correction; retrospective unnamed case. Costolo corroborates the independent rubric, and Kopelman identifies a concrete error: a 45-minute presentation outweighed months of execution. First Round revised the rubric, demonstrating that explicit process can be debugged. The company is unnamed, so neither the miss nor its counterfactual value can be verified. Evidence note.

Recurring frameworks and operating models

1. Seed capital buys information

• [investor-stated · strong] The plan is a hypothesis, and the seed round should validate, de-risk, or disprove it quickly and capital-efficiently. Yale states the model directly; Wharton explains the falling-cost conditions that made it a distinct financing product. Yale, 07:47–08:34; Wharton, capital-gap answer.
• [researcher inference · moderate] A modern underwriting memo should specify the uncertainty, cheapest discriminating test, expected evidence date, and abandonment condition. Otherwise “experimentation” can excuse undirected burn.

2. Underwrite navigation, not map accuracy

• [investor-stated · strong] Across Yale, Colossus, Wharton, and LDV, Kopelman favors adaptable, nonconforming founders over people whose original forecasts merely sound precise. Yale, 05:35–06:32; Colossus, 11:50–14:05; LDV, 05:29–07:40.
• [investor-stated · moderate] Product choices provide pre-pitch evidence of judgment; later operating evidence should outweigh presentation polish. LDV, 05:29–07:40; Vintage, 06:13–08:17.
• [researcher inference · moderate] The strongest observable version is not “founder pivots often,” but “founder changes belief after relevant evidence while preserving a coherent objective.”

3. Separate decision quality from outcome quality

• [investor-stated · strong] Decision Education and Uncapped reject “resulting”—judging a decision solely by its later outcome—and treat private marks as particularly weak intermediate evidence. Decision Education, resulting/PMF turns; Uncapped, 17:37–18:20.
• [investor-stated · strong] First Round records independent votes, rubric scores, funded and passed companies, and later evidence as institutional “game tape.” Decision Education, IC/game-tape turns.
• [researcher inference · provisional] This should improve auditability and learning. No source demonstrates higher realized returns versus an otherwise comparable investment committee.

4. Disagreement is data

• [investor-stated · strong] Members score before discussion, every level submits, two-thirds approval replaces individual veto, and a non-investing moderator looks for process quality. Decision Education, IC turns; 20VC, 12:43–13:51.
• [investor-stated · strong] Areas of score dispersion—not merely mean enthusiasm—direct the committee to hidden assumptions. Decision Education, confidence-dispersion turn.
• [researcher inference · moderate] The process reduces visible hierarchy and anchoring risks, but sponsor effects can migrate into rubric design, deal access, or post-vote interpretation.

5. Prefer present pain to forecasted themes

• [investor-stated · moderate] Colossus and LDV favor a current problem, useful application, and narrow wedge over a stack of thematic predictions. Uber is framed as making an existing market faster, easier, and cheaper. Colossus, 19:36–20:45; LDV, 14:28–16:42; Yale, 20:00–20:59.
• [researcher inference · moderate] “Anti-thesis” does not mean market-blind. The test is whether a company can show user value before every macro forecast must be true.
• [researcher inference · moderate] Deep infrastructure and regulated markets may require underwriting a future state before a narrow user wedge is economically visible.

6. Complete fund math beats logo counting

• [investor-stated · strong] 20VC, Upfront, and Uncapped repeatedly connect selection with entry price, initial ownership, dilution, reserves, exit value, and duration. 20VC, 07:48–10:36; Upfront, 11:12–12:49; Uncapped, 43:23–46:20.
• [retrospective witness · moderate] Twitter and Dropbox show the cost of price discipline; the talks do not establish that a more permissive policy would have improved the portfolio after all losing deals. Yale, 14:20–15:29; 20VC, 07:48–10:36.
• [researcher inference · strong] Any case claim should report the vehicle, check, round, ownership path, follow-ons, attribution, and cash realization—or explicitly mark each unknown.

7. Stage specialization is a marginal-advantage claim

• [investor-stated · strong] From 2015 through 2025, Kopelman locates First Round’s product in the company’s earliest 24–36 months: product-market fit, early team, culture, pricing, and positioning. FundersClub, 07:38–07:42; Wharton, mandate answer; Uncapped, 44:48–45:12.
• [investor-stated · strong] The firm stayed stage-focused but moved from mostly initial-check deployment to a majority reserve allocation. FundersClub, 07:38–07:42; Uncapped, 44:26–44:42.
• [researcher inference · moderate] The defensible claim is specialization in first-round selection and early support, not literal absence from later rounds.

8. The venture firm is a company

• [investor-stated · strong] Upfront defines founders as customers, LPs as shareholders, and firm experiences, knowledge, software, network, and decision speed as products. Upfront, 05:09–07:30.
• [investor-stated · moderate] Colossus and FundersClub extend this through portfolio community and distributed networks. Colossus, 43:29–47:28; FundersClub, 08:08–12:30.
• [researcher inference · moderate] Platform value requires usage, trust, density, and measured outcomes. Service breadth alone can become agency overhead that dilutes investing attention.

9. Boards improve founder decisions through questions

• [investor-stated · strong] 20VC, Colossus, and LDV consistently say the founder drives; a new board member should context-load, then ask questions rather than import an operator recipe. 20VC, 14:12–15:25; Colossus, 30:28–32:18; LDV, 09:48–13:56.
• [researcher inference · moderate] Question-led governance protects agency in ordinary conditions but needs a separate escalation rule for ethics, solvency, or fiduciary emergencies.

10. Institution before hero attribution

• [investor-stated · strong] Kopelman criticizes attribution culture, describes partnership voting and handoffs, and credits Uber to Rob Hayes and Blue Apron board work to Phin Barnes. 20VC, 12:43–15:25 and 30:20–31:20; Upfront, 02:03–03:40.
• [researcher inference · strong] Firm portfolio membership cannot establish Kopelman’s sourcing, selection, board role, ownership, or personal return. Roblox and other examples remain firm-level unless a primary record names the partner.

Intellectual chronology, evolution, and contradictions

Contradictions that should remain visible

• Price discipline versus outlier capture: Twitter/Dropbox show that price filters can reject extraordinary companies. The corpus never supplies the losing-deal denominator needed to choose a better threshold. 20VC, 07:48–10:36.
• Seed purity versus reserves: “Stay at seed” is consistent with a first-check mandate, but not with a literal reading once a majority of capital is reserved. FundersClub, 07:38–07:42; Uncapped, 44:26–44:42.
• Storytelling versus operating evidence: Storytelling appears in the 2024 rubric; the 2026 correction says the committee overweighted the pitch. The synthesis is to measure storytelling without allowing a meeting to erase longitudinal evidence. Decision Education, rubric turn; Vintage, 06:13–08:17.
• Anti-theme versus market knowledge: Kopelman rejects top-down vision theses but still uses existing-problem quality, market awareness, and timing. “No thesis” cannot mean no causal market model. LDV, 14:28–16:42; Colossus, 19:36–20:45.
• Community scale versus trust: Larger networks create more possible matches and more noise. None of the talks reports a causal platform-performance comparison. Colossus, 43:29–47:28.
• Process quality versus returns: Better artifacts make errors legible; they do not prove better cash outcomes.
• Collaboration versus accountability: Institution-first credit reduces hero narratives, but can obscure who sourced, sponsored, voted, served, or made follow-on choices unless those roles are explicitly logged.

Biography, deal cases, and attribution boundaries

• Infonautics: Retrospective sources describe the right problem/wrong initial solution and declining capital needs. The start year conflicts—1991 in Decision Education and 1992 on First Round’s profile—and should stay unresolved without corporate records. Decision Education, biography turns.
• Half.com: Used to show recombination of existing marketplace designs and lower startup capital. It is operator history, not proof of venture selection skill. Decision Education, Half.com turn; Wharton, biography answer.
• TurnTide: Wharton gives a recalled $750,000 to first product. Transaction history belongs in the investments phase, not as a talk-derived return. Wharton, biography answer.
• Twitter: Kopelman repeatedly calls the price-based pass a major miss. The case supports humility and price/outlier tension; it does not prove the ex-ante rule was irrational. Yale, 14:20–15:29; FundersClub, 14:50–17:34.
• Square: The Twitter relationship is said to have contributed to later Square access. This is relationship option value, not a complete investment attribution or return record. Yale, 14:20–15:29.
• Uber: The Yale lecture discusses company-market fit, while Upfront explicitly credits partner Rob Hayes. Preserve firm/company and partner roles separately. Yale, 20:00–20:59; Upfront, 02:03–03:40.
• Roblox: Colossus uses it as a seed-snapshot/adaptability case; Uncapped mentions later exposure. Without the original memo, vehicle, check, and named partner, keep attribution at firm level. Colossus, 16:27–17:01; Uncapped, 43:23–46:20.
• Blue Apron: 20VC records a board handoff to Phin Barnes, direct evidence that board ownership and public firm association can differ. 20VC, 30:20–31:20.
• Dorm Room Fund: The FundersClub talk reports student ICs, small checks, and a firm veto. Treat the counts as retrospective self-report until reconciled to program records. FundersClub, 08:08–12:30.

Modern VC translation — as of 2026-08-24

What transfers

• [researcher inference · strong] Record a falsifiable pre-mortem, independent scores, confidence, key disagreement, price/ownership case, and next evidence date before partner discussion.
• [researcher inference · strong] Review both investments and passes. For each, distinguish decision quality at time zero from subsequent luck, financing markets, and execution.
• [researcher inference · moderate] Use product telemetry, customer behavior, build cadence, and changed beliefs as founder-decision evidence; presentation is one signal, not the evidence set.
• [researcher inference · strong] Maintain a role ledger: source, sponsor, vote, board, reserve decision, handoff, vehicle, ownership, and realization.

What can fail

• [researcher inference · strong] First Round’s apparent process edge may be largely selection access, brand, network, and accumulated reputation. The decision system could document advantages it did not cause.
• [researcher inference · moderate] A rubric can institutionalize yesterday’s pattern and penalize unfamiliar founders more consistently, especially when dimensions such as “killer instinct” or “storytelling” are not behaviorally anchored.
• [researcher inference · moderate] Fast decisions and founder-service breadth can trade diligence and investing focus for conversion and brand.

Falsifiable edge

[researcher inference · provisional] Relative to the firm’s prior process or a pre-specified comparison group, rubric use should improve out-of-sample calibration of 6-, 18-, and 36-month operating forecasts for both funded and passed companies. A proper test needs timestamped pre-discussion forecasts, locked metrics, the rejected set, partner identifiers, and no substitution of financing marks for operating outcomes.

Misuse warnings

• Do not convert “navigator” into post-hoc praise for any pivot.
• Do not call a large logo a good investment without entry price, ownership, dilution, reserves, exit, duration, and vehicle.
• Do not treat a firm community event as platform value without usage and outcome evidence.
• Do not use “founder autonomy” to avoid fiduciary escalation.
• Do not equate a polished investment-committee artifact with a calibrated forecast.

Listen to these five first

1. Decision Education, 2024 — best complete account of the decision operating system and its intended safeguards.
2. Yale, 2014 — earliest clear formulation of founder adaptability, hypothesis financing, and present-problem underwriting.
3. Colossus, 2020 — strongest broad synthesis of navigation, boards, platform/community, and seed evolution.
4. Vintage CEO Summit, 2026 — best evidence that the rubric is revised after a diagnosed decision error.
5. Uncapped, 2025 — most useful current statement of complete fund math, stage advantage, reserves, and uncertainty.

For portfolio mechanics and attribution specifically, read 20VC immediately after the five.

Retrieval and access ledger

These fourteen distinct, lower-priority appearances complete the 25-item denominator after the eleven noted appearances. They were discovered and triaged but not represented as deeply read source notes.

Version and provenance controls

• Colossus: https://colossus.com/episode/kopelman-the-past-present-and-future-of-seed-investing/ is canonical; the joincolossus.com route is a redirect, not another appearance.
• 20VC: https://www.thetwentyminutevc.com/joshkopelman is canonical and dated January 27, 2019. It offers show notes and audio, not an official transcript.
• Decision Education: the canonical page’s explicit footer date, December 11, 2024, controls over conflicting December 4/12 header and mirror displays.
• Upfront: the TechCrunch article dated February 3 and YouTube upload dated April 3 document one Summit interview.
• YouTube notes: Yale, FundersClub, Upfront, LDV, Uncapped, and Vintage use complete public automatic captions. Quotes are limited to clear passages; caption files and full transcripts are not committed.
• Partial notes: 20VC uses official audio/show notes, Colossus uses official chapters plus a third-party transcript cross-check, and Origins uses official audio plus third-party chapter/clip routes.

Decision-use summary

Kopelman’s most defensible edge is not clairvoyance. It is an attempt to turn uncertain seed judgment into an inspectable institutional process: define the hypothesis, assess the founder’s learning system, preserve independent views, surface disagreement, price the ownership path, record attribution, and revisit both wins and passes. The strongest evidence shows that this process exists and changes. The missing evidence is whether it produces better calibrated forecasts or realized cash outcomes than access, reputation, and portfolio luck would produce without it.

