Bruce Dunlevie

1. Snapshot and testable edge thesis

The thesis is deliberately narrower than “great picker.” It applies to an institutional conversion system in a specific era, not every Benchmark partner, every vehicle, or Dunlevie's philanthropy and trustee roles. It is a process hypothesis with moderate support, while repeatable personal performance remains [researcher inference · insufficient public record].

2. Core philosophy

Dunlevie's clearest direct doctrine appears in one 2022 Rice talk: reject a rigid category strategy, do broad fieldwork, choose a very small number from a large funnel, emphasize people because plans change, prefer a small fund, and govern a minority position through trust and persuasion. Rice, 12:24–19:59, 48:07–50:08, and 64:25–65:09 [investor-stated · strong] An Amity attendee relayed the same small-fund, openness, and founder-selection themes a year later, but without transcript authority. Amity attendee note [retrospective witness · provisional]

The bounded synthesis is:

1. Search without topical blinders, inside a competence boundary. [researcher inference · moderate] Dunlevie rejects strategy yet avoids biotech/FDA and expresses uncertainty about Web3. Rice, 12:24–14:16, 32:00–33:35, 65:26–65:54
2. Treat adaptability and trust as necessary, never sufficient. [investor-stated · strong] Plans change and minority investors need candor; Google and WeWork force a two-sided test for personality bias. Rice, 14:16–16:23 and 52:46–54:17
3. Make partnership and vehicle serve the work. [researcher inference · moderate] Small scale, equal economics, independent founder exposure, and board capacity form one organizational hypothesis; the private partnership agreement and causal return test are missing. Acquired Part I · Acquired Part II
4. Separate insight from capture. [researcher inference · strong] A true behavior or technology change must still pass company, implementation, price, financing, and control tests. WeWork, Tabula, EnVerv, and Impresse make the boundary observable. writings synthesis
5. Use trust first, then enforce precommitted boundaries. [researcher inference · strong] Relationship governance can preserve information and cooperation, but private dissent without rights can become acquiescence. Special Committee statement

This is reconstructed from direct words, documented behavior, and adverse cases. It is not a recovered Benchmark manual.

3. Sourcing

The public record supports a relationship-and-fieldwork funnel rather than a sector map. Dunlevie described roughly 1,000 or 2,000 opportunities and about ten investments in changing scopes, and emphasized meeting counterintuitive founders instead of filtering by a fixed theme. Rice, 12:24–19:59 and 65:26–65:54 The rounded figures cannot produce a career conversion rate. [investor-stated · moderate]

Supported access routes include:

• Operating and technical networks. [documented behavior · strong] Geoworks led to Hawkins and Palm; specialist Valley relationships reached Rambus and Kalpana. Hawkins oral history, printed pp. 12–13
• Repeat-founder relationships. [documented behavior · moderate] Palm's team later formed Handspring and attracted Benchmark II. Handspring funding statement
• Partner networks. [retrospective witness · moderate] Benchmark pooled Kagle, Rachleff, Harvey, Vaden, and Dunlevie, then later partners; shared access does not identify the originating individual. Marquardt oral history, printed pp. 56–58
• Direct demand checks. [retrospective witness · moderate] Dunlevie's eBay account moved from dismissal to customer calls and visible growth, while Kagle retained deal credit. Rice, 29:12–31:42
• Institutional networks. [researcher inference · moderate] Rice, Stanford, Goldman, MPAE, trustee roles, and Getty widen possible contact surfaces, but no opportunity log measures their contribution. profile influence map

A repeatable sourcing audit would record introducer, prior relationship, why this team is non-obvious, information unavailable to an ordinary investor, domain competence owner, conflicts, originating partner, sponsor, dissent, and later pass review. Without those fields, access and selection are confounded. ARM is excluded because contemporary formation records identify Acorn, Apple, and VLSI and no searched primary source connects Dunlevie or MPAE. Los Angeles Times · Tech Monitor [researcher inference · strong]

4. Picking

The people screen is the strongest recurring motif, but it needs an evidence harness.

No public record supplies a stable weighting formula, valuation ceiling, rejected-founder denominator, demographic-validity study, or complete reference process. An auto-captioned Rice passage about founder ages and a sociopathic trait was not audio-verified and has no demonstrated predictive validity; it is treated as a risk and excluded from the operating model. Rice, 14:43–15:16 [investor-stated · provisional]

5. Deal and portfolio mechanics

Public mechanics are filing snapshots, not a private portfolio policy.

The three Pareto views must remain separate: One Medical leads the directional economic ranking, Palm/Handspring the historical ranking, and WeWork the learning ranking. Pareto analysis Removing One Medical does not produce a defensible next-dollar order because remaining measures are incompatible. [researcher inference · strong]

6. Vehicle and incentive fit

Fit test — [researcher inference · strong]. Before investment, name the eligible entity, mandate, check, expected ownership, reserve need, partner time, board authority, conflicts, capital duration, likely follow-on providers, and exit/liquidity path. A vehicle mismatch is not automatically a bad company, and a trust filing is not a Benchmark investment.

7. Team attribution and work with founders

The operating contribution pattern is relationship-intensive and often board-visible: challenge product scope, recruit or add stage capability, connect commercial resources, and preserve candor. [researcher inference · moderate] Presence is not impact. A useful attribution ledger has separate columns for introduction, thesis, diligence, sponsor, negotiation, approval, vehicle, security, board labor, executive recruitment, reserves, stop/exit decision, narration, and economics.

8. Views on the craft

• Power law. [investor-stated · moderate] Dunlevie uses extraordinary first-fund magnitude to argue that a small number matters, but four incompatible accounts cannot be averaged: Rice — reported: 100x; Amity — reported: 80x-plus; Forbes — reported: 91.8x (about, from the $85 million/$7.8 billion headline); Metrick–Yasuda — reported: 42x on $113 million. Rice, 12:09–12:25 · Amity attendee post · Forbes · Metrick–Yasuda, pp. 89–90
• Fund size. [investor-stated · strong] He ties smaller scale to MOIC and partner attention. The claim is falsifiable: comparable net results must survive vintage and outlier controls. Rice, 12:53–13:39
• Access versus selection. [researcher inference · strong] Institutions and relationships shape the opportunity set; complete opportunity logs are needed to prove that people/trust gates add value beyond privileged access.
• Governance. [investor-stated · strong] Minority investors depend on trust and persuasion. WeWork proves the need for a separate control and escalation doctrine. Rice, 48:07–50:08 · Special Committee statement
• Exits. [researcher inference · insufficient public record] No universal sale, IPO, distribution, hold, or stop rule survives.
• Great investor. [researcher inference · moderate] The supported ideal is open but competent search, behavior-grounded people judgment, independent deliberation, vehicle discipline, intensive service, explicit control thresholds, honest attribution, and learning from false positives and false negatives. A famous portfolio and coherent retrospective narrative are not enough.

9. Signature bets and decision process

Use the frozen-date cases rather than retelling outcomes:

1. Palm to Handspring — formation access, product challenge, repeat-team capability, platform dependence, partner friction, and separate vehicles.
2. Rambus — formation-era board work, disclosed MPAE V security, long tenure, and governance controversies without personal guilt inference.
3. Genesys — investor service converted into a fund warrant; 9.9% fund block versus 10.1% broader total including personal options.
4. One Medical and ServiceSource — similar long board visibility but different appointment evidence, different shareholder and transaction outcomes, and no personal-return inference.
5. WeWork — the central no-hindsight failure lab for category, founder, price, duration, control, and later-capital power.
6. Tabula and Toys “R” Us — technical milestone failure versus an organizational-truth stop.

For each lab: freeze known/unknown/disputed facts; state must-be-true conditions; decide invest/pass, security, price, ownership, authority, milestones, and next check; write falsifiers and dissent; reveal outcome only after the decision; score reasoning separately from company outcome and fund return. [researcher inference · strong]

10. Mental models and maxims

The reconciled inventory contains exactly eight distinct supportable rules; the headings below match the field guide one-for-one.

1. Search broadly, but stay within competence.
2. Back adaptable people, then run the two-sided evidence test.
3. Separate category truth from company capture, economics, price, financing, and control.
4. Preserve independent partner judgment before advocacy.
5. Size the fund and team to board-service capacity, then align economics.
6. Start with trust and persuasion, but precommit formal escalation.
7. Match the company stage with operating capability and real authority.
8. Stage capital against milestones and stop when organizational truth fails.

Repeat-team investing is absorbed into Model 2; trend, implementation, economics, and financing remain separate Models 3 and 8 because their kill conditions differ; no-pre-sell is Model 4; equal economics and board capacity are Model 5. Age/demographic captions, universal price/ownership/reserve/exit rules, international expansion, philanthropy, and the provisional CarbonDrop identity are excluded as models because the record lacks a reusable decision rule.

11. Evolution and contradictions

Central contradictions remain productive rather than resolved away: no strategy versus domain limits; people-first versus Google/WeWork personality errors; persuasion versus enforceable control; small-fund advocacy versus incompatible return bases; and no-pre-sell intuition versus decision memory. A modern operating system needs explicit gates because Dunlevie's public doctrine leaves the transition points private.

12. Failures and limits

• WeWork — favorable intuition, failed system. [researcher inference · strong] Community demand could be real while leases, price, cash burn, supervotes, related parties, and SoftBank power defeated capture and control. Forbes · Chapter 11 8-K
• Tabula — adjectives without staged falsification. [researcher inference · strong] A large market, breakthrough claim, and proven team did not substitute for technical, customer, burn, and follow-on milestones. Tabula release · shutdown
• Google — false negative. [retrospective witness · moderate] A working product was visible; interpersonal aversion dominated the collective decision. The roughly $10 billion counterfactual is not realizable return. Rice, 52:46–54:17
• ServiceSource — accountability boundary. [documented behavior · strong] The board retained a director who failed to receive majority support; reasons and later process change remain private. 2019 8-K
• Toys “R” Us — stop under truth failure. [retrospective witness · moderate] Reporting says Benchmark canceled despite a guaranteed payout after cooperation claims proved unreliable; contract and minutes are unavailable. New Yorker
• Impresse and EnVerv — financing does not equal proof. [contemporaneous record · moderate] Large financing preceded Impresse's shutdown/asset sale; EnVerv's selected assets sold for $4.9 million. Impresse shutdown · Semtech 10-Q
• Marin — causal humility. [researcher inference · strong] Chapter 11 eight years after Dunlevie's board departure is an adverse company outcome, not proof of his causal failure. Marin Chapter 11 index
• Selection and survivorship. [researcher inference · strong] SEC visibility overweights board roles and public companies; prepared statements overweight confidence; famous stories overweight wins and vivid failures; quiet deals and passes remain unseen.

No public evidence shows a complete private write-off schedule, opportunity log, personal return, or explicit process change after most failures. The strongest practical response is to preserve decision-date forecasts and exceptions, not to manufacture a clean learning arc.

13. Historical operating environment

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

All translations below are [researcher inference · strong] unless a row states otherwise. They are analyst applications, not claims about Dunlevie's current practice.

15. Comparison with completed peers

The valid completed peers as of 2026-08-22 are Andy Rachleff, Fred Wilson, and Ann Winblad. Rachleff is the same-firm institutional and attribution control; Wilson is a public thesis/venture-math and failure-transparency counterpoint; Winblad is a same-era operator-to-investor, board-intensive technology specialist. Their vehicles and cash flows are not performance-comparable. [researcher inference · strong]

The comparison sharpens Dunlevie's distinctive claim: less explicit than Rachleff on product/market proof, less public than Wilson on math and errors, less specialized than Winblad on software/customer evidence, but unusually coherent on small equal-partner architecture plus trust-based long-board service. [researcher inference · moderate] That distinctiveness is not proof of superiority.

16. The strongest case against greatness

The skeptical thesis is that Dunlevie's reputation combines privileged institutional access, team economics, one spectacular early fund, long board visibility, and retrospective narrative selection—not independently demonstrated personal alpha.

• Outlier and luck. [researcher inference · strong] Benchmark I's incompatible reported magnitudes may be dominated by eBay, whose introduction and board work belong to Kagle. Later Benchmark IV evidence prevents a simple persistent-fund story. Forbes · Metrick–Yasuda
• Access versus selection. [researcher inference · strong] Rice, Stanford, Goldman, MPAE, Benchmark, board work, and later trusteeships create a plausible high-quality access network; no rejected-opportunity cohort shows that Dunlevie's gates outperform access alone. Rice, 05:31–09:20 and 19:30–19:45 · Getty institutional biography · searched denominator
• Team misattribution. [documented behavior · strong] Kagle/eBay and Gurley/Uber are explicit controls; Palm product credit belongs to Hawkins, Dubinsky, Colligan, and their teams; shared legal control and equal economics obscure personal labor elsewhere. Acquired Part I, eBay reconstruction · Gurley transcript · Hawkins oral history · SEC reporting-person index
• Missing denominator. [researcher inference · strong] Nine sufficiently reconstructed relationships cannot support a career hit rate; 12 aggregator-only leads and unknown private investments, failures, follow-ons, and passes could change the visible balance. denominator reconciliation · uncertainty ledger
• Weak economics. [researcher inference · strong] Company transaction values, public holdings, one LP's MPAE V record, and incompatible Benchmark I headlines cannot produce personal DPI, TVPI, MOIC, IRR, or carry. UC Regents MPAE V schedule, p. 6 · Metrick–Yasuda, pp. 89–90 · Pareto/economic boundary
• Era effects. [researcher inference · moderate] Formation-era specialist computing, 1990s IPO liquidity, the post-2000 contraction, cheap 2010s private capital, and filing-visible public boards shaped both opportunity and outcome. Rambus prospectus · Forbes, 2001 · Irish Times, 2001 · SoftBank commitment record
• Contradictory behavior. [researcher inference · strong] Google challenges people judgment; WeWork challenges charisma screening and persuasion; ServiceSource challenges accountability; Tabula challenges technical milestone discipline; boom-era statements challenge later cycle wisdom. Rice, Google and WeWork · ServiceSource vote 8-K · Tabula financing/shutdown · Irish Times
• Non-repeatability. [researcher inference · insufficient public record] No current audited Dunlevie portfolio or public 2026 investment-policy record establishes that the architecture transfers beyond its original scale, geography, partnership, and market regime; the latest direct public investing talk is retrospective, Trinity 2026 was off record, and the current CarbonDrop identity remains provisional. talk saturation and retrievals · source-map coverage · Rice carbon lead, 61:10–62:44

Evidence deciding between the favorable and skeptical views would include complete fund cash flows and write-offs; opportunity/pass logs; dated partner scores and votes; partner labor/carry allocation; board minutes showing interventions and accepted/rejected advice; entry securities, dilution, reserves, and exits; and outcomes matched to ex-ante scores across wins, ordinary cases, and failures. The favorable thesis should win only if process variables predict capital-weighted net outcomes after controlling for access, partner, vehicle, vintage, sector, and the dominant outlier.

17. Analyst study guide, glossary, and go deeper

Recommended path

1. Read the biography and its falsifiable edge before hearing the hero stories.
2. Watch Rice 2022, then use talks to mark direct voice versus retrospective recollection.
3. Read the writings source-by-source record, beginning with the restricted-case boundaries and adverse WeWork account.
4. Complete the Palm/Handspring, WeWork, and Tabula/Toys labs without outcomes.
5. Work all eight mental models, including Model 4 and Model 8's exception ledgers.
6. Return to the skeptical thesis and write what new evidence would change your view.

Glossary

Reusable sourcing checklist

• Record source channel, introducer, relationship age, unique information, conflicts, partner owner, domain competence, and why the opportunity is non-obvious.
• Preserve passes and duplicates; separate company family, legal entity, round, follow-on, and non-portfolio control.
• Ask which access advantage is repeatable and whether an ordinary fund could obtain the same evidence.

Reusable diligence and IC checklist

• Freeze category change, company capture, founder adaptation/candor, product proof, unit economics, price, security, ownership, financing dependency, regulation, and control as separate votes.
• Run the two-sided personality test: evidence discounted because of discomfort; evidence excused because of charisma.
• Collect independent partner scores before discussion; preserve the post-discussion thesis, dissent, kill criteria, and next-check milestone.
• Name the eligible vehicle, reserve burden, partner time, authority, conflicts, and outcomes that would invalidate the decision.

Reusable governance checklist

• Write a role/authority matrix for founder, executive team, board, investors, controller, committees, and material counterparties.
• Define information, conduct, capital, runway, milestone, compliance, conflict, leadership, and resignation triggers before crisis.
• At every board cycle record constraint, decision, owner, deadline, dissent, and whether advice changed action; board tenure is not value-add evidence.
• Separate fiduciary action, fund economics, personal/trust economics, founder conduct, and counterparty conduct.

Best sources and why

1. Rice Business Partners 2022 — only full public direct-voice investing talk; strongest philosophy source, with retrospective and caption boundaries.
2. Hawkins oral history and Dubinsky oral histories — best operator controls on Palm formation, product credit, and partnership friction.
3. Acquired Part I and Part II — best public institutional reconstruction; partly relays an unreleased private interview.
4. Genesys prospectus — cleanest source on investor labor, fund warrant, ownership, and personal-option separation.
5. WeWork S-1, Chancery opinion, and bankruptcy 8-K — strongest control, process, and outcome stress test.
6. Metrick–Yasuda — essential counterweight to first-fund return folklore and cross-vintage extrapolation.
7. New Yorker — strongest skeptical narrative; retrospective and adversarial, so pair it with filings and court records.
8. Complete mapped bibliography — 153 processed and priority routes with reliability grades and access boundaries.

Go deeper through the unresolved ledger in investments: MPAE/Benchmark LP records, opportunity logs, partner votes, Palm/Kalpana cap tables, WeWork minutes and cash flows, Tabula milestones, and licensed E148/E218 could materially change the assessment.

Pareto 80/20 — what drove the record

Public evidence supports a power-law reading of the vehicles around Bruce Dunlevie, but not a literal 80/20 calculation for his personal record. No complete opportunity set, portfolio schedule, cash-flow ledger, ownership history, write-off schedule, or carry allocation is public. The rankings below are therefore directional. Company values are not fund proceeds; shared legal control is not personal ownership; a board seat is not proof of origination; and Benchmark's equal partnership makes team credit essential.

Ranked by likely economic contribution

Dominant-outlier test. Removing One Medical does not produce a defensible next-dollar ordering: Genesys, Kalpana, Rambus, and Palm/Handspring use incompatible measures—company value, one reported fund-dollar figure, a pre-IPO block, and company consideration. [researcher inference · strong] The honest finding is not “one deal produced X%,” but that the public record cannot tell whether one or several outliers dominated Dunlevie's vehicle economics.

Ranked by historical importance

Ranked by decision-learning value

Evidence boundary and searched denominator

Scope and inclusion rules

The searched universe covers Dunlevie name variants (Bruce Dunlevie, Bruce W. Dunlevie, Bruce Wall Dunlevie); MPAE/Merrill Pickard entities; numbered Benchmark funds, founders' funds, management companies, and alternative vehicles; 1989–2026; the public candidate list; Dunlevie's SEC reporting-person index; issuer filings; company/acquirer releases; court/regulator records; oral histories; books and press; and the already processed written/spoken corpus. A company enters the in-scope candidate ledger if a public source associates Dunlevie with investment advocacy, formation, a board/designation right, investor consulting, a personal/foundation venture, or an aggregator lead. Follow-ons are one company family, Palm and Handspring remain separate legal investments, and Raza Foundries/RMI is one linked but unresolved entity family rather than two assumed cash-flow lines.

Legal reporting status is a discovery tool, not a led-deal count: the SEC query returned 177 filings and at least 27 manually deduplicated issuer families. The public aggregator supplied 24 candidate companies. Their union with source-discovered names produced 33 searched candidate records resolving to 32 company/venture families because Raza Foundries and RMI are separate legal-name searches but one unresolved economic relationship. Five exclusions/non-portfolio controls bring the search ledger to 38 records. SEC reporting-person index · Foundersuite discovery page

Reconciliation

Coverage fraction: not computable. Visible numerators can be named—nine sufficiently reconstructed relationships, 20 supported relationships, or 33 searched candidate records—but the denominator of all Dunlevie opportunities, investments, follow-ons, write-offs, or company families is not public. Rounded recollections of roughly 1,000 or 2,000 opportunities with about ten investments use changing scopes and cannot repair that missing term. Rice video, 19:30–19:45 and 65:26–65:54 [researcher inference · strong] Selection and survivorship bias remain severe.

Historical market, regulatory, and base-rate context

Vehicle economics and incentives

Vehicle map

Incentive comparison with a current LP-backed fund

People, capital, and influence map

Unknown fields remain unknown: no public record consistently identifies analyst work, reference callers, term negotiators, reserve authorizers, board observers, dissenting partners, or distribution decisions across the portfolio.

Comprehensive deal ledger

Material and priority ledger

Discovery candidates retained with unresolved attribution

These 12 rows complete the set of aggregator-only names. Each appeared in the 24-company discovery list but lacks independent evidence for a Dunlevie-specific decision chain. [researcher inference · insufficient public record] They are unverified promotional/discovery leads, not investments, and are retained rather than discarded. Unless stated otherwise, stage, vehicle, check, security, ownership, sourcing, diligence, approval, board work, follow-ons, dilution, outcome, proceeds, and retrospective lesson are not publicly established in this record. Foundersuite discovery page

Explicit exclusions and non-portfolio controls

Ex-ante forensic case studies and decision labs

Case 1: Palm to Handspring — relationship compounding with partnership friction

Freeze: Palm formation, c. 1990–1992. Known: Hawkins was working on pen/mobile-computing ideas and had a relationship path through Geoworks; Tandy-related structuring was possible. Unknown: final device, demand, standards, manufacturing economics, or an exit. Dunlevie challenged Hawkins and Dubinsky to form an independent company and eventually to build their own smaller computer after the larger Zoomer concept struggled. [retrospective witness · strong] Product simplification belongs to Hawkins and the operating team, not Dunlevie. Hawkins, printed pp. 12–13 · Dubinsky part 1, printed pp. 33–34

Freeze: Handspring, 1998. Known: Hawkins, Dubinsky, and Colligan had built Palm; 3Com would license the platform; an application ecosystem existed. Unknown: check, ownership, hardware margins, channel economics, license dependence, and competitive response. Benchmark and KPCB backed the repeat team, with Dunlevie and Doerr named board representatives. 1998 funding release

Outcome revealed. Palm sold to U.S. Robotics for reported company consideration near $44m. Handspring later combined with Palm; Benchmark II held 4,533,214 Handspring shares while Dunlevie held 381,086 separately. Dubinsky also recalls that Dunlevie could not always get partner support, disconfirming a lone-decider narrative. Palm sale · Handspring 13D

Decision lab — analyst-created, no hindsight.

• Would you invest then? Require a view on whether the repeat team is learning faster than platform owners and hardware incumbents.
• What would kill the deal? No sharp scope reduction, unacceptable platform-license dependence, or partner refusal to reserve follow-on capital.
• What ownership/terms fit the risk? Seek enough ownership and a board seat to justify company-building work; protect IP/license continuity without assuming control over 3Com.
• What evidence unlocks the next check? Working prototype, usage frequency, developer pull, unit economics, manufacturing milestones, and explicit partner vote.
• Actual decision and outcome: MPAE supported Palm; Benchmark II/KPCB supported Handspring with board representation; both generated strategic transactions, but cash-on-cash returns remain unproved.

Case 2: Rambus — deep governance and a missing realization ledger

Freeze: 1990–1993. Known: a founder-led semiconductor-memory interface company required patient technical development and ecosystem adoption; Dunlevie joined at formation. By the 1993 Series C, MPAE V, Mohr Davidow, and KPCB each bought 322,628 shares at $3. Unknown: licensing adoption, standards conflict, time to revenue, future capital, and exact board thesis. Rambus prospectus, related-party section

Outcome revealed. MPAE V held 2,491,516 shares before the 1997 IPO; Dunlevie served about 21 years and became chair. The company later restated results over options-accounting problems. A complaint's personal allegations were dismissed as conclusory/unsupported; the FTC standards matter named Rambus, not Dunlevie. Rambus 2007 10-K · Kelley dismissal · FTC docket

Decision lab — analyst-created, no hindsight.

• Would you invest then? Only if technical differentiation and licensing enforceability survive independent expert review.
• What would kill the deal? Standards dependence without defensible adoption, unbounded litigation strategy, or governance controls unable to monitor option/accounting practices.
• What ownership/terms fit the risk? Milestone-linked financing, board/audit rights, and enough ownership for a long development cycle.
• What evidence unlocks the next check? Validated interfaces, licensees, standards strategy, cash runway, and governance-control testing.
• Actual decision and outcome: MPAE V invested and Dunlevie served long-term; public equity value was created, but the fund's realization and the governance counterfactual are unavailable.

Case 3: Genesys — consulting labor became fund consideration

Freeze: 1996. Known: Genesys was raising preferred rounds; Benchmark could provide both capital and hands-on consulting. Unknown: internal sponsor, exact operating work, commercial pipeline, price discipline, and downside memo. Benchmark bought Series A and B; the company granted Benchmark a 420,282-share warrant, exercisable at $5.9483, in exchange for Dunlevie's consulting services. Genesys prospectus, printed pp. 54–57

Outcome revealed. Benchmark's 1,897,368-share post-IPO block was 9.9%; the broader 1,947,368-share/10.1% Dunlevie-attributed total included 50,000 personal options. Alcatel later announced an approximately $1.5b company acquisition. [researcher inference · strong] The structure supports an active-service mechanism more directly than a generic board biography, but cannot reveal whether consulting caused the outcome or what the fund realized. Genesys prospectus

Decision lab — analyst-created, no hindsight.

