Reid Dennis

1. Snapshot and testable edge thesis

Predictions

[researcher inference · moderate] If the edge thesis is true, unseen investment memoranda and board files should show recurring technical-to-customer questions, attributable referrals, explicit job-relevant people tests, plan revisions after new evidence, and vehicle/capacity checks before commitments. Those patterns should appear in ordinary Dennis-led cases, not only Ampex, Recognition, or Seagate, and should predict better outcomes after price, ownership, sector, vintage, partner work, and access are controlled. Full prediction and falsifier statement · searched denominator

Falsifiers

[researcher inference · moderate] The thesis fails or narrows if complete files show that access selected the opportunity set, partners supplied the decisive judgment, Dennis's technical screening did not distinguish wins from false positives, board intervention did not change matched outcomes, or ex-outlier cash flows were ordinary. Reference, Apple, public securities held to zero, Collagen's failed first use, person-selection errors, fund dispersion, and Marquardt's transferability critique are already disconfirming observations—not footnotes to be explained away. Failure postmortems · role attribution · portfolio forensics · Marquardt, printed pp. 23–30

2. Core philosophy

What Dennis actually said

• Entrepreneurs own creation. [investor-stated · strong] Dennis explicitly said venture investors do not create companies; entrepreneurs do. Investor finance, imagination, support, and influence are participating inputs. Raw transcript 2, pp. 6–7
• Capital is necessary but insufficient. [investor-stated · strong] His 1979 prepared statement treats qualified managers and key employees as a second scarce input and employee equity as compensation for startup cash and security risk. Prepared statement, printed pp. 1060–1062
• People judgment dominates, but has no formula. [investor-stated · moderate] Dennis preferred relevant experience, trust, vision, and small teams, while crediting luck and later calling people mistakes his largest errors. Raw transcript 2, pp. 26–29 · CHM video, 01:06:29–01:10:44
• Plans change. [investor-stated · moderate] Financing begins with a hypothesis; boards learn how the team behaves, recruit or support where possible, and may replace leadership when repeated evidence requires it. Raw transcript 2, pp. 7–10 and 27–28
• Technology, company necessity, price, and timing are separate. [investor-stated · strong] His January 2000 warning expected continued Internet adoption while rejecting prevailing prices and refusing to predict the correction date. Almanac, body ¶¶4 and 11–19
• Company building is not transaction completion. [investor-stated · moderate] Dennis contrasted operating consequences with investment-banking transactions and admired long-lived company building, but acknowledged that LP liquidity can force action. Raw transcript 2, pp. 29–32

Synthesis

[researcher inference · moderate] The integrated sequence is: use technical literacy and networks to find a live problem; test whether a differentiated company and role-relevant team can solve it at the offered price; name the person and vehicle making the decision; finance within the fund's capacity and duration; treat the plan as revisable; help without claiming the entrepreneur's work; and keep company viability, investor liquidity, and realized fund economics separate. This is a reconstruction across decades, not a recovered Dennis playbook. The detailed procedures and counterexamples live in mental models.

3. Sourcing

Referral translation

[retrospective witness · moderate] Ampex came through a friend of Dennis's brother, and brokers and engineers later routed technical founders to him because he could interpret technical claims for a financial audience. Dennis oral history, printed pp. 13–18 [researcher inference · moderate] The repeatable mechanism is not “warm intros win”; trusted nodes reduce search cost and route specialized claims to someone able to interrogate them, while family access and employer reputation remain confounders. Network model and exceptions

Syndication and sponsor ownership

[retrospective witness · moderate] The Group turned member-sourced leads into lunches, pooled checks, follow-on calls, and one member's board responsibility. Inc., ¶¶4–5 [researcher inference · strong] Measurex shows that source, capital, and outcome were distributed among Draper, Wythes, Bancroft/Bessemer, Bossen, Reid, Peggy, and others; ROLM makes a different sponsor boundary legible because McMurtry sourced, disclosed and recused, others approved, and McMurtry boarded. Measurex decision lab · ROLM decision lab · McMurtry, printed pp. 30–32