Model 1 — Adaptive founders with falsification gates

• Rule — Back founders who can update the route without losing the problem, but fund adaptation only through predeclared falsification gates: a named hypothesis, cheapest discriminating test, dated evidence threshold, runway limit, and stop or pivot decision. Persistence without new evidence is escalation, not adaptability. [investor-stated · strong]
• Evidence status — Kopelman repeatedly states the founder-adaptability criterion; cheap hypothesis testing is also direct voice. Application evidence is retrospective and selected, so the mechanism is strongly attributed but only moderately demonstrated across outcomes. [investor-stated · strong attribution / moderate application support]
• Provenance and origin — The earliest processed articulation is the 2014 Yale lecture: assess how founders adapt because the initial plan will change, and use seed money to validate or disprove a hypothesis quickly. Colossus later compresses the idea into “navigators, not cartographers,” while Decision Education describes the first 18–24 months as hypothesis testing. Yale lecture, 05:35–08:34 · Colossus, 11:50–14:05 · Decision Education, transcript discussion of the imagine-if stage [investor-stated · strong]
• Mechanism and reasoning — Seed forecasts decay faster than operating evidence accumulates. A founder who notices disconfirming signals, designs a narrow test, and changes course before cash runs out converts uncertainty into information; a gate prevents the same language from rationalizing endless pivots. [investor-stated · strong]
• Historical and vehicle context — Falling software-startup costs let First Round use minority seed checks as experiments; its 2011 SEC letter reported typical initial checks of $250,000–$750,000 and possible follow-ons. That structure creates option value only if reserve decisions respond to evidence. SEC comment letter, pp. 1–3 [contemporaneous record · strong]
• When to use — Use for seed software and marketplace companies where customer, product, or technical feedback can arrive inside one financing runway. Do not use it to waive clinical, regulatory, hardware, or infrastructure dependencies that require longer tests; there the gate must match the actual evidence clock. [researcher inference · strong]
• How to apply it — Before investment: write the current problem, founder's falsifiable premise, cheapest test, signal/noise boundary, deadline, burn, alternate explanation, and kill condition. Before every reserve: compare the forecast with observed evidence, ask what changed, and choose continue, pivot, recapitalize, sell, or stop. IC questions: “What result would make the founder abandon this route?” “How many tests fit in runway?” “Is the next check buying information or merely time?” [researcher inference · strong]
• Examples — Four distinct cases ground the model:
  • Documented application — Upstart. Founder Dave Girouard recalls weak demand for income-share agreements, a Kopelman board question about customer pull, and a rapid move to loans when demand appeared. The team—not the investor—executed the pivot; formal vote and reserve history are unknown. First Round Review, board-question and pivot sections [retrospective witness · strong]
  • Documented application — Infonautics. Kopelman describes pursuing the right information-access problem with an early solution that proved wrong; this founder experience is a retrospective origin illustration, not proof it predicted later investing skill. Decision Education, Infonautics discussion [retrospective witness · moderate]
  • Researcher-applied illustration — Notion. First Round says Notion rebuilt four times before traction. The record supports repeated change but not the milestone, reserve, or board gates that distinguished learning from escalation. First Round manifesto, Notion passage · Notion decision lab [researcher inference · moderate]
  • Researcher-applied illustration — Knewton counterexample. More than $180 million of reported financing, a pivot toward open educational resources, funding stress, and an undisclosed Wiley sale show why adaptation without public learning and reserve gates cannot be presumed valuable. Josh/FRC attribution remains provisional. Inside Higher Ed, 2019-05-07 article body [researcher inference · provisional attribution / moderate outcome support]
• Evidence — Direct support comes from the Yale lecture's adaptation and experiment passages, Colossus's navigator criterion, Decision Education's hypothesis-testing frame, and the 2026 Vintage account of revising a rubric after it overweighted a pitch relative to accumulated execution. Yale, 05:35–08:34 · Colossus, 11:50–14:05 · Decision Education, founder/rubric discussion · Vintage CEO Summit, 06:13–08:17 [retrospective witness · strong]
• Limits, failure modes, and exception ledger — “Adaptable” can be assigned after success, confuse noise-chasing with learning, reward polished narratives, and excuse reserve escalation. Notion is a survivor; Knewton and AltSchool show that pivots can follow heavy capital consumption. The public record lacks the rejected founder set, contemporaneous hypothesis logs, reserve votes, and failed founders who adapted equally well. Capital-intensive or regulated tests also violate the fast-feedback assumption. [researcher inference · strong]
• Evolution and contradictions — The 2014 “heat-seeking missile” metaphor became a more explicit 2024–2026 system: independent rubrics, intermediate product-market-fit states, decision game tape, and a rubric reweighted toward operating work. That is evolution from trait language toward observable evidence, but no calibration study shows better returns. Decision Education, rubric and PMF discussion · Vintage CEO Summit, 04:27–08:17 [documented behavior · moderate]
• Modern VC translation (as of 2026-08-24) — Preserve the adaptive-founder test, but make it auditable: store the pre-investment premise, confidence, evidence deadline, cost-to-learn, regulatory/technical dependency, and stop rule; compare updates with cohort outcomes by partner and vintage. In AI applications, rapid prototype changes count only when paid retention, workflow completion, unit cost, or defensibility improves; model swapping without user evidence is activity, not learning. [researcher inference · strong]

Model 2 — Start with a present problem, not a forecasted theme

• Rule — Prefer a painful problem observable now and a narrow usable wedge over a company whose value requires several forecasted market layers; add a theme only after the present user, transaction, and why-now mechanism are concrete. This is a rebuttable prior, not a ban on category creation. [investor-stated · moderate]
• Evidence status — Provisional model. Kopelman states the rule and supplies memorable retrospective illustrations, but the landed record contains only two usable cases, no contemporaneous First Round thesis memos, and no matched themed-versus-problem-led denominator. [investor-stated · strong attribution / provisional application support]
• Provenance and origin — Colossus contains the clearest formulation—“It starts with an existing problem”—and uses search as an example; Yale later supplies an existing-market description of Uber, while LDV records the related claim that First Round did not organize computer-vision applications around a top-down “vision thesis.” Colossus, 19:36–20:45 · Yale lecture, 20:00–20:59 · LDV Vision Summit, 14:28–16:42 [investor-stated · moderate]
• Mechanism and reasoning — A present problem yields inspectable behavior, an atomic job, and a short feedback loop; a grand theme can hide simultaneous assumptions about adoption, infrastructure, regulation, distribution, and timing. The wedge reduces joint-probability error without assuming the eventual market remains narrow. [investor-stated · strong]
• Historical and vehicle context — First Round formed as web startup costs fell and small checks could finance a customer test before a large category was proven. That setting favors application evidence, but network, infrastructure, biotech, and regulated businesses may need a longer why-now chain. Wharton transcript, capital-gap answer · SEC comment letter, p. 1 [retrospective witness · strong]
• When to use — Use when a seed pitch depends on a new behavior, workflow, marketplace, or AI capability and a real user can already demonstrate pain. Do not use it to reject enabling infrastructure merely because end demand is indirect; require a present bottleneck, committed design partner, or measurable cost instead. [researcher inference · strong]
• How to apply it — Write the user's current workaround, frequency, cost, urgency, buying authority, atomic wedge, why now, and first falsifying cohort. Decompose every theme claim into necessary layers and assign evidence/confidence to each. IC questions: “What happens if the theme arrives five years late?” “Who pays before the ecosystem is complete?” “Which single layer can be tested this quarter?” [researcher inference · strong]
• Examples — The exhausted record supports only two cases:
  • Documented application — Google search illustration. Kopelman uses Google's original search problem as the exemplar of beginning with an existing need rather than underwriting the later advertising and platform empire. It is an investor-stated external illustration, not a First Round decision. Colossus, 19:36–20:45 [investor-stated · moderate]
  • Documented application — Uber existing-market illustration. Kopelman describes Uber as making an existing market faster, easier, and cheaper. The example is retrospective and First Round credits the deal to Rob Hayes, so it supports the problem frame but not Kopelman-specific selection credit. Yale lecture, 20:00–20:59 · First Round companies, partner filters [retrospective witness · moderate]
• Evidence — The supporting works are Colossus's existing-problem chapter, Yale's Uber passage, LDV's anti-theme discussion, and the Wharton account of capital-light early testing. Colossus, 19:36–20:45 · Yale, 20:00–20:59 · LDV, 14:28–16:42 · Wharton, founder/capital-gap questions [investor-stated · moderate]
• Limits, failure modes, and exception ledger — A “present problem” can be redescribed after success; strong themes can reveal latent demand, and a narrow wedge can cap ambition or ignore platform timing. Exact exhausted-route ledger: the landed record processed Colossus's official chapters (full transcript gated; third-party ASR used only to check short locators), Yale and LDV videos with automatic captions, the current and archived First Round company/profile routes, the 36-company ledger and SEC filings, and the seven Redeye priority routes through canonical URL, feed, title variants, and parser/archive index. The Redeye bodies remained inaccessible; no decision-era First Round memo, rejected theme set, partner vote, or themed-versus-present-problem outcome cohort was located. Sources coverage assessment [researcher inference · strong]
• Evolution and contradictions — The rule sits in tension with First Round's broad “imagine if” language and investments whose market had to be created. LDV's “no vision thesis” narrows the claim from anti-vision to anti-top-down theme selection; implicit sector concentration may still arise through networks and partner expertise. Decision Education, imagine-if discussion · LDV, 14:28–16:42 [researcher inference · moderate]
• Modern VC translation (as of 2026-08-24) — For AI, climate, defense, and regulated software, require a present buyer, costly workaround, deployment path, and wedge metric before granting a thematic premium. Keep a separate upside tree for category expansion, but discount every unproven dependency. Misuse is equating a familiar current pain with venture-scale economics or rejecting infrastructure whose present customer is a developer rather than an end user. [researcher inference · strong]

Model 3 — Independent judgment and the three clocks

• Rule — Run every investment on three separate clocks: freeze independent judgment before group discussion; score operating evidence at predeclared intermediate milestones; and judge economic outcome only when ownership-adjusted cash or a clearly bounded terminal value exists. Never let a later mark rewrite the original information set. [investor-stated · strong]
• Evidence status — Independent pre-discussion rubrics, disagreement surfacing, a non-investing moderator, intermediate product-market-fit states, and decision “game tape” are investor-stated. Combining them into three clocks is a strongly supported researcher model; no controlled study establishes return improvement. [researcher inference · strong]
• Provenance and origin — Crystal Balls vs. Calculators separated uncertain forecasts from arithmetic and warned that quarterly marks are paper. Decision Education later made decision/outcome separation explicit and described independent scores, a moderator, four PMF states, and stored funded/passed decisions. Crystal Balls, lines 31–77 · Decision Education, transcript sections on decision quality, PMF states, moderator, and game tape [investor-stated · strong]
• Mechanism and reasoning — The decision clock preserves information before hierarchy and advocacy create anchoring; the evidence clock supplies feedback before a long venture outcome; the outcome clock prevents valuation, survival, or one exit from laundering a poor process. Comparing all three exposes whether the error was thesis, execution, follow-on, timing, or attribution. [researcher inference · strong]
• Historical and vehicle context — Seed funds decide with sparse evidence, then wait years while private marks, follow-ons, and dilution obscure cash performance. First Round's two-thirds partnership vote and pooled vehicles make both independent judgment and later attribution institutional rather than personal. 20VC, 12:43–13:51 · Upstart S-1, Principal Stockholders footnote [contemporaneous record · strong]
• When to use — Use for every IC decision, reserve, board escalation, exit, and postmortem. Do not replace qualitative judgment with a score or treat a milestone as universal; calibrate clocks to stage, regulation, and the evidence latency of the business. [researcher inference · strong]
• How to apply it — Before discussion, each voter records probability, rationale, dissent, key unknown, price/ownership, and falsifier. Set a dated evidence milestone and reserve decision. At review, preserve the original record, score what arrived, and separate company outcome from fund cash. Ask: “What did we know then?” “Which signal moved?” “Was the process good even if the outcome was bad?” “Is this mark realizable?” [researcher inference · strong]
• Examples — Four distinct cases show the clocks:
  • Documented application — unnamed rubric miss. Kopelman says First Round overweighted a 45-minute presentation and underweighted months of execution, then revised the rubric. The company is unnamed, so the original scores and counterfactual return cannot be checked. Vintage CEO Summit, 06:13–08:17 [retrospective witness · moderate]
  • Documented application — Upstart. Weak demand for income-share agreements and immediate loan demand were evidence-clock signals; the later IPO is a separate outcome clock and does not prove the seed process by itself. First Round Review, board/pivot sections · Upstart lab [retrospective witness · strong]
  • Researcher-applied illustration — Notion. Four rebuilds belong to the evidence clock; a high private valuation belongs to a later, still-unrealized outcome clock. Missing board and reserve records prevent a decision-clock audit. First Round manifesto, Notion passage · Notion lab [researcher inference · moderate]
  • Researcher-applied illustration — Clover Health. The announced $3.7 billion SPAC EV, later allegations, shareholder settlement metadata, and 2024 SEC staff disposition are distinct later observations; none reveals the Series A memo or Josh's cash return. SEC transaction announcement, Transaction Overview · Clover 8-K, Item 8.01 [contemporaneous record · strong]
• Evidence — Decision Education supplies the full process description; 20VC supplies the partnership threshold and truth-seeking norm; Vintage supplies an observed error and stated rubric revision; 10 Years shows an earlier attempt to inspect portfolio data while removing a dominant outlier and disclosing limits. Decision Education · 20VC, 12:43–13:51 · Vintage, 04:27–08:17 · 10 Years, methodology and Findings [investor-stated · strong]
• Limits, failure modes, and exception ledger — Rubrics can formalize shared bias, moderators can suppress necessary advocacy, PMF labels can create false precision, and later reviewers can score the old record with hindsight. The record exposes no rubric versions, voter-level forecasts, pass outcomes, calibration curve, or net-return comparison. A fast clock is especially dangerous in clinical, hardware, or regulated products. [researcher inference · strong]
• Evolution and contradictions — The record evolves from 2015 fund arithmetic and observational portfolio analysis to 2024 decision artifacts and 2026 rubric revision. “Purely on the pick” conflicts with the three-clock view unless picking is understood to include price, ownership, reserves, and later process. Upfront Summit, 02:03–03:40 · Uncapped, 50:14–52:21 [researcher inference · moderate]
• Modern VC translation (as of 2026-08-24) — Store immutable pre-discussion forecasts, dissent, source channel, ownership, and falsifiers; define stage-specific evidence clocks; and maintain a cash-based outcome layer distinct from private marks. Use AI to retrieve comparable past decisions, not to generate retrospective certainty. Audit calibration by partner, rubric version, channel, sector, and vintage, including passes and write-offs. [researcher inference · strong]