• Would you invest then? Yes only if the fund can specify which operating bottleneck its consulting can remove.
• What would kill the deal? Service demands exceeding partner capacity, compensation misaligned with all shareholders, or no measurable commercial milestone.
• What ownership/terms fit the risk? Separate cash investment economics from service consideration; approve related-party terms transparently.
• What evidence unlocks the next check? Customer conversion, product reliability, hiring outcomes, and a record of the consulting deliverables.
• Actual decision and outcome: Benchmark invested; Dunlevie's services earned a fund warrant; the company went public and was sold, with final fund proceeds unknown.

Case 4: One Medical and ServiceSource — long board service, different appointment evidence and outcomes

Freeze: 2004–2007 entries. Known: Benchmark V obtained large positions; ServiceSource designated Dunlevie to its board, while One Medical shows him as a Benchmark-affiliated director from 2007 and later lead independent but does not state his appointment route. ServiceSource sold renewal-management services; One Medical sought to redesign primary care. Unknown: initial checks, price, ownership targets, IC votes, One Medical's appointment route, and exit plans. Later prospectuses document 19.4% ServiceSource and 13.0% One Medical positions. ServiceSource prospectus · 1Life prospectus

Decision path. At ServiceSource, Dunlevie became lead independent and participated in a CEO transition; in 2019 more votes opposed than supported him, yet the board declined his conditional resignation. At One Medical, he supported adding Carlyle's sector-specific growth capital; Amazon later acquired the company. ServiceSource transition release · 2019 vote · Carlyle announcement

Decision lab — analyst-created, no hindsight.

• Would you invest then? Compare recurring-revenue quality, customer concentration, regulatory exposure, and whether the board can recruit the missing capability.
• What would kill the deal? ServiceSource: weak renewal economics or customer concentration; One Medical: clinical/regulatory failures or expansion that breaks care quality.
• What ownership/terms fit the risk? A meaningful minority stake plus designation rights, with explicit CEO-transition and later-capital procedures.
• What evidence unlocks the next check? Cohort retention, unit economics, customer/patient outcomes, leadership bench, and dilution-adjusted ownership.
• Actual decision and outcome: Benchmark V held meaningful stakes and Dunlevie governed both; One Medical achieved a large strategic sale, while ServiceSource sold for about $131m after public governance criticism. Fund-level comparative returns remain unknown.

Case 5: WeWork — category insight without an economic or control bridge

Freeze: 2012. Known: WeWork locations displayed rapid growth, visible user energy, and community; Dunlevie compared the encounter to eBay's emergent behavior. Unknown: durable unit economics, recession occupancy, lease-duration mismatch, build-out capital, price, and control trajectory. The analogy generated a hypothesis but did not identify a transferable network-effect mechanism. [researcher inference · strong] Forbes, paragraphs beginning “Three more locations”

Freeze: 2014–2016. Known: reported growth supported demand; Regus offered a public valuation comparison; long leases, short commitments, cash use, competition, and supervoting governance were visible. Dunlevie reportedly opposed Neumann's 2014 supervote proposal but voted for it, and publicly argued user expectations had changed. Bloomberg compilation, PDF p. 172 · New Yorker, governance passages

Freeze: 2019–2021. The IPO filing exposed losses, related-party arrangements, Benchmark's shared minority block, and Neumann control. After the IPO failed, the board removed Neumann, formed a two-member special committee, and later disputed SoftBank's tender. A Delaware opinion documents authority and alternatives; settlement later produced a tender, not a merits judgment. S-1 · Chancery opinion, pp. 7 onward · 2023 proxy, settlement description

Outcome revealed. Dunlevie resigned in May 2023; WeWork filed Chapter 11 in November. A real coworking demand trend may have survived while the company's capital structure and governance failed. [researcher inference · moderate] resignation 8-K · bankruptcy 8-K

Failure decision lab — analyst-created, no hindsight.

• Would you invest then? Only after separating category demand, company capture, unit economics, price, lease duration, and control.
• What would kill the deal? Negative mature-location cash economics, uncontrolled related-party transactions, supervotes without sunset, or financing that removes early-investor leverage.
• What ownership/terms fit the risk? Board/designation rights, supervote sunset, related-party approval, leverage/lease guardrails, milestone financing, information rights, and independent-committee authority.
• What evidence unlocks the next check? Mature cohorts through a downturn, contribution margins after build-out, churn, cash conversion, conflict audit, and control-compliant budgets.
• Actual decision and outcome: Benchmark VII invested and Dunlevie governed for almost eleven years; objections did not prevent supervotes, later SoftBank capital changed power, a committee litigated, and the company entered bankruptcy. Exact fund/trust loss or recovery remains unproved.

Case 6: Tabula and Toys “R” Us — two ways to stop

Freeze: Tabula, March 2011. Known: a $108m Series D, a claimed $70b market, purported breakthrough architecture, and experienced management. Unknown: Benchmark's check, performance versus incumbents, customer conversion, gross margins, burn, board rights, and maximum reserve. The round was led by Crosslink and DAG; Benchmark participated. financing release

Tabula outcome. A 2015 specialist report said the company planned to close after raising more than $200m and failing to make significant market impact. The record cannot isolate technical feasibility, timing, productization, sales, price, or financing as the decisive error. [researcher inference · insufficient public record] EE Times

Failure decision lab — analyst-created, no hindsight.

• Would you invest then? Only after independent benchmarks and signed customer-development evidence justify the capital intensity.
• What would kill the deal? Missed performance/power/yield targets, no design wins, customer delays, or a financing need above the precommitted reserve ceiling.
• What ownership/terms fit the risk? Tranched capital, technical milestones, board information rights, and downside preferences sized to a hardware shutdown path.
• What evidence unlocks the next check? Tape-out results, production yield, design wins, customer revenue, gross margin, and runway to the next financing-independent milestone.
• Actual decision and outcome: Benchmark followed into the Series D; the company later shut down. Benchmark's exact capital/recovery is unknown.

Freeze: Toys “R” Us online, April 1999. Known: an incumbent brand, stores, vendors, planned 500,000-square-foot fulfillment, a separate Silicon Valley subsidiary, and a publicly described major Benchmark commitment. Unknown: internal cooperation, actual funded capital, security, ownership, rights, and truthfulness of operating representations. 1999 release

Stop decision and outcome. The New Yorker later reports that internal resistance and misleading cooperation claims led Dunlevie to cancel despite a guaranteed $100m payout. [retrospective witness · moderate] The account supports willingness to stop on organizational integrity, not exact economics. New Yorker, Toys passage

Stop-decision lab — analyst-created, no hindsight.

• Would you invest then? Only after testing whether the incumbent will cede people, data, inventory, and decision rights to the digital unit.
• What would kill the deal? False operating assurances, vetoes by legacy management, or inability to align inventory/fulfillment incentives.
• What ownership/terms fit the risk? Staged funding, representations, termination rights, control over the digital subsidiary, and compensation tied to cross-channel execution.
• What evidence unlocks the next check? Named accountable executives, inventory/data access, pilot fulfillment metrics, vendor integration, and board-certified cooperation.
• Actual decision and outcome: Benchmark announced a large commitment; Dunlevie reportedly cancelled. No primary agreement proves funded capital, guarantee mechanics, or realized opportunity cost.

Portfolio-level forensics

The observed denominator is 33 researched candidate records resolving to 32 company/venture families, not a full career portfolio. Nine relationships are sufficiently reconstructed; 11 are partial or specifically unresolved, 12 are unverified aggregator-only leads, and the visible set is biased toward public filings, public advocacy, and known outcomes.

Fund-level controls

• For one UC LP through 2003-03-31, MPAE V — reported: 5.51x and reported: 46.1% net IRR; period and net basis are known, fund contribution by deal is not. UC schedule
• Benchmark I claims—Metrick–Yasuda reported: 42x; the Amity attendee reported: 80x+; Forbes reported: 91.8x (approximately, at the headline level); Rice reported: 100x—cannot be averaged or reconciled. [researcher inference · strong] They use incompatible bases, not four measurements of one audited number. Metrick–Yasuda · Amity attendee note · Forbes · Rice
• The same academic source gives Benchmark IV reported: 0.2% net IRR on its stated basis, warning against extending first-fund fame across vintages or to one partner. Metrick–Yasuda, printed pp. 89–90
• No entry/exit valuation shortcut is used. Without ownership, dilution, cash flows, timing, fees, and distributions, a company multiple is not a fund MOIC.

Failure, miss, and anti-portfolio postmortems

Verified non-attributions also discipline the record: eBay is Kagle-led, Uber is Gurley-led, and ARM has no supported Dunlevie/MPAE link. This is not modesty theater; it prevents partnership fame from manufacturing an individual hit rate.

Uncertainty and retrieval ledger

The remaining uncertainty is structural, not an invitation to interpolate. Until these records appear, the defensible conclusion is that Dunlevie's public record shows strong early access, unusually durable board work, repeat-founder relationship capital, and several important outcomes—alongside missing economics, team-credit ambiguity, and failures that expose the limits of people-first selection and persuasion-first governance. [researcher inference · strong]

Coverage and evidence base

Deterministic reconciliation of the mapped corpus

The WRITINGS denominator is the 30 records under Direct voice, interview-derived cases, and written record in the source map. It is deliberately broader than “things Dunlevie wrote”: no public essay, letter, memo, book, or shareholder/LP communication demonstrably authored by him surfaced. The mapped set instead contains reported interviews, edited or relayed interview material, teaching cases, prepared financing and board statements, a joint legal position, a jointly signed policy declaration, and two low-priority philanthropic statements. Calling this an authored corpus would overstate both voice and control.

Priority reconciliation is also exact: 15 ★★★, 13 ★★, and 2 ★ records. All 28 ★★★/★★ items have notes and a subsection; the two ★ controls are included so that the denominator cannot quietly discard values evidence that does not fit the venture narrative. An adversarial cross-source control is incorporated throughout the synthesis rather than counted as a 31st source.

Reliability and attribution boundary

The safest unit of direct evidence is a clearly attributed sentence, not the surrounding publication. Reported interviews are investor-stated only for Dunlevie’s exact words; journalistic causation and context remain contemporaneous record or retrospective witness. Financing and personnel-release quotes are prepared advocacy. The Cato declaration is joint; the 2020 WeWork position belongs to a two-member special committee and its counsel; the philanthropic statements involve joint family capital. Acquired’s hosts relay an unpublished private interview, and the Stanford cases were written by case authors. None can be recast as a Dunlevie-authored essay.

The corpus is selected by publication incentives: organizations announce raises, hires, disputes, and gifts, but rarely publish rejected deals, internal dissent, pricing limits, follow-on refusals, or quiet failures. Repeated emphasis on people, large markets, or growth is therefore a prepared-statement motif, not proof of a stable investing system. Each source analysis below tests rhetoric against contemporaneous risks or later outcomes where the record permits.

Source-by-source analysis

1. Benchmark Capital in Europe: Bringing Silicon Valley Venture Capital to the Continent, E218 (2006; revised 2008)

• Context, authorship, access — contemporaneous record · strong: Stanford identifies Katherine Bose and John Glynn Jr. as case authors and frames Dunlevie as protagonist in an early-2007 choice among further expansion, retaining the United Kingdom/Israel structure, or making those offices independent. Only the official synopsis and publication metadata were accessible; the 19-page licensed body and exhibits were not. Stanford catalog, synopsis and metadata
• Decision problem — researcher inference · moderate: The named alternatives make internationalization a vehicle-and-governance problem, not simply a market-size thesis. Decision rights, partner cohesion, local access, board-service capacity, and brand control are plausible variables, but none may be attributed to Dunlevie without the case body. Stanford E218 catalog, synopsis
• Evidence and assumptions — contemporaneous record · provisional: The catalog establishes the geographies, alternatives, and Dunlevie’s centrality; it does not expose his preferred option, partner objections, economics, portfolio results, or resolution. A 2006 publication date paired with an “early 2007” setting also needs title-page verification. Stanford E218 catalog, synopsis and publication metadata
• Operational implication — researcher inference · provisional: A modern IC considering a satellite office should compare local sourcing gains against coordination cost, authority ambiguity, partner-quality dilution, and the option of an independent affiliate. This is a decision exercise derived from the synopsis, not Dunlevie doctrine. Stanford E218 catalog, named alternatives
• Evolution and retrieval boundary — researcher inference · strong: E218 appears to revisit E148’s European experiment, but its existence alone proves neither failure nor changed belief. A licensed copy is required to map each interview-derived claim and exhibit by printed page. Stanford E218 catalog, availability and synopsis

2. Benchmark Capital in Europe, E148 (2003)

• Context, authorship, access — contemporaneous record · strong: Stanford’s catalog names Jamie Earle and John Glynn Jr. as authors and calls Dunlevie the Benchmark partner who “conceived and championed” the European practice. The official abstract was available; the licensed body was not. Stanford catalog, abstract and availability
• Argument available to the public — contemporaneous record · moderate: The catalog says the case examines why Benchmark opened in Europe, how the operation was organized, and how the decision played out. That is a problem statement, not a readable reasoning chain or verdict. Stanford E148 catalog, abstract
• Attribution and assumptions — researcher inference · strong: “Conceived and championed” supports personal sponsorship, not sole authority, authorship, or credit for every European deal. The inaccessible body may combine interviews, author synthesis, and outside research; those layers must be separated before use. Stanford E148 catalog, authorship and abstract
• Decision implication — researcher inference · provisional: The high-value question is whether Benchmark’s edge was portable or depended on a dense Silicon Valley network and small-team rituals. The abstract cannot answer it, so no success/failure lesson is assigned. Stanford E148 catalog, abstract and access boundary
• Evolution and retrieval boundary — researcher inference · strong: E148 is the earliest mapped institutional treatment of Dunlevie’s international initiative; E218 later reopens structure. Legitimate page-complete access is needed for dates, votes, staffing, economics, exhibits, and counterfactual opportunity cost. Stanford E148 catalog · Stanford E218 catalog

3. Finding the Next Starbucks, pp. 251–252 (2006 hardcover)

• Context, authorship, access — contemporaneous record · strong: Michael Moe authored the book; bibliographic metadata routes Dunlevie material to printed pp. 251–252 in the 2006 hardcover, while current publisher pages also show a 2007 paperback/e-book. The relevant pages did not render in the legitimate preview. Google Books, routed p. 251 and edition metadata
• Argument and evidence — researcher inference · insufficient public record: No interview question, answer, example, or decision rule was readable. Moe’s “four Ps” framework and publisher description are Moe’s ideas, not a proxy for Dunlevie’s. Google Books, routed pp. 251–252 and preview boundary
• Attribution — researcher inference · strong: Even a future readable passage would be edited interview material inside another author’s book, not a Dunlevie-authored work. Search snippets and retailer summaries are excluded. Google Books, author and edition metadata
• Potential decision value — researcher inference · provisional: A 2006 interview could test Dunlevie’s view after Benchmark’s first decade and the dot-com correction, but that is a retrieval rationale only. Google Books, routed pp. 251–252
• Evolution and retrieval boundary — researcher inference · strong: The 2006/2007 dates are edition-specific rather than contradictory. A licensed hardcover or page-complete edition must supply the interview heading, exact pages, editing context, and decision examples before this record can shape the synthesis. Google Books, hardcover route and edition metadata

4. How Venture Capitalists Are Deforming Capitalism (2020)

• Context and access — contemporaneous record · strong: Charles Duhigg’s reported investigation is fully accessible and combines documents, interviews, former employees, and Dunlevie’s phone responses. It is adversarial journalism, not board minutes or a Dunlevie essay. New Yorker, paragraphs beginning “Doing so” through “Another former”
• Argument — investor-stated · strong: Dunlevie describes venture board work as persuasion and says investors are expected to remain engaged with difficult founders. The mechanism is relationship continuity rather than command-and-control; the article’s account of WeWork tests its limits. New Yorker, paragraph beginning “Another former”
• Evidence and contradiction — documented behavior · moderate: The report says he opposed Neumann’s 2014 supervoting proposal, invoked the corrupting effects of power, then voted with the board to grant it. Voiced objection and exercised formal authority diverged. New Yorker, paragraphs beginning “Doing so” and “Another former”
• Decision rule and limit — researcher inference · moderate: Staying “in the ring” preserves influence only while explicit control, disclosure, capital, and conduct thresholds remain enforceable. After repeated breaches, persuasion can become delay. The rule is an analyst safeguard, not his stated escalation policy. New Yorker, paragraphs beginning “Doing so” and “Another former”
• Biography and evolution — retrospective witness · moderate: The article contrasts reported willingness to abandon the Toys “R” Us arrangement on principle with continued WeWork service, S-1 approval, Neumann’s removal, and later special-committee litigation. Different facts may justify different choices; the absent published threshold is the unresolved contradiction. New Yorker, Toys “R” Us and WeWork governance passages

5. Benchmark Capital, Part I (2022)

• Context, authorship, access — retrospective witness · moderate: Acquired’s full episode and source list are public, and the hosts say they interviewed Dunlevie. His underlying interview audio/transcript is not released, so host paraphrases and embedded recollections are not verifiable direct voice. Acquired, 00:28:53–00:37:32 and source list
• Argument — retrospective witness · moderate: The hosts reconstruct a founding logic of more board capacity, broader portfolio diversification, and reduced personal risk, then describe Dunlevie recruiting Kevin Harvey for founder-like temperament. Acquired, 00:30:34–00:37:32
• Mechanism and assumptions — researcher inference · moderate: Equal economics can reduce internal ownership of deals and invite mutual help only when partner selection, monitoring, shared workload, and anti-free-rider norms work. Compensation design alone is not causal proof. Acquired, 01:37:00–01:38:48
• Decision and biography evidence — retrospective witness · moderate: The episode supports Dunlevie’s role in team formation and later European expansion, but eBay narration remains primarily a Bob Kagle/firm story and should not be reassigned to Dunlevie. Acquired, 00:58:09–00:59:52 and about 02:15:11
• Contradiction and limit — researcher inference · provisional: The polished origin story and fund-return superlatives risk survivor selection. Private interviews, ambiguous succession dates, and absent fund cash flows prevent treating the episode as audited economics or a literal founding memo. Acquired Part I, source list and 00:28:53–00:37:32

6. Benchmark Part II: The Dinner (2022)

• Context and attribution — retrospective witness · moderate: Dunlevie is not a guest. Host Ben Gilbert relays his rule against pre-selling deals; the underlying interview remains unpublished. Current partners’ descriptions establish 2022 practice, not automatic 1995 authorship. Acquired, 12:16–15:14
• Argument and mechanism — researcher inference · moderate: The rule aims to let each partner encounter the founder and evidence before a champion creates an anchor or coalition. One-table dinner, open-ended Monday debate, and avoidance of conventional memos are presented as complementary anti-hierarchy practices. Acquired, 06:32–15:14
• Evidence and assumptions — documented behavior · moderate: Later-generation partners describe these rituals, but the episode does not show how exceptions, forecasts, dissent, or postmortems are preserved, nor who originated each norm. Acquired Part II, 06:32–15:14
• Decision implication — researcher inference · moderate: Ban advocacy before independent founder exposure, then preserve written forecasts, kill criteria, and a decision log. Otherwise debiasing the room can destroy institutional memory. Acquired Part II, 12:16–15:14
• Evolution and contradiction — researcher inference · moderate: The anti-hierarchy account coexists with selective access, a tiny partnership, and succession pipelines. Equal voice is an aspiration, not proof that power and information are perfectly equal. Acquired Part II, 06:32–19:08

7. Venture capital industry facing cash crunch and investor reluctance (2001)

• Context and access — contemporaneous record · strong: The Irish Times interviewed Dunlevie during the post-dot-com financing contraction and juxtaposed Benchmark trophy companies with weak industry-wide outcomes. Irish Times, opening Benchmark-office paragraphs
• Argument — investor-stated · strong: His “alchemy” description acknowledges uncertainty; his satire of implausibly narrow dot-coms links excess capital to weak company formation, visible failure, and LP disillusionment. Irish Times, paragraph beginning “Look, people were setting up dotcoms”
• Evidence and assumptions — contemporaneous record · moderate: Industry totals and return comparisons are secondary reporting, while the eBay appreciation omits vehicle, dilution, distributions, fees, and timing. Winner artifacts cannot substitute for a portfolio denominator. Irish Times, opening office and industry-return sections
• Decision rule — researcher inference · moderate: In a funding glut, require a durable economic reason for segmentation and discount novelty created only by cheap capital. The precise screen is inferred because no pricing, reserves, or follow-on policy appears. Irish Times, paragraph beginning “Look, people were setting up dotcoms”
• Evolution and contradiction — researcher inference · provisional: Post-crash ridicule is easier than ex-ante rejection. The record does not show whether Benchmark tightened its process or whether its own weak boom-era deals matched the critique. Irish Times, Dunlevie passages and reported portfolio context

8. Joe Blow StakesHisClaim.com (1999)

• Context and access — contemporaneous record · strong: Fortune captured Dunlevie during the Internet land-grab, when company formation accelerated and stable category expertise was scarce. Fortune, paragraph beginning “Passion, in fact”
• Argument — investor-stated · strong: He prioritizes leadership, cognitive strength, and real-time strategic adaptability over prior Internet tenure. The logic is conditional: in a truly new domain, learning rate can be more informative than credentials. Fortune, same paragraph
• Evidence and assumptions — contemporaneous record · moderate: The surrounding boom explains the need for adaptability but supplies no reliable measure. Charisma, improvisation, and undisciplined pivots can mimic learning. Fortune, 1999 boom context and paragraph beginning “Passion, in fact”
• Decision rule — researcher inference · moderate: Test adaptability through prior course corrections, reference checks, honest updating, recruiting, and evidence-based changes—not a founder’s self-description. Fortune, paragraph beginning “Passion, in fact”
• Evolution and contradiction — researcher inference · moderate: The people-first rule precedes the 2001 category critique and WeWork. It survives only with market and control tests; founder brilliance cannot compensate for every economic or authority defect. Fortune, 1999 Dunlevie paragraph · New Yorker, 2020 WeWork governance passages

9. Inside the Phenomenal Rise of WeWork (2014)

• Context and access — contemporaneous record · strong: Forbes reports that Dunlevie flew to New York in 2012 and compared the live WeWork phenomenon with his 1997 eBay encounter. The 2014 record limits later hindsight. Forbes, paragraph beginning “Three more locations went up in 2012”
• Argument — investor-stated · strong: The attraction was an emergent behavior he could not fully name: visible user energy/community → possible durable category advantage → willingness to invest before the mechanism was fully legible. Forbes, same paragraph
• Evidence and assumptions — contemporaneous record · strong: Rapid growth supported demand, while long leases, short member commitments, capital intensity, service dilution, competition, recession exposure, and valuation versus Regus were observable risks in the same source. Forbes, closing risk discussion
• Decision rule — researcher inference · strong: Analogy is a hypothesis generator: identify the causal invariant and falsifier before it becomes evidence. eBay’s network effects and asset-light structure did not transfer automatically. Forbes, eBay analogy and closing WeWork risk discussion
• Evolution and contradiction — documented behavior · moderate: Early enthusiasm coexists with later profitability doubts and objections. The source does not reveal original unit-economics thresholds, terms, partner vote, or follow-on conditions. Forbes, 2012 reaction · New Yorker, WeWork governance passages

10. Is This the Office of the Future or a $5 Billion Waste of Space? (2015)

• Context and custody — contemporaneous record · moderate: Andrew Rice’s Bloomberg article is page-complete only in a third-party compilation; the accurate locator is PDF p. 172, right column, final paragraph. Bloomberg compilation, PDF p. 172
• Argument — investor-stated · strong: Dunlevie argues that younger users’ exposure to communal flexible work produced a durable shift in expectations: cohort experience → persistent preference → demand. Bloomberg compilation, PDF p. 172, right column
• Evidence and assumptions — contemporaneous record · moderate: Reported member/location growth supports demand; Regus, capital consumption, long obligations, and the $5 billion valuation expose the missing bridge from preference to economics. Bloomberg compilation, PDF p. 172
• Decision rule — researcher inference · strong: Separate whether behavior is changing, whether this company captures it, and whether price already assumes successful capture. Dunlevie’s statement addresses only the first. Bloomberg compilation, PDF p. 172, right column
• Evolution and contradiction — researcher inference · moderate: WeWork’s later Chapter 11 does not by itself disprove coworking demand; likewise, demand evidence cannot validate the 2015 valuation or capital structure. Bloomberg compilation, PDF p. 172 · WeWork 2023 Form 8-K

11. Adam Neumann’s $16 Billion Neo-Utopian Play to Turn WeWork Into WeWorld (2016)

• Context and access — contemporaneous record · moderate: The public page exposes title, dek, issue metadata, image caption, author, and opening paragraphs; the remainder, including the mapped Dunlevie passage, is subscriber-restricted. Fast Company, public opening and issue metadata
• Argument — researcher inference · insufficient public record: The accessible opening establishes a reported $16 billion valuation and expansion beyond offices, but no Dunlevie rationale, objection, or change of view can be recovered responsibly. Fast Company, public opening and issue metadata
• Evidence and assumptions — contemporaneous record · insufficient public record: The accessible page does not expose the mapped Dunlevie passage, so any summary of his skepticism or on-site reaction remains a retrieval lead rather than evidence. Fast Company, subscriber boundary after the public opening
• Decision implication — researcher inference · provisional: A licensed passage could distinguish whether skepticism concerned valuation, scope, founder behavior, or economics and whether it preceded a vote. No rule is inferred from metadata. Fast Company, public metadata and access boundary
• Evolution and retrieval boundary — researcher inference · strong: Retrieve issue No. 204 through a licensed route and verify the reported print p. 98 locator. Search snippets cannot substitute. Fast Company, issue metadata