Model 4 — Do the fund math before falling in love with the company

• Rule — Before company enthusiasm becomes commitment, calculate whether check size, entry ownership, dilution, reserves, time, and plausible cash exit can move the specific fund. A great company at the wrong price or allocation can be a poor fund investment; a famous logo is not a return. [investor-stated · strong]
• Evidence status — Kopelman directly connects price, ownership, and fund construction and publicly names price-based misses. Public data do not expose full First Round cash flows, so the arithmetic rule is strong while any claim of superior implementation is unproved. [investor-stated · strong attribution / insufficient performance evidence]
• Provenance and origin — Crystal Balls vs. Calculators reduces the first pass to entry and exit arithmetic, reports a pace slowdown, and warns that marks are unrealized; 20VC says ownership is largely set in the first check; Upfront and Uncapped attack logo-based return narratives. Crystal Balls, lines 31–77 · 20VC, 07:48–10:36 and ownership show-note question · Uncapped, 43:23–46:20 [investor-stated · strong]
• Mechanism and reasoning — Venture outcomes are power-law distributed, but fund contribution still equals cash returned relative to fund size after ownership, dilution, reserves, fees, time, and loss rate. Performing the sensitivity analysis before advocacy exposes the exit value and allocation required for the thesis to matter. [investor-stated · strong]
• Historical and vehicle context — First Round's fund sizes, initial-check norms, minority positions, and follow-on rights changed across vintages. The 2011 letter reported a roughly $135 million current fund and sub-20% typical initial ownership; a 2026 Form D reports $325 million offered for X-F. These are different vehicles and cannot share one ownership assumption. SEC comment letter, pp. 1–2 · Fund X-F Form D, Offering and Related Persons [contemporaneous record · strong]
• When to use — Use before term-sheet authorization, allocation changes, reserve checks, opportunity vehicles, secondaries, and holding/sale decisions. Do not let the model substitute for company diligence or use a single exit value without loss, dilution, timing, and probability scenarios. [researcher inference · strong]
• How to apply it — Build base/downside/upside cases for check, post-money ownership, option-pool expansion, pro rata, dilution, reserve, exit timing, gross proceeds, and fund contribution. Calculate required exit for 0.5x, 1x, and target fund impact; show concentration and opportunity cost. IC questions: “What allocation makes this material?” “What must be true after two dilutive rounds?” “What cash—not mark—returns to which vehicle?” [researcher inference · strong]
• Examples — Four distinct cases expose the arithmetic boundary:
  • Documented application — Twitter pass. Kopelman says First Round declined a later Twitter seed opportunity on price; the later outcome makes it a visible miss but does not supply the offered allocation, fund model, or counterfactual proceeds. Yale lecture, 14:20–15:29 · 20VC, 07:48–10:36 [retrospective witness · moderate]
  • Documented application — Dropbox pass. 20VC's show notes and audio attribute another price miss to First Round; no term sheet, allocation, vote, or decision memo survives, so “lost return” is not computable. 20VC, 07:48–10:36 and show notes [retrospective witness · provisional]
  • Researcher-applied illustration — Upstart ownership snapshot. The S-1 shows FRC III entities at 3,220,682 shares and 5.2% pre-offering under shared GP control. It is a useful ownership checkpoint, not initial cost, realized sale, or Josh economics. Upstart S-1, Principal Stockholders and footnote 14 [contemporaneous record · strong]
  • Researcher-applied illustration — Flatiron acquisition. Roche's $1.9 billion fully diluted transaction value is observable, but First Round's check, ownership at sale, dilution, escrow, and distributions are absent; transaction value cannot become an investor multiple. Roche, 2018-04-06 closing release · Flatiron lab [contemporaneous record · strong]
• Evidence — The framework rests on the 2015 LP communication, 2019 price/ownership interview, 2020 fund-math discussion, 2025 complete-path critique, and primary ownership/vehicle filings. Crystal Balls, lines 31–77 · 20VC · Upfront, 02:03–03:40 and 11:12–12:00 · Uncapped, 43:23–46:20 [investor-stated · strong]
• Limits, failure modes, and exception ledger — The public framework can overfocus on entry and exit, while actual return needs ownership, dilution, follow-ons, fees, carry, time, losses, secondaries, and liquidity. Price discipline can become a socially acceptable story for access or conviction failures. No complete rejected set, fund cash-flow series, valuation policy, or partner allocation is public. [researcher inference · strong]
• Evolution and contradictions — First Round moved from a 2015 claim of stable pace/fund size to a 2025 admission that reserves grew from a minority to a majority of capital. The first-check ownership rule survives, but larger reserves increase the importance of follow-on skill that Kopelman says he is unsure the firm has. Uncapped, 44:26–45:06 [investor-stated · strong]
• Modern VC translation (as of 2026-08-24) — Require an auditable fund-return bridge in every memo, including SAFEs, option pools, pro rata, structured rounds, secondary liquidity, recycling, and reserve opportunity cost. For high-priced AI seeds, stress revenue-quality, compute cost, two extra rounds, and a 50% exit-multiple contraction. Prohibit valuation-to-return shortcuts in portfolio reporting. [researcher inference · strong]

Model 5 — Make financing dependency and reserve gates explicit

• Rule — Every round must buy a named fact that can reduce risk before the next financing; underwrite a no-funding case, reserve only against new evidence, and stop when the next dollar merely postpones the same uncertainty. [investor-stated · strong]
• Evidence status — The Watney partner letter directly states the financing-dependency problem; the SEC letter documents follow-ons, and Kopelman later acknowledges uncertainty about reserve skill. Company-level reserve decisions are absent, so the risk rule is strong but implementation support is moderate. [investor-stated · strong attribution / moderate application support]
• Provenance and origin — The 2011 SEC letter documents initial checks, follow-ons, ownership maintenance, and later-stage debt/equity rounds. The 2016 Watney letter tells companies to model runway, milestones, unit economics, syndicate capacity, and self-sufficiency when future capital is not assured. SEC comment letter, pp. 1–3 · Watney Rule, embedded partner letter [contemporaneous record · strong]
• Mechanism and reasoning — Financing markets are an exogenous dependency. A milestone-gated plan converts runway into information and preserves choice; automatic pro rata or rescue checks can compound sunk-cost bias, while premature refusal can abandon a positive update. The reserve decision must therefore compare evidence gained per dollar with new-company opportunity cost. [researcher inference · strong]
• Historical and vehicle context — The rule emerged after a long growth-over-burn market began repricing. First Round's pooled funds face finite reserves, ownership goals, fund lives, and cross-company competition; unlike an evergreen balance sheet, each rescue consumes an option elsewhere. Specific reserve ratios, LP terms, and fund extensions remain private. [researcher inference · strong]
• When to use — Use in initial runway planning, every bridge or pro rata, down rounds, regulatory delays, pivots, and post-failure asset decisions. Do not apply a software-speed milestone to clinical or hardware evidence; model the correct latency and irreversibility. [researcher inference · strong]
• How to apply it — Model no financing, delayed financing, and flat/down financing. For each tranche define cash date, reversible costs, evidence milestone, decision owner, follow-on amount, ownership effect, syndicate capacity, and stop/sale path. Ask: “What uncertainty disappears?” “Can insiders fund the full path?” “What evidence failed to arrive?” “Would we make this check as a new investment today?” [researcher inference · strong]
• Examples — Four distinct cases show why the gate matters:
  • Documented application — Upstart. Girouard recalls weak pull, roughly six months of cash, and a rapid pivot when the loan product drew immediate demand. The evidence supports a time-bounded product decision, although First Round's reserve action is unknown. First Round Review, founder board/pivot account [retrospective witness · strong]
  • Researcher-applied illustration — Notion. Four rebuilds, layoffs, a move to Japan, and a reported $150,000 family loan show a financing boundary around product learning; the public record does not reveal First Round reserve refusals or gates. First Round manifesto, Notion passage · Forbes Midas Touch, funding-history paragraphs [researcher inference · moderate]
  • Researcher-applied illustration — Knewton. Reporting describes more than $180 million raised, debts, funding drying up, a pivot, and an undisclosed Wiley acquisition. Without reserve, board, or waterfall records, the lesson is to require adoption and renewal evidence—not to declare a total loss. Inside Higher Ed, article body [researcher inference · moderate]
  • Researcher-applied illustration — AltSchool. More than $174 million of reported capital, roughly $30 million annual burn, school closures, and a software shift make operated-school and software milestones separable; FRC's capital response and governance are unknown. Forbes, capital/burn/closure account · EdSurge, rename/leadership account [researcher inference · moderate]
• Evidence — Watney provides the operating rule; the 2011 letter supplies fund-level mechanics; Uncapped records that reserves grew from minority to majority and Kopelman is unsure of follow-on skill; the failures ledger supplies bounded company tests. Watney, embedded letter · SEC letter, pp. 1–3 · Uncapped, 44:26–45:06 · Failure postmortems [investor-stated · strong]
• Limits, failure modes, and exception ledger — Milestone finance can force short-term optimization, penalize long evidence clocks, and create signaling spirals. Defensive reserves can preserve value through a cycle or become escalation after failure. The anonymous Watney case has no outcome; company reserve schedules, refusals, write-downs, opportunity cost, and partner votes are missing. TechForward is an exception: capital reportedly funded enforcement after operations ceased, so a residual legal asset can justify a different gate. TechForward Fortune account [researcher inference · strong]
• Evolution and contradictions — First Round's early model emphasized many small first checks; Watney added defensive financing discipline; the current model accepts much larger reserves while questioning follow-on skill. Seed specialization and majority reserve allocation coexist only if reserves defend seed-originated advantages rather than turning the vehicle into an unacknowledged multi-stage fund. Uncapped, 43:23–46:20 [investor-stated · strong]
• Modern VC translation (as of 2026-08-24) — Add a financing-dependency register to every AI, climate, hardware, and regulated deal: next fact, full cost, latency, insider capacity, financing alternatives, covenant/regulatory dependencies, dilution, and stop owner. Separate compute or customer-acquisition spending that buys a repeatable signal from spend that buys temporary volume. Re-underwrite reserves against today's opportunity set, not original cost. [researcher inference · strong]