12. Publishing in the Remix Era (2011)

• Context and access — contemporaneous record · strong: Digiday’s full article concerns Bo.lt—not Klip—and describes a product for copying, modifying, and redistributing pages amid live copyright, attribution, advertising, and publisher-control risk. Digiday, closing section
• Argument — investor-stated · strong: Dunlevie says users increasingly want to forward, comment, and author rather than consume passively; the reporter says he expected the model to change. Trend conviction plus proven builders supported product iteration. Digiday, paragraph beginning “The service has $5 million in backing”
• Evidence and assumptions — contemporaneous record · moderate: The $5 million backing and team history are reported; ownership, milestones, legal tolerance, outcome, and value-capture mechanism are absent. A correct participation trend does not validate Bo.lt. Digiday, paragraph beginning “The service has $5 million in backing”
• Decision rule — researcher inference · moderate: Back team-plus-trend optionality only when price, runway, learning milestones, and legal risk allow finite experiments. “The model will change” without stop criteria is unfalsifiable. Digiday, product-risk discussion and closing paragraph
• Evolution and contradiction — researcher inference · moderate: This non-trophy case controls category hindsight: the article supports a participation thesis while leaving Bo.lt’s company outcome unresolved. Digiday, closing section

13. The WeWork Manifesto (2018)

• Context and access — contemporaneous record · strong: The WRAL archive preserves David Gelles’s syndicated New York Times feature and names multiple adjacent initiatives visible before the failed IPO. WRAL/NYT syndication, paragraph beginning “Is it all a bit much”
• Argument — investor-stated · strong: Dunlevie says he argued that the widening ambition could be too much while describing founders like Neumann as resistant to investor direction; he still calls the workplace shift an early major phenomenon. WRAL, “Is it all a bit much” and “But so far nothing” paragraphs
• Evidence and assumptions — contemporaneous record · moderate: WeLive, Rise, Meetup, Flatiron School, WeGrow, valuation criticism, and durability questions show adjacent-expansion risk was public by 2018. The article does not identify votes, funding conditions, or economics. WRAL/NYT, adjacent-initiative and criticism sections
• Decision rule — researcher inference · strong: Separate category conviction from company scope, governance, and capital allocation; assign explicit authority and stop criteria when adjacent bets can threaten the core. WRAL/NYT, paragraphs beginning “Is it all a bit much” and “But so far nothing”
• Evolution and contradiction — documented behavior · moderate: Public skepticism and continued category conviction are compatible; continued support after persuasion failed requires an unpublished explanation of how risk was bounded. WRAL/NYT, Dunlevie passages · New Yorker, later governance account

14. Eastside Story (2000)

• Context and access — contemporaneous record · strong: Stanford Magazine profiled Eastside College Preparatory while Dunlevie served on its board, providing a mission-driven rather than financial test of his people judgment. Stanford Magazine, paragraphs beginning “Financial backing has not come easily”
• Argument — investor-stated · strong: He values Chris Bischof’s entrepreneurial caliber, commitment, and mission while warning that an eighteen-hour-day founder/faculty model is not sustainable. Admiration and key-person risk coexist. Stanford Magazine, same and following paragraph
• Evidence and assumptions — retrospective witness · moderate: Fundraising, family trust, high expectations, and institution-building support the profile; a subjective top-ten ranking has no named denominator or independent longitudinal comparison. Stanford Magazine, Eastside operating context and Dunlevie paragraphs
• Decision rule — researcher inference · strong: Treat heroic labor as early momentum and a scaling liability; require institutionalized processes, delegation, and succession before assuming durability. Stanford Magazine, Dunlevie sustainability warning
• Evolution and boundary — researcher inference · moderate: The nonprofit example broadens founder evaluation beyond profit but cannot establish a venture rule; capital, regulation, outcomes, and exit logic differ. Stanford Magazine, school/startup analogy and board context

15. Declaration of Independence: End Corporate Welfare (1998)

• Context, authorship, access — contemporaneous record · strong: T. J. Rodgers authored the policy paper; Dunlevie appears among signatories to the appended declaration. Signature proves support for the joint pledge, not authorship or agreement with every surrounding example. Cato paper, printed/PDF pp. 20–21
• Argument — investor-stated · moderate: The pledge supports a substantial, evenhanded reduction in corporate subsidies even if a signer’s own organization loses funding. The narrow principle is symmetrical rule application. Cato paper, joint declaration, printed/PDF p. 20
• Evidence and assumptions — contemporaneous record · provisional: The document contains no evidence about solicitation, drafting, Benchmark deliberation, portfolio subsidy exposure, or later consistency. Cato paper, declaration and signatory list, printed/PDF pp. 20–21
• Decision rule — researcher inference · moderate: Prefer generally applicable capital-allocation rules and test willingness to accept downside on one’s own side. This does not answer how to advise a portfolio company already using public programs. Cato paper, joint declaration, printed/PDF p. 20
• Evolution and contradiction — researcher inference · insufficient public record: No later Dunlevie policy record was processed, so the 1998 position cannot establish a durable macro philosophy. Cato paper, 1998 joint declaration and signatory list

16. Handspring funding and 3Com licensing statement (1998)

• Context and access — contemporaneous record · strong: A historical reproduction preserves the company release announcing Benchmark/Kleiner Perkins financing and a 3Com Palm-platform license. It does not disclose checks, securities, ownership, or vehicles. Original-release section, opening transaction paragraphs
• Argument — investor-stated · moderate: Dunlevie’s prepared statement backs the operators associated with Palm’s category creation, an existing application ecosystem, the platform license, and consumer handheld computing—not merely a new product specification. Original-release section, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · moderate: The release identifies Hawkins, Dubinsky, and Colligan as the operating team and names Dunlevie as Benchmark’s board representative. Forecasts of category leadership remain promotion. Handspring release, operating-team and Dunlevie paragraphs
• Decision rule — documented behavior · moderate: Prior trust and demonstrated category-building ability can support repeat-founder underwriting; platform rights and ecosystem access are material enabling evidence. Handspring release, transaction and Dunlevie paragraphs
• Evolution and contradiction — researcher inference · strong: The release says the investors were working again with Palm’s team but does not identify Palm’s earlier vehicle. A later Schedule 13D assigns Handspring/Palm shares to Benchmark II entities, while no processed primary record assigns Palm to MPAE; Palm’s vehicle therefore remains unresolved rather than pooled with Benchmark II. Handspring release, Dunlevie paragraph · Handspring/Palm Schedule 13D, reporting-person ownership tables

17. Impresse raises more than $62 million (1999)

• Context and access — contemporaneous record · strong: The financing report announces more than $62 million of combined equity and committed debt in a third round, but does not split the instruments or Benchmark’s contribution. InternetNews, financing paragraphs
• Argument — investor-stated · moderate: Dunlevie’s prepared screen combines seasoned management, deep technology, domain expertise, employee quality, and a growing B2B workflow. It is a conjunction, not a market-size slogan. InternetNews, Dunlevie financing paragraph
• Evidence and assumptions — contemporaneous record · provisional: “Undisputed leader” and a hot market are promotional assertions; no independent market share, retention, unit economics, valuation, or board attribution appears. InternetNews, financing and Dunlevie paragraphs
• Decision rule — researcher inference · moderate: At scale-up, require specialist team quality plus commercial proof and a capital-to-milestone plan; technical depth and domain language alone do not prove durable demand. InternetNews, financing size, uses, and Dunlevie screen
• Evolution and contradiction — contemporaneous record · moderate: Impresse shut down in 2001 and its IP/customer relationships were purchased from an assignee for creditors, contradicting leadership rhetoric without isolating cause or fund loss. Shutdown report · asset-purchase report

18. Raza Foundries formation (2000)

• Context and access — contemporaneous record · strong: EDN describes an operating-intensive platform for communications ventures, with Benchmark as lead investor and eight other reported partner companies. No check, fund, security, or ownership is disclosed. EDN, incubator comparison and closing paragraphs
• Argument — investor-stated · moderate: Dunlevie favors hard technical projects when an expert shared platform can impose execution discipline and shorten time to market. EDN, Dunlevie closing statement
• Evidence and assumptions — contemporaneous record · provisional: Engineering, management, domain staff, and an early portfolio sale support the launch story, but the “exactly right” idea/timing/market claim is promotional and the cross-company resource economics are undisclosed. EDN, platform description, early sale, and Dunlevie paragraph
• Decision rule — researcher inference · moderate: Underwrite a studio/incubator only after allocating staff, IP, conflicts, follow-ons, and accountability by company; shared expertise can reduce technical risk while creating governance complexity. EDN, shared operating-platform description
• Evolution and contradiction — retrospective witness · moderate: A later specialist account describes post-bubble difficulty and consolidation into RMI, directly challenging launch certainty without supporting blanket failure or return math. Light Reading, “RMI’s Long Strange Trip”

19. Tabula secures $108 million (2011)

• Context and access — contemporaneous record · strong: Tabula’s issuer release says Crosslink and DAG led a $108 million Series D and names Benchmark as one of five existing participants. Dunlevie is a public sponsor, not proven round lead, source, or board member. PR Newswire, news highlights and Dunlevie paragraph
• Argument — investor-stated · moderate: His prepared screen joins an asserted $70 billion market, technical discontinuity, and proven management to justify continued support of a capital-intensive semiconductor bet. Tabula release, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: The market is undefined and unsourced; breakthrough, dominance, and inevitability claims lack independent technical validation, customer milestones, price/performance, burn, or kill criteria. Tabula release, news highlights and management forecasts
• Decision rule — researcher inference · strong: Define served market, benchmark technical advantage, identify qualification/customer-conversion gates, and precommit a follow-on ceiling before using market/technology/team as an approval checklist. Tabula release, announced uses and Dunlevie screen
• Evolution and contradiction — contemporaneous record · strong: EE Times reported planned closure in 2015 after more than $200 million raised and little market impact, falsifying dominance rhetoric but not identifying Benchmark’s cash loss or the precise failure mechanism. EE Times, 2015 shutdown report

20. Aquifi changes the computing interface (2014)

• Context and access — contemporaneous record · strong: TechCrunch reports Benchmark and angels invested $9 million in 2012 in a Canesta-veteran team pursuing software-based 3-D tracking through commodity sensors. It does not allocate the round or vehicle. TechCrunch, financing and Dunlevie paragraph
• Argument — investor-stated · moderate: Dunlevie’s prepared statement treats accumulated team learning and a long-lived technical vision as underwriting advantages; the implicit bet is that sensor economics and software capability reached an adoption inflection. TechCrunch, financing and Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: Developer engagement and 2015 enabled devices were forecasts. No accuracy, latency, customer contract, OEM adoption, unit economics, board role, or realized outcome is disclosed. TechCrunch, product roadmap and financing sections
• Decision rule — researcher inference · moderate: Revisit a failed or early technical vision when component cost and implementation architecture change, but require objective performance and channel milestones. TechCrunch, Canesta history, commodity-sensor approach, and roadmap
• Evolution and contradiction — researcher inference · insufficient public record: No independent acquisition, shutdown, or continuing-operation evidence was located, so the record cannot label Aquifi a win, loss, or proof of the pattern. TechCrunch, article boundary and announced 2015 roadmap

21. One Medical/Carlyle investment announcement (2018)

• Context and access — contemporaneous record · strong: Carlyle announced a significant minority commitment of up to $350 million, including primary proceeds for expansion; “up to” is a ceiling, not funded cash. Carlyle release, transaction and Dunlevie paragraphs
• Argument — investor-stated · moderate: Dunlevie’s prepared board statement emphasizes matching expansion capital with an investor possessing healthcare operating experience, not merely maximizing available dollars. Carlyle release, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: Company footprint and customer claims come from promotion; primary/secondary split, Benchmark sell-down, valuation, dilution, milestones, and rights are absent. Carlyle release, company footprint and transaction sections
• Decision rule — researcher inference · moderate: Add a stage-specific capital partner when its domain capabilities and governance contribution fill a real gap; define capital use, authority, expansion gates, and dilution beforehand. Carlyle release, incoming-investor and use-of-proceeds sections
• Evolution and contradiction — contemporaneous record · moderate: Amazon completed the One Medical acquisition in 2023, but company consideration cannot establish Benchmark cost, distributions, or Dunlevie economics. Amazon completion announcement

22. WeWork leadership changes (2019)

• Context and access — contemporaneous record · strong: During the failed-IPO crisis, WeWork announced Neumann’s CEO departure and Minson/Gunningham as co-CEOs; this was crisis succession, not routine transition. WeWork newsroom, opening leadership paragraphs
• Argument — investor-stated · moderate: Dunlevie’s prepared board position endorses continuity, scale, and experienced internal operators. The cited run-rate revenue is annualized activity—not profit, liquidity, cash flow, or audited annual revenue. WeWork leadership statement, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: The release proves public support for the handoff but not sole responsibility, internal probabilities, board alternatives, lease analysis, or governance repair. WeWork leadership statement, succession announcement
• Decision rule — researcher inference · moderate: A crisis succession should specify control changes, interim mandate, runway, governance repair, profitability gates, and permanent-CEO criteria; optimistic continuity language is insufficient. WeWork leadership statement, announced roles and omitted operating terms
• Evolution and contradiction — contemporaneous record · strong: Sandeep Mathrani replaced the co-CEOs in February 2020, and WeWork entered Chapter 11 in 2023. Those facts falsify a durable-succession or broad-trajectory inference without isolating causation. Mathrani appointment · 2023 Form 8-K

23. WeWork Special Committee authority statement (2020)

• Context, authorship, access — contemporaneous record · strong: The statement belongs jointly to Dunlevie and Lew Frankfort; the attached letter was written and signed by counsel for the committee. Legal argument is not Dunlevie-authored prose or an adjudication. Business Wire, opening statement and authority sections
• Argument — contemporaneous record · strong: The committee says the board determined both members independent of SoftBank and Neumann, delegated full board authority, and empowered enforcement of the Master Transaction Agreement for minority holders. Business Wire, opening statement and “The Committee Is Fully Empowered”
• Evidence and assumptions — contemporaneous record · moderate: Board resolutions, contract text, and asserted tendered employee stock support advocacy; committee independence, personal/fund holdings, complete absence of conflict, and SoftBank liability remain distinct questions. Business Wire, authority and tendered-stock sections
• Decision rule — researcher inference · moderate: When a controller is counterparty, use disinterested directors, explicit delegated power, independent advisers, veto/enforcement rights, sequencing, and a written conflict record. Business Wire, “The Committee Is Fully Empowered” and MTA sections
• Evolution and contradiction — contemporaneous record · strong: The parties settled; WeWork’s proxy reports a 2021 tender and dismissal. Negotiated leverage is not a merits judgment validating every committee claim. SoftBank settlement · WeWork 2023 proxy, tender discussion

24. Toys “R” Us online venture statement (1999)

• Context and access — contemporaneous record · strong: The release announces a separate Silicon Valley subsidiary, a strategic Benchmark relationship, and large fulfillment capacity; capital, security, vehicle, governance, and funding status are omitted. Press-release mirror, “Partnership with Benchmark Capital”
• Argument — investor-stated · moderate: Dunlevie calls it Benchmark’s largest commitment and combines incumbent brand, stores, distribution, and suppliers with a separately incentivized e-commerce operation and Benchmark expertise. Press-release mirror, Dunlevie and partnership paragraphs
• Evidence and assumptions — contemporaneous record · provisional: Store, country, vendor, and capacity figures are management claims. The thesis assumes organizational separation can overcome incumbent resistance while preserving legacy advantages. Press-release mirror, company-scale and online-venture paragraphs
• Decision rule — researcher inference · moderate: In incumbent/digital partnerships, make operating cooperation, truthful reporting, authority, incentives, and withdrawal rights explicit; assets alone do not create execution. Press-release mirror, separate-subsidiary and Benchmark-partnership sections
• Evolution and contradiction — retrospective witness · moderate: The New Yorker later reports that Dunlevie cancelled after internal resistance and forfeited a guaranteed payout. Without agreements, funded capital and recovery remain unknown; later retailer bankruptcy is not the outcome of the cancelled 1999 arrangement. New Yorker, Toys “R” Us passage

25. Advanced Equities emerges as a big player (2007)

• Context and access — contemporaneous record · strong: InvestmentNews/Bloomberg describes a private-placement channel for high-risk securities after specialist banking capacity and small-company IPO attractiveness contracted. InvestmentNews, “ready and renewable source of capital” and closing paragraphs
• Argument — investor-stated · moderate: Dunlevie evaluates the channel on timeliness and scale. This is financing-infrastructure judgment, not a company-picking doctrine. InvestmentNews, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: Placement totals and forecasts came from company management; high commissions, suitability, disclosure quality, cap-table effects, and counterparty conflicts are live risks. InvestmentNews, placement totals, forecasts, and business-model discussion
• Decision rule — researcher inference · strong: Assess a financing intermediary on delivery capacity plus disclosure controls, investor suitability, economics, conflicts, reputation, and cap-table consequences—not speed alone. InvestmentNews, capital-channel and economics discussion
• Evolution and contradiction — contemporaneous record · moderate: A 2012 SEC order found misstatements, omissions, and supervisory failures in a later offering. It does not name Dunlevie, prove the conduct existed in 2007, or identify Benchmark involvement, but it is material counterparty-risk evidence. SEC Release No. 9362

26. EnVerv raises $12 million (2011)

• Context and access — contemporaneous record · strong: The reproduced company release says Benchmark led a $12 million Series B with NEA and Walden, but gives no Benchmark check, vehicle, security, ownership, or board role. Release, opening financing paragraph
• Argument — investor-stated · moderate: Dunlevie’s prepared thesis joins a mission-critical smart-grid need with the team’s rapid delivery of multiple power-line-communications systems. Release, first Dunlevie statement
• Evidence and assumptions — contemporaneous record · provisional: Claimed superiority, worldwide trials, and broad acceptance are promotional. Scaling depended on qualification, standards, customer conversion, manufacturing, and support. EnVerv release, product and market-acceptance claims
• Decision rule — researcher inference · moderate: At a technical scale-up round, translate “mission critical” and rapid design into independent product benchmarks, design wins, conversion, gross-margin, qualification, and manufacturing gates. EnVerv release, Dunlevie and product paragraphs
• Evolution and contradiction — contemporaneous record · strong: Semtech bought select assets for $4.9 million in 2015 and called subsequent revenue immaterial, counterevidence to momentum but not a cash-flow-complete Benchmark loss. Semtech 2015 Form 10-K, Note 3, printed p. 77/PDF p. 82

27. Euclid Series B statement (2013)

• Context and access — contemporaneous record · strong: The release says Benchmark led a $17.3 million Series B and that Dunlevie joined Euclid’s board; individual sourcing, check, vehicle, ownership, and designation terms remain unknown. WebWire, opening financing paragraphs
• Argument — investor-stated · moderate: Dunlevie emphasizes easy implementation, scalability, and measurable operating effectiveness for physical-retail analytics—an adoption screen rather than technical novelty alone. WebWire, closing Dunlevie quotation
• Evidence and assumptions — contemporaneous record · provisional: Device counts, retailer adoption, privacy, and effectiveness come from the company; customer retention, accuracy, revenue, and unit economics are absent. WebWire, customer, deployment, and product-claim paragraphs
• Decision rule — researcher inference · moderate: For an analytics tool, require low deployment friction, validated measurement, customer-action linkage, retention, and privacy controls before treating scale claims as value creation. WebWire, implementation and effectiveness claims
• Evolution and contradiction — contemporaneous record · moderate: WeWork acquired Euclid in 2019 and redirected the technology toward workplace analytics; terms and board/recusal process are undisclosed, so the deal cannot prove the retail thesis or a fund return. WeWork acquisition announcement

28. WeWork appoints Artie Minson (2015)

• Context and access — contemporaneous record · strong: WeWork’s release appoints Minson president/COO across operations, expansion, finance, legal, IT, and people functions at a reported 42 locations. PR Newswire, role and company-scale paragraphs
• Argument — investor-stated · moderate: Dunlevie’s prepared endorsement treats experienced operational leadership as the appropriate augmentation for a founder-led company entering international scale. PR Newswire, Dunlevie paragraph
• Evidence and assumptions — contemporaneous record · provisional: The quote proves public support, not that Dunlevie sourced the executive, led the search, set authority, or viewed expansion pace as a risk. PR Newswire, appointment and Dunlevie paragraphs
• Decision rule — researcher inference · moderate: Add a scaling operator before complexity outruns the founder, then define authority over finance, systems, pace, and people plus explicit board escalation triggers. PR Newswire, described remit and scale-up rationale
• Evolution and contradiction — contemporaneous record · strong: Minson later became co-CEO during Neumann’s removal, but another CEO followed and WeWork later entered Chapter 11. Hiring expertise did not itself solve control, capital allocation, or lease exposure. Mathrani appointment · WeWork 2023 Form 8-K

29. Rice football endowment statement (2018)

• Context, attribution, access — contemporaneous record · strong: Rice announces Bruce and Elizabeth Dunlevie’s endowment of the head coaching position and records Bruce’s prepared comments; gift size and restrictions are absent, and capital attribution remains joint. Rice, opening gift and later Dunlevie paragraphs
• Argument — investor-stated · moderate: The statement values intellect, integrity, discipline, and positive culture and treats permanent institutional support as more durable than one-time operating help. Rice, Dunlevie statement
• Evidence and assumptions — contemporaneous record · provisional: Rice’s transformative/catalytic language is promotion; no payout policy, governance, hiring authority, academic trade-off, or outcome measure appears. Rice, institutional and gift-description paragraphs
• Decision rule — researcher inference · moderate: For institutional giving, define the role, restriction, governance owner, payout, continuity logic, and outcome measures. Transferring this to venture founder selection would be analogy, not documented doctrine. Rice, endowment terms disclosed and omitted
• Evolution and boundary — researcher inference · strong: The record adds values evidence but cannot establish investment skill, Benchmark proceeds, or a personal capital-allocation return. Rice, article scope

30. Stanford Medicine maternal-fetal health gift (2021)

• Context, attribution, access — contemporaneous record · strong: Stanford announces an $80 million gift from Elizabeth and Bruce Dunlevie across maternal-fetal research, clinical programs, facilities, and access. Their institutional roles and statements are separately attributed. Stanford Medicine, gift description and Dunlevie paragraphs
• Argument — investor-stated · moderate: Bruce links personal care experience to funding both discovery and delivery; Elizabeth separately emphasizes access. The record supports joint mission capital, not an investing framework. Stanford Medicine, Bruce and Elizabeth Dunlevie statements
• Evidence and assumptions — contemporaneous record · provisional: Planned research, recruitment, translation, dissemination, and facilities are uses and forecasts, not completed outcomes; timing, asset form, restrictions, and payout are undisclosed. Stanford Medicine, planned uses and program descriptions
• Decision rule — researcher inference · moderate: For mission capital, articulate the affected population, research-to-delivery pathway, facilities/people funded, access objective, milestones, and governance owner. Stanford Medicine, gift-allocation and access sections
• Evolution and boundary — researcher inference · strong: Personal experience can create informed commitment or selection bias. No cost-effectiveness comparison or investment-return implication is supportable. Stanford Medicine, personal-experience and program-use paragraphs

Cross-source synthesis

Founder and team judgment is the strongest recurring motif

The most durable signal is not a sector doctrine but a people screen. In 1999 Dunlevie favored leadership, intelligence, and rapid adaptation when Internet experience was scarce; the Handspring release then paired prior team performance with a platform ecosystem; later statements add domain depth, accumulated technical learning, and stage-appropriate operators. Fortune, “Passion, in fact” paragraph · Handspring release, Dunlevie paragraph · Aquifi, financing paragraph · One Medical/Carlyle release, Dunlevie paragraph

Cross-source conclusion — researcher inference · moderate: he appears to treat learning capacity, repeat-team evidence, and relevant operating depth as substitutes or complements depending on category maturity. That is stronger than saying “back great founders,” but still incomplete: the public corpus never exposes a consistent scoring method, references, rejected-founder denominator, price boundary, or how a team signal changes after missed milestones. Eastside adds the important exception that exceptional commitment can be a scaling liability when the institution depends on heroic hours. Stanford Magazine, Dunlevie paragraphs

See a behavioral or technical shift, then separate it from company capture

WeWork’s observed community, persistent workplace expectations, Bo.lt’s participatory-media premise, and EnVerv’s smart-grid need show repeated attraction to a changing behavior or enabling condition. Forbes, WeWork 2012 reaction · Bloomberg compilation, PDF p. 172 · Digiday, closing paragraph · EnVerv release, first Dunlevie statement

Cross-source conclusion — researcher inference · strong: the evidence repeatedly proves the trend less well than it proves Dunlevie’s attraction to it, and it proves company value capture least of all. WeWork’s lease mismatch, Bo.lt’s unsettled model/legal exposure, EnVerv’s small asset sale, and Tabula’s shutdown show why category truth, implementation, economics, financing, and entry price require separate votes. Forbes, closing risk discussion · Semtech 2015 Form 10-K, Note 3 · EE Times, Tabula shutdown

Operating help matters, but authority determines whether it can work

Raza Foundries promises pooled operating discipline, Euclid emphasizes implementation, One Medical adds a healthcare specialist, and WeWork recruits experienced executives. EDN, Dunlevie paragraph · Euclid, Dunlevie paragraph · One Medical/Carlyle release · Minson appointment

Cross-source conclusion — researcher inference · moderate: Dunlevie consistently treats experienced people and operating infrastructure as risk reducers, but the record repeatedly withholds their decision rights. WeWork is the decisive counterexample: an operator can be “right” for the moment yet remain unable to repair founder control, capital allocation, or balance-sheet exposure. The lesson is authority-plus-capability, not capability alone. New Yorker, WeWork governance passages · WeWork 2023 bankruptcy 8-K

Relationship governance shifts from persuasion to formal enforcement

The 2018 and 2020 records show two different modes. Dunlevie publicly accepts that Neumann resisted his arguments; after the 2019 crisis, the special committee invokes explicit independence, delegated authority, contract enforcement, and litigation. WRAL/NYT, “Is it all a bit much” paragraph · Business Wire, authority sections

Cross-source conclusion — documented behavior · moderate: the record shows escalation in formal mechanism, but not a published rule for timing it. The 2020 enforcement posture protected minority-holder bargaining after the founder crisis; it does not erase the earlier vote to grant control or prove every legal allegation. A modern board needs thresholds that trigger escalation before value destruction makes formal authority reactive.