Model 6 — Match stage specialization to vehicle fit

• Rule — Claim a stage edge only when fund size, initial check, ownership target, reserve policy, partner time, governance, service model, liquidity horizon, and LP expectations all fit that stage; reject attractive adjacent deals that require a different operating system. [investor-stated · strong]
• Evidence status — Kopelman consistently states a first-24-month specialization and explains the original financing gap; filings document vehicle constraints. The design is well supported, but its causal contribution to net returns is not public. [researcher inference · strong design / insufficient performance support]
• Provenance and origin — Wharton describes falling startup capital needs colliding with larger venture checks; FundersClub records a deliberate decision to remain seed-focused; Origins calls the earliest 18–24 months the “forever job.” Wharton, capital-gap and stage answers · FundersClub fireside, 07:38–07:42 · Origins, third-party chapter around 16:14–22:13 [retrospective witness · strong]
• Mechanism and reasoning — A focused vehicle sees more comparable decisions, can price service and partner time to the company's maturity, and avoids ownership conflicts with later vehicles. Misfit creates check-size pressure, signaling, disclosure conflicts, service dilution, or a portfolio too small/large for the stated edge. [researcher inference · strong]
• Historical and vehicle context — The original advantage arose when web company formation costs fell before traditional fund economics adjusted. By 2020 seed capital was abundant, shifting the claimed edge from capital-gap arbitrage toward access, picking, and service; by 2025 reserves had expanded. Upfront, 02:03–05:38 · Uncapped, 43:23–46:20 [investor-stated · strong]
• When to use — Use when designing a new fund, changing fund size, adding opportunity/growth vehicles, revising reserves, entering a geography, or deciding whether a deal's capital and governance needs exceed the team's specialty. Do not equate a “seed” label with fit; inspect actual ownership, duration, support load, and follow-on needs. [researcher inference · strong]
• How to apply it — Map capital source, fund size/life, check range, ownership target, concentration, reserve ratio, follow-on authority, board capacity, service cost, geographic reach, liquidity, key-person/succession, and LP constraints. Stress a representative company across three rounds. Ask: “Which capability is truly advantaged?” “What breaks if check or reserve doubles?” “Will later vehicles compete with founders or the seed fund?” [researcher inference · strong]
• Examples — Four distinct vehicle cases ground the model:
  • Documented application — First Round's 2011 fund model. The SEC letter reports pre-revenue early-stage equity, $250,000–$750,000 typical initial checks, under-20% typical initial stakes, a roughly $135 million current fund, follow-ons, LP limits, and minority control. SEC comment letter, pp. 1–4 [contemporaneous record · strong]
  • Documented application — Dorm Room Fund. Kopelman describes approximately $20,000 student-selected checks, local student committees, and a retained First Round veto—an ecosystem vehicle placing decisions near campus information while bounding governance. Return and veto data are missing. FundersClub fireside, 08:08–12:30 [retrospective witness · moderate]
  • Researcher-applied illustration — Upstart/FRC III. The S-1 shows a pooled FRC III position and three-member GP committee, demonstrating that seed sponsorship, legal ownership, and control sit at different layers. Initial check, fund contribution, and Josh economics remain unknown. Upstart S-1, Principal Stockholders footnote 14 [contemporaneous record · strong]
  • Researcher-applied illustration — First Round Capital X-F. The 2026 Form D reports $325 million offered and names Josh as a related executive person. It documents current institutional scale, not old-fund terms, stage performance, carry, or a company investment. Fund X-F Form D, Offering and Related Persons [contemporaneous record · strong]
• Evidence — The Wharton and seed-boom accounts state the market gap; FundersClub and Origins state durable stage intent; the SEC letter and Forms D expose supported vehicle dimensions; Uncapped states current reserve tension. Wharton · Seed funding boom, surviving page sections · SEC letter · Uncapped, 43:23–46:20 [contemporaneous record · strong]
• Limits, failure modes, and exception ledger — Specialization can become branding, reject attractive adjacencies, conceal implicit multi-stage exposure, or leave founders without later support. Lower formation costs attracted seed competition and weakened the original capital-gap edge. Public evidence omits LPAs, fees/carry, exact fund life, reserve ratios, realized fund returns, partner capacity, key-person clauses, and successor authority. [researcher inference · strong]
• Evolution and contradictions — “Stay seed-focused” in 2015 coexists with a current majority reserve allocation. The defensible reconciliation is stage-of-origin specialization with selective ownership defense; it fails if later checks consume the marginal advantage or create multi-stage conflicts. Kopelman's stated uncertainty about follow-ons keeps the contradiction open. FundersClub, 07:38–07:42 · Uncapped, 44:26–45:06 [investor-stated · strong]
• Modern VC translation (as of 2026-08-24) — For any new seed, AI, geography, or opportunity vehicle, publish an internal fit memo: customer stage, check/ownership, evidence latency, service promise, reserves, cross-vehicle allocation, governance, liquidity, and LP/key-person constraints. Re-test the claimed edge after fund-size changes; “we have more capital” is not a new capability. [researcher inference · strong]

Model 7 — Design access, then measure the channel

• Rule — Treat access as a product and a measurable funnel: deliberately create non-referral routes, preserve source-channel metadata through decision and ownership, and judge a channel on unique evidence and fund outcomes after cost—not on inbound volume or famous exceptions. [investor-stated · strong]
• Evidence status — First Round's portfolio study links non-referral observations to a stated Open App change; later sources document active sourcing roles, distributed campus access, and process-speed measurement. Selection-adjusted channel performance remains unavailable. [documented behavior · moderate]
• Provenance and origin — 10 Years analyzed roughly 300 companies, removed Uber, disclosed representativeness limits, and says the non-referral finding motivated an Open App. The current investor-role post reframes access as sustained outreach, events, follow-up, coffees, and relationship maintenance. 10 Years, Finding 9 and methodology · Investor role post, ¶1–9 [investor-stated · moderate]
• Mechanism and reasoning — Referral-only funnels reproduce incumbents' networks and social proof. Designed routes can surface orthogonal founders and evidence, but more candidates add false positives and labor; channel instrumentation is needed to distinguish reach from conversion, ownership, and outcomes. [researcher inference · strong]
• Historical and vehicle context — Seed capital commoditized while founder access and service became differentiators. A platform fund can finance sourcing labor and programs that an angel cannot, but that cost belongs in channel economics and may favor geographies near the firm's offices. Upfront, firm-as-company and process sections 05:09–07:30 [investor-stated · strong]
• When to use — Use when adding inbound applications, scouts, campus programs, outbound investors, conferences, geography expansion, founder referrals, or AI prospecting. Do not scale a channel until attribution, consent, screening cost, allocation, and false-negative/false-positive definitions exist. [researcher inference · strong]
• How to apply it — Capture first source, introducer, relationship age, prior affiliation, response time, meeting conversion, independent score, decision, price, ownership offered/obtained, partner time, founder experience, follow-on, and cash outcome. Compare matched channel cohorts by vintage and sector. Ask: “What unique evidence did this route create?” “Who never enters?” “Does conversion survive cost and selection controls?” [researcher inference · strong]
• Examples — Four distinct access designs ground the model:
  • Documented application — Open App. First Round says non-referral observations in its 10-year portfolio analysis led it to open an application route. The firm-selected sample, mixed marks, and missing raw data make the result hypothesis-generating rather than causal. 10 Years, Finding 9 and lines 81–96 [documented behavior · moderate]
  • Documented application — First Search. The 2017 project converted portfolio recruiting queries into deliberate retrieval and reusable access rather than relying only on partner contacts. Hiring and outcome effects were not published. First Search thread, full archived thread [investor-stated · strong]
  • Documented application — Dorm Room Fund. Student committees placed sourcing and selection near campus peer networks while First Round retained a veto. The appearance reports program scale but no matched access, diversity, or return cohort. FundersClub fireside, 08:08–12:30 [retrospective witness · moderate]
  • Documented application — Yale cold inbound. Kopelman says Nick Shalek's July 2007 cold email led Yale to anchor First Round's first institutional venture fund, illustrating that designed receptivity matters on the LP side too. Commitment terms and the precise “first” legal claim remain unavailable. Yale anchor post, ¶1–5 [retrospective witness · strong chronology / insufficient terms]
• Evidence — The 2015 observational essay supplies the behavior change; the investor-role post supplies the current active-sourcing rubric; Upfront records decision-time measurement; FundersClub and Yale supply distributed/cold-access cases. 10 Years · Investor role post · Upfront, 06:40–07:30 · Yale anchor post [investor-stated · moderate]
• Limits, failure modes, and exception ledger — Open routes can produce volume without quality; outbound can encode the researcher's bias; referrals can contain valuable trust; “builders enjoy them” can select social similarity. The portfolio study excludes Uber and recent companies, mixes marks with exits, and has no raw data. No channel-cost ledger, declined set, source persistence, allocation, demographic audit, or realized-return replication is public. [researcher inference · strong]
• Evolution and contradictions — The record moves from personal networks and cold LP inbound, to Open App and Dorm Room Fund, to First Search, platform/community, and a trained investor role. Founder referrals still create a Matthew-effect loop, so the institution is simultaneously broadening access and benefiting from accumulated prestige. Uncapped, 15:45–16:34 [investor-stated · moderate]
• Modern VC translation (as of 2026-08-24) — Instrument human and AI sourcing as auditable channels. Preserve consent, provenance, geographic/demographic reach, time cost, unique evidence, independent score, allocation, and eventual cash outcome; run holdout or matched-cohort reviews where feasible. Do not optimize meetings booked. Optimize access to otherwise unseen evidence without automating pedigree bias. [researcher inference · strong]

Model 8 — Turn the portfolio into a trusted network and retrieval system

• Rule — A portfolio becomes institutional leverage only when founders trust it enough to contribute and a retrieval system can deliver the right peer, candidate, precedent, or operating artifact at the decision moment; measure use and decision change, not logo count. [investor-stated · moderate]
• Evidence status — Kopelman directly describes community and platform as firm infrastructure, while First Search and repeated founder surveys are documented programs. Public evidence shows intent and activity, not causal improvement in hiring, PMF, survival, or return. [documented behavior · moderate]
• Provenance and origin — First Round's platform expansion included tools such as HackPR and later First Search; Colossus describes the founder community as designed institutional leverage. The 2015 and 2016 State of Startups surveys show repeated attempts to learn from founders inside and outside the portfolio. HackPR launch, 2012-10-09 · First Search thread · Colossus, 43:29–47:28 [investor-stated · moderate]
• Mechanism and reasoning — Dense, trusted peer exchange reduces search cost and transfers tacit pattern knowledge; structured retrieval can find relevant help faster than a partner's memory. Value depends on contribution quality, permission, contextual fit, and whether the retrieved item changes a decision rather than merely creating engagement. [researcher inference · strong]
• Historical and vehicle context — As seed checks commoditized, a pooled platform fund could finance recruiting, knowledge, software, and events across many companies. Scale creates retrieval density but also privacy, conflict, selection, and service-cost risks; small angels may offer intimacy without comparable searchable breadth. Upfront, firm-as-company section 05:09–05:38 [investor-stated · strong]
• When to use — Use for hiring, go-to-market, vendor/counsel choice, incident response, pricing, fundraising, board preparation, and precedent retrieval where portfolio members consent. Do not use cross-company data where confidentiality, competition, health/financial data, privilege, or antitrust boundaries are unclear. [researcher inference · strong]
• How to apply it — Tag each request by company stage, function, urgency, sensitivity, and decision; retrieve permissioned peers/artifacts; record response time, match quality, founder cost, action taken, and result. Audit participation concentration and conflicts. Ask: “Who benefits and who contributes?” “Would the founder share this outside the room?” “Did retrieval change a decision or only save search time?” [researcher inference · strong]
• Examples — Four distinct program cases ground the model:
  • Documented application — First Search. Kopelman's 2017 thread describes turning portfolio hiring needs and candidate relationships into a reusable search capability. The public thread supplies design intent, not hire quality, retention, or causal company outcomes. First Search thread, full archived thread [documented behavior · strong intent / insufficient outcome]
  • Documented application — founder community. Kopelman describes First Round's community and platform as a deliberate support system during the earliest company-building period. No founder-level usage or matched outcome comparison is disclosed. Colossus, 43:29–47:28 [investor-stated · moderate]
  • Documented application — 2015 State of Startups. The survey reached more than 500 venture-backed founders, with fewer than one quarter from First Round's own community, showing an attempt to widen feedback beyond portfolio members. The wrapper does not expose the full chart packet or causal effects. 2015 article, body [investor-stated · strong]
  • Documented application — 2016 State of Startups. The follow-up reached more than 700 founders inside and outside First Round and added lead-investor and diversity questions. Voluntary response and missing report internals limit inference. 2016 article, opening paragraphs [investor-stated · strong]
• Evidence — First Search and HackPR document service experiments; Colossus and Upfront state the platform/community theory; the two survey gateways document founder-feedback loops. First Search · HackPR · Colossus, 43:29–47:28 · 2016 survey gateway [investor-stated · moderate]
• Limits, failure modes, and exception ledger — Community can become marketing, consensus, adverse-selection camouflage, or compulsory sharing; retrieval can surface stale, non-comparable, or conflicted advice. Surveys omit unfunded founders and suffer nonresponse. The record lacks service adoption, contribution distribution, founder burden, privacy controls, hire retention, decisions changed, support cost, and outcome controls. [researcher inference · strong]
• Evolution and contradictions — The firm evolves from point tools and surveys to a platform/community narrative and current decision-memory system. That broadens “portfolio value” from services to knowledge, but it also raises the cost and complexity that a stage-specialist vehicle must justify. Uncapped, 50:14–52:21 [investor-stated · moderate]
• Modern VC translation (as of 2026-08-24) — Build permissioned, provenance-aware retrieval over portfolio requests, expert responses, operating artifacts, and decision outcomes. Use least-privilege access, expiration, conflict checks, and human context review; never train across founder-confidential data by default. Report time-to-answer, action rate, outcome proxy, contributor concentration, and founder burden—not community membership. [researcher inference · strong]

Model 9 — Treat legal, regulatory, data, and enforcement rights as product risk

• Rule — If a product depends on legal classification, license, data right, bank/clinical partner, disclosure, or enforceable contract, underwrite that dependency as part of product-market fit: obtain independent analysis, define operating controls and board escalation, and fund only through jurisdiction/data/compliance milestones. [researcher inference · strong]
• Evidence status — This is not preserved as a single Kopelman maxim. It is a strongly supported researcher model drawn from multiple regulated/data deals, a litigation-support account, filings, pleadings, and procedural dispositions; Josh-specific diligence and board actions remain mostly unknown. [researcher inference · strong case support / insufficient individual attribution]
• Provenance and origin — TechForward is the earliest processed public case in which Kopelman describes funding enforcement of residual contractual/IP rights after operations ceased. The portfolio later spans clinical data, consumer credit, Medicare insurance, ad privacy, and auction/gambling classification. Fortune TechForward account · Best Buy 10-K, Item 3/Legal Proceedings [researcher inference · strong event / bounded motive]
• Mechanism and reasoning — A legal or data dependency can kill distribution, invalidate unit economics, block evidence access, create liability, or destroy exitability even when users like the product. Early mapping makes the dependency falsifiable; governance and stop rules prevent “counsel said okay” from becoming a blanket growth license. [researcher inference · strong]
• Historical and vehicle context — First Round's strategy expanded from web software into credit, insurance, clinical-data, advertising, and consumer mechanics while regulation and platform/data rules matured. Minority board and information rights may detect but not control every risk; finite reserves create pressure to distinguish remediable compliance work from a broken product premise. [researcher inference · strong]
• When to use — Use before investment and every launch/market expansion in fintech, health, insurance, AI/data, marketplaces, gaming, defense, employment, privacy, and regulated infrastructure. Also use after operating failure when a contract or IP claim is a residual asset. Do not confuse legal feasibility with customer demand or merits probability with collectible cash. [researcher inference · strong]
• How to apply it — Map jurisdictions, regulator, license/classification, data provenance/consent, counterparties, related parties, model/fairness, required disclosures, complaints, audit owner, board escalation, insurance, indemnity, enforcement venue, and stop event. Commission independent specialist review; attach dates and assumptions. Ask: “What right makes the product possible?” “Who can revoke it?” “Which metric reveals harm?” “Can we geo-fence or unwind?” [researcher inference · strong]
• Examples — Four distinct cases ground the model:
  • Researcher-applied illustration — Project Fair Bid / BigDeal. The $4.5 million financing and Kopelman board association are reported contemporaneously; a later complaint alleged RICO and illegal-gambling theories. Those are allegations only. Investor defendants were voluntarily dismissed without prejudice in Washington amid jurisdiction issues, and the refiled California case was dismissed without prejudice after settlement notice; neither disposition was a merits finding, and public settlement terms are absent. TechCrunch, financing and board report · Complaint exhibit, allegations · W.D. Wash. order, 2011-08-11 · N.D. Cal. order, 2013-12-23 [contemporaneous record · strong fence / insufficient merits and investor-action evidence]
  • Researcher-applied illustration — Clover Health. The case makes actuarial, related-party, disclosure, coding, and regulatory controls part of product diligence. Hindenburg's assertions remain allegations; Clover later reported that SEC staff did not intend to recommend enforcement, which is not blanket exoneration or a disposition of every matter. Josh-specific compliance votes or interventions are not public. Hindenburg, allegations · Clover 8-K, Item 8.01 [contemporaneous record · strong fence / insufficient partner attribution]
  • Researcher-applied illustration — Upstart. The S-1 identifies bank-partner concentration, fair-lending/model, usury, and “true lender” risks that could impair distribution or economics despite demonstrated borrower demand. It does not reveal what Kopelman knew at seed, counsel's advice, or later board votes. Upstart S-1, Risk Factors [contemporaneous record · strong later-risk evidence / insufficient seed-state attribution]
  • Researcher-applied illustration — Flatiron Health. The formation record supports a fragmented oncology-data problem and Roche's acquisition confirms a strategic endpoint; it does not establish the seed-era legality, provenance, completeness, interoperability, or clinical utility of the underlying data. Those become diligence gates, not facts inferred from the outcome. Forbes, founder/data-gap account · Roche, 2018-04-06 closing release [researcher inference · moderate / insufficient seed-rights evidence]
• Evidence — Upstart's S-1, Clover's 8-K, Project Fair Bid's complaint and two procedural orders, and Best Buy's disclosed TechForward litigation provide company-specific legal records; founder/firm/reporting routes supply product and role context. They support the risk categories and dispositions, not a Josh-authored framework or undisclosed diligence. Upstart S-1 · Clover 8-K · Project Fair Bid W.D. Wash. order · Best Buy 10-K, Item 3 [researcher inference · strong cases / insufficient doctrine attribution]
• Limits, failure modes, and exception ledger — Specialist advice can create false certainty; regulators, facts, and product behavior change. A complaint is not a finding, dismissal without prejudice is not merits exoneration, and a no-enforcement recommendation is not universal clearance. Missing evidence includes legal opinions, diligence memos, board minutes, regulator correspondence, complaints, loss reserves, control testing, remediation cost, and Josh-specific votes. [researcher inference · strong]
• Evolution and contradictions — TechForward shows enforcement as a residual asset after operating failure, whereas the later regulated cases support moving legal/data gates before scale. Question-led founder autonomy yields to explicit escalation when law, fiduciary duty, or user harm is implicated; no public record proves Kopelman made that evolution himself. Fortune TechForward account [researcher inference · moderate]
• Modern VC translation (as of 2026-08-24) — Add a rights-and-enforcement graph to product diligence: data/model provenance, consent, licenses, regulator, accountable partner, audit signal, disclosure owner, jurisdiction, insurance, remedy, and kill switch. For AI, test training and output rights plus model bias and auditability. Maintain an allegation/disposition table so adverse claims, findings, settlements, and open matters never collapse into one narrative. [researcher inference · strong]

Model 10 — Map roles before assigning credit

• Rule — Before attributing a deal or outcome, map who sourced, sponsored, diligenced, approved, negotiated, held the board role, recruited, authorized follow-ons, influenced exit, and later narrated the case; map the vehicle and founders separately. Treat every unmapped role as unknown. [researcher inference · strong]
• Evidence status — Kopelman directly criticizes venture attribution culture and supplies several role-specific corrections; firm cards and filings identify some sponsors, vehicles, and legal controllers. The full partner-labor record, dissent, carry allocation, and company-level causal contribution remain unavailable. [researcher inference · strong boundaries]
• Provenance and origin — 20VC documents two-thirds partnership approval, no individual veto, and a Blue Apron board handoff; Upfront explicitly credits Uber to Rob Hayes. First Round's company filters assign named partners, while Upstart's S-1 assigns the economic position to FRC III entities controlled by a three-person GP committee. 20VC, 12:43–15:25 and 30:20–31:20 · Upfront, 02:03–03:40 · Upstart S-1, Principal Stockholders footnote [contemporaneous record · strong]
• Mechanism and reasoning — Role mapping blocks the lone-hero fallacy, separates legal ownership from labor and narration, and makes disagreement, succession, and incentives auditable. It also protects founder/operator agency: access, advice, capital, and a board question can matter without creating the product, pivot, or outcome. [researcher inference · strong]
• Historical and vehicle context — First Round evolved from a founder-led seed franchise into a multi-partner manager of pooled numbered funds. A partner may sponsor a company while LP capital bears the investment, a GP committee controls securities, another partner assumes the board role, and founders retain operational control. SEC 2011 comment letter, pp. 1–4 [contemporaneous record · strong]
• When to use — Use in every IC memo, reserve decision, board review, exit narrative, track-record claim, LP attribution, ranking, case study, and succession decision. Re-open the map when board seats, partner ownership, vehicle rights, or company leadership change. [researcher inference · strong]
• How to apply it — Build a dated RACI-style table for source, sponsor, diligence, vote, dissent, terms, vehicle, board/observer seat, recruiting, follow-on, exit, and narrator. Attach the primary artifact for each cell; distinguish firm, vehicle, partner, founder, co-investor, and company. Never infer work from a signature, card, board title, or pooled stake alone. [researcher inference · strong]
• Examples — Four distinct attribution cases ground the model:
  • Documented application — Uber. Kopelman discusses Uber as a company-market-fit example, but Upfront explicitly credits First Round partner Rob Hayes. It is valid firm history and invalid Josh deal credit absent a separate record. Upfront, 02:03–03:40 · First Round, partner/company filters [contemporaneous record · strong]
  • Documented application — Roblox. Kopelman uses Roblox to illustrate seed uncertainty and adaptation, while First Round's portfolio record assigns the company to Chris Fralic. Josh is a later narrator of a firm case, not thereby its sponsor. Colossus, 16:27–17:01 · First Round, partner/company filters [investor-stated · strong]
  • Documented application — Blue Apron. Kopelman says he handed the board role to Phin Barnes, directly separating initial association from later governance labor. The public episode does not allocate sourcing, vote, follow-on, exit work, or economics. 20VC, 30:20–31:20 [investor-stated · strong / incomplete labor map]
  • Researcher-applied illustration — Upstart. First Round identifies Kopelman as the Seed partner and the founder attributes a pivotal demand question to him, but the S-1 places 3,220,682 shares and 5.2% pre-offering ownership in FRC III entities under shared three-person GP control. The team executed the pivot; neither pooled economics nor the IPO belongs to Josh alone. First Round Review, board/pivot account · Upstart S-1, Principal Stockholders [researcher inference · strong role fence]
• Evidence — 20VC and Upfront supply direct role corrections; First Round cards supply current sponsor labels; the SEC letter and Upstart S-1 expose pooled-vehicle and governance boundaries. Together they prove that roles can diverge, not a complete career contribution ledger. 20VC · First Round companies · Upstart S-1 [researcher inference · strong principle / incomplete denominator]
• Limits, failure modes, and exception ledger — Official cards can name a sponsor without exposing labor; board seats do not prove attendance or influence; filings show legal control, not internal work; narrators have self- and survivorship bias. Missing are IC memos, CRM/source logs, diligence assignments, votes/dissent, term negotiation, board minutes, recruiting records, reserve approvals, exit work, carry allocation, and founder/co-investor counteraccounts. [researcher inference · strong]
• Evolution and contradictions — Public history moves from a founder-of-the-firm narrative toward explicit distributed partnership, voting, handoffs, and the current search for a younger investment-team “apprentice.” This supports institutionalization but does not reveal the complete labor ledger or eliminate key-person influence. First Round, current Josh profile [contemporaneous record · moderate]
• Modern VC translation (as of 2026-08-24) — Require a living contribution ledger with source, sponsor, voter, dissenter, diligence owner, negotiator, board steward, follow-on owner, exit contributor, founder/team credit, vehicle, and artifact. Publish partner track records only from that ledger; show firm and vehicle results separately, preserve handoffs, and label any missing cell unknown. [researcher inference · strong]

Model 11 — Use questions to improve board decisions, then escalate when duty requires

• Rule — In early board work, context-load and ask decision-changing questions before importing answers: expose the founder's prediction, alternatives, falsifier, owner, and date. Question-led governance preserves operator agency; it does not excuse passivity when legality, fiduciary duty, solvency, or user harm requires action. [researcher inference · strong rule / provisional application record]
• Evidence status — Provisional model. Kopelman repeats the listen-and-question doctrine across three appearances, but the processed record supports only one company-specific question and one adverse-case illustration. There is no board-minute or comparable outcome cohort. [investor-stated · strong attribution / provisional behavioral support]
• Provenance and origin — Colossus says the best board members ask questions that help founders reach decisions; 20VC says new board members should listen through the first meetings; LDV contrasts founder vision with investors importing operating “recipes.” Colossus, 30:28–32:18 · 20VC, 14:12–15:25 · LDV, 09:48–13:56 [investor-stated · strong]
• Mechanism and reasoning — Questions make assumptions observable, reduce advice transfer from a non-comparable operating context, and let founders own the update. A dated prediction creates evidence for the next decision. Escalation rules prevent “founder-friendly” listening from becoming negligence or consensus avoidance. [researcher inference · strong]
• Historical and vehicle context — First Round's seed strategy emphasizes minority positions during the first 24 months of company building; the SEC letter describes management assistance and board-seat boundaries without proving control. In that setting, inquiry can be high leverage, but pooled capital, follow-ons, board duties, and regulated risk still require explicit decisions. SEC 2011 comment letter, pp. 2–4 [contemporaneous record · strong context]
• When to use — Use during board onboarding, product/model pivots, hiring, runway and reserves, financing, compliance, market entry, and postmortems—especially when the investor's experience is adjacent rather than identical. Switch from inquiry to explicit recommendation, vote, independent review, or stop when law, duty, solvency, safety, or irreversible exposure crosses a pre-agreed threshold. [researcher inference · strong]
• How to apply it — Spend the first meetings building context; then record: “What must be true?” “What would change your mind?” “What alternatives were tested?” “Who owns the next test?” “By what date?” “What is the stop/escalation event?” At the next board, compare prediction with evidence and document founder response, dissent, action, and unresolved risk. [researcher inference · strong]
• Examples — The exhausted record supports only two cases:
  • Documented application — Upstart. Founder Dave Girouard recalls Kopelman asking whether customers were “at the gate” for the original income-share product. A loan test produced immediate pull, and the team killed the original model within days. This supports one diagnostic question and founder execution; formal vote, alternatives, dissent, and counterfactual outcome are unknown. First Round Review, board-question and pivot sections [retrospective witness · strong attribution / survivor-limited]
  • Researcher-applied illustration — Project Fair Bid / BigDeal. The case illustrates questions a board should ask about consumer loss, disclosure, payments, jurisdiction, and auction/gambling classification before launch. No public minutes show that Kopelman asked or failed to ask them. Later allegations were never adjudicated on the merits; the Washington investor dismissals and California settlement dismissal were without prejudice. TechCrunch, financing and board report · W.D. Wash. order, 2011-08-11 · N.D. Cal. order, 2013-12-23 [researcher inference · strong fence / insufficient board-behavior evidence]
• Evidence — Three direct-voice appearances establish the doctrine; Upstart supplies the sole preserved company-specific board question; Project Fair Bid supplies only a fenced governance stress test. That is sufficient for a provisional rule and insufficient for a general causal claim. 20VC · Colossus · Upstart founder account [researcher inference · moderate application evidence]
• Limits, failure modes, and exception ledger — Questions can delay action, hide investor conviction, create performative governance, or preserve power asymmetry while appearing founder-friendly. Exact exhausted-route ledger: 136 deduplicated source rows were processed; board-relevant retrieval covered Colossus's official chapter page (full transcript gated; third-party ASR used only to check locators), 20VC's official audio/show notes (no official full transcript), LDV video/automatic captions, the Upstart founder Review, current and archived First Round company/profile cards, SEC filings and the 2011 comment letter, Project Fair Bid's complaint and two orders, and targeted company/board-role routes. No board minutes, attendance logs, question logs, contemporaneous founder response, dissent, comparable board outcomes, or Josh-specific vote was located. Seven priority Redeye bodies also remained unavailable after canonical URL, FeedBurner, title-variant, citation-trail, and parser/archive-index attempts. Sources coverage assessment · Writings access ledger [researcher inference · strong]
• Evolution and contradictions — The 2019–2020 listen/question language becomes more operational when paired with later decision rubrics, pre-discussion scoring, and game-tape review. Question-led governance is compatible with independent judgment only if the investor records a view and escalation threshold rather than outsourcing the decision to the room. Decision Education [investor-stated · moderate]
• Modern VC translation (as of 2026-08-24) — Use a board decision record with context, prediction, alternatives, falsifier, evidence deadline, named owner, founder response, dissent, and escalation trigger. AI can retrieve precedents and track commitments, but it should not generate authoritative advice from confidential cross-company data or replace counsel, fiduciary judgment, or accountable votes. [researcher inference · strong]