Benchmark’s small equal partnership is institution evidence, not personal philosophy

Acquired relays Dunlevie’s role in forming and extending the team and reports the no-pre-sell norm, while current partners describe rituals that maintain shared context. Acquired Part I, 00:30:34–00:37:32 and 01:37:00–01:38:48 · Acquired Part II, 06:32–15:14

Cross-source conclusion — researcher inference · moderate: board-load capacity, equal economics, founder-like partner selection, and independent founder exposure form a coherent institutional design only as a bundle. Yet Dunlevie’s private interview is unreleased and Stanford’s Europe cases are restricted. Individual origin, exceptions, underperformance management, geographic portability, and causal performance remain unproven.

Cycle humility appears after excess, but pricing discipline remains absent

The 1998 Cato pledge favors evenhanded rules, the 2001 interview diagnoses capital-driven junk formation, and the 2007 financing-channel interview notes how regulation and post-bubble market structure changed capital supply. Cato, printed pp. 20–21 · Irish Times, Dunlevie paragraphs · InvestmentNews, closing post-bubble paragraphs

Cross-source conclusion — researcher inference · provisional: Dunlevie recognizes that financing regimes alter company quality and survival, but the corpus supplies no valuation formula, ownership target, reserve rule, or exit discipline. Prepared statements therefore cannot establish price sensitivity or portfolio construction.

Intellectual chronology, evolution, and contradictions

The main unresolved tension is not simply “founder friendly versus governance.” It is a sequence: qualitative founder/trend recognition permits an early bet; relationship persuasion preserves cooperation; but formal controls must activate before a charismatic founder or capital-intensive model exhausts the company’s options. The corpus documents all three stages separately yet never provides Dunlevie’s trigger between them.

Decision playbook

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

These translations are researcher inference unless explicitly marked. They use hypothetical 2026 IC situations because the mapped record supplies no audited current Dunlevie portfolio or public 2026 policy statement.

Founder learning over credentials

• What transfers — researcher inference · strong: In a new AI-native or technical workflow, demonstrated learning and recruiting can outsignal a conventional title; the historical rule remains the 1999 adaptability test. Fortune, Dunlevie paragraph
• What does not transfer / changed conditions — researcher inference · strong: The 1999 internet-company example did not depend on an opaque, rapidly changing third-party general-purpose model. In 2026, industry produces more than 90% of notable AI models while frontier disclosure has fallen, and EU rules require model documentation for downstream providers; diligence must therefore test model/vendor dependence, data rights, compliance ownership, and dated product learning. Stanford AI Index 2026, R&D highlights · European Commission, GPAI obligations
• Most likely misuse — researcher inference · strong: treating speed, charisma, or repeated narrative changes as learning, or ignoring accumulated expertise where safety, regulation, hardware, enterprise integration, or downstream-model duties matter. Fortune, 1999 adaptability passage · European Commission, downstream-provider information

Trend recognition versus company capture and price

• What transfers — researcher inference · strong: Separate user-behavior change, company moat, business-model durability, financing structure, and entry price. WeWork demonstrates why a valid category intuition cannot carry all five. Forbes, contemporaneous risks · WeWork bankruptcy 8-K
• What does not transfer / changed conditions — researcher inference · strong: WeWork joined category intuition to a lease-heavy company; a 2026 AI application instead sits above concentrated model, compute, and cloud suppliers, while venture capital is itself concentrated in AI and mega-rounds. Test paid retention, gross-margin sensitivity to model costs, vendor switching, proprietary data rights, and whether price already assumes platform-level capture. Stanford AI Index 2026, R&D concentration · NVCA/PitchBook Q2 2026 Venture Monitor
• Most likely misuse — researcher inference · strong: saying “this is the next eBay” without naming the causal invariant, balance-sheet and supplier differences, and kill evidence. Forbes, Dunlevie analogy and contemporaneous WeWork risks · Stanford AI Index 2026, infrastructure concentration

Technical scale-up and follow-on capital

• What transfers — researcher inference · strong: Market/team/technology is a search filter, not an approval memo. Tabula and EnVerv require independent benchmarks, customer qualification, manufacturing or deployment proof, and a precommitted capital ceiling. Tabula launch · EnVerv asset outcome
• What does not transfer / changed conditions — researcher inference · strong: Tabula and EnVerv were semiconductor/productization bets, while a 2026 deep-tech or AI-infrastructure company also faces highly concentrated compute, foundry, energy, and model dependencies. Stage checks against reproducible performance, design wins, yield/reliability, energy and deployment constraints, gross margin, and financing runway rather than a total-addressable-market slide. Stanford AI Index 2026, compute, foundry, and energy evidence
• Most likely misuse — researcher inference · strong: reading a large follow-on as technical validation or continuing because prior capital is sunk when independent milestones remain unmet. Tabula financing claim · NVCA/PitchBook Q2 2026, mega-round concentration

Operator augmentation and authority

• What transfers — researcher inference · moderate: Add a stage-fit operator before systems fail, but write the authority matrix at hire. Minson’s case shows that experience alone cannot constrain founder control or capital allocation. Minson appointment · Mathrani succession
• What does not transfer / changed conditions — researcher inference · strong: A 2015 property-company operator mandate did not allocate AI-model regulatory responsibilities. If a 2026 company is itself a GPAI provider—and, for incident-reporting and cybersecurity duties, a provider of a systemic-risk model—the authority matrix should assign the applicable documentation, copyright, downstream-information, incident, and cybersecurity work alongside hiring, budget, fundraising, product scope, risk, and board reporting. European Commission, scope and 2026 GPAI obligations
• Most likely misuse — researcher inference · strong: recruiting a prestigious executive as investor reassurance while leaving consequential operating, capital, information, and legally applicable compliance rights unchanged. Minson appointment and stated remit · European Commission, role-specific GPAI duties

Partnership debiasing and decision memory

• What transfers — researcher inference · moderate: Independent founder exposure and limited pre-selling can reduce sponsor anchoring; later structured documentation can preserve forecasts and dissent. Acquired Part II, 12:16–15:14
• What does not transfer / changed conditions — researcher inference · strong: Benchmark’s reported one-table, no-pre-sell ritual was built for a tiny co-located partnership. For a 2026 partnership that is distributed, preserve independent judgment with pre-discussion scores, then freeze the thesis, ownership target, kill criteria, and next-check evidence after debate; the case is especially important in an AI/mega-round market whose headline volume masks concentration. Acquired Part II, 06:32–15:14 · NVCA/PitchBook Q2 2026 Venture Monitor
• Most likely misuse — researcher inference · strong: using “no memo” as permission for intuition-only decisions, weak audit trails, or opaque attribution in a fast, concentrated financing market. Acquired Part II, 12:16–15:14 · NVCA/PitchBook Q2 2026 concentration

Founder-friendly governance with escalation

• What transfers — researcher inference · strong: Relationship persuasion is useful only alongside pre-agreed triggers for formal action. The WeWork sequence shows private objection, supportive formal votes, crisis succession, and later enforcement. New Yorker, governance account · Special Committee statement
• What does not transfer / changed conditions — researcher inference · strong: The 2014 WeWork control vote predated Delaware’s current statutory framework. A 2026 board must first classify controller status and transaction type, then apply the current fact-specific §144 route and its conditions: non-going-private controller transactions have alternative committee, disinterested-stockholder, or fairness routes, while controller going-private transactions require both process approvals or fairness. That framework is not a universal disinterested-approval mandate or a retroactive safe harbor; the board should separately predefine information, conduct, capital, runway, milestone, leadership, and resignation triggers. Delaware Code, Title 8 §144 · New Yorker, reported 2014 vote
• Most likely misuse — researcher inference · strong: equating founder friendliness with waived controls, or citing private disagreement after voting for a harmful structure; current statutory process does not replace fact-specific fiduciary analysis. New Yorker, WeWork governance account · Delaware Code, Title 8 §144

Cycle-aware underwriting

• What transfers — researcher inference · moderate: Capital availability is part of product and survival risk. Ask whether the company exists because financing is cheap and whether it survives when follow-on and exit channels close. Irish Times, 2001 Dunlevie passages
• What does not transfer / changed conditions — researcher inference · strong: The 2001 evidence describes a broad post-bubble cash crunch; Q2 2026 instead shows record capital and exits whose headline strength is concentrated in AI, mega-rounds, and established funds. Stress runway without a new round, unsubsidized demand, financing-channel concentration, and lower-multiple exits because aggregate abundance is not company-level access. Irish Times, 2001 · NVCA/PitchBook Q2 2026 Venture Monitor
• Most likely misuse — researcher inference · strong: using “cycle discipline” only after a crash, treating record aggregate funding as universal liquidity, or rejecting every capital-intensive company without separating strategic financing from dependency. Irish Times, 2001 Dunlevie passages · NVCA/PitchBook Q2 2026 concentration

Best writings and unresolved record

Read these five first

1. How Venture Capitalists Are Deforming Capitalism (2020) — the richest direct and adversarial governance evidence, including persuasion, formal-vote contradiction, Toys “R” Us, and WeWork; read as reported investigation, not board transcript. New Yorker
2. Joe Blow StakesHisClaim.com (1999) — the cleanest contemporaneous founder-selection rule, before Benchmark’s famous outcomes could fully reshape the story. Fortune
3. Inside the Phenomenal Rise of WeWork (2014) — the best decision-near record of qualitative attraction and visible downside in the same piece. Forbes
4. Venture capital industry facing cash crunch and investor reluctance (2001) — the strongest contemporaneous cycle correction and denominator warning, albeit without a disclosed Benchmark process change. Irish Times
5. The WeWork Manifesto (2018) — the clearest pre-IPO coexistence of category conviction, scope skepticism, and admitted limits of persuasion. WRAL/NYT syndication

This is a ranking of evidentiary usefulness, not Dunlevie authorship. The 2020 special-committee statement is the strongest formal primary position but is narrower and jointly/counsel authored; the Acquired episodes are the richest partnership narrative but depend on an unpublished interview.

Unresolved record that could change the synthesis

• Direct voice: no public authored essay, memo, letter, LP communication, long transcript, or released oral history surfaced. Release the underlying Acquired interview, if authorized, with date and exact locators.
• Restricted cases: obtain licensed E148 and E218 bodies/exhibits and map speaker, date, argument, partner dissent, economics, and resolution page by page. Their contents could materially change the internationalization and vehicle-design analysis.
• Restricted book passage: obtain Finding the Next Starbucks hardcover pp. 251–252; do not infer from Moe’s framework or snippets.
• Restricted magazine passage: retrieve Fast Company issue No. 204 and verify print p. 98, the exact skepticism, timing, and what evidence reportedly changed Dunlevie’s view.
• Public company/board records: no investment memos, board minutes, contemporaneous downside cases, written forecasts, or explicit pricing/ownership/reserve/exit rules surfaced. Those gaps prevent a complete “philosophy” claim.
• Authorship: establish who drafted each prepared statement and which beliefs were Dunlevie’s versus company, board, family, committee, counsel, or communications-team positions.
• Outcome denominator: Aquifi and Bo.lt remain unresolved; Impresse, Raza Foundries, Tabula, EnVerv, Euclid, and WeWork lack cash-flow-complete Benchmark vehicle economics. Promotional consistency cannot be tested against a full set of gains and losses.
• Process change: no direct record says what Benchmark or Dunlevie changed after the dot-com bust, Tabula/EnVerv adverse outcomes, WeWork control concessions, failed IPO, or bankruptcy.
• Governance threshold: the corpus never states when persuasion should yield to veto, financing refusal, leadership action, independent committee, litigation, resignation, or exit. That is the most decision-relevant missing rule.
• Saturation boundary: public Medium profile/feed routes yielded no authored stories; public case/catalog, press-release, interview, and archive routes were processed or bounded. Legitimately accessible body text remains outstanding only through the named licensed/restricted records above.

What the spoken record can prove

Bruce Dunlevie's public spoken corpus is narrow but no longer empty. Rice Business Partners links a 68:15 official recording with direct Dunlevie voice, and its full automatic-caption track was inspected. Rice Business event archive, October 11, 2022 entry · Rice video Evidence label: contemporaneous record · strong for custody and investor-stated · strong for audible claims, with automatic-caption limits on exact wording.

That single direct appearance supplies the spoken evidence for his views on small funds, uncertainty, founder selection, fieldwork, trust, minority-investor influence, eBay, WeWork, Google, and a new carbon-sequestration lead. Rice video, 09:20–65:14 It remains a 2022 retrospective by a successful firm founder: return figures are unaudited, famous cases are selected, partner attribution is incomplete, and later knowledge can reshape the stories. Evidence label: investor-stated · strong for the fact he made the claims; retrospective witness · moderate for historical reconstruction and performance.

Acquired published two episodes drawing on one private Dunlevie interview, but did not release his audio or transcript. Part I · Part II The hosts' words are valuable relays, not public Dunlevie quotations, and Part II's dinner guests are later Benchmark partners. Evidence label: retrospective witness · moderate. NYU, Amity, and Trinity establish three more appearances with content barriers; Rice's 2018 award video is narrated by someone else. The corpus therefore cannot support a claim that five or seven direct Dunlevie talks survive.

Coverage denominator and access reconciliation

As of 2026-08-22, the exact note denominator is 7/7 processed source artifacts representing 6 underlying appearance families. Acquired Parts I and II are separate public episodes but derive Dunlevie material from one underlying private interview.

The access view is 4 full public artifacts—one direct recording, two relay episodes, and one narrated institutional video—plus 1 partial/proxy record at Amity and 2 content-unavailable records at NYU and Trinity. “Full” describes access to the artifact, not directness or reliability. Evidence label: researcher inference · strong, mechanically reconciled against the seven note files and the routes below.

The source family links are: Rice 2022, Acquired I, Acquired II, NYU 2020, Amity 2023, Trinity 2026, and Rice 2018.

Source-by-source guide

2022-10-11 — Rice Business Partners: Venture Capital in America

Context, voice, and access. Rice's event archive links the unlisted official Rice Business upload; the recording is public, 68:15, and carries English automatic captions but no creator transcript. Rice event archive, October 11 entry and video link · official recording The earlier metadata-only assessment was wrong. Evidence label: contemporaneous record · strong.

Central argument. Dunlevie treats venture as an uncertainty-and-people craft rather than a forecasting exercise: stay open to implausible companies, see a very large opportunity funnel, select a few ambitious founders, and use trust because a minority investor cannot command the company. Rice video, 12:24–19:59 and 64:25–65:09 Evidence label: investor-stated · strong for the framework, investor-stated · moderate for rounded funnel and first-money percentages.

Vehicle return boundary. Dunlevie at Rice — reported: 100x for Fund I; Forbes — reported: $85 million/$7.8 billion, about 91.8x before basis adjustments; Metrick–Yasuda — reported: 42x on $113 million. Rice, 12:09–12:25 · Forbes · Metrick–Yasuda, pp. 89–90 Evidence label: investor-stated · provisional. The figures lack a common basis.

Picking and process. The auto-captioned founder screen appears to compare ages 28, 38, and 48 and to describe a sociopathic trait as a feature; exact wording was not audio-verified. Rice, 14:43–15:16 The heuristic creates exclusion, legal, and reputational risk and has no demonstrated predictive validity. The deal process moves selected founders to a Monday partnership meeting and a same-day decision. Rice, 23:24–24:49 His anti-diligence phrasing occurs around an automatic-caption gap; it should be read as a challenge to false precision, not evidence that Benchmark skipped references, market work, legal review, or technical verification. Evidence label: investor-stated · provisional for the sensitive captioned wording and researcher inference · strong for the risk boundary.

Service and governance. Dunlevie calls venture a service industry and investors stagehands, says founder succession risk should be disclosed before investment, and describes candid private coaching after board meetings. Rice video, 37:24–40:20 and 58:04–58:58 The model is relationship-based influence; the talk does not supply enforceable triggers for when persuasion should yield to formal action. Evidence label: investor-stated · strong for the stated practice; researcher inference · moderate for the missing-control diagnosis.

Decision evidence. The eBay account supports shared sourcing/diligence, while Bob Kagle did the deal and retains primary deal/board credit. Rice, 29:12–31:42 · Gurley transcript Evidence label: retrospective witness · moderate.

New retrieval lead. Dunlevie reports an unnamed carbon-sequestration venture. Rice, 61:10–62:44 Evidence label: investor-stated · moderate. Identifying it as CarbonDrop is an inference from matching filings, not his claim. 2023 Form D · 2026 Form D Evidence label: researcher inference · provisional.

2022-09-28 — Acquired: Benchmark Part I

Context and boundary. The hosts identify Dunlevie among private interview sources but do not publish his interview. Their founding chronology is therefore an attributed relay combined with Stanford cases, books, press, and other partner interviews. Acquired Part I, episode and source list Evidence label: retrospective witness · moderate.

Contribution. The episode attributes to Dunlevie a role in enlarging the partnership, recruiting Kevin Harvey for founder-like temperament, and helping establish Europe. Acquired Part I, 00:28:53–00:37:32 and about 02:15:11 It links small-team size to board-service capacity and portfolio breadth, then presents equal economics as an incentive against deal ownership. Acquired Part I, 00:30:34–00:35:20 and 01:37:00–01:38:48 Evidence label: retrospective witness · moderate. No private-interview sentence is quote-verified, and the episode does not allocate every idea among Kagle, Dunlevie, Rachleff, Harvey, and later partners.

2022-10-17 — Acquired: Benchmark Part II — The Dinner

Context and boundary. Dunlevie is not a dinner guest. At 12:16–12:30, host Ben Gilbert relays a Dunlevie rule against pre-selling a deal; the private source is unavailable. Acquired Part II, 12:16–12:30 · uListen transcript, 12:16 locator Evidence label: retrospective witness · moderate.

Institutional mechanism. Later Benchmark partners connect the rule to founder-first exposure, no conventional memo, open-ended Monday discussion, and a single-conversation dinner designed to resist hierarchy. Acquired Part II, 06:32–15:14 Evidence label: documented behavior · moderate for 2022 practice; researcher inference · provisional for 1995 continuity and causal efficacy. The unresolved trade-off is whether less pre-selling reduces advocacy bias enough to offset weaker written forecasts, dissent records, and postmortem accountability.

2020-09-25 — NYU Law: Governance in Smaller Valley Companies

NYU's agenda schedules Dunlevie with Penny Herscher, Drew Hambly, and Thomas Kamei, moderated by Lisa Stimmell, from 11:00 to 11:50 a.m. ET. NYU agenda, printed/PDF p. 1 The institute's 2020 and 2021 reports corroborate the program. 2020 report · 2021 report Evidence label: contemporaneous record · strong for occurrence; researcher inference · insufficient public record for content.

The post-WeWork timing makes the panel a high-value governance retrieval route, but no recording, transcript, captions, slides, chat, or attributed notes surfaced. Timing cannot establish that Dunlevie discussed WeWork, founder control, minority holders, independence, or escalation. Evidence label: researcher inference · strong for the access boundary.

Fall 2023 — Amity Ventures annual meeting

One attendee identifies Dunlevie in a fireside with CJ Reim and paraphrases a small-fund/MOIC rule, direct founder outreach, fieldwork, a very small annual founder-selection set, anti-strategy openness, and Fund I reported: 80x+. Pantoja post, body bullets Evidence label: retrospective witness · provisional.

No recording or transcript surfaced, the event's exact date is not stated, and memorable attendee notes may compress speaker, moderator, and context. Amity — reported: 80x+; Rice — reported: 100x. The figures cannot be reconciled mechanically because neither source supplies a common date or gross/net/carry basis. Evidence label: researcher inference · strong. Their main analytical value is consistency with Rice's small-fund and anti-strategy claims, and consistency is not proof of return causation.

2026 — Trinity Venture Academy

Three public posts independently identify Dunlevie among the eight-week program's weekly VC speakers. Vogeser · Cho · Atherton Cho explicitly describes the VCs as speaking off the record. Cho, program description Evidence label: retrospective witness · moderate for occurrence and contemporaneous record · strong for the stated access condition.

No topic, claim, deal, or quotation from the session is used here. Multiple occurrence witnesses do not override an off-record agreement. Evidence label: researcher inference · strong.

2018-05-31 — Rice Alumni Gold Medal video

The official 1:40 video recognizes Dunlevie as a Gold Medal honoree but uses third-party narration throughout; its full automatic-caption track contains no Dunlevie speech. Rice Alumni video, 00:00–01:40 Evidence label: contemporaneous record · strong for custody and award; retrospective witness · moderate for narrated biography.

The narration's statement that he formed his own computer company conflicts with filing-based chronology that identifies Everex as an operating role. It is preserved as a discrepancy, not a verified founder claim. Evidence label: researcher inference · moderate. The artifact contributes context and an archival lead, not investing philosophy.

Recurring frameworks and stories

Stay open, but build an opportunity funnel

Dunlevie rejects category strategy because it creates blinders, urges exposure to counterintuitive founders, and describes large opportunity volume supporting a small investment set. Rice video, 12:24–19:59 The Amity attendee relays the same anti-strategy and small-number selection themes. Pantoja post, strategy and founder-selection bullets Evidence label: investor-stated · strong for Rice; retrospective witness · provisional for Amity.

The bounded rule is not indiscriminate openness: Dunlevie says he avoids biotech/FDA investing and remains mystified by Web3. Rice video, 32:00–33:35 and 65:26–65:54 Evidence label: investor-stated · strong. A usable interpretation is to remove topical blinders inside domains where the partnership can actually evaluate people and evidence, while escalating outside expertise where it cannot. Evidence label: researcher inference · moderate.

Put people first without letting charisma waive falsifiers

The Rice founder screen weights people heavily and, in an auto-captioned passage not audio-verified, appears to prefer age 28 over 38 or 48 and to frame a sociopathic trait as a feature; this creates exclusion, legal, and reputational risk without demonstrated predictive validity. Rice video, 14:43–15:16 Evidence label: investor-stated · provisional. Comparing Google as a personality-driven false negative with WeWork as a potential charisma-driven false positive is an analyst diagnostic, not Dunlevie’s stated framework. Rice, Google 52:46–54:17 · Rice, WeWork 24:46–27:02 Evidence label: researcher inference · strong.

The paired diagnostic is: What product or user evidence is personality causing us to discount, and what governance or economic evidence is personality causing us to excuse? Rice, Google and WeWork retrospectives · New Yorker, contemporaneous-governance counterevidence Evidence label: researcher inference · strong. The cases do not establish a numerical people/product weight that transfers across sectors.

Treat trust as a selection criterion and a governance mechanism

Dunlevie says plans change and minority investors cannot rely on control, so trust must begin before the investment and continue through candid founder coaching. Rice video, 15:37–16:23, 48:07–50:08, and 58:04–58:58 Evidence label: investor-stated · strong. Ben Shewry separately recalls Dunlevie emailing at the pandemic's onset with advice to lead positively, while explicitly qualifying the exact wording as approximate. Shewry transcript, printed/PDF pp. 7–8 Evidence label: retrospective witness · moderate, not verbatim Dunlevie voice.

Trust cannot substitute for enforceable authority when incentives diverge. Dunlevie's own WeWork story says influence declined as external valuation rose, while public records document later formal governance and litigation. Rice video, 24:46–27:02 · Special Committee statement Evidence label: documented behavior · strong for escalation and researcher inference · strong for the trust/control boundary.

Make the fund and partnership serve the work

Rice links small fund size to MOIC and diseconomies of scale; Acquired links team size to board capacity and equal economics; Part II links the incentive structure to anti-pre-sell discussion norms. Rice video, 12:53–13:39 · Acquired Part I, 00:30:34–00:35:20 and 01:37:00–01:38:48 · Acquired Part II, 06:32–15:14 Evidence labels: investor-stated · strong for Dunlevie’s fund preference; retrospective witness · moderate for the Part I institutional reconstruction; documented behavior · moderate for later partners’ 2022 practice.

The sources support a coherent design hypothesis—small vehicle, selective partnership, equal sharing, intensive boards, independent deliberation—but not its causal return contribution. Rice, small-fund and board-service claims · Acquired Part I, equal-economics reconstruction · Forbes, first-fund outcome report eBay, market regime, access, partner skill, and one outlier fund can explain much of the observed narrative. Evidence label: researcher inference · moderate.

Biography and intellectual evolution

Rice's biography places Dallas, Rice, Andersen, Stanford, Goldman, a technology operating role, venture entry in 1989, and Benchmark's 1995 formation into one retrospective sequence. Rice video, 05:31–09:20 Evidence label: retrospective witness · moderate. Chronology alone does not prove that humanities study caused people-first investing, that operating exposure caused anti-forecasting, or that Goldman caused rapid decisions.

The only clearly repeated post-2022 themes are small-fund MOIC, fieldwork, founder selectivity, and anti-strategy openness. Rice video, 12:24–19:59 · Pantoja post Evidence label: investor-stated · strong at Rice and retrospective witness · provisional at Amity. The surviving talks do not document when these views originated or what Dunlevie changed after the dot-com bust, Tabula, EnVerv, WeWork, or the Google miss.

Deal rationales and contemporaneous controls

eBay — demand evidence rescued a pass, but credit is shared

Dunlevie describes an initial dismissal, then cites profitability, rapid growth, and customer checks; this supports shared diligence, while Bob Kagle did the deal and retains primary credit. Rice, 29:12–31:42 · Gurley transcript Evidence label: retrospective witness · moderate.