Evidence gaps and candidate models

• Inventory reconciliation. The retained inventory is exactly 11 models. Models 1 and 3–10 are full models with four distinct cited cases each; Models 2 and 11 are explicitly provisional with two cited cases and an exact exhausted-route ledger each: 40 examples total. This count is evidence architecture, not a claim that the models are independent or complete. [researcher inference · strong]
• Merged candidates. “Grit” and founder empathy are inputs to Model 1; price/ownership belongs in Model 4; product-market-fit states belong in Model 3; founder access and firm community separate into Models 7 and 8; regulation belongs in Model 9; and attribution belongs in Model 10. Merging prevents one anecdote from masquerading as a new model. [researcher inference · strong]
• Provisional or excluded candidates. “Present problem, not theme” and “question-led boards” remain provisional because each has only two cases. Litigation as a residual asset is a Model 9 edge case, not a standalone model; raw persistence lacks a failure-adjusted base rate; “speed” lacks a quality distribution; “what would David say?” lacks an LPA or repeated vehicle-decision record; and “relevance”/AI renewal lacks enough observed decisions to retain separately. [researcher inference · strong]
• Material missing evidence. The record lacks seven priority Redeye bodies, original investment and rejection memos, raw decision rubrics, a matched passed-versus-funded set, board minutes and votes, reserve schedules, partner labor records, LPAs and cash flows, complete dilution/distribution histories, a career deal denominator, founder-survey instruments, and comparable outcome cohorts. Sources coverage assessment [researcher inference · strong]
• Falsifiers and next retrieval. Downgrade or merge a model if decision-era artifacts contradict the public doctrine, if failure-adjusted matched cohorts show no discriminating value, or if role records move the behavior to another partner. Highest-value retrieval is permissioned IC/rejection memos, dated rubrics and subsequent scores, board decision logs, reserve models, founder/co-investor counteraccounts, Redeye exports, and fund cash-flow/ownership records. Until then, public outcomes remain illustrations—not returns or causal validation. [researcher inference · strong]

What the evidence supports

The strongest supportable conclusion is narrower than a “great picker” claim. Across 2004–2026, Kopelman and First Round publicly describe a coherent seed operating system: widen access, assess founder adaptability, preserve private judgments before group discussion, use a small first check to purchase decisive evidence, separate company quality from price and ownership, and build portfolio/community infrastructure. Decision Education, 32:58–46:19 · SEC letter, pp. 1–4 [researcher inference · strong]

The record also supports genuine institutional evolution. A 2011 filing establishes firm mechanics; the 2015 writings add arithmetic and portfolio analysis; later talks expose independent scoring, moderation, process “game tape,” and a 2026 rubric correction. Written chronology · Spoken chronology [researcher inference · strong] These sources establish stated and documented process more strongly than they establish realized results.

The deal package supports seven bounded reconstructions, six no-hindsight labs, a 36-row retained ledger, and several adverse outcomes. It does not support Josh-specific DPI, TVPI, IRR, MOIC, hit rate, or personal proceeds. Pareto boundary · Denominator [researcher inference · strong]

What could falsify the thesis

The favorable thesis predicts measurable evidence that is currently private: non-referral opportunities should outperform matched referrals; rubric scores, confidence, and dispersion should forecast operating states; moderation should reduce repeated error; first checks should outperform later reserves after risk and time adjustment; and results should persist across partners and vintages after removing the top outcome. Edge thesis [researcher inference · strong]

The thesis would be weakened or overturned if those channel and rubric effects disappear, if rules were written after outcomes, if one partner/vintage/outlier explains the record, if later reserves equal or beat initial checks, or if matched firms without the claimed system perform similarly. Access, brand, LP network, timing, partner labor, and founder execution are live alternatives rather than footnotes. Skeptical case [researcher inference · strong]

Attribution and denominator audit

The arithmetic reconciles: 62 searched company families = 36 retained + 26 excluded; 36 retained = 7 sufficiently reconstructed + 29 unresolved. First Round's “100+” claim is a lower bound, not an enumerated career universe, so coverage is not computable. Reconciliation [researcher inference · strong]

Attribution is role-specific. Official company cards identify Josh for selected deals, but pooled vehicles, co-investors, founders, and other First Round partners retain their own roles. Upstart's S-1 assigns the holding to FRC III entities under shared GP control; Uber, Square, and Roblox remain assigned to other named partners absent independent Josh evidence. Upstart S-1, Principal Stockholders · First Round company filters [contemporaneous record · strong]

Transaction consideration, enterprise value, financing valuation, and an IPO ownership snapshot are outcome markers, not proceeds. The dossier never derives return from entry and exit valuations because cost, ownership, dilution, sales, distributions, time, fees, carry, and taxes are unavailable. [researcher inference · strong]

Transfer limits for modern VCs

The transferable unit is a decision record, not the Kopelman legend: source channel, independent forecast, decisive premise, alternatives, legal/data gates, ownership and vehicle fit, milestone, reserve rule, governance owner, and dated postmortem. Modern translation [researcher inference · strong]

Transfer fails when a capital-intensive or regulated company cannot buy decisive proof with a small check; when AI demo velocity hides rights, reliability, security, or unit cost; when a larger fund requires outcomes incompatible with seed ownership; or when “question-led” governance becomes an excuse to avoid fiduciary action. The 11 models preserve these boundaries and keep two under-evidenced rules provisional. Mental-model field guide [researcher inference · strong]

Unresolved evidence

• Complete opportunity, pass, loss, ownership, dilution, reserve, sale, distribution, fee, carry, tax, IC-vote, board, partner-labor, and personal-economics records.
• Full bodies and exact dates for seven priority Redeye essays.
• Raw 2015 portfolio analysis, State of Startups instruments, response/nonresponse data, marks, write-offs, and later replications.
• Original memos and vote records for the six decision labs; complete deal responsibility and cash-flow chains.
• TechForward cost, vehicle, appeal/collection sequence, and net distribution.
• Matched evidence that First Search, community, channel design, moderation, or rubric revisions changed operating outcomes or net returns.
• Exact succession, key-person, allocation, reserve, compensation, LP, and current-fund terms.

These are bounded public-record limits, not permission to fill gaps with inference. They determine what future evidence could change the assessment. [researcher inference · strong]

Central high-risk claim audit

Reproducible ordinary-claim audit

Two ordinary claims were selected from each of the eight reader documents by taking an early and late substantive section after freezing the final section order.

Ordinary audit result: 16/16 PASS. [researcher inference · strong]

Independent biography cross-checks

Independent deal and team cross-checks

Mental-model inventory audit

All 40 example rows carry claim-local citations and canonical two-axis labels; the two below-default models are provisional and document attempted routes rather than padding. Evidence gaps [researcher inference · strong]

Twenty-point acceptance evidence

All twenty acceptance conditions pass in the inspected state. The conclusion remains asymmetric: T054 supports a coherent, unusually inspectable seed decision system more strongly than it supports repeatable Kopelman-specific economic alpha. [researcher inference · strong]

Biography and firm

• ★★★ | First Round profile — Josh Kopelman | https://www.firstround.com/team/investing/josh-kopelman | current; accessed 2026-08-24 | official biography | Role, history, selected investments, and direct-voice index.
• ★★★ | First Round investing team | https://www.firstround.com/team/ | current; accessed 2026-08-24 | official team page | Current roles and team-attribution boundary.
• ★★★ | First Round companies directory | https://www.firstround.com/companies/ | current; accessed 2026-08-24 | official portfolio index | Company and named-partner attribution checks.
• ★★★ | First Round manifesto | https://www.firstround.com/news/manifesto | current; accessed 2026-08-24 | firm manifesto | Firm-authored stage, support, and case narrative.
• ★★★ | Josh Kopelman LinkedIn profile | https://www.linkedin.com/in/jkopelman | current; accessed 2026-08-24 | first-person profile/index | Canonical eight-article and legacy-blog index.
• ★★ | Redeye VC canonical route | http://redeye.firstround.com | legacy; attempted 2026-08-24 | legacy first-person blog | Canonical archive boundary for 2006–2008 writing.
• ★★ | Redeye VC FeedBurner feed | http://feeds.feedburner.com/redeyevc | legacy; attempted 2026-08-24 | RSS route | Feed/title recovery attempt; bodies unavailable.
• ★★★ | Inquirer board-chair transition | https://www.inquirer.com/business/philadelphia-inquirer-board-lisa-kabnick-josh-kopelman-20240605.html | 2024-06-05 | reported institutional record | Specific 2016–2024 civic-governance route.
• ★★★ | Philadelphia Inquirer leadership | https://about.inquirer.com/leadership/ | current; accessed 2026-08-24 | official leadership page | Current roster and stale-claim check.
• ★★ | Philadelphia mayoral-race spending context | https://www.inquirer.com/politics/election/helen-gym-jeff-yass-philly-mayors-race-20230508.html | 2023-05-08 | reported political context | Local adverse context, not investment evidence.
• ★★ | Technology's next wave of disruption | https://www.phillymag.com/business/2017/08/05/josh-kopelman-first-round-capital/ | 2017-08-05 | profile/interview | Philadelphia geography and civic/career context.
• ★★★ | Infonautics S-4 | https://www.sec.gov/Archives/edgar/data/909494/000091205701513070/a2047106zs_4.pdf | 2001-04-20 | SEC filing PDF | Company history, roles, financing, and public outcome.
• ★★★ | eBay 2002 Form 10-K | https://www.sec.gov/Archives/edgar/data/1065088/000089161803001538/f88540e10vk.htm | 2003-03-27 | SEC filing | Issuer record for Half.com integration.
• ★★ | FTC early-termination notice 20003733 | https://www.ftc.gov/legal-library/browse/early-termination-notices/20003733 | 2000 | regulatory notice | Primary transaction chronology route.
• ★★ | eBay to close Half.com | https://www.latimes.com/archives/la-xpm-2003-mar-07-fi-ebay7-story.html | 2003-03-07 | reported outcome | Contemporaneous post-acquisition context.
• ★★ | Symantec buys TurnTide | https://www.theregister.com/2004/07/13/symantec_buys_anti_spam_firm_turntide/ | 2004-07-13 | reported transaction | Rapid acquisition record; not personal proceeds.
• ★★★ | First Round SEC comment letter | https://www.sec.gov/comments/s7-37-10/s73710-80.pdf | 2011-02-10 | primary letter PDF | Check ranges, follow-ons, conflicts, and team boundary.
• ★★★ | First Round Capital IX Form D | https://www.sec.gov/Archives/edgar/data/1926484/000192648422000001/xslFormDX08/primary_doc.xml | 2022 | SEC Form D | Vehicle and related-person evidence.
• ★★★ | First Round Capital X Form D | https://www.sec.gov/Archives/edgar/data/2086435/000208643525000001/xslFormDX08/primary_doc.xml | 2025 | SEC Form D | Current vehicle scale and role.
• ★★★ | First Round Capital X-F Form D | https://www.sec.gov/Archives/edgar/data/2147875/000214787526000001/xslFormDX01/primary_doc.xml | 2026 | SEC Form D | Current formal role; economics undisclosed.
• ★ | US Patent 8,600,826 | https://patents.google.com/patent/US8600826B2/en | 2013-12-03 | patent | Marketplace-invention context only.
• ★ | US Patent 7,813,967 | https://patents.google.com/patent/US7813967B2/en | 2010-10-12 | patent | Half.com-era co-inventor record.
• ★ | US Patent 7,983,953 | https://patents.google.com/patent/US7983953B2/en | 2011-07-19 | patent | Marketplace-system operating evidence.
• ★ | US Patent 7,953,641 | https://patents.google.com/patent/US7953641B2/en | 2011-05-31 | patent | Telephone-listing invention record.