WeWork — real behavioral insight did not solve governance or economics

In 2022 Dunlevie recalls exceptional founder talent, early margins, and lost influence after a SoftBank investment; the automatic caption renders $10 million, but the spoken unit was not audio-verified. SoftBank later reported more than $14.25 billion committed, while the Dunlevie Living Trust’s separate PIPE was $10 million. Rice, 24:46–27:02 · SoftBank commitment record · WeWork 2022 proxy, trust PIPE His roughly $3.5 billion 2022 revenue forecast exceeded reported $3.245 billion revenue. WeWork 2022 10-K Evidence label: retrospective witness · moderate.

Google — interpersonal aversion overrode visible product evidence

Dunlevie says Benchmark collectively passed despite product evidence and estimates a roughly $10 billion counterfactual; it is neither a personal loss nor a realizable return. Rice, 52:46–54:17 Evidence label: retrospective witness · moderate.

Carbon sequestration — direct current lead, not yet a case

Dunlevie reports an unnamed carbon-sequestration venture. Rice, 61:10–62:44 Evidence label: investor-stated · moderate. Identifying it as CarbonDrop is an inference from matching filings, not his claim. 2023 Form D · 2026 Form D Evidence label: researcher inference · provisional.

Contradictions and counterevidence

1. Small-fund return magnitude. Rice — reported: 100x; Forbes — reported: 91.8x (about, from $85 million/$7.8 billion); Metrick–Yasuda — reported: 42x on $113 million. Rice, 12:09–12:25 · Forbes · Metrick–Yasuda, pp. 89–90 Researcher inference · strong: no common basis.
2. People first versus people overreach. Google shows personality bias suppressing product evidence; WeWork shows founder magnetism coexisting with governance and balance-sheet failure. Rice video, Google and WeWork Evidence label: retrospective witness · moderate. The viable rule must test both false negatives and false positives.
3. Persuasion versus control. Dunlevie treats trust and minority ownership as core, yet says Benchmark lost influence as WeWork's valuation rose and later participated in formal committee enforcement. Rice video, 24:46–27:02 and 64:25–65:09 · Special Committee statement Evidence label: documented behavior · strong. The spoken corpus does not state an escalation threshold.
4. No strategy versus domain limits. He warns that strategy creates blinders but avoids biotech/FDA and expresses uncertainty about Web3. Rice video, 12:24–14:16, 32:00–33:35, and 65:26–65:54 Evidence label: investor-stated · strong. The tension resolves only if “no strategy” means openness inside a competence-and-evidence boundary rather than universal scope. Evidence label: researcher inference · moderate.
5. Anti-diligence rhetoric versus decision accountability. Fast founder exposure may reduce false precision, but no-pre-sell/no-memo customs can weaken durable forecasts and dissent. Rice video, 23:24–24:49 · Acquired Part II, 12:16–19:08 Evidence label: researcher inference · moderate. The corpus lacks losing-deal IC records needed to judge the trade-off.

Forecasts, mechanisms, and falsifiers

The favorable edge thesis is therefore testable: a small equal partnership may convert unusual-founder access, independent deliberation, and high-trust minority governance into superior selection and service. Rice, founder funnel and minority governance · Acquired Part II, decision process · Special Committee, formal-enforcement counterexample Evidence label: researcher inference · provisional. It weakens if vehicle results are one-outlier dominated, if partner/opportunity records do not show the claimed process, if founder heuristics add no predictive value, or if persuasion routinely fails without formal rights.

Listen or inspect these five first

Only the first item is public Dunlevie voice; items two and three relay one private interview, item four is an attendee proxy, and item five is narrated institutional context.

1. Rice Business Partners 2022 — the only full public direct-voice investing appearance; start at 09:20 for philosophy, 24:48 for WeWork, 29:11 for eBay, and 52:45 for Google.
2. Acquired Part I — best public reconstruction of team formation, partner selection, equal economics, and Europe; Dunlevie's private interview remains relayed.
3. Acquired Part II — best institutional-process complement, especially 06:32–15:14; later partners plus one Dunlevie relay, not a Dunlevie guest episode.
4. Amity attendee notes — useful for post-Rice consistency on fund size, fieldwork, and openness; reliability C and no quotation authority.
5. Rice Gold Medal video — a short contextual biography and access audit; third-party narration only, with a disputed computer-company formulation.

Unresolved and excluded routes

No source above is counted merely because it mentions Dunlevie. An appearance requires occurrence evidence; a content note additionally preserves whether the available speech is direct, relayed, narrated, paraphrased, metadata-only, or off record.

Saturation and highest-value retrievals

The discovery pass used exact-name and name-variant searches across 1989–1994, 1995–1999, 2000–2004, 2005–2009, 2010–2013, 2014–2017, 2018–2021, and 2022–2026; podcast indexes and transcript vendors; YouTube exact-name and Benchmark searches plus official Rice channel/event routes; Vimeo; NYU, Amity, Trinity, Stanford, CHM, Berkeley, conference, and web-archive routes; and caption/transcript queries. Evidence label: researcher inference · strong for the bounded search statement. Three final source-class sweeps added no further substantive direct-voice appearance.

The YouTube surface search did not reveal the Rice 2022 recording because it is unlisted; the official Rice events archive did. Rice events archive This means future saturation checks must crawl institution pages and hidden outbound links, not rely on video search alone. Evidence label: documented behavior · strong for the retrieval route.

Highest-value next retrievals are:

1. Request a creator transcript or corrected captions for Rice 2022, then verify any desired quotation against audio.
2. Ask whether Acquired can release the private Dunlevie interview date, audio, transcript, or permissioned excerpts with sentence-level provenance.
3. Request NYU/Cvent recording, slide, chat, or transcript inventory without inferring content from the panel title.
4. Inspect Rice's 2001 and 2019 folders for media inventories, acceptance remarks, or transcript-bearing files before requesting reproductions.
5. Identify the carbon-sequestration company and foundation from corporate, nonprofit, scientific, and financing records.

Under the present public record, the defensible conclusion is precise: one full direct investing talk exists; two public episodes relay one private interview; three appearances are access bounded; and one official contextual video contains no Dunlevie voice.

1. Search broadly, but stay within competence

• Rule — Build a wide opportunity funnel and meet counterintuitive founders before imposing a theme, but require a named person who can evaluate the category's technical, commercial, regulatory, and financing evidence.
• Evidence status — Dunlevie's rejection of rigid strategy and his explicit biotech/FDA and Web3 boundaries are [investor-stated · strong]; the competence-bounded operating rule is [researcher inference · moderate].
• Provenance and origin — The earliest complete public articulation is Rice in 2022: he warns that strategy creates blinders, describes a large funnel feeding few investments, and then names areas he avoids or does not understand. Rice, 12:24–19:59, 32:00–33:35, and 65:26–65:54 An Amity attendee relayed the same anti-strategy theme in 2023. Amity attendee post
• Mechanism and reasoning — Fixed themes can screen out non-obvious markets before live evidence arrives; broad fieldwork increases variation in people and problems observed. Competence prevents openness from becoming random exposure: someone must be able to detect technical impossibility, false customer proof, regulatory barriers, capital dependence, and adverse selection. [researcher inference · strong]
• Historical and vehicle context — MPAE and early Benchmark operated in a dense U.S. technology network where direct founder/board contact and a small partnership could make fieldwork cheap relative to the option value of a rare outlier. Palm emerged through Geoworks relationships, while Stanford's restricted cases show that exporting the institution to Europe/Asia raised a different authority-and-portability question. Hawkins oral history, printed pp. 12–13 · Stanford E218 catalog
• When to use — Use at sourcing and first screen in uncertain, formation-stage markets where category labels are unstable. Do not use it to waive domain diligence in regulated health, safety-critical systems, semiconductors, legal-financial infrastructure, or any area where the partnership lacks an evidence owner.
• How to apply it — (1) log every channel and topical filter; (2) ask what surprising observation earned a meeting; (3) assign a competence owner or independent specialist; (4) state unknowns and disqualifying evidence; (5) compare the opportunity with passes outside the current fashion; (6) advance only when someone can test the mechanism. IC questions: “What would our strategy have hidden?” “Who can prove us wrong?” Kill if no competent owner can independently verify the critical risk before capital is committed.
• Examples —
  • Palm/Geoworks — documented formation behavior, not a proven application of the 2022 maxim. A Geoworks relationship led Dunlevie to approach Hawkins and help create an option around a new company; the operators later changed product scope. Hawkins oral history, printed pp. 12–13 [documented behavior · strong]
  • eBay — researcher-applied illustration. Dunlevie initially dismissed it, then customer calls, profitability, and growth changed the partnership view; Kagle retains deal credit. Broad exposure mattered only because demand evidence corrected the first reaction. Rice, 29:12–31:42 [retrospective witness · moderate]
  • Carbon sequestration — boundary case. In 2022 Dunlevie described an unnamed current venture; matching Form Ds make CarbonDrop plausible but not proved. Openness survives late in his career, while identity, role, vehicle, and outcome remain unresolved. Rice, 61:10–62:44 · 2026 Form D [researcher inference · provisional]
  • Biotech/Web3 — explicit non-application. Dunlevie names limits rather than claiming universal range. Rice, 32:00–33:35 and 65:26–65:54 [investor-stated · strong]
• Evidence — Independent support comes from direct Rice doctrine, the Hawkins primary oral history, the Stanford case metadata, the 2023 attendee relay, and SEC filings for the provisional current lead. The rounded Rice opportunity counts use changing scopes and cannot establish a conversion rate. talks, forecast table
• Limits, failure modes, and exception ledger — Open search can reward novelty theater, drown a small team, reproduce referral bias, or hide a de facto strategy in who gets meetings. The public record lacks channel-level opportunity logs, pass reasons, geographic funnel data, and returns by source. International expansion, philanthropy, and CarbonDrop are not elevated into separate models because repeatable decision rules are missing.
• Evolution and contradictions — Early behavior looks broad across devices, semiconductors, enterprise services, health, ads, workspace, and climate; only 2022 direct voice names the rule. “No strategy” conflicts with explicit domain avoidance and resolves only as openness within testable competence. [researcher inference · moderate]
• Modern VC translation (as of 2026-08-22) — AI reduces build cost while concentrating frontier models, compute, foundries, energy, and cloud dependencies. Stanford AI Index 2026, R&D [researcher inference · strong] Keep a broad funnel, but score model/vendor dependence, data rights, compliance ownership, gross-margin sensitivity, and who can test each. Most likely misuse: branding undisciplined inbound as “anti-strategy.”

2. Back adaptable people, then run the two-sided evidence test

• Rule — Prefer founders who demonstrate fast, evidence-based adaptation and candor, then ask both: “What product evidence is personal discomfort making us discount?” and “What governance or economic evidence is charisma making us excuse?”
• Evidence status — Adaptability over credentials is [investor-stated · strong]; the paired false-negative/false-positive diagnostic is [researcher inference · strong] grounded in Dunlevie's Google and WeWork retrospectives.
• Provenance and origin — In 1999 Dunlevie emphasized leadership, intelligence, and rapid adaptation when Internet experience was scarce. Fortune, “Passion, in fact” paragraph In 2022 he said plans change and weighted people heavily. Rice, 14:16–16:23
• Mechanism and reasoning — At formation, the plan is a low-confidence forecast; learning rate, recruiting, truthfulness, and revision quality determine whether the team can discover the real company. The mechanism fails when polish, affinity, demographic proxies, or founder magnetism substitute for observable learning, or when interpersonal discomfort vetoes strong product/customer evidence. [researcher inference · strong]
• Historical and vehicle context — In 1999 Dunlevie said experienced Internet people were scarce, making transferable learning more useful; his later technical and health statements still emphasize domain experience. Fortune, 1999 · Tabula release · One Medical/Carlyle In regulated, hardware, health, or safety-critical domains, accumulated expertise remains necessary. A minority board investor also needs candor because formal control is limited. Rice, 48:07–50:08
• When to use — Use in founder selection, references, pivots, recruiting, and follow-on decisions. Do not use “adaptability” to excuse repeated narrative changes without outcomes, or people judgment to waive product, price, control, technical, or legal tests.
• How to apply it — (1) request three dated belief changes; (2) identify evidence, speed, action, and result; (3) compare self-account with independent references; (4) score candor under conflict and bad news; (5) run the two questions separately before group discussion; (6) write which evidence would change the score. Kill for material dishonesty or uncorrected blame shifting; stage when learning is plausible but outcome proof remains thin.
• Examples —
  • Palm/Handspring — documented repeat-team illustration. The Palm operators learned through Zoomer, narrowed the product, and later formed Handspring; Benchmark backed transferable capability, while the operators retain product credit. Dubinsky part 1, printed pp. 33–34 · Handspring release [documented behavior · moderate]
  • eBay — researcher-applied positive update. Profitability, rapid growth, and customer checks reversed an initial people/product reaction; the evidence changed the decision. Rice, 29:12–31:42 [retrospective witness · moderate]
  • Google — false negative. Dunlevie says a working product impressed the partnership, but aversion to the founders helped produce a collective pass. The roughly $10 billion hindsight figure is not realizable return. Rice, 52:46–54:17 [retrospective witness · moderate]
  • WeWork — false-positive/control boundary. Exceptional founder talent and visible community coexisted with supervotes, related parties, capital intensity, and bankruptcy. Rice, 24:46–27:02 · WeWork S-1 [researcher inference · strong]
• Evidence — Direct doctrine comes from Fortune and Rice; application/counterevidence comes from Dubinsky, the Handspring statement, Rice deal recollections, WeWork filings, and the bankruptcy record. WeWork bankruptcy 8-K
• Limits, failure modes, and exception ledger — No rejected-founder cohort, scorecard, reference protocol, demographic-validity evidence, or company outcomes matched to founder scores survives. The auto-captioned age/sociopathic passage is not audio-verified and creates exclusion, legal, and reputation risk; it is excluded from the model. Rice, 14:43–15:16 The model can overfit famous anecdotes and underweight cofounder/team systems.
• Evolution and contradictions — People language appears in 1999 and remains in 2022, but WeWork and Google show opposite error modes. No public Dunlevie postmortem says the screen changed after either case. The model evolves here through analyst-added symmetry, not a claimed investor revision. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — For AI-native and regulated companies, learning must include vendor-model dependence, data/copyright provenance, downstream documentation, safety, and named compliance ownership. European Commission, GPAI obligations [researcher inference · strong] Most likely misuse: treating speed, charisma, or repeated pivots as learning while ignoring domain expertise and measured action.

3. Separate category truth from company capture, economics, price, financing, and control

• Rule — Give six independent votes: is the behavior/technology change real; can this company capture it; are unit economics durable; is the entry price compatible with fund return; can financing survive; and do governance/control rights preserve action?
• Evidence status — Dunlevie's attraction to changing behaviors and technologies is [investor-stated · moderate]; the six-vote framework is [researcher inference · strong] reconstructed from repeated adverse outcomes.
• Provenance and origin — WeWork is the richest origin: Dunlevie recognized community energy in 2012–2014, while contemporaneous coverage disclosed lease duration and cash-use risks. Forbes, 2014 Bo.lt, EnVerv, and Tabula statements repeat behavior/technology-change attraction without publishing capture or price rules. Digiday · EnVerv release
• Mechanism and reasoning — Category truth attracts competitors and capital; value may accrue to suppliers, customers, landlords, standards owners, or later financiers. A company can grow while destroying equity value through negative margins, duration mismatch, dilution, or control transfer. Separating votes prevents a vivid macro analogy from carrying company and security underwriting. [researcher inference · strong]
• Historical and vehicle context — 1990s IPO liquidity rewarded early ownership; 2010s private-capital abundance enabled long private duration and mega-rounds. WeWork's short member commitments, long leases, founder supervotes, and SoftBank's more than $14.25 billion of commitments show how financing and control can dominate early insight. SoftBank commitment record · WeWork S-1
• When to use — Use for every category thesis, especially fashionable markets and capital-intensive models. Do not collapse the votes because user demand is visible, a prestigious investor marks up a round, or a founder has previously succeeded.
• How to apply it — (1) write falsifiable category evidence; (2) map who captures each dollar and critical dependency; (3) test contribution margin and duration under stress; (4) model security, dilution, ownership, and exit cases; (5) remove the next financing and test survival; (6) map votes, supervotes, information, conflicts, reserved matters, and stronger counterparties. Kill if any non-negotiable vote fails; stage when evidence can be bought cheaply.
• Examples —
  • WeWork — documented category, adverse company/security boundary. Community demand was observable, while long leases, supervotes, related parties, and SoftBank power defeated a simple category-to-return story; Chapter 11 completed the company outcome, not the fund cash-flow calculation. Forbes · bankruptcy 8-K [researcher inference · strong]
  • Tabula — researcher-applied technology/capital boundary. Large-market and breakthrough claims plus a large Series D did not prevent shutdown. Tabula release · EE Times [researcher inference · strong]
  • EnVerv — researcher-applied capture boundary. Mission-critical smart-grid need and product execution claims preceded Semtech's $4.9 million selected-asset purchase; the record does not prove equity value. EnVerv release · Semtech 10-Q [researcher inference · moderate]
  • One Medical — favorable but bounded. A large pre-IPO Benchmark V stake and strategic sale make it the clearest directional economic candidate, yet costs, sell-downs, dilution, and proceeds are missing. 1Life prospectus · Amazon completion [researcher inference · moderate]
• Evidence — The model uses independent contemporaneous journalism, company statements, primary SEC ownership/control filings, acquirer records, and bankruptcy outcomes. The deal record supplies the vehicle and denominator boundary. portfolio forensics
• Limits, failure modes, and exception ledger — The public corpus contains no Dunlevie valuation formula, ownership target, reserve curve, or exit rule. Analyst arithmetic can create false precision when entry securities and cash flows are absent. A true category can still support several winners; a failed company does not falsify the category itself.
• Evolution and contradictions — Dunlevie's 2001 cycle diagnosis recognizes financing risk after the Internet crash; 2014–2018 WeWork statements preserve category conviction while concerns grow; the record never shows a general six-vote rule or when price/control became separate vetoes. Irish Times · WRAL/NYT [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — A 2026 AI application can have explosive use yet depend on concentrated models, compute, energy, and clouds. Stanford AI Index 2026 [researcher inference · strong] Add model-cost sensitivity, vendor switching, proprietary-data rights, security/compliance, paid retention, gross margin, dilution, and control. Most likely misuse: using “platform shift” to justify any price or dependency.

4. Preserve independent partner judgment before advocacy

• Rule — Give each decision-maker direct exposure to the founder and base evidence before sponsor advocacy; then preserve a concise dated record of independent scores, group reasoning, dissent, falsifiers, and the next decision.
• Evidence status — The no-pre-sell norm is [retrospective witness · provisional] because Acquired relays one unreleased Dunlevie interview family and later partners' practice; the written-memory adaptation is [researcher inference · moderate].
• Provenance and origin — Acquired Part II says the sponsor did not pre-sell a company and partners met founders independently before discussion; the Dunlevie sentence is relayed rather than public audio. Acquired Part II, 12:16–15:14 Part I connects a small team and board capacity to firm architecture. Acquired Part I, 00:30:34–00:35:20
• Mechanism and reasoning — Sponsor narratives anchor attention and turn independent partners into confirmation resources. Direct exposure creates partially independent observations; equal economics can reduce champion ownership. A short post-debate record keeps the benefit without turning “no pre-sell” into “no evidence trail.” [researcher inference · moderate]
• Historical and vehicle context — The reported ritual emerged in a tiny co-located equal partnership where every partner could meet a founder and board load limited deal volume. Acquired Part I, 00:30:34–00:35:20 and 01:37:00–01:38:48 · Acquired Part II, 06:32–15:14 It may not scale to a distributed platform, multi-stage committees, specialist subteams, or fast mega-rounds without structured asynchronous evidence; Q2 2026 venture activity is concentrated in AI and mega-rounds rather than uniformly liquid. NVCA/PitchBook Q2 2026 [researcher inference · strong]
• When to use — Use before IC, partner admission, executive selection, follow-ons, and conflicted board decisions. Do not use when independent exposure would duplicate costly diligence without adding information, or when safety/legal decisions require shared expert analysis first.
• How to apply it — (1) circulate neutral facts and risk questions, not the sponsor thesis; (2) obtain direct founder/product/customer exposure; (3) record pre-discussion scores; (4) debate with roles and conflicts visible; (5) write the final thesis, dissent, ownership/terms, kill criteria, and next-check milestone; (6) audit forecast calibration later. Kill the process if coalition-building or attribution pressure prevents independent judgment.
• Examples —
  • eBay — researcher-applied illustration, not documented use of the norm. Dunlevie initially dismissed it, then customer evidence changed the collective view; Kagle brought and led the deal. The case shows the value of independent updating, but no contemporaneous IC record proves the ritual. Rice, 29:12–31:42 [retrospective witness · moderate]
  • Google — researcher-applied counterexample. The partnership saw a functioning product and still passed amid interpersonal aversion. Independent exposure alone does not remove shared personality bias. Rice, 52:46–54:17 [retrospective witness · moderate]
  • Palm financing — researcher-applied internal-dissent case. Dubinsky recalls that Dunlevie could not always persuade his partners to support financing. Dissent is visible, but the decision record and whether the norm applied are absent. Dubinsky part 2, printed p. 4 [retrospective witness · moderate]
  • WeWork — boundary case. A small-partnership decision process did not prevent later supervotes or escalation delay; entry vote and later control votes require separate audit. New Yorker [researcher inference · strong]
• Evidence — Four claim-supporting placements above draw from Acquired, Rice, Dubinsky, and adverse WeWork reporting; however, Acquired's two episodes partly share one private interview and do not count as two independent Dunlevie sources. The support grade remains provisional.
• Limits, failure modes, and exception ledger — The surviving record lacks a contemporaneous partnership manual, IC memos, pre-vote scores, losing-deal records, partner exceptions, and proof that the norm improved decisions. Attempted routes: Acquired public episodes/transcripts yielded relayed norms but no private audio; Rice yielded general decision anecdotes, not the ritual's origin; restricted Stanford E148/E218 catalogs yielded metadata only; Benchmark's 1997 Wayback route timed out; partner interviews and issuer filings did not expose IC procedure. This is an exhausted public-record shortfall, not an evidence waiver.
• Evolution and contradictions — Later partners in 2022 describe inherited practice, but no dated source establishes its 1995 form. The norm fights advocacy anchoring while the absence of formal memos can weaken decision memory and attribution. The adaptation adds records after debate without claiming Dunlevie required them. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — In a distributed, AI- and mega-round-concentrated market, use independent pre-discussion scores, then freeze a concise post-discussion thesis, security, ownership, dissent, kill criteria, and next-check evidence. NVCA/PitchBook Q2 2026 [researcher inference · strong] Most likely misuse: “no memo” as permission for intuition, weak audit trails, or celebrity-partner opacity.

5. Size the fund and team to board-service capacity, then align economics

• Rule — Raise only the capital that a small, high-trust partner team can deploy into material positions it can actively serve; align partner economics so sourcing, dissent, and company work are not distorted by internal deal ownership.
• Evidence status — Dunlevie's small-fund preference is [investor-stated · strong]; equal economics and board-capacity architecture are [retrospective witness · moderate]; causal outperformance is [researcher inference · provisional].
• Provenance and origin — Rice links smaller funds to MOIC and diseconomies of scale. Rice, 12:09–13:39 Acquired and Marquardt reconstruct the founding partner set, equal sharing, partner-selection care, and board-load limits. Acquired Part I, 00:30:34–00:37:32 and 01:37:00–01:38:48 · Marquardt oral history, printed pp. 56–58
• Mechanism and reasoning — A smaller fund lowers required check size and exit scale; material ownership makes fixed board labor economically relevant; limited partner count preserves shared context; equal economics can reduce hoarding and reward work after the check. The same design can blur accountability, under-resource specialized help, and let one outlier create a false causal story. [researcher inference · moderate]
• Historical and vehicle context — Benchmark I formed amid 1990s IPO liquidity and produced extraordinary but incompatible outcomes: Rice — reported: 100x; Amity — reported: 80x-plus; Forbes — reported: about 91.8x from the $85 million/$7.8 billion headline; Metrick–Yasuda — reported: 42x on $113 million. Rice, 12:09–12:25 · Amity attendee post · Forbes · Metrick–Yasuda, pp. 89–90 MPAE V's one-LP record is reported: 5.51x and reported: 46.1% net IRR on a different vehicle; Benchmark numbered funds, founders' funds, trusts, and management entities cannot be pooled. UC Regents schedule, p. 6
• When to use — Use in fund formation, fundraising, partner admission, portfolio pacing, board allocation, and succession. Do not assume “small” is optimal where specialist regulatory, security, recruiting, or global resources require a platform, or where ownership is too low for service to matter.
• How to apply it — (1) define opportunity set and minimum useful ownership; (2) calculate check and reserve load under dilution; (3) cap active boards per partner; (4) list work only partners can do; (5) allocate economics, decision rights, and key-person succession; (6) stress net returns excluding the dominant outlier and under lower exits. Kill fundraising growth that lacks a matched opportunity, ownership, or service-capacity increase.
• Examples —
  • Benchmark I — institutional positive with outlier caveat. Four accounts cannot be averaged: Dunlevie at Rice — reported: 100x; an Amity attendee — reported: 80x-plus; Forbes — reported: $85 million becoming $7.8 billion, about 91.8x at the headline level; Metrick–Yasuda — reported: 42x on $113 million. Rice, 12:09–12:25 · Amity attendee post · Forbes · Metrick–Yasuda, pp. 89–90 Kagle's eBay attribution prevents a Dunlevie/small-fund causal claim. [researcher inference · strong]
  • Genesys — documented service-to-vehicle alignment. Dunlevie's consulting generated a Benchmark warrant; the 1,897,368-share fund block was 9.9% after offering, separate from 50,000 personal options. Genesys prospectus, pp. 47 and 54–57 [documented behavior · strong]
  • One Medical/ServiceSource — board-capacity comparison. Dunlevie served long board tenures behind material Benchmark V stakes; One Medical reached a large sale, while ServiceSource produced a failed director vote and smaller sale. 1Life prospectus · ServiceSource vote [documented behavior · strong] for stakes/roles, not causal value.
  • Benchmark Europe/Balderton — portability boundary. International expansion tested whether culture, partner economics, authority, and networks could travel; restricted cases prevent a full result. Stanford E218 · ScienceBusiness [researcher inference · provisional]
• Evidence — Independent sources include direct Rice remarks, Acquired's institutional reconstruction, Marquardt oral history, SEC issuer filings, one LP schedule, academic fund data, and Forbes. Evidence strongly supports design elements and selected service; it does not support causal net alpha.
• Limits, failure modes, and exception ledger — No partnership agreement, full fund cash flows, fee/carry schedule, partner capital account, board-load time series, opportunity denominator, or equal-economics exception survives. Model-specific attempted routes and results: Rice yielded the direct small-fund rationale and reported: 100x but no cash-flow basis or contemporaneous policy; Acquired yielded retrospective equal-economics and board-capacity claims but no public private-interview audio or governing document; Marquardt's oral history yielded formation and partner-selection recollection but no signed agreement; the UC schedule yielded one LP's MPAE V reported: 5.51x and reported: 46.1% net IRR but no Benchmark or deal decomposition; Metrick–Yasuda yielded secondary Benchmark I reported: 42x and Benchmark IV reported: 0.2% net IRR on different bases and dates; issuer filings yielded selected stakes, boards, and follow-ons but no complete vehicle cash flows or board-load capacity; restricted Stanford E148/E218 catalogs yielded metadata only. Small funds may sacrifice follow-on capacity; equal economics may diffuse accountability; active boards can become overreach; eBay or market regime may explain Fund I.
• Evolution and contradictions — Dunlevie's 2022 doctrine presents the mature rationale; founding sources are retrospective. Benchmark later added partners and international affiliates, testing scale and cultural continuity. Metrick–Yasuda's Benchmark IV result challenges universal persistence. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — Q2 2026 venture activity is concentrated in AI, mega-rounds, and established managers, so aggregate abundance does not prove that every larger vehicle has a serviceable opportunity set. NVCA/PitchBook Q2 2026 [researcher inference · strong] Analyst-created hypothetical: a multi-stage AI platform considering a larger successor fund should proceed only if the incremental check sizes, ownership, specialist capacity, reserves, and partner attention map to a demonstrably larger eligible funnel; this application is [researcher inference · moderate], not a Dunlevie case. Observable applicability signals are active-board load, response time, ownership after dilution, reserve coverage, specialist bottlenecks, succession coverage, and basis-consistent returns excluding the top company. Most likely misuse: “small is virtuous” without net cash-flow, service, or outlier evidence.