Writings

• ★★★ | 15 Rejections That Changed My Life | https://www.linkedin.com/pulse/15-rejections-changed-my-life-josh-kopelman-tk7de | 2026-02-11 | first-person essay | Full Infonautics, rejection, and contingency retrospective.
• ★★★ | First Round investor role post | https://www.linkedin.com/posts/jkopelman_investor-activity-7415106013462175744-WCbM | date provisional from activity ID; accessed 2026-08-24 | first-person post | Current hiring rubric; no stable page date.
• ★★★ | Yale anchor and Prospect Fellowship post | https://www.linkedin.com/posts/jkopelman_when-we-started-first-round-capital-in-2004-activity-7239379933758705665-fHKa | date provisional from activity ID; circa 2024 | first-person post | Direct 2007 Yale relationship and influence account.
• ★★★ | 10 Years of Investing, 10 Lessons Learned | https://www.linkedin.com/pulse/ten-year-project-josh-kopelman | 2015-07-30 | first-person essay | Ten explicit seed-investing lessons.
• ★★★ | Crystal Balls vs. Calculators | https://www.linkedin.com/pulse/crystal-balls-vs-calculators-josh-kopelman | 2015-06-29 | first-person essay | Pre-evidence judgment versus spreadsheet precision.
• ★★★ | The Watney Rule for Startups | https://www.linkedin.com/pulse/watney-rule-startups-return-old-normal-josh-kopelman | 2016-03-04 | republished partner letter | Collective First Round financing-cycle letter.
• ★★ | What It's Like to Lead a Startup According to 700+ Founders | https://www.linkedin.com/pulse/what-its-like-lead-startup-according-700-founders-josh-kopelman | 2016-12-01 | article/report gateway | Full framing; report has separate boundary.
• ★★ | What does it mean to be a startup entrepreneur in 2015? | https://www.linkedin.com/pulse/what-does-mean-startup-entrepreneur-2015-josh-kopelman | 2015-12-03 | article/report gateway | Short survey framing; comments excluded.
• ★★ | What the Seed Funding Boom Means | https://www.inc.com/first-round-review/josh-kopelman/what-the-seed-funding-boom-means.html | 2013; exact date unresolved | edited essay | Seed-specialization argument with date limit.
• ★★★ | The defunct startup that beat Best Buy in court | https://fortune.com/2012/12/05/the-defunct-startup-that-beat-best-buy-in-court/ | 2012-12-05 | first-person reported narrative | Litigation-financing account requiring docket checks.
• ★★ | Yet another VC blog | http://redeye.firstround.com/2006/03/yet_another_vc_blog.html | 2006-03; metadata only | inaccessible legacy essay | Body unavailable after direct/feed/archive attempts.
• ★★ | As a little kid | http://redeye.firstround.com/2006/03/as_a_little_kid.html | 2006-03-10; citation date | inaccessible legacy essay | Citation-level survival only.
• ★★★ | Shrink a Market! | http://redeye.firstround.com/2006/04/shrink_a_market.html | 2006-04; date incomplete | inaccessible legacy essay | Widely cited thesis route; fragments only.
• ★★★ | 53,651 | http://redeye.firstround.com/2006/05/53651.html | 2006-05-11 | inaccessible legacy essay | Audience-selection route; metadata/fragments only.
• ★★ | The M&A Lotto | http://redeye.firstround.com/2006/09/the_ma_lotto.html | 2006-09; date incomplete | inaccessible legacy essay | Outcome-language route; body unavailable.
• ★★★ | The First Penny Gap | http://redeye.firstround.com/2007/03/the_first_penny.html | 2007-03; date incomplete | inaccessible legacy essay | Freemium/conversion route; body unavailable.
• ★★ | After the TechCrunch | http://redeye.firstround.com/2008/01/after-the-techc.html | 2008-01; date incomplete | inaccessible legacy essay | Post-launch route; metadata only.
• ★ | Happy Holidays from First Round! | https://www.linkedin.com/pulse/happy-holidays-from-first-round-josh-kopelman | 2016-12-20 | poem/video post | Cultural artifact, not substantive philosophy.
• ★ | Happy Holidays from First Round Capital | https://www.linkedin.com/pulse/happy-holidays-from-first-round-capital-josh-kopelman | 2015-12-17 | poem/video post | Cultural artifact, not an investing model.
• ★ | First Search launch thread | https://threadreaderapp.com/thread/936261180380930048.html | 2017-11-30 | thread mirror | Preserves direct product-discovery writing.
• ★ | State of Startups 2015 | http://stateofstartups2015.firstround.com | 2015 | legacy firm report | Underlying survey route and access boundary.
• ★ | State of Startups 2016 | http://stateofstartups.firstround.com | 2016 | legacy firm report | Underlying 700-plus-founder report route.
• ★ | First Round Review author index — Josh Kopelman | https://review.firstround.com/articles/josh-kopelman/ | current; accessed 2026-08-24 | author index | Exact-name corpus route.
• ★ | First Round Review author index — Josh Kopleman typo | https://review.firstround.com/articles/josh-kopleman/ | current; accessed 2026-08-24 | index variant | Typo sweep prevents corpus loss.
• ★ | Second-time founder dinner post | https://www.linkedin.com/posts/jkopelman_theres-plenty-of-advice-floating-around-activity-7039283799033020416-7O_d | 2023; platform-relative | repost/context | Lower-priority operating context.
• ★ | Uncapped episode reflection post | https://www.linkedin.com/posts/jkopelman_how-to-make-money-in-venture-josh-kopelman-activity-7323768369906688000-PDi9 | 2025; platform-relative | first-person post | Josh supplies preferred video locators.
• ★ | Simon Zhou joining First Round | https://www.linkedin.com/posts/jkopelman_excited-to-have-simon-zhou-joining-us-at-first-round-capital-activity-7485732576654069760-GStM | 2026-08-23 | first-person post | Current team-language check.

Spoken appearances

• ★★★ | Imagine If with Josh Kopelman | https://alliancefordecisioneducation.org/podcasts/episode-032-imagine-if-with-josh-kopelman/ | 2024-12-11 | podcast transcript | Full transcript; header/footer date conflict retained.
• ★★★ | 20VC — Josh Kopelman | https://www.thetwentyminutevc.com/joshkopelman | date on page; accessed 2026-08-24 | podcast/show notes | Price, attribution, boards, ownership, and misses.
• ★★★ | Past, Present, and Future of Seed Investing | https://colossus.com/episode/kopelman-the-past-present-and-future-of-seed-investing/ | 2020 | podcast transcript | Seed history, community, and case evidence.
• ★★ | Colossus legacy redirect | https://joincolossus.com/episode/kopelman-the-past-present-and-future-of-seed-investing/ | 2020; redirect checked 2026-08-24 | legacy route | Exact redirect retained for parity.
• ★★★ | How To Make Money in Venture — Uncapped Ep. 8 | https://www.youtube.com/watch?v=7VCSi7F1j0o | 2025-05-01 | video interview | Fund math, disequilibrium, and firm design.
• ★★★ | What Are Early-stage VCs Looking for? | https://knowledge.wharton.upenn.edu/podcast/knowledge-at-wharton-podcast/new-companies-get-off-ground/ | 2018-06-21 | edited transcript/audio | Founder selection and career chronology.
• ★★ | Yale entrepreneurial lessons lecture | https://www.youtube.com/watch?v=5Br8WmQavMw | 2014-07-01 | video lecture | University-hosted historical direct voice.
• ★★ | Vintage CEO Summit conversation | https://www.youtube.com/watch?v=YHNLQTk1qAs | 2026-08-17 | video | Recent rubric-evolution check.
• ★ | FundersClub/UproUnd conversation | https://www.youtube.com/watch?v=zaa2Cl1dopk | 2015-10-30 | video | Contextual direct voice.
• ★ | Philly Who interview | https://www.youtube.com/watch?v=CI2U6WvYRW0 | 2018-12-12 | video | Philadelphia and ecosystem context.
• ★ | LDV conversation | https://www.youtube.com/watch?v=5vP1Mu4Ycq8 | 2022-05-09 | video | Contextual full interview.
• ★★ | Upfront companion video | https://www.youtube.com/watch?v=0Q09iCYJ_Uo | 2020-04-03 upload | video | Companion to edited event article.
• ★★ | How to start and run an early-stage fund | https://techcrunch.com/2020/02/03/josh-kopelman-on-how-to-start-and-run-an-early-stage-fund-in-a-hypercompetitive-market/ | 2020-02-03 | edited event transcript | Brand, stage, fundraising, and firm construction.
• ★ | Origins — Josh Kopelman | https://podcasts.apple.com/us/podcast/josh-kopelman-first-round-capital/id1111792048?i=1000483153046 | 2020-07-07 | official audio/show notes | Career and firm-building chronology.
• ★ | Founder Factory appearance | https://www.youtube.com/watch?v=2O_bmTzG62k | date on platform | video | Exact parity route.
• ★ | Penn/Wharton appearance | https://www.youtube.com/watch?v=KDRNKm4om08 | date on platform | video | Exact parity route.
• ★ | Forbes interview | https://www.youtube.com/watch?v=oA3A84-lN4g | date on platform | video | Contextual triangulation.
• ★ | TechCrunch Disrupt NY interview | https://techcrunch.com/video/josh-kopelman-first-round-capital/ | 2012-05-23 recording | event video | Historical internationalization route.
• ★ | Off the Sidelines | https://technical.ly/startups/off-the-sidelines-josh-kopelman-first-round-capital-advice-investors-founders/ | 2020-02-05 | podcast article/audio | Investor motivation and founder help.
• ★ | Friday Q&A with Josh Kopelman | https://technical.ly/startups/friday-qa-josh-kopelman-of-first-round-capital/ | 2010 | edited interview | Historical Philadelphia/access evidence.
• ★ | Josh Kopelman on the changing VC landscape | https://vator.tv/2014-01-02-josh-kopelman-on-the-changing-vc-landscape/ | 2014-01-02 | edited interview | Lower-priority direct answers on seed-fund scale and portfolio community; reported figures are not return evidence.
• ★ | Startup Grind Philadelphia — democratization of entrepreneurship | https://technical.ly/startups/josh-kopelman-democratization-entrepreneurship-video/ | 2014-11-17 | event article/video route | Selected remarks and embedded-video route; no stable full caption transcript.
• ★ | Wharton Alumni Network at Its Best | https://magazine.wharton.upenn.edu/digital/wharton-alumni-network-at-its-best/ | 2015-12-02; original post 2015-10-23 | institutional event recap | Selected dinner quotations and Nat Turner chronology; not a full speech transcript.
• ★ | Technical.ly Slack AMA — five takeaways | https://technical.ly/startups/5-takeaways-technical-ly-slack-ama-first-round-capitals-josh-kopelman/ | 2016-02-16 | edited written AMA | Direct written answers; retained as a spoken-sweep boundary item, not time-aligned audio.
• ★ | TechCrunch — SPACs and rolling funds | https://techcrunch.com/2020/09/10/vc-josh-kopelman-isnt-so-sure-about-spacs-but-he-thinks-so-called-rolling-funds-could-prove-powerful/ | 2020-09-10 | lightly edited interview | Vehicle, liquidity, succession, and geography context; publication is not a time-aligned transcript.