6. Start with trust and persuasion, but precommit formal escalation

• Rule — Select for candor and use relationship persuasion while it preserves information and cooperation; before investing, define evidence that moves the board to formal information rights, veto, leadership change, financing refusal, conflict committee, enforcement, or resignation.
• Evidence status — Trust and persuasion are [investor-stated · strong]; the need for precommitted escalation is [researcher inference · strong] supported by WeWork, Toys “R” Us, and ServiceSource.
• Provenance and origin — Dunlevie says minority investors cannot rely on control and must trust founders through changing plans. Rice, 15:37–16:23, 48:07–50:08, and 58:04–58:58 WeWork's later special committee publicly asserted independence, delegated authority, and enforcement. Special Committee statement
• Mechanism and reasoning — Trust lowers information friction and makes bad-news disclosure possible; persuasion preserves founder agency and collaboration. When incentives diverge, valuations rise, or a controller becomes counterparty, relational capital may lose force. Precommitted triggers keep patience from becoming acquiescence and make formal action less personality-dependent. [researcher inference · strong]
• Historical and vehicle context — Minority venture boards depend more on consent than command, but public-company fiduciary duties, supervotes, controller transactions, and conflicted financing introduce formal process. Benchmark VII, Dunlevie's family trust, SoftBank, Neumann, and the WeWork committee had different economics and powers. WeWork 2022 proxy · Chancery opinion
• When to use — Use from references through every board cycle. Do not rely on persuasion alone after misrepresentation, repeated milestone breach, obstruction of information, controller conflict, insolvency risk, compliance failure, or leadership incapacity.
• How to apply it — (1) test bad-news candor before investment; (2) write reserved matters and information paths; (3) define conduct, capital, runway, milestone, conflict, and leadership triggers; (4) map disinterested decision-makers/advisers; (5) document advice, dissent, vote, and follow-through; (6) escalate at the threshold, then reassess engagement or resignation. Kill for deliberate material deception; do not confuse loyalty with fiduciary passivity.
• Examples —
  • Palm/Handspring — relationship-positive illustration. Dunlevie remained engaged across company formation, product reset, sale, and a repeat-founder company; partner financing friction shows trust did not eliminate institutional dissent. Dubinsky parts 1–2 [documented behavior · moderate]
  • WeWork — decisive counterexample. Dunlevie says influence declined as valuation and SoftBank capital rose; the record later moves to succession, committee authority, litigation, resignation, and bankruptcy. Rice, 24:46–27:02 · bankruptcy 8-K [documented behavior · strong] for escalation, not timing efficacy.
  • Toys “R” Us online — stop case. Adverse reporting says Benchmark canceled after internal cooperation claims proved unreliable despite a guaranteed payout. New Yorker [retrospective witness · moderate]
  • ServiceSource — accountability case. Dunlevie failed to receive majority support; the board declined his conditional resignation and retained him. 2019 vote 8-K [documented behavior · strong] Reasons and later change remain private.
• Evidence — Direct Rice doctrine, founder oral histories, SEC filings, a court opinion, a committee statement, and adversarial reporting give independent support across relationship, control, stop, and accountability. The sources do not reveal a general Dunlevie escalation schedule.
• Limits, failure modes, and exception ledger — Formal rights can harm trust, arrive too late, be unavailable, or be exercised by conflicted people. A committee claim is advocacy, not a merits finding; a shareholder vote is accountability evidence, not proof of illegality; bankruptcy is not a complete fund-loss ledger. The model must distinguish advice, vote, recusal, enforcement, and resignation.
• Evolution and contradictions — Public language moves from 2014 category enthusiasm to 2018 acknowledged disagreement, 2019 succession, and 2020 enforcement. No direct source says Dunlevie changed his general governance rule. The contradiction between persuasion and control remains unresolved without board minutes. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — Apply current Delaware §144 only after classifying the parties and transaction. For a controller's non-going-private transaction, §144(b) offers alternative fact-specific routes: a properly empowered committee of at least two board-determined disinterested directors acting in good faith and without gross negligence, an informed and uncoerced majority vote of votes cast by disinterested stockholders if the deal was conditioned that way, or fairness to the corporation and stockholders. A controller going-private transaction instead follows §144(c): both §144(b)(1) and (b)(2), or fairness. Section 144(a) separately covers specified director/officer-interest transactions through its own alternative routes. A WeWork-like fact pattern cannot be assigned a current safe-harbor route without first establishing controller/control-group status, transaction type, interests, delegation, disclosures, and approvals; the 2020 dispute predated this framework. Delaware Code, Title 8 §144(a)–(e) [researcher inference · strong; not legal advice] Most likely misuse: mandating disinterested approval for every transaction, calling waived rights founder-friendly, or treating a committee title as proof of independence.

7. Match the company stage with operating capability and real authority

• Rule — Identify the company's next binding bottleneck, recruit or add the operator/capital partner with relevant capability, and transfer the decisions, budget, information, and accountability required to solve it.
• Evidence status — Dunlevie's repeated praise for experienced teams and stage operators is [investor-stated · moderate]; the authority requirement is [researcher inference · moderate] and WeWork is the central counterexample.
• Provenance and origin — Genesys provides early primary evidence that Dunlevie himself performed consulting work. Genesys prospectus, related-party transactions Later prepared statements emphasize pooled operating discipline at Raza, implementation at Euclid, healthcare expertise at One Medical/Carlyle, and an experienced president/COO at WeWork. Raza · Euclid
• Mechanism and reasoning — Stage transitions change the constraint—from product definition to repeatable sales, finance, regulation, organizational scale, or turnaround. Relevant experience can compress learning, but only when the person has authority and incentives to change the constraint. Titles without budget, information, founder backing, or board enforcement are theater. [researcher inference · strong]
• Historical and vehicle context — Early venture boards sometimes supplied direct operating labor, as the Genesys warrant-for-consulting record demonstrates. Genesys prospectus, related-party transactions Later companies can add specialist growth capital and executives, as One Medical/Carlyle and WeWork/Minson show, but supervoting founders, new investors, and complex boards split authority. One Medical/Carlyle · Minson appointment · WeWork S-1 An investor-affiliated or investor-appointed director is not the operator and must not absorb management credit. [researcher inference · strong]
• When to use — Use at financing, product-market transition, regulatory entry, international expansion, finance/control buildout, crisis, and succession. Do not add an executive merely for investor signaling, or assume prior title substitutes for current-stage capability.
• How to apply it — (1) name the single next bottleneck; (2) list decisions and information needed; (3) map current owner and gap; (4) choose hire, adviser, board member, or capital partner; (5) write decision rights, budget, metrics, and review date; (6) remove overlap and define escalation. Kill the hire if the authority gap cannot be resolved or the founder will not transfer the necessary scope.
• Examples —
  • Genesys — documented direct work. A company warrant went to Benchmark for Dunlevie's consulting, cleanly separating individual labor from fund consideration. Genesys prospectus, pp. 47 and 54–57 [documented behavior · strong]
  • One Medical/Carlyle — favorable stage partner. The announcement framed Carlyle as healthcare capital and expertise for the next growth phase; One Medical later reached a strategic sale. The source does not prove which intervention caused value. Carlyle announcement · Amazon completion [researcher inference · moderate]
  • Euclid — implementation claim. Dunlevie emphasized experience and implementation in a physical-analytics company; WeWork later acquired Euclid, but board-role overlap and transaction approval remain incomplete. Euclid Series B [investor-stated · moderate]
  • WeWork/Minson — authority failure boundary. An experienced president/COO joined with a stated remit, yet founder control and capital allocation remained decisive; later succession did not prevent bankruptcy. Minson appointment · Mathrani succession [researcher inference · strong]
• Evidence — Independent primary/near-primary sources cover Genesys's security, Euclid and Minson statements, Carlyle's role, succession, Amazon's outcome, and bankruptcy. These establish capability additions and outcomes, not a general causal effect.
• Limits, failure modes, and exception ledger — Titles can create role conflict, prestige bias, or founder triangulation; new investors may add incentives not expertise; operator success is stage- and domain-specific. No public authority matrices, executive scorecards, rejected-hire denominator, or causal board records survive. Euclid–WeWork approval/recusal evidence remains unresolved.
• Evolution and contradictions — The corpus moves from Dunlevie performing work directly at Genesys to public praise for operators and specialist capital. WeWork shows that augmentation without authority cannot repair control or balance-sheet problems. No direct postmortem states that he changed the model. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — For an applicable GPAI provider, assign the general documentation, downstream-information, copyright-policy, and training-summary duties alongside product, hiring, capital, and risk authority. Assign incident reporting and adequate cybersecurity as additional duties only when the provider's GPAI model has systemic risk. European Commission GPAI guidance [researcher inference · strong] Most likely misuse: recruiting a celebrated executive while consequential rights remain unchanged.

8. Stage capital against milestones and stop when organizational truth fails

• Rule — Make every check purchase a named technical, customer, economic, regulatory, or organizational proof; re-underwrite independently before follow-on; stop when critical evidence fails or the company cannot supply truthful cooperation, even if prior capital or a guaranteed payment creates sunk-cost pressure.
• Evidence status — The combined rule is [researcher inference · provisional]. Tabula, EnVerv, Impresse, and Marin support the need for milestones; Toys “R” Us provides one retrospective stop account. No direct Dunlevie milestone/reserve manual survives.
• Provenance and origin — Dunlevie's prepared statements at Tabula, EnVerv, and Impresse emphasize market, technology, team, and product speed, while later outcomes expose missing proof. Tabula Series D · Impresse financing The New Yorker later reports the Toys stop after cooperation claims failed. New Yorker
• Mechanism and reasoning — Capital buys information only when a milestone is observable and tied to a decision. Without a precommitted threshold, large rounds become social proof and prior capital becomes a reason to continue. Organizational truth is upstream of every milestone: a board cannot adapt to evidence it does not receive. [researcher inference · strong]
• Historical and vehicle context — Semiconductors require technical/productization and customer-qualification proof; Tabula's financing and shutdown expose the missing bridge. Tabula release · EE Times shutdown Enterprise services require retention and margin; Impresse's financing and asset outcome show why. Impresse financing · Impresse asset purchase Dot-com ventures needed real demand after easy financing; Dunlevie's 2001 post-crash account supplies the cycle boundary. Irish Times Closed-end funds add reserve opportunity cost and fund-life constraints, but Dunlevie's reserve policy is private. vehicle boundary [researcher inference · strong]
• When to use — Use at initial term sheet, each follow-on, bridge, turnaround, covenant waiver, and board crisis. Do not use a generic revenue milestone across different business models, or stop solely because the market turned when company-specific evidence remains favorable and the vehicle can fund it.
• How to apply it — (1) decompose technical, customer, economic, regulatory, financing, and organization risks; (2) choose one observable proof per risk; (3) define owner, date, evidence source, capital budget, and threshold; (4) state stop/change/follow-on action in advance; (5) assign an independent re-underwriter; (6) record truth/cooperation breaches separately. Kill on deliberate material deception; stop or restructure when the next proof cannot justify reserve opportunity cost.
• Examples —
  • Tabula — researcher-applied technical/capital failure. A $108 million Series D and breakthrough claims preceded closure after more than $200 million reportedly raised; exact Benchmark exposure and failure mechanics remain unknown. Tabula release · EE Times [researcher inference · strong]
  • EnVerv — researcher-applied commercialization boundary. Rapid system-on-chip execution claims preceded a $4.9 million selected-asset purchase; product speed did not establish equity value or utility adoption. EnVerv release · Semtech 10-Q [researcher inference · moderate]
  • Impresse — researcher-applied financing boundary. More than $62 million financing preceded shutdown and creditor-assignee asset sale; team and B2B market claims lacked a surviving retention/margin bridge. shutdown · asset purchase [researcher inference · moderate]
  • Toys “R” Us online — retrospective documented stop. Benchmark reportedly canceled after organizational cooperation claims proved unreliable, even with a guaranteed payout. Contract, funded amount, and minutes are absent. New Yorker [retrospective witness · moderate]
  • Marin — exception to simple outcome scoring. Benchmark VI follow-ons are visible, but Chapter 11 occurred eight years after Dunlevie left; later failure cannot reveal the correct earlier milestone. Marin prospectus · Chapter 11 index [researcher inference · strong]
• Evidence — The examples contain more than four claim-supporting placements across company releases, independent trade reporting, SEC filings, and adverse press. They establish outcomes and missing bridges; only Toys connects a stop action to Dunlevie/Benchmark, so the model remains provisional.
• Limits, failure modes, and exception ledger — Attempted routes: Tabula releases and EE Times exposed financing/outcome but not board decks, tape-outs, design wins, recovery, or sponsor; EnVerv releases and Semtech filings exposed claim/asset price but not equity ownership or milestones; Impresse financing/shutdown/asset reports lacked retention, board, and recovery; Marin filings exposed follow-ons/board dates but not reserve decisions; Toys release/adverse narrative lacked contract and minutes. No public Benchmark LP letters, IC memos, reserve schedule, follow-on rejects, liquidation waterfalls, or Dunlevie postmortem were located. The shortfall is exhausted-record, not padded certainty.
• Evolution and contradictions — The 2001 post-bubble diagnosis recognizes capital dependence, but later Tabula/EnVerv statements still use affirmative market/team/technology language. Toys suggests a truth-based stop; no dated evidence shows it became a general investment rule. [researcher inference · strong]
• Modern VC translation (as of 2026-08-22) — AI and deep-tech mega-round concentration can disguise sunk-cost escalation. NVCA/PitchBook Q2 2026 [researcher inference · strong] Stage reproducible benchmark, design win, paid retention, yield/reliability, energy, compliance, margin, and runway; require an independent re-underwrite. Most likely misuse: reading a follow-on or valuation mark as milestone proof.

Evidence gaps and candidate models

• Universal price discipline, ownership target, reserve ratio, concentration rule, or exit discipline — excluded. Selected securities and stakes exist, but no reusable Dunlevie formula or complete cash-flow history does. deal mechanics [researcher inference · insufficient public record]
• International expansion — excluded as a distinct model. Stanford E148/E218 are purchase-restricted and Balderton's emergence does not reveal a Dunlevie decision rule, authority, or causal result. Stanford E218 [researcher inference · insufficient public record]
• Philanthropic institution building — excluded. Rice and Stanford statements reveal values and governance service, not a venture rule with grounded portfolio applications. [researcher inference · strong]
• Carbon sequestration/CarbonDrop — excluded as a model. Dunlevie names an unnamed venture; matching filings support only a provisional identity inference, with no repeatable underwriting rule. Rice · 2026 Form D [researcher inference · provisional]
• Demographic founder heuristics — excluded. The auto-captioned age/trait passage is not audio-verified and lacks predictive, legal, and fairness support. Rice, 14:43–15:16 [investor-stated · provisional]
• Standalone cycle timing — absorbed into Models 3 and 8. The 2001 diagnosis and 2026 translation support financing-regime stress, but no buy/sell timing rule or private application ledger survives. Irish Times [researcher inference · moderate]

The eight-model inventory reconciles every recurring decision signal in the notes, writings, talks, investments, and philosophy index without splitting synonyms or elevating slogans into unsupported rules.

What the evidence supports

The defensible case for Bruce Dunlevie is narrower than the reputation and more useful than a rank. The public record supports a formation-era venture practitioner who combined broad relationship access, unusually durable board work, repeat-founder backing, and a small equal-partnership design. It does not support a personal career return, a hit rate, or a claim that his judgment alone caused Benchmark's best outcomes. The package therefore treats the edge as a [researcher inference · moderate] process hypothesis: broad fieldwork can surface non-obvious founders; adaptability and candor can matter when plans change; independent partner exposure can reduce sponsor anchoring; and a small fund/team can make long service economically rational. Profile thesis and falsifiers · Philosophy snapshot

Three cases carry the favorable thesis without requiring a personal-return fiction. Palm/Handspring links a Geoworks relationship, company formation, a product-scope challenge, repeat-team financing, and governance, while Jeff Hawkins, Donna Dubinsky, Ed Colligan, and their teams retain product and operating credit. Hawkins oral history, printed pp. 12–13/PDF pp. 16–17 · Dubinsky part 1, printed pp. 33–34 Genesys is the cleanest Dunlevie-specific labor record: a 420,282-share warrant at $5.9483 went to a Benchmark entity for consulting services he provided, while 50,000 personal options remain outside the 1,897,368-share fund block. Genesys prospectus, printed pp. 47 and 54–57 One Medical shows long board service and a filing-visible Benchmark V stake of 13,629,610 shares before the IPO, followed by a strategic sale; cost, dilution, distributions, and personal economics remain unknown. 1Life prospectus, management and principal-stockholder tables · Amazon completion

The same record supplies genuine disconfirmation. Google is a personality-driven false negative in Dunlevie's retrospective account. Tabula pairs large-market, breakthrough-technology, and proven-team promotion with a later shutdown after reported capital of more than $200 million. ServiceSource's 2019 vote—32,176,515 shares for and 37,603,679 against—tests board accountability because the board declined his conditional resignation. WeWork tests whether category intuition, trust, and persuasion can survive supervoting control, related parties, capital-duration mismatch, and a dominant later financier. Rice, Google at 52:46–54:17 · Tabula financing · Tabula shutdown · ServiceSource 2019 8-K, Item 5.07 · WeWork 2019 S-1, control and related-party sections

The economics survive only as bounded records:

• A UC schedule reports one LP's MPAE V experience through 2003-03-31: $10.0 million committed, $9.658 million paid in, $52.892 million distributed, $0.345 million NAV, $53.236 million total value, reported: 5.51x net multiple, and reported: 46.1% net IRR. It is not a whole-fund gross result, a company attribution, or Dunlevie's personal return. UC Regents schedule, p. 6
• Benchmark I has incompatible bases: Dunlevie at Rice — reported: 100x in 2022; Forbes — reported: $85 million becoming $7.8 billion, about 91.8x before basis adjustment; Metrick–Yasuda — reported: 42x on a $113 million fund. The academic source also gives Benchmark IV reported: 0.2% net IRR as of March 2007. These are not observations to average. Rice, 12:09–12:25 · Forbes, 2001 · Metrick–Yasuda, printed pp. 89–90
• Removing One Medical, the clearest directional economic candidate, does not yield a defensible next-dollar rank. Genesys is expressed through an IPO block and a company sale; Kalpana through a single reported MPAE dollar figure and company share consideration; Rambus through a pre-IPO block; and Palm/Handspring through company consideration and merger shares. Pareto analysis [researcher inference · strong]

The source and extraction base is unusually broad for a discreet investor. The final map contains 153 processed source records—85 ★★★, 55 ★★, and 13 ★—while preserving body-access limits. The 30-record writings denominator contains 26 full records, one public fragment, and three metadata-only records; it does not pretend Dunlevie authored the reported interviews, cases, releases, joint statements, or relayed material. The spoken corpus contains seven processed artifacts across six underlying appearance families: one direct public Rice talk, two public Acquired relays from one unreleased private interview, three access-bounded appearances, and one narrated institutional video. Source-map inventory · Writings reconciliation · Talks denominator

What could falsify the thesis

The favorable hypothesis predicts more than a coherent story. In unseen contemporaneous records, it should produce broad but competence-bounded category exposure; direct founder contact before consensus; independent partner judgments before advocacy; dated evidence of founder adaptation and candor; explicit ownership, service, and escalation choices; and better capital-weighted net outcomes for cases that pass those gates than for rejected or weakly rated cases. Profile, prediction and weakening evidence Evidence that access alone generated the opportunity set, that partner decisions were pre-sold, that founder scores did not discriminate ordinary outcomes, that intensive boards did not alter decisions, or that larger vehicles preserved the same service and compatible net performance would weaken or overturn the process claim. [researcher inference · strong]

The strongest skeptical explanation is not that Dunlevie had no skill. It is that Rice, Stanford, Goldman, MPAE, Benchmark, founders, and boards created privileged access; equal economics socialized team outcomes; one exceptional first fund generated a halo materially powered by Bob Kagle's eBay work; long public-board tenures created filing visibility; and missing opportunity, write-off, and cash-flow denominators allowed a polished retrospective philosophy to be fitted to selected cases. Profile, skeptical explanation · Philosophy, strongest case against greatness [researcher inference · strong]

That skeptical view performs at least as well as the favorable view on the public data:

The dominant-outlier test remains unresolved rather than passed by assertion. One Medical can be removed from the visible Dunlevie deal ranking, but incompatible measures prevent a replacement dollar order. Benchmark I can be discussed excluding eBay only if its cash flows are obtained; the public sources do not disclose the required residual series. The package correctly says “not computable” rather than substituting company values, current share prices, or hypothetical Google ownership. Investments, fund-level controls

Attribution and denominator audit

Mechanical reconciliation

Deal, team, and vehicle cross-check

The legal/economic units also reconcile. Palm is linked only to MPAE with the numbered vehicle unknown; Rambus and Geoworks are MPAE V; Handspring is Benchmark II plus a separate personal position; ServiceSource and One Medical are Benchmark V; Marin is Benchmark VI; WeWork is Benchmark VII/AIV plus a later separate trust PIPE; CarbonDrop is a private operating-company lead, not a verified Benchmark investment. Vehicle map Shared control, a board designation, and an individual's pecuniary interest are different facts. [researcher inference · strong]

Central-claim citation audit

Ordinary-claim citation audit

The ordinary sample uses a reproducible spread: two non-thesis anchors in each of the seven inherited core documents—the first factual assertion after the first H2 and the first factual assertion in a named middle section—plus the first mechanical count assertion in this review. Claims that also carried a central conclusion were excluded from this table and audited above.

Transfer limits for modern VCs

What transfers is the discipline of separating decisions, not the prestige of the historical actors.

The completed peers sharpen the transfer boundary. Andy Rachleff makes product/market evidence and non-consensus change more explicit; Fred Wilson publishes more venture math and failure updates; Ann Winblad supplies a more software/customer-specific operating system. Dunlevie's distinct contribution is the proposed coupling of open fieldwork, small equal-partner architecture, and trust-based long-board service. Distinctiveness is not superiority, and the peers' vehicles and cash flows are not performance-comparable. [researcher inference · moderate]

Unresolved evidence

The remaining gaps are decision-relevant and cannot be closed by more confident prose:

• Fund and personal economics: complete MPAE/MPAE V and numbered-Benchmark capital calls, distributions, fees, carry, write-offs, NAV policy, and partner allocations. These would resolve the Fund I bases, dominant-outlier question, company contribution, and personal-return boundary.
• Opportunity and pass denominator: dated CRM/deal logs, source channels, pass reasons, founder scores, follow-on decisions, and rejected cohorts. These would test access versus selection and make a hit rate possible.
• Partner and team labor: source, sponsor, diligence, vote, negotiation, board work, recruiting, reserve, stop/exit, narration, and carry for every material deal. This is the only clean way to distinguish equal economics from equal work.
• Board efficacy and escalation: minutes and advice records for WeWork, ServiceSource, One Medical, Rambus, Palm/Handspring, and other long boards. WeWork committee authority, controller conflict, and the timing of escalation require complete process records, not only advocacy and court summaries.
• Company securities and realizations: Palm/Handspring terms and distributions; Genesys warrant exercise and sale holdings; Kalpana/Pure Digital cash flows; Rambus sell-downs; One Medical/ServiceSource distributions; WeWork fund/trust claims and recovery; Marin realization; Tabula liquidation.
• Licensed and archival narrative sources: page-complete E148/E218, Finding the Next Starbucks, eBoys, Done Deals, the Dartmouth MPAE V box, and any surviving partnership/board collections. Metadata remains a retrieval route, not body evidence. Source-map access ledger
• Current identity and activity: CarbonDrop/CarbonLock capitalization, the match to the unnamed Rice carbon venture, Dunlevie's ownership and authority, and whether any current investing role exists. Form D amounts do not establish cumulative proceeds, valuation, or Benchmark involvement. 2023 Form D · 2026 Form D

No reasonable public route identified in the source ledger remains silently unprocessed. The high-value next evidence is private, licensed, archival, or LP/company record material. Until it appears, the favorable edge should remain [researcher inference · moderate], the skeptical alternative [researcher inference · strong], and personal alpha [researcher inference · insufficient public record].