Deals and outcomes

• ★★★ | Project Fair Bid complaint exhibit | https://cases.justia.com/federal/district-courts/washington/wawdce/2%3A2011cv00809/175727/1/0.pdf | 2011 | court filing PDF | Pleaded allegations only.
• ★★ | Flatiron financing report | https://fortune.com/2014/05/01/google-ventures-leading-nine-figure-investment-in-flatiron-health/ | 2014-05-01 | reported financing | Decision-era financing and syndicate context.
• ★★★ | AT&T 2018 annual report | https://investors.att.com/~/media/Files/A/ATT-IR-V2/financial-reports/annual-reports/2018/complete-2018-annual-report.pdf | 2019 | issuer report PDF | Official AppNexus acquisition disclosure.
• ★★★ | Project Fair Bid dismissal order | https://law.justia.com/cases/federal/district-courts/washington/wawdce/2%3A2011cv00809/175727/23/ | 2012 | court order | Procedural disposition, not merits exoneration.
• ★★★ | Upstart founder board/pivot account | https://review.firstround.com/fresh-off-ipo-upstarts-ceo-shares-why-the-startup-isnt-a-typical-success-story/ | 2021 | edited founder interview | Specific board question; survivor boundary.
• ★★ | AT&T acquisition of AppNexus | https://www.axios.com/2018/06/25/att-acquisition-appnexus-time-warner-advertising-ad-tech | 2018-06-25 | reported transaction | Reported consideration cross-check.
• ★★★ | First Round card — Clover Health | https://www.firstround.com/companies/?selected=clover-health | current; accessed 2026-08-24 | official portfolio card | Series A partner attribution.
• ★★★ | First Round card — Flatiron Health | https://www.firstround.com/companies/?selected=flatiron-health | current; accessed 2026-08-24 | official portfolio card | Seed partner attribution.
• ★★★ | First Round card — Notion | https://www.firstround.com/companies/notion | current; accessed 2026-08-24 | official company page | Named partner, seed check, and rebuild history.
• ★★ | Notion funding history | https://www.forbes.com/newsletters/midas-touch/2021/10/11/how-vcs-got-into-notions-new-round-newarks-new-fund-and-a-womens-angel-investor-group-hits-a-milestone/ | 2021-10-11 | reported financing | Funding and paper-value context.
• ★★ | Flatiron founders and data gap | https://www.forbes.com/sites/matthewherper/2018/11/14/at-24-two-entrepreneurs-took-on-cancer-at-32-theyre-worth-hundreds-of-millions/ | 2018-11-14 | founder profile | Formation, market problem, outcome.
• ★★ | Wiley buys Knewton | https://www.insidehighered.com/digital-learning/article/2019/05/07/wiley-buys-knewton-adaptive-learning-technology-company | 2019-05-07 | reported acquisition | Failure/downside route; association provisional.
• ★★★ | Roche completes Flatiron acquisition | https://www.roche.com/media/releases/med-cor-2018-04-06 | 2018-04-06 | issuer release | Official transaction value; not proceeds.
• ★★★ | Upstart S-1 | https://www.sec.gov/Archives/edgar/data/1647639/000119312520285895/d867925ds1.htm | 2020-11-05 | SEC filing | Vehicle ownership, shared control, and risks.
• ★★★ | Clover support agreement | https://www.sec.gov/Archives/edgar/data/1801170/000110465920112403/tm2032431d1_ex10-3.htm | 2020 | SEC exhibit | FRC V signature and public-transition support.
• ★★★ | Clover transaction announcement | https://www.sec.gov/Archives/edgar/data/1801170/000110465920112403/tm2032431d1_ex99-1.htm | 2020-10-06 | SEC exhibit | SPAC terms and valuation boundary.
• ★★★ | Clover 8-K — investigation conclusion | https://www.sec.gov/Archives/edgar/data/1801170/000180117024000195/clov-20240926.htm | 2024-09-30 | SEC filing | No-enforcement-recommendation notice.
• ★★★ | Venrock AppNexus memo | https://www.venrock.com/wp-content/uploads/2018/06/Click-here-to-read-our-original-investment-memo-.pdf | 2007–2008; published 2018 | co-investor memo | Decision-era thesis; not First Round memo.
• ★★ | Project Fair Bid raises $4.5M | https://techcrunch.com/2009/07/23/project-fair-bid-raises-45-million-for-stealth-auction-platform/ | 2009-07-23 | reported financing | Financing and board-role report.
• ★★ | First Round filter — Upstart | https://www.firstround.com/companies/?selected=upstart | current; accessed 2026-08-24 | official portfolio query | Attribution-parity route.
• ★★ | First Round filter — AppNexus | https://www.firstround.com/companies/?selected=appnexus | current; accessed 2026-08-24 | official portfolio query | Discovery/attribution route.
• ★ | First Round filter — Knewton | https://www.firstround.com/companies/?selected=knewton | current; accessed 2026-08-24 | official portfolio query | Discovery and negative check.
• ★ | First Round filter — Project Fair Bid | https://www.firstround.com/companies/?selected=project-fair-bid | current; accessed 2026-08-24 | official portfolio query | Alias and attribution check.
• ★ | First Round filter — Drop | https://www.firstround.com/companies/?selected=drop | current; accessed 2026-08-24 | official portfolio query | Massdrop/Drop alias route.
• ★★ | First Round filter — Pomelo Care | https://www.firstround.com/companies/?selected=pomelo-care | current; accessed 2026-08-24 | official portfolio query | Exact cited query variant for Josh attribution; canonical entity page also exists.
• ★★ | First Round filter — Loyal | https://www.firstround.com/companies/?selected=loyal | current; accessed 2026-08-24 | official portfolio query | Exact cited query variant for Josh attribution; canonical entity page also exists.
• ★★ | First Round filter — Parallel | https://www.firstround.com/companies/?selected=parallel | current; accessed 2026-08-24 | official portfolio query | Exact cited query variant; firm directory attributes both Todd Jackson and Josh Kopelman.
• ★★ | Startups Launch to Help Web 1.0 Sites | https://techcrunch.com/2006/03/07/startups-launch-to-help-web-10-sites/ | 2006-03-07 | contemporaneous financing/launch report | Names First Round and Kopelman in Bazaarvoice backing; no check, ownership, or proceeds evidence.
• ★★★ | Enova to acquire OnDeck | https://www.enova.com/newsroom/enova-to-acquire-ondeck-to-create-a-leading-fintech-company-serving-consumers-and-small-businesses/ | 2020-07-28 | issuer transaction release | Official approximately $90m transaction terms; does not establish First Round or Josh proceeds.
• ★★★ | Roblox S-1 | https://www.sec.gov/Archives/edgar/data/1315098/000119312520298230/d87104ds1.htm | 2020-11-19 | SEC filing | Pooled FRC II ownership and GP control evidence; does not overwrite Chris Fralic's partner sponsorship.

Adverse and regulatory

• ★★★ | TechForward v. Best Buy docket | https://dockets.justia.com/docket/california/cacdce/2%3A2011cv01313/494448 | 2011–2012 | court docket | Procedural record for narrative checks.
• ★★★ | Best Buy fiscal 2012 10-K | https://www.sec.gov/Archives/edgar/data/764478/000076447812000035/bby-201210k.htm | 2012 | SEC filing | Issuer-side legal record.
• ★★ | Why First Round funded a lawsuit | https://news.ycombinator.com/item?id=4878687 | 2012-12-06 | discussion thread | Reaction only; comments not facts.
• ★★ | Best Buy litigation scholarship | https://www.nyujlb.org/_files/ugd/716e9c_76e03a8b79ae4d768eaf0496a7358019.pdf | date on PDF | legal scholarship PDF | Independent doctrinal treatment.
• ★★ | TechForward wins trade-secrets case | https://techcrunch.com/2012/12/05/techforward-wins-27m-in-lawsuit-against-best-buy-over-stolen-trade-secrets/ | 2012-12-05 | reported litigation | Outcome and financing context.
• ★★★ | Hindenburg report on Clover | https://hindenburgresearch.com/clover/ | 2021-02-04 | adversarial short report | Allegations fenced from findings.
• ★★★ | Clover response to short report | https://www.sec.gov/Archives/edgar/data/1801170/000119312521029637/d66346dex991.htm | 2021-02-05 | SEC exhibit | Interested-party counterevidence.
• ★★ | Clover securities complaint | https://www.cloverhealthsecuritieslitigation.com/admin/api/connectedapps.cms.extensions/asset?id=eb7304ce-850d-4ad4-abb4-877b2b6a8efe&inline=true&languageId=1033 | 2021–2023 | court filing PDF | Pleaded allegations only.
• ★★ | Project Fair Bid California order | https://cases.justia.com/federal/district-courts/california/candce/4%3A2013cv00759/263455/57/0.pdf | 2013 | court order PDF | Additional procedural disposition.
• ★★ | We Have Met the Enemy | https://www.kauffman.org/wp-content/uploads/2012/05/we_have_met_the_enemy_venture_capital_report_kauffman_foundation.pdf | 2012-05 | research report PDF | Skeptical venture economics context.
• ★★ | Can AltSchool save itself? | https://www.forbes.com/sites/susanadams/2019/01/30/can-altschoolthe-edtech-startup-with-174m-from-billionaires-like-zuckerberg-and-thielsave-itself-from-failure/ | 2019-01-30 | failure analysis | Sector comparator, not Josh attribution.
• ★★ | RIP Path | https://techcrunch.com/2018/09/17/rip-path/ | 2018-09-17 | reported shutdown | Failure context with attribution boundary.
• ★★ | Best Buy 2013 annual report mirror | https://www.annualreports.com/HostedData/AnnualReportArchive/b/NYSE_BBY_2013.pdf | 2013 | issuer-report mirror | Litigation disclosure cross-check.
• ★ | Patent-industry TechForward verdict report | https://generalpatent.com/techforward-wins-22-million-trade-secrets-lawsuit-against-best-buy.html | 2012-11-26 | industry report | Secondary chronology check.
• ★ | Courthouse News — Project Fair Bid complaint summary | https://www.courthousenews.com/class-calls-auction-website-a-rico-fraud/ | 2011; page date not exposed | litigation reporting | Complaint summary only; plaintiff allegations remain fenced from findings and procedural disposition.
• ★★ | Stanford Clover securities case tracker | https://securities.stanford.edu/filings-case.html?id=107633 | 2021–2023 | secondary docket index | Shareholder-action and settlement metadata; not an adjudication of every allegation.
• ★★ | Chronicle — Wiley buys Knewton | https://www.chronicle.com/article/another-big-move-hits-higher-ed-publishing-as-wiley-buys-knewton/ | 2019-05-06; access may be partial | reported acquisition | Outcome and industry-characterization route; terms and Josh/FRC attribution remain unresolved.
• ★★ | AltSchool becomes Altitude Learning | https://www.edsurge.com/news/2019-06-28-altschool-gets-an-alt-name-and-new-leadership | 2019-06-28 | reported failure/restructure | Rename and leadership-change evidence; Josh-specific vehicle, governance, and economics remain unverified.

Archive and context

• ★★ | Venture Capital Journal profile PDF | https://assets.fenwick.com/legacy/FenwickDocuments/VCJ%20June%202013_cover%20story.pdf | 2013-06 | magazine PDF | Historical firm/industry framing.
• ★★ | Private Equity at a Crossroads | https://knowledge.wharton.upenn.edu/wp-content/uploads/2014--05-06-Private-Equity-at-Crossroads.pdf | 2014 | conference PDF | Institutional-capital context.
• ★★ | Steve Anderson as one-man deal machine | https://www.forbes.com/sites/ryanmac/2016/03/23/steve-anderson-baseline-ventures-instagram-venture-capital-raver/ | 2016-03-23 | peer profile | Early-seed ecosystem context.
• ★★ | First Round launches HackPR | https://techcrunch.com/2012/10/09/first-round-capital-launches-a-platform-for-startups-to-hack-pr/ | 2012-10-09 | firm-initiative report | Platform services beyond capital.
• ★ | CRV QuickStart discussion index | https://www.techmeme.com/061101/p14 | 2006-11-01 | media aggregation | Redeye metadata, lead only.
• ★ | Alternatives to Dancing on a Bubble | https://www.washingtonpost.com/archive/business/2006/10/24/alternatives-to-dancing-on-a-bubble/7d06012f-b253-483b-a99c-0390e0461f25/ | 2006-10-24 | market context | Decision-era financing-cycle context.
• ★ | Josh Kopelman reference page | https://en.wikipedia.org/wiki/Josh_Kopelman | current; accessed 2026-08-24 | compiled reference | Discovery only.
• ★ | First Round Capital reference page | https://en.wikipedia.org/wiki/First_Round_Capital | current; accessed 2026-08-24 | compiled reference | Discovery/alias index only.
• ★ | Josh Kopelman compiled profile | https://yespress.io/josh-kopelman | 2026; accessed 2026-08-24 | compiled profile | Lead only; promotional figures unsupported.
• ★ | Redeye VC parser archive | https://parsers.vc/startup/redeye.firstround.com/ | current; accessed 2026-08-24 | third-party archive index | Title/date discovery only.
• ★ | AngelMatch Josh Kopelman profile | https://angelmatch.io/investors/josh-kopelman | current; accessed 2026-08-24 | compiled dynamic database | Discovery lead only; no stable company-level locator, vehicle, terms, ownership, or outcome verification.
• ★ | Mercury Josh Kopelman investor profile | https://mercury.com/investor-database/josh-kopelman | current; accessed 2026-08-24 | compiled dynamic database | Discovery lead only; company associations require primary confirmation.

Coverage assessment

• 136 unique source rows: 24 biography/firm + 27 writings + 25 spoken appearances + 30 deals/outcomes + 18 adverse/regulatory + 12 archive/context = 136.
• Writing denominator: 27 items = 17 priority (10 accessible/deep, 7 inaccessible Redeye) + 10 lower priority. The two holiday posts are lower-priority cultural artifacts, not substantive investing essays.
• Access boundaries: seven priority Redeye bodies remained unavailable after canonical, feed, title-variant, and archive-index attempts. LinkedIn activity dates are provisional unless explicitly exposed.
• Attribution: Josh-authored, collective partner, firm, issuer, court, adversarial, and compiled sources remain distinct. Company outcomes and transaction values are not investor returns.
• Hard gaps: no complete Josh deal ledger, cash-flow series, rejected-set denominator, decision memos, partner votes, or full Redeye archive.