Twenty-point acceptance evidence

Direct voice, interview-derived cases, and written record

• ★★★ | Benchmark Capital in Europe: Bringing Silicon Valley Venture Capital to the Continent, E218 | https://www.gsb.stanford.edu/faculty-research/case-studies/benchmark-capital-international-expansion-silicon-valley-venture | 2006-09-07; revised 2008-11-09 | official catalog metadata A; retrospective case narrative B; body purchase-restricted and unprocessed | Dunlevie quotations count as direct voice only after licensed page-complete verification; catalog says 19 pages.
• ★★★ | Benchmark Capital in Europe, E148 | https://www.gsb.stanford.edu/faculty-research/case-studies/benchmark-capital-europe | 2003 | official catalog metadata A; retrospective narrative B; body purchase-restricted and unprocessed | Earlier version control only; quotations cannot be treated as verified direct voice yet.
• ★★★ | Finding the Next Starbucks, pp. 251–252 | https://books.google.com/books?id=FCMbUYFSZVQC&pg=PA251 | 2006-12-28 | book-interview metadata; pages not readable in inspected preview; access partial | Metadata-only retrieval lead; no Dunlevie argument or quotation processed.
• ★★★ | How Venture Capitalists Are Deforming Capitalism | https://www.newyorker.com/magazine/2020/11/30/how-venture-capitalists-are-deforming-capitalism | 2020-11-23 | reported phone interview and independent critique; reliability A for attributed passages, B for narrative; access full | Governance, persuasion, WeWork, and limits of controlling difficult founders.

Relayed private-interview material

• ★★★ | Benchmark Capital, Part I | https://www.acquired.fm/episodes/benchmark-capital | 2022-09-28 | podcast with host transcript based partly on private Dunlevie interview; reliability B; access full | Rich founding chronology and source list; hosts relay rather than publish Dunlevie's interview.
• ★★★ | Benchmark Part II: The Dinner | https://www.acquired.fm/episodes/benchmark-part-ii-the-dinner | 2022-10-17 | podcast; official audio/show notes and third-party navigation transcript; reliability B/C; access mixed | Relays the rule against pre-selling deals; original Dunlevie audio remains unpublished.

Reported interviews, prepared statements, and joint records

• ★★★ | Venture capital industry facing cash crunch and investor reluctance | https://www.irishtimes.com/business/venture-capital-industry-facing-cash-crunch-and-investor-reluctance-1.323061 | 2001-08-17 | newspaper interview; reliability A for attributed passages, B for report; access full | Best contemporaneous post-bubble discussion of venture uncertainty and excess.
• ★★★ | Joe Blow StakesHisClaim.com | https://fortune.com/article/internet-business-plan-startups/ | 1999-06-07 | reported interview; reliability A for attributed passage, B for report; access full/variable | Decision-date founder criteria: leadership, adaptability, and intelligence over prior Internet experience.
• ★★★ | Inside the Phenomenal Rise of WeWork | https://www.forbes.com/sites/alexkonrad/2014/11/05/the-rise-of-wework/ | 2014-11-05 | reported interview; reliability A/B; access full/variable | Initial eBay analogy and inability to articulate the pattern precisely; useful ex-ante WeWork evidence.
• ★★★ | Is This the Office of the Future or a $5 Billion Waste of Space? | https://strategygroove.com/wp-content/uploads/2011/08/bb-may-nov-all-issues-v2.pdf | 2015-06-19 | Bloomberg profile in unofficial compilation; PDF p. 172, right column, final paragraph; A for exact wording, B for report, C for mirror custody; access full | Ex-ante workplace-expectations thesis with explicit custody limit.
• ★★★ | Adam Neumann's $16 Billion Neo-Utopian Play to Turn WeWork Into WeWorld | https://www.fastcompany.com/3057415/adam-neumanns-16-billion-neo-utopian-play-to-turn-wework-into-wewo | 2016-03-14 | reported profile; access subscriber restricted | Restricted retrieval lead; public fragment establishes article context, not Dunlevie's reasoning.
• ★★ | Publishing in the Remix Era | https://digiday.com/media/publishing-in-the-remix-era/ | 2011-04-21 | trade interview; A only for verbatim wording, B for journalistic container; full page checked 2026-08-22, final section | Consumer participation and evolving-model thesis applied to Bo.lt.
• ★★★ | The WeWork Manifesto | https://www.wral.com/archive/17348308/ | 2018-02-18 | syndicated New York Times report; reliability A/B; access full | Contemporaneous admission that he opposed expansion but lacked leverage with Neumann.
• ★★ | Eastside Story | https://stanfordmag.org/contents/eastside-story | 2000-07 | institutional magazine interview; reliability A/B; access full | Founder and mission evaluation outside a conventional venture context.
• ★★ | Declaration of Independence: End Corporate Welfare | https://www.cato.org/sites/cato.org/files/pubs/pdf/bp-037.pdf | 1998-04-27 | jointly signed primary document, printed/PDF p. 20; reliability A for signature; access full | Supports an evenhanded cut; not proof he individually authored every sentence.
• ★★ | Handspring funding and 3Com licensing release | https://www.penbasedcomputing.com/timeline/palmpilot-organizer-creators-secure-funding-and-sign-licensing-agreement-with-3com/ | 1998-11-05 | historical press-release reproduction; reliability A for prepared quote, B for custody; access full | Decision-date confidence in the Palm team and consumer-computing opportunity.
• ★★ | Impresse raises more than $62 million | https://www.internetnews.com/it-management/impresse-raises-more-than-62-million/ | 1999-11-23 | funding report with prepared quote; reliability A/B; access full | B2B procurement thesis and management/domain-expertise screen.
• ★★ | Raza Foundries formation | https://www.edn.com/ex-amd-executive-raza-forms-technology-development-company/ | 2000-05-31 | contemporary report with prepared quote; reliability A/B; access full | Operational-discipline thesis for complex technology; not outcome evidence.
• ★★★ | Tabula secures $108 million | https://www.prnewswire.com/news-releases/tabula-secures-108-million-in-the-largest-round-awarded-to-a-private-semiconductor-company-over-the-last-decade-118764824.html | 2011-03-28 | company release; A only for exact prepared wording, C as promotional thesis/outcome support; full page checked 2026-08-22, Dunlevie paragraph | Large-market, breakthrough-technology, proven-management screen frozen before shutdown.
• ★★ | Aquifi changes the computing interface | https://techcrunch.com/2014/04/28/aquifi-changes-the-computing-interface-with-a-wave-of-the-hand/ | 2014-04-28 | funding report with prepared quote; reliability A/B; access full | Team learning and technical accumulation thesis; outcome unmeasured.
• ★★ | One Medical/Carlyle investment announcement | https://www.carlyle.com/ja/node/30011 | 2018-08-22 | official release; A only for exact prepared wording, C as promotional thesis/outcome support; full page checked 2026-08-22, Dunlevie paragraph | Expansion thesis and preference for healthcare-experienced growth capital.
• ★★★ | WeWork leadership changes | https://www.wework.com/newsroom/weworks-board-of-directors-announces-leadership-changes | 2019-09-24 | official company statement; reliability A for contemporaneous prepared words; access full | Public board position during IPO collapse; test against later records.
• ★★★ | WeWork Special Committee authority statement | https://www.businesswire.com/news/home/20200420005803/en/WeWork-Special-Committee-Sets-the-Record-Straight-Regarding-its-Authority-to-Represent-the-Best-Interests-of-Minority-Shareholders | 2020-04-20 | joint Dunlevie/Frankfort statement; reliability A; access full | Primary voice in SoftBank tender dispute; preserve joint attribution.
• ★★ | Toys “R” Us online venture statement | https://www.rebelscum.com/tramu0427.htm | 1999-04-27 | press-release mirror; A for exact prepared wording, C for promotion; full page checked 2026-08-22, Dunlevie paragraph | Then-largest Benchmark investment and brick-and-mortar/e-commerce thesis.
• ★★ | Advanced Equities emerges as big player | https://www.investmentnews.com/fintech/advanced-equities-emerges-as-big-player-in-venture-capital/11561 | 2007-10-01 | reported interview; A for verbatim wording, B for container; full page checked 2026-08-22, Dunlevie paragraphs | Timeliness and scale in private-company financing.
• ★★ | EnVerv raises $12 million | https://www.aa.com.tr/en/archive/smart-grid-system-on-chip-innovator-enverv-raises-12m-round-of-financing/388739 | 2011-12-20 | release reproduction; A for exact prepared wording, C for promotion; full page checked 2026-08-22 | Mission-critical connectivity and management-execution thesis.
• ★★ | Euclid Series B statement | https://www.webwire.com/ViewPressRel.asp?aId=170446 | 2013-02-21 | release reproduction; A for exact prepared wording, C for promotion; full page checked 2026-08-22, closing quote | Retail measurement as scalable analytics.
• ★★ | WeWork appoints Artie Minson | https://www.prnewswire.com/news-releases/wework-announces-artie-minson-as-president-and-chief-operating-officer-300091571.html | 2015-06-01 | company release; A for exact prepared wording, C for promotion; full page checked 2026-08-22 | Operational-leadership scaling thesis.
• ★ | Rice football endowment statement | https://riceowls.com/news/2018/8/21/rice-announces-endowment-of-head-football-coaching-position | 2018-08-21 | joint prepared statement; A for exact joint wording, C for investment relevance; full page checked 2026-08-22 | Values/philanthropy record, not investing doctrine.
• ★ | Stanford Medicine donation statement | https://med.stanford.edu/news/all-news/2021/02/80-million-dollar-donation-to-improve-health-of-mothers-and-babies.html | 2021-02-16 | prepared philanthropic quote; A for exact wording, C for investment relevance; full page checked 2026-08-22 | Personal/philanthropic voice only.

Talks, events, and archive routes

• ★★★ | Venture Capital in America — official Rice Business video | https://www.youtube.com/watch?v=MzsvTFQ7y4w | event 2022-10-11; upload 2022-10-26 | official unlisted video, 68:15; reliability A for recorded words/custody, B for retrospective history, C for unaudited returns; access full; auto-captions inspected | Only full public direct-voice investing appearance; timestamped note complete.
• ★★ | Rice Business event photo album | https://www.flickr.com/photos/ricebusiness/albums/72177720303053029 | 2022-10 | institutional album; reliability A for occurrence; media only | Confirms context but carries no analyzable transcript.
• ★★★ | Governance in Smaller Valley Companies — NYU Law bootcamp | https://www.law.nyu.edu/sites/default/files/09_15_20%20Bootcamp%20Agenda%20for%20Cventv2.pdf | 2020-09-25 | official agenda; A for occurrence; content unavailable/ungraded; no recording/transcript | Governance panel access ledger only.
• ★★ | Amity Ventures annual meeting attendee notes | https://www.linkedin.com/posts/marcelinopantoja_at-the-annual-meeting-for-amity-ventures-activity-7143005611683692545-Z8eR | 2023 | attendee paraphrase; reliability C; access variable | Leads on small funds and founder outreach; no quote or return is treated as verified.
• ★ | Trinity Venture Academy participant post | https://www.linkedin.com/posts/michael-vogeser_super-excited-to-have-won-the-venture-academy-activity-7445157419627843584-64eN | 2026 | participant metadata; reliability C; explicitly off record | Confirms appearance but supplies no publishable content.
• ★ | Rice Alumni Gold Medal video | https://www.youtube.com/watch?v=ZdVaRPIsXcw | 2018-05-31 | official short video; captions checked | Third-party narration only; excluded from direct voice.
• ★★ | Rice archive — Bruce W. Dunlevie '79, 2019 | https://archives.library.rice.edu/repositories/2/archival_objects/317998 | 2019 | finding aid, Box 24 Folder 72; reliability A for route; by request | Likely alumni-award material; contents not inspected.
• ★★ | Rice archive — Bruce Wall Dunlevie '79, 2001 | https://archives.library.rice.edu/repositories/2/archival_objects/102391 | 2001 | finding aid, UA 0003 Box 22 Folder 18; reliability A for route; by request | Highest-value biographical archival request route.
• ★ | Bruce Dunlevie Medium profile | https://medium.com/@brucedunlevie | checked 2026-08-22 | creator-profile route; A for custody | No story items surfaced.
• ★ | Bruce Dunlevie Medium RSS | https://medium.com/feed/@brucedunlevie | checked 2026-08-22 | official feed route; A for route | Feed returned zero story items.
• ★ | Bruce Dunlevie LinkedIn profile | https://www.linkedin.com/in/bruce-dunlevie-4285603/ | checked 2026-08-22 | identity/social route; login/fetch barrier | No indexable authored post surfaced; profile body not treated as processed.
• ★★ | Benchmark historical website — 1997 snapshot | https://web.archive.org/web/19970222220811/http://www.benchmark.com/ | 1997 | archive route; reliability A/B; repeated timeout | Potential historical team material; not treated as read.
• ★★★ | Rice Business Partners events archive — video discovery route | https://business.rice.edu/companies/rice-business-partners/events | checked 2026-08-22 | official institutional archive; reliability A; access full | October 11, 2022 entry exposes “Watch the video presentation” link to the unlisted official upload.
• ★★ | uListen transcript — Benchmark Part II | https://ulisten.ai/channels/acquired/benchmark-part-ii-the-dinner_ugTUtpkhBDw/transcript | 2022-10-17 | third-party navigation transcript; reliability C for custody; access full | Locates 12:16 Dunlevie relay; official audio controls.
• ★★ | NYU ICGF 2020 annual report | https://www.law.nyu.edu/sites/default/files/ICGF_Report_2020_opt.pdf | 2020 | official institutional report; reliability A for occurrence | Corroborates governance bootcamp; no content.
• ★★ | NYU ICGF 2021 annual report | https://www.law.nyu.edu/sites/default/files/ICGF%20Annual%20Report%20-%202021.pdf | 2021 | official institutional report; reliability A for program record | Corroborates occurrence; no content.
• ★★ | Ben Shewry Disrupt Yourself transcript | https://whitneyjohnson.com/wp-content/uploads/2020/10/DisruptYourselfPodcast_185BenShewry.pdf | 2020 | official transcript, printed/PDF pp. 7–8; retrospective witness moderate | Relays pandemic-era Dunlevie email and qualifies exact wording; not direct voice.
• ★ | Ben Shewry episode metadata | https://thedisruptionadvisors.com/ben-shewry/ | 2020 | official episode metadata; reliability A for episode | Custody control for Shewry relay.
• ★ | Trinity Venture Academy — Bernard Cho confirmation | https://www.linkedin.com/posts/bernardphcho_medarcade-ventureacademy-cambridge-activity-7448408486729760768-hcwS | 2026 | participant post; reliability C | Confirms Dunlevie and explicitly says speakers were off record.
• ★ | Trinity Venture Academy — Emily Atherton confirmation | https://www.linkedin.com/posts/emily-atherton-029205268_last-month-these-brilliant-venture-fellows-activity-7445394563512287232-FGR3 | 2026 | organizer/participant post; reliability C | Independent occurrence confirmation; no content.
• ★ | Untraced “real Venture Capital” fragment | https://www.linkedin.com/posts/eduardo-casarini-b4201a54_this-is-bruce-dunlevie-founder-of-vc-firm-activity-7057677902015799297-Jcw0 | 2023 | secondary social paraphrase; reliability C | Original venue/date/media not recovered; excluded unresolved lead.
• ★ | Trinity Fellow Benefactor admission liturgy | https://trinitycollegechoir.com/wp-content/uploads/2021/12/2017-03-17.pdf | 2017-03-17 | official prepared ceremonial text; reliability A for ceremony | No investing content or audio; excluded.

Biography, formation, and institutional records

• ★★★ | Rambus 1999 proxy | https://s202.q4cdn.com/680194126/files/doc_financials/annual/1999/SEC-RMBS-1012870-98-3188.pdf | 1999 | SEC-filed proxy, printed pp. 10–11; reliability A; access full | Board date, career, education, MPAE/Benchmark chronology, and committee membership.
• ★★★ | Handspring 2003 Form 10-K | https://www.sec.gov/Archives/edgar/data/1091822/000089161803004948/f92645a2e10vkza.htm | 2003 | SEC annual filing; reliability A; access full | Best filing biography for 1995 Benchmark formation, MPAE tenure, Everex, Goldman, and Handspring board.
• ★★★ | Follow the Money | https://www.wired.com/1995/03/money-11/ | 1995-03 | contemporaneous report; reliability B; access full | TVI/MPAE split and expected Kagle-Dunlevie-Rachleff combination.
• ★★★ | David Marquardt CHM/NVCA oral history | https://archive.computerhistory.org/resources/access/text/2021/05/102792229-05-01-acc.pdf | interviews 2011–2012 | eyewitness oral history, printed pp. 56–58; reliability A/B; access full | Kagle's recruitment of Dunlevie, formation, equal economics, and team differentiation; BCG recollection uncertain.
• ★★★ | T. Bondurant French CHM/NVCA oral history | https://archive.computerhistory.org/resources/access/text/2021/09/102792280-05-01-acc.pdf | retrospective | LP/industry witness, printed pp. 66–68; reliability A/B; access full | TVI/MPAE genealogy, Benchmark formation, and loyalty/governance caution.
• ★★ | Burt McMurtry Bancroft oral history | https://digicoll.lib.berkeley.edu/nanna/record/218493/files/mcmurtry_burt.pdf?registerDownload=1&version=1&withMetadata=0&withWatermark=0 | retrospective | eyewitness oral history around 03-00:22:37; reliability A/B; access failed 403 | Formation route preserved but not treated as fully accessible.
• ★★★ | The New Way to Start Up in Silicon Valley | https://keithkrach.com/wp-content/uploads/2021/06/The-New-Way-to-Start-Up-in-Silicon-Valley.pdf | 1998-03-02 | Fortune scan; reliability B; access full | Period EIR model and Palm/Rambus narrative; fund economics reported, not audited.
• ★★ | Rice distinguished alumnus profile | https://news2.rice.edu/2001/05/10/distinguished-alumni-recognized-3/ | 2001-05-10 | institutional retrospective; reliability A/B; access full | Education, Andersen, career, Palm, and Benchmark; “own computer company” conflicts with filings.
• ★★ | University of Florida Dunlevie endowment profile | https://www.uff.ufl.edu/giving-opportunities/012357-elizabeth-wood-dunlevie-honors-term-professorship-endowment/ | current | institutional profile; reliability A/B; access full | Rice/Stanford dates and Andersen relationship history.
• ★★ | Getty appointment | https://www.getty.edu/news/bruce-w-dunlevie-joins-j-paul-getty-board-trustees/ | 2017 | institutional announcement; reliability A/B; access full | Education and trustee chronology.
• ★★ | Getty board | https://www.getty.edu/about/leadership-governance/board/ | current; checked 2026-08-22 | current institution page; reliability A | Current committee roles; not current Benchmark evidence.
• ★★ | Trinity Annual Record 2019 | https://www.trin.cam.ac.uk/download/trinity-college-annual-record-2019/?refresh=68a39b7ae58691755552634&wpdmdl=29299 | 2019 | official record, printed p. 19; reliability A; access full | 1977 history notation and 2017 Fellow Benefactor admission; not a degree record.
• ★★ | Stanford trustee roundup | https://news.stanford.edu/stories/2015/12/trustee-roundup-dec-121415 | 2015-12 | official report; reliability A; access full | Trustee term and Land & Buildings chair role.
• ★★ | Stanford Management Company CEO appointment | https://news.stanford.edu/stories/2015/03/wallace-lead-smc-032315 | 2015-03-23 | official report; reliability A; access full | Direct evidence Dunlevie chaired SMC's board.
• ★★ | Center for Entrepreneurial Studies gift | https://www.gsb.stanford.edu/experience/news-history/center-entrepreneurial-studies-sparks-bright-ideas | 2008-10-15 | Stanford statement; reliability A for prepared words; access full | Education/governance and program-design context.
• ★★ | West Coast VC sets up new European fund | https://sciencebusiness.net/news/72063/West-Coast-VC-sets-up-new-European-fund | 2007-06-20 | reported interview/firm-transition record; A for exact Dunlevie wording, B for container; full page checked 2026-08-22 | Benchmark Europe became Balderton and Dunlevie's stated rationale.

Palm, Handspring, and direct witness records

• ★★★ | Jeff Hawkins Computer History Museum oral history | https://archive.computerhistory.org/resources/text/Oral_History/Hawkins_Jeff/Hawkins_Jeff_1.oral_history.2007.102658113.pdf | 2007-07-30 | founder eyewitness transcript, printed pp. 12–13/PDF pp. 16–17; reliability A/B; access full | Approach through Geoworks, possible Tandy structure, and product-scope challenge before Pilot.
• ★★★ | Donna Dubinsky oral history, part 1 | https://archive.computerhistory.org/resources/access/text/2025/01/102792830-05-01-acc.pdf | retrospective | founder eyewitness, printed pp. 33–34; reliability A/B; access full | Dunlevie urged an independent device and later advised against a sale, but could not secure partner funding.
• ★★★ | Donna Dubinsky oral history, part 2 | https://archive.computerhistory.org/resources/access/text/2025/01/102792866-05-01-acc.pdf | retrospective | founder eyewitness, printed p. 4; reliability A/B; access full | Dunlevie willingness versus partner reluctance around Handspring financing.
• ★★★ | From initial idea to unique advantage | https://www.researchgate.net/publication/3228143_From_initial_idea_to_unique_advantage_the_entrepreneurial_challenge_of_constructing_a_resource_base | 2001 | academic Palm case; reliability B; access variable | Capital and product chronology; $8.5 million aggregate cash does not allocate MPAE economics.
• ★★★ | U.S. Robotics acquires Palm | https://www.penbasedcomputing.com/timeline/us-robotics-acquires-palm-computing-inc/ | 1995-09-01 | announcement reproduction; reliability A/B; access full | Roughly $44 million company consideration; not MPAE proceeds.
• ★★★ | Handspring/Palm S-4 | https://www.sec.gov/Archives/edgar/data/1100389/000119312503053454/ds4a.htm | 2003 | SEC merger filing; reliability A; access full | Dunlevie-attributed Handspring shares and 3.4% stake before combination.
• ★★★ | Handspring/Palm Schedule 13D | https://www.sec.gov/Archives/edgar/data/1091822/000119312503010276/dsc13d.htm | 2003 | SEC ownership filing; reliability A; access full | Separates Benchmark II fund shares from Dunlevie direct/trust shares.
• ★★★ | Handspring merger 8-K | https://www.sec.gov/Archives/edgar/data/1091822/000089161803002945/f90805e8vk.htm | 2003 | SEC current report; reliability A; access full | Combined-company share and Dunlevie/Doerr board appointment.

Deal, ownership, vehicle, and outcome records

• ★★★ | Bruce Dunlevie SEC reporting-person index, CIK 0001190211 | https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001190211&type=&dateb=&owner=include&count=200 | checked 2026-08-22 | regulator index; reliability A for filings, not leadership; access full | 177 filings returned in the queried result set; manual issuer-name deduplication produced at least 27 families; legal-control denominator only.
• ★ | Foundersuite Bruce Dunlevie profile | https://foundersuite.com/investors/bruce-dunlevie | checked 2026-08-22 | aggregator/discovery lead; reliability C; full page | Current page names 24 candidate companies; each requires independent verification and none is credited from this row alone.
• ★★★ | Rambus 1997 final prospectus | https://www.sec.gov/Archives/edgar/data/917273/0000950109-97-004098.txt | 1997-05-14 | SEC prospectus, printed pp. 52–57; reliability A; access full | MPAE V's $3 Series C block, pre-IPO shares, shared-control attribution, and disclaimer; incomplete realization.
• ★★★ | Rambus 2007 Form 10-K | https://www.sec.gov/Archives/edgar/data/917273/000119312507201474/d10k.htm | 2007 | SEC annual filing; reliability A; access full | Founding-era board service and confirmed company options-review/restatement facts.
• ★★ | Rambus chairman appointment | https://www.rambus.com/bruce-dunlevie-appointed-chairman-of-rambus-board-of-directors/ | 2008-05 | company release; reliability A/C; access full | Chair date and retrospective career claim.
• ★★ | Rambus retirement | https://www.rambus.com/bruce-dunlevie-steps-down-from-rambus-board-of-directors/ | 2011-06-10 | company release; reliability A; access full | Bounds 21-year board service.
• ★★★ | Genesys 1997 final prospectus | https://www.sec.gov/Archives/edgar/data/1036436/000092962497000749/0000929624-97-000749.txt | 1997-06-17 | SEC prospectus, printed pp. 47 and 54–57; reliability A; access full | Benchmark financing, consulting-services warrant, board role, shared voting, and ownership disclaimer.
• ★★ | Genesys/Alcatel acquisition report | https://www.latimes.com/archives/la-xpm-1999-sep-29-fi-15215-story.html | 1999-09-29 | contemporaneous report; reliability B; access full | Approximate $1.5 billion company value; not Benchmark proceeds.
• ★★★ | Geoworks 1996 proxy | https://www.sec.gov/Archives/edgar/data/922285/0000891618-96-001154.txt | 1996-07-09 | SEC proxy; reliability A; access full | Board tenure and separation of MPAE V, direct, and option shares.
• ★★★ | Geoworks 2004 Form 10-K | https://www.sec.gov/Archives/edgar/data/922285/000092189504001133/form10k05716_03312004.htm | 2004 | SEC annual filing; reliability A; access full | Operating cessation/recapitalization; Dunlevie's role at cessation is not established.
• ★★ | Kalpana — Buried Treasures | https://www.forbes.com/asap/1998/1005/067_print.html | 1998-10-05 | reported retrospective; reliability B; access variable | Attributes Kalpana to Dunlevie and reports ~$60 million for MPAE; distributions unavailable.
• ★★★ | Cisco announces Kalpana acquisition | https://newsroom.cisco.com/c/r/newsroom/en/us/a/y1994/m10/cisco-to-acquire-kalpana-leading-ethernet-switching-company.html | 1994-10-24 | acquirer announcement; reliability A; access full | 7,480,114-share company consideration; not MPAE proceeds.
• ★★★ | ServiceSource 2011 final prospectus | https://www.sec.gov/Archives/edgar/data/1310114/000119312511077440/d424b4.htm | 2011-03-24 | SEC prospectus; reliability A; access full | Board/customer rationale, Benchmark V stake, designation right, shared control, and disclaimer.
• ★★★ | ServiceSource CEO transition release | https://www.sec.gov/Archives/edgar/data/1310114/000113626114000376/exhibit99-1.pdf | 2014 | SEC-filed release; reliability A/B; access full | Lead-independent role and interim leadership/CEO search.
• ★★★ | ServiceSource 2019 voting 8-K | https://www.sec.gov/Archives/edgar/data/1310114/000119312519150610/d751642d8k.htm | 2019-05-17 | SEC current report; reliability A; access full | Exact failed-election totals and board's refusal of resignation.
• ★★★ | ServiceSource 2019 Form 10-K | https://www.sec.gov/Archives/edgar/data/1310114/000131011420000007/srev-12312019x10k.htm | filed 2020-02-20 | SEC annual filing, signatures; reliability A; access full | Dunlevie signed as director on 2020-02-19, preventing a false 2019 board-tenure endpoint.
• ★★★ | Concentrix acquires ServiceSource | https://ir.concentrix.com/news/news-details/2022/Concentrix-Announces-Acquisition-of-ServiceSource-a-Global-B2B-Digital-Sales-Company-05-09-2022/default.aspx | 2022-05-09 | acquirer announcement; reliability A; access full | ~$131 million including net cash; not Benchmark proceeds.
• ★★★ | 1Life Healthcare 2020 final prospectus | https://www.sec.gov/Archives/edgar/data/1404123/000119312520021820/d806726d424b4.htm | prospectus dated 2020-01-30; filed 2020-02-03 | SEC prospectus, management/principal-stockholder sections; reliability A; access full | Board since 2007, lead-independent role, Benchmark affiliation and V stake, and shared-control boundary; appointment route not stated.
• ★★★ | Amazon–One Medical agreement | https://www.aboutamazon.com/news/company-news/amazon-and-one-medical-sign-an-agreement-for-amazon-to-acquire-one-medical | 2022-07-21 | acquirer announcement; reliability A; access full | $18 per share and ~$3.9 billion including net debt.
• ★★★ | Amazon completes One Medical acquisition | https://press.aboutamazon.com/2023/2/one-medical-joins-amazon-to-make-it-easier-for-people-to-get-and-stay-healthier | 2023-02-22 | acquirer announcement; reliability A; access full | Confirms completion, not investor proceeds.
• ★★★ | Marin Software 2013 final prospectus | https://www.sec.gov/Archives/edgar/data/1389002/000119312513122202/d450382d424b4.htm | 2013-03-21 | SEC prospectus; reliability A; access full | Board/designation, selected Benchmark VI cost, and shares; original Series B cost missing.
• ★★★ | Marin Chapter 11 filing index | https://www.sec.gov/Archives/edgar/data/1389002/000095017025093994/0000950170-25-093994-index.html | 2025-07-03 | SEC filing index; reliability A; access full | Long-tail outcome eight years after board departure.
• ★★ | Pure Digital acquisition report | https://venturebeat.com/social/cisco-buys-pure-digital-for-590m-claims-market-leadership-in-video-capture | 2009-03-19 | reported transaction/board attribution; reliability B; access full | ~$590 million company value plus incentives; no Benchmark cash flows.
• ★★★ | WeWork 2019 S-1 | https://www.sec.gov/Archives/edgar/data/1533523/000119312519220499/0001193125-19-220499.txt | 2019-08-14 | SEC registration statement; reliability A; access full | Benchmark designation/board/stake, Series A right, shared control, and Neumann supervotes.
• ★★★ | WeWork 2021 stockholders agreement | https://www.sec.gov/Archives/edgar/data/1813756/000119312521308162/d188107dex102.htm | 2021 | SEC-filed agreement, Article II §2.1(c); reliability A; access full | One Benchmark-designated director, initially Dunlevie.
• ★★★ | WeWork 2022 proxy | https://www.sec.gov/Archives/edgar/data/1813756/000119312522098597/d150307ddef14a.htm | 2022 | SEC proxy; reliability A; access full | Separates Dunlevie Living Trust's $10 million PIPE from Benchmark VII.
• ★★★ | WeWork 2023 proxy | https://www.sec.gov/Archives/edgar/data/1813756/000114036123021868/ny20007007x2_def14a.htm | filed 2023-05-01 | SEC proxy, director biography; reliability A; access full | Dates lead-independent service from October 2021 and bounds board roles.
• ★★★ | WeWork 2023 Schedule 13D/A | https://www.sec.gov/Archives/edgar/data/1611050/000119312523074781/d488748dsc13da.htm | 2023 | SEC ownership filing; reliability A; access full | Separates Benchmark VII vehicle, trust, and direct shares.
• ★★★ | WeWork Special Committee decision | https://courts.delaware.gov/Opinions/Download.aspx?id=314180 | 2021 | Delaware Chancery opinion, pp. 7 onward; reliability A; access full | Committee formation, authority, alternatives, and SoftBank dispute.
• ★★★ | WeWork Dunlevie resignation 8-K | https://www.sec.gov/Archives/edgar/data/1813756/000181375623000046/we-20230508.htm | 2023-05-11 | SEC current report; reliability A; access full | Resignation effective May 9, no disagreement, and designation right.
• ★★★ | WeWork Chapter 11 8-K | https://www.sec.gov/Archives/edgar/data/1813756/000119312523271902/d522028d8k.htm | 2023-11-07 | SEC current report; reliability A; access full | Bankruptcy filing; not a complete fund/trust loss statement.
• ★★★ | Tabula shutdown | https://www.eetimes.com/tabula-set-for-march-shutdown/ | 2015-01-30 | specialist report; reliability B; access full | $200 million-plus reportedly raised and planned closure; ex-post pair for 2011 thesis.
• ★★★ | CarbonDrop 2023 Form D | https://www.sec.gov/Archives/edgar/data/1895123/000189512323000002/primary_doc.xml | filed 2023-11-22 | SEC exempt-offering notice; reliability A for filed fields; access full | Dunlevie CEO/director, related persons, and $13.5 million sold of $14 million offered; not investor identity, valuation, or audited proceeds.
• ★★★ | CarbonDrop 2026 Form D | https://www.sec.gov/Archives/edgar/data/1895123/000189512326000002/xslFormDX08/primary_doc.xml | filed 2026-04-21 | SEC exempt-offering notice; reliability A for filed fields; access full | Dunlevie CEO/director, related persons, and $16.175 million sold of $16.425 million offered; continuity and cumulative economics remain unproved.
• ★★★ | UC Regents alternative-investments schedule | https://regents.universityofcalifornia.edu/regmeet/nov03/603spreadsheetnov.pdf | through 2003-03-31 | LP schedule, p. 6; reliability A; access full | MPAE V commitment, cash in/out, NAV, reported: 5.51x net multiple, reported: 46.1% net IRR; one LP's net experience.
• ★★ | Dartmouth MPAE V archive record | https://archives-manuscripts.dartmouth.edu/repositories/2/archival_objects/391308 | 2003-12-31 | finding aid, box 12237; reliability A for route; by request | Potential vehicle archive.
• ★★ | Benchmark Capital Holdings IAPD | https://adviserinfo.sec.gov/firm/summary/160352 | current | regulator adviser page; reliability A; access full | Current registration route; no historical returns.
• ★★★ | The Gilt-Edged Venture Capitalist | https://www.forbes.com/2001/02/06/0207VC.html | 2001-02-06 | skeptical contemporaneous report; reliability B; access variable | Reports first-fund return and failures without audited methodology.
• ★★★ | Benchmark Capital: All Sizzle, No Burn | https://www.forbes.com/asap/2000/0529/121_print.html | 2000-05-29 | contemporaneous firm denominator; reliability B; access variable | 1999 capital, company, pace, board, selection, and IPO figures; snapshot, not career performance.

Outcome controls linked from the written corpus

• ★★★ | In the Matter of Advanced Equities | https://www.sec.gov/litigation/admin/2012/33-9362.pdf | 2012 | SEC administrative order; reliability A; access full | Control against treating speed/scale praise as adequate intermediary diligence; entity/respondent findings are not Dunlevie findings.
• ★★★ | Semtech Form 10-Q — EnVerv selected-asset acquisition | https://www.sec.gov/Archives/edgar/data/88941/000162828015006974/smtc-07262015x10q.htm | filed 2015 | SEC quarterly filing; reliability A; access full | $4.9 million selected-assets transaction, not proof of an EnVerv equity acquisition or Benchmark return.
• ★★ | WeWork acquires Euclid | https://www.wework.com/ideas/newsroom-landing-page/wework-acquires-spatial-analytics-leader-euclid | 2019-02-11 | acquirer announcement; reliability A for transaction announcement, C for promotion; access full | Technology repositioning toward workplace spatial analytics; no price disclosed.
• ★★ | Impresse shuts down | https://whattheythink.com/news/11448-impresse-shuts-down-details-scarce/ | 2001 | trade report; reliability B; access full | Adverse company outcome; incomplete investor economics.
• ★★ | PrintCafe buys Impresse intellectual property | https://whattheythink.com/news/11221-more-details-printcafe-buys-impresse-intellectual/ | 2001 | trade report; reliability B; access full | Asset purchase must not be described as a successful company exit.
• ★★ | RMI's Long, Strange Trip | https://www.lightreading.com/semiconductors/rmi-s-long-strange-trip | retrospective | specialist report; reliability B; access full | Raza/RMI entity and strategy history; do not collapse Foundries entities without records.
• ★★ | WeWork appoints Sandeep Mathrani as CEO | https://www.wework.com/newsroom/wework-appoints-sandeep-mathrani-as-ceo | 2020-02-01 | company announcement; reliability A for appointment, C for promotion; access full | Later leadership outcome controlling the 2019 co-CEO endorsement.
• ★★★ | SoftBank settles WeWork litigation | https://group.softbank/en/news/press/20210227 | 2021-02-27 | counterparty announcement; reliability A for settlement terms disclosed, C for characterization; access full | Settlement on confidential terms; the source supplies no merits adjudication and therefore controls the committee’s advocacy.
• ★★★ | SoftBank Group — WeWork commitments to date | https://group.softbank/en/news/press/20200402_0 | 2020-04-02 | counterparty announcement; reliability A for commitments disclosed; access full | More than $14.25 billion committed as of April 2020; controls the automatic-caption unit without proving Dunlevie’s exact spoken word.

Current 2026 translation controls

• ★★★ | Stanford AI Index 2026 — Research and Development | https://hai.stanford.edu/ai-index/2026-ai-index-report/research-and-development | 2026 | research-institute report with disclosed chapter methods; reliability A/B; access full | Current AI model, compute, foundry, disclosure, and infrastructure concentration; used only for dated translation, not Dunlevie attribution.
• ★★★ | European Commission — GPAI provider obligations | https://digital-strategy.ec.europa.eu/en/faqs/guidelines-obligations-general-purpose-ai-providers | current through 2026-08-22 | official regulatory guidance; reliability A for Commission position; access full | Current model-provider and downstream-information obligations; guidance is not a court interpretation.
• ★★★ | Delaware Code, Title 8 §144 | https://delcode.delaware.gov/title8/c001/sc04/ | current through 2026-08-22 | official statutory text; reliability A; access full | Current controlling-stockholder transaction framework; used as a translation boundary, not retroactively applied to WeWork.
• ★★★ | NVCA/PitchBook Q2 2026 Venture Monitor | https://nvca.org/pitchbook-nvca-venture-monitor/ | Q2 2026 | trade-association/data-provider report; reliability A/B; access full | Record aggregate activity with AI, mega-round, manager, and exit concentration; headline capital is not universal company access.

Adverse, legal, and attribution controls

• ★★★ | Kelley v. Rambus dismissal order | https://law.justia.com/cases/federal/district-courts/california/candce/5%3A2007cv01238/189677/263/ | 2008-12-09 | federal order mirror; reliability A/B; access full | Pleaded backdating allegation involving Dunlevie called conclusory/unsupported; dismissal granted.
• ★★★ | Rambus 2009 Form 10-K litigation summary | https://www.sec.gov/Archives/edgar/data/917273/000091727310000014/f10k2009.htm | 2010 | SEC annual filing; reliability A; access full | Records Ninth Circuit affirmation; issuer description is not independent analysis.
• ★★★ | FTC Rambus matter | https://www.ftc.gov/legal-library/browse/cases-proceedings/011-0017-rambus-inc-matter | 2002–2009 | regulator docket; reliability A; access full | Corporate standards case ultimately dismissed; Dunlevie was not respondent.
• ★★★ | Vernet v. WeWork complaint | https://assets.bwbx.io/documents/users/iqjWHBFdfxIU/rWo.bu8AmUc4/v0 | 2020-06-03 | filed complaint; reliability A for allegations only; access full | Names Dunlevie and alleges director responsibility; no merits finding.
• ★★ | Vernet docket mirror | https://dockets.justia.com/docket/california/candce/4%3A2020cv03686/360438 | through 2021-04 | docket index; reliability A/B; incomplete | Records consolidation but not final merits disposition; PACER needed.
• ★★★ | Finger v. 1Life Healthcare docket | https://dockets.justia.com/docket/new-york/nysdce/1%3A2022cv07451/585635 | 2022 | docket index; reliability A/B; access full for public events | Merger complaint voluntarily dismissed before answer; no merits finding.
• ★★★ | ServiceSource board criticism | https://corpgov.com/more-blood-in-the-water-servicesources-board-ignores-vote-to-oust-two-directors/ | 2019-06 | governance commentary; reliability B; access full | Independent critique of retention, compensation advice, and underperformance; SEC filing controls vote.
• ★★★ | eBoys review — Salon | https://www.salon.com/2000/05/24/venture/ | 2000-05-24 | contemporary critical review; reliability B; access full | Critiques machismo, insularity, equal-partnership narrative, and luck-versus-skill anxiety.
• ★★★ | ARM launch report | https://www.latimes.com/archives/la-xpm-1990-11-28-fi-4993-story.html | 1990-11-28 | contemporaneous report; reliability B; access full | Acorn, Apple, and VLSI are formation parties; contradicts unsupported Dunlevie attribution.
• ★★ | ARM formation corroboration | https://www.techmonitor.ai/technology/acorns_risc_design_into_venture_with_apple_vlsi/ | 1990 | contemporaneous industry report; reliability B; access full | Independent formation-party control; no Dunlevie/MPAE link.
• ★★★ | Metrick & Yasuda venture-capital text excerpt | https://depot.som.yale.edu/icf/papers/fileuploads/2689/original/2011_ICF_WPS_The_Best_Venture_Capitalists_-_Metrick.pdf | through 2007-03 | academic text, printed pp. 89–90; reliability B; access full | Benchmark I — reported: 42x; Benchmark IV — reported: 0.2% net IRR on a different basis; do not collapse vintages.

Books, one-remove testimony, and completed-peer controls

• ★★★ | eBoys — publisher page | https://www.penguinrandomhouse.com/books/174866/eboys-by-randall--e-stross/ | 2000 | embedded-reporter book catalog; reliability B; licensed/restricted | High-value inside narrative; conversations are Stross reconstructions, not transcript.
• ★★ | eBoys — author catalog | https://www.randallstross.com/ | current | author catalog; reliability A/B; access full | Legal route and edition metadata; not a substitute for body text.
• ★★ | Done Deals — catalog | https://books.google.com/books/about/Done_Deals.html?id=ZdG3wAEACAAJ | 2000 | book catalog; reliability B; access limited | Kagle chapter offers firm-origin testimony; body not openly processed.
• ★★★ | Bill Gurley on Tim Ferriss, transcript #651 | https://tim.blog/2023/01/25/bill-gurley-transcript/ | 2023-01-25 | official transcript of partner testimony; reliability A for Gurley, B for attribution; access full | “What could go right?” and relationship-sourcing lineage; not Dunlevie's direct publication.
• ★★ | Bill Gurley X attribution | https://x.com/bgurley/status/572255806051131392 | 2015-03-02 | original Gurley post; A for Gurley's wording, B for Dunlevie attribution; current rendering/login barrier checked 2026-08-22 | Earlier public “What could go right?” lineage; text was identified but live rendering was not reliably retrievable.
• ★★★ | Andy Rachleff interview — democratizing investment | https://www.gsb.stanford.edu/insights/andy-rachleff-democratizing-investment-business | 2013-06-25 | institutional interview; reliability A for Rachleff, B for attribution; access full | Trust-first negotiation framework.
• ★★★ | Andy Rachleff — Learning Leader transcript | https://podscripts.co/podcasts/the-learning-leader-show-with-ryan-hawk/291-andy-rachleff-what-do-you-uniquely-offer-that-people-desperately-want | episode 291 | third-party navigation transcript, 00:13:35–00:16:57; A for Rachleff audio when verified, B for Dunlevie attribution, C for transcript custody | Fuller trust-first rule and boundary for exploitative counterparties; original audio remains controlling.
• ★★★ | Invest in Relationships, Not Transactions | https://medium.com/aleph-vc/invest-in-relationships-not-transactions-f4d6cf9f363b | 2021 | Michael Eisenberg recollection; reliability B; access full | Dunlevie's “low market share” framing and relationship sourcing.
• ★★ | Core Product Value and Entrepreneurial Success | https://25iq.com/2017/03/24/core-product-value-and-entrepreneurial-success/ | 2017-03-24 | private-conversation recollection; reliability B; access full | Strongest route for pattern-recognition attribution; original conversation unpublished.
• ★★ | On board management — Michael Smerklo | https://www.nextcoastventures.com/blog/2017-01-29-on-board-management-wise-words-from-4-silicon-valley-legends | 2017-01-29 | founder witness essay; reliability B; access full | Information-gap lesson for directors; one-remove governance evidence.
• ★★ | Andy Rachleff — completed source package | https://github.com/abhishekgarg8/venture-midas-cannon/blob/main/investors/031-andy-rachleff/sources.md | accessed 2026-08-22 | completed same-firm peer map; reliability varies | Comparator for formation/equal economics/eBay attribution; underlying sources control.
• ★★ | Reid Dennis — completed source package | https://github.com/abhishekgarg8/venture-midas-cannon/blob/main/investors/004-reid-dennis/sources.md | accessed 2026-08-22 | completed peer map; reliability varies | Early-vintage vehicle attribution and archival gaps.
• ★★ | Arthur Rock — completed source package | https://github.com/abhishekgarg8/venture-midas-cannon/blob/main/investors/006-arthur-rock/sources.md | accessed 2026-08-22 | completed peer map; reliability varies | Board influence versus personal-return evidence.
• ★★ | Peter Crisp — completed source package | https://github.com/abhishekgarg8/venture-midas-cannon/blob/main/investors/008-peter-crisp/sources.md | accessed 2026-08-22 | completed peer map; reliability varies | Institutional chronology, denominator, and uncertainty comparator.

Coverage assessment

Inventory and access

• Inventory: 153 annotated source rows: 30 mixed voice/written, 22 talk/archive, 16 biography/formation/institutional, 8 Palm/Handspring, 48 deal/vehicle/outcome, 4 current translation controls, 11 adverse/attribution, and 14 book/testimony/peer routes.
• Quality: 85 ★★★, 55 ★★, and 13 ★ routes. Tier 1 custody includes SEC filings, court opinions, regulator dockets, official institution records, company/acquirer announcements, LP schedules, and oral histories.
• Access: E148/E218, eBoys, Done Deals, and the Moe interview require purchase/licensed access; Rice 2022 is full official direct video with auto-captions; NYU and Amity lack public transcripts; Wayback and Berkeley routes failed. A processed metadata row is never treated as body evidence.
• Search ledger: Variants included Bruce Dunlevie, Bruce W. Dunlevie, Bruce Wall Dunlevie, the pleading error Bruce S. Dunlevie, and punctuated/unpunctuated Merrill Pickard/MPAE forms; searches were banded across 1980s–1994, 1995–2001, 2002–2009, 2010–2017, and 2018–2026. Venues included Benchmark, Stanford, Rice, NYU, CHM, Berkeley, Congress/GovInfo, SEC, courts, YouTube, Vimeo, Medium/RSS, LinkedIn, X, podcast indexes, and web archives. Barriers are logged row by row.
• Historical context: Wired 1995 and Fortune 1998 cover formation/EIR structure; Forbes 2000/2001 and Irish Times 2001 freeze the dot-com cycle; Metrick–Yasuda provides vintage controls. These sources contextualize rather than prove personal alpha.
• Saturation: Three final year/source-class breadth sweeps produced no new substantive direct-voice class; remaining gaps require private, licensed, archival, or LP records.

Corpus reconciliation

• Writings: No Dunlevie-authored essay, newsletter, book, blog corpus, or Medium story was located. The mapped 30-row record separates 2 restricted Stanford cases, 1 partial two-page book interview, 2 relayed Acquired appearances, fully accessed reported passages, and prepared/joint statements; Fast Company remains restricted. writings.md must reconcile every row and label each mode rather than call the whole set direct voice.
• Talks: The TALKS denominator is 7/7 note artifacts across 6 families: 1 full public direct Rice appearance, 2 Acquired host relays from 1 private interview, 3 access-bounded appearances, and 1 narrated contextual video; Trinity was off record.
• Maxims: “What could go right?”, the trust-first rule, “low market share,” and pattern recognition are one-remove partner/witness attributions and must retain that lineage.

Deal and vehicle denominators

• Discovery versus control: Foundersuite names 24 candidate companies. The queried SEC result set returned 177 filings and manual issuer-name deduplication produced at least 27 families as of 2026-08-22. Each is discovery/legal-control evidence, not a led-deal count.
• Primary decision-right/origination core — researcher inference · provisional: Palm, Handspring, Rambus, Genesys, Geoworks, ServiceSource, One Medical, Marin, and WeWork. Primary filings or strong witness records establish an origin, board, designation, or consulting role; One Medical establishes Benchmark affiliation and long board service, not a designation or appointment route.
• Weaker attributable cases — researcher inference · provisional: Tabula has a Benchmark investment plus Dunlevie thesis quote but no formal decision right; Kalpana has reported deal attribution only; Pure Digital has secondary board attribution only.
• Vehicle map — researcher inference · moderate: MPAE with vehicle unspecified is connected to Palm; MPAE V is explicit for Rambus/Geoworks; Forbes reports unnumbered MPAE economics for Kalpana; Benchmark II maps to Handspring/Palm, V to ServiceSource/One Medical, VI to Marin, VII/AIV to WeWork, and Genesys filings name unnumbered Benchmark entities. Do not pool them.
• Outcome boundary: MPAE V's UC schedule is one LP's net experience. Company sale values do not reveal fund proceeds. First-fund figures are reported/unaudited. No public record establishes Dunlevie's career IRR, MOIC, hit rate, opportunity denominator, write-off rate, or carry share.

Adverse and attribution controls

• ARM: Formation reports identify Acorn, Apple, and VLSI. No primary or contemporary source connected Dunlevie/MPAE; remove the unsupported roster claim.
• WeWork: Preserve Neumann's supervotes, Benchmark VII, Dunlevie's board/special-committee work, SoftBank, and the separate family-trust PIPE. Bankruptcy is real; fund/trust recovery is incomplete.
• Rambus: The company confirmed options-accounting failures and a restatement. A pleaded personal allegation was dismissed as conclusory/unsupported; the FTC standards case named Rambus, not Dunlevie.
• ServiceSource: SEC voting evidence and independent criticism support a direct accountability case without implying illegality.
• Tabula, Geoworks, Marin: Use shutdown, operating cessation/recapitalization, and later bankruptcy. No cash-flow record proves Dunlevie-specific write-offs, and Marin failed long after his exit.
• Negative search: No Dunlevie-specific SEC enforcement action, criminal proceeding, or regulator order was located in searched public routes. This is bounded, not a clearance statement.

Highest-value unresolved retrievals

1. MPAE partnership/fund files, LP statements, memoranda, Palm/Kalpana cap tables, and distributions.
2. Benchmark agreements, fund cash flows, board books, IC records, partner labor/carry, and complete opportunity/portfolio/write-off schedules.
3. Licensed E148/E218, eBoys, Done Deals, and Finding the Next Starbucks with page-level audit.
4. A creator transcript, corrected captions, or audio-level quote verification for Rice; original recordings/transcripts for NYU and Amity; and Acquired’s private interview audio/transcript.
5. Final PACER/court disposition for consolidated WeWork investor litigation and private recovery data.

Source notes