Geographic proximity as operating capacity

[investor-stated · moderate] Dennis said closeness helped investors stay in contact and spot problems. [contemporaneous record · moderate] Exabyte was in Boulder, and a 1983 report describes investments outside the Valley. Christian Science Monitor, body ¶¶7–9 [researcher inference · moderate] Travel time is a board-capacity cost, not a universal radius rule; use it only when the work actually requires frequent presence and remote substitutes are inadequate. Direct spoken synthesis

Funnel boundary

[retrospective witness · provisional] Dennis later claimed IVP funded roughly 1.5%–2.5% of opportunities, but intake unit, years, stages, duplicates, and definition are absent; The Group's twenty-five-company denominator is likewise incomplete. CHM video, 01:15:31–01:17:05 · reported Group outcomes [researcher inference · moderate] The Leza account exposes a second risk: familiar networks can improve trust while narrowing who reaches the funnel. A modern reconstruction must record source, sponsor, rejection reason, identity-independent job evidence, and false negatives; no cold-inbound or complete pass ledger survives here. Network model and funnel audit · SFGate, ¶¶3–14

4. Picking

Technology must reach a present customer at investable economics

[researcher inference · moderate] Dennis's defensible screening advantage begins with technical understanding but ends in a commercial/security judgment. [retrospective witness · moderate] Exabyte reused video hardware in data storage; Reference's 14-inch optical system was technically plausible but too early and cumbersome; Collagen's burn indication failed before a different use emerged; Apple was passed because the remembered price/revenue relationship lay outside Dennis's experience. Technology-to-economics model · Exabyte and Reference, printed pp. 52–56 · Apple, raw transcript 2, pp. 39–40

[researcher inference · moderate] The practical gate asks four different questions: does the technology work; is there a current, paid and sufficiently large customer problem; is the company differentiated from later or unnecessary entrants; and do valuation, preferences, ownership, dilution, and time permit an investor return? A “yes” at one level cannot substitute for another. Technology-to-economics model · Almanac, body ¶¶11–19

People are tested by role-relevant evidence, not resemblance

[investor-stated · moderate] Dennis preferred entrepreneurs with relevant company experience, advised aspiring investors to gain operating experience, called people his largest errors, acknowledged Google-like exceptions, and repeated an acquaintance's anti-overweight-CEO rule. Raw transcript 2, pp. 26–32 and 37–38 [researcher inference · moderate] Seagate's team, Wilf Corrigan at LSI, and Exabyte's application shift fit the preference, but the body-size heuristic is rejected as stigmatizing and unsupported; Leza's allegation and Dennis's response further show how familiarity and credentials can become closed-network proxies. People-judgment evidence and limits · SFGate, ¶¶1–14

[researcher inference · moderate] The transferable test is falsifiable: name the job to be done; identify evidence of learning, execution, integrity, and complementary teamwork; pre-register what reference or work-sample evidence would change the judgment; and audit comparable candidates who were never referred. [investor-stated · moderate] Dennis's retirement admission supplies the reciprocal rule for investors—hand off when the market exceeds the team's competence. Role-relevant competence model · raw transcript 2, pp. 41–42

Red flags and kill criteria

• [researcher inference · moderate] A technology thesis without current use, form-factor fit, or willingness to pay: Reference. Reference decision lab
• [researcher inference · moderate] A fashionable category without necessity or differentiation: Dennis's 2000 late/unnecessary Internet-company warning. Almanac, body ¶¶11–19
• [researcher inference · moderate] A good company at a security price outside a defined return case: Apple, while preserving that the pass may still be wrong. Apple decision lab
• [researcher inference · strong] Person confidence based on familiarity, appearance, or prestige rather than role evidence: the body-size rule and access dispute. People-judgment evidence and limits
• [researcher inference · moderate] A team or market beyond the partnership's current knowledge: Dennis's succession admission. Raw transcript 2, pp. 41–42
• [researcher inference · strong] A process that cannot name source, voter, objector, vehicle, board actor, and cash-flow outcome: much of the surviving selected-company record. Material-case responsibility matrix · searched denominator

5. Deal and portfolio mechanics

Portfolio objective and denominator

[investor-stated · moderate] Dennis disliked designing around only one or two wins in ten and remembered a higher viable-company frequency. Inc., ¶¶13–14 [researcher inference · strong] That is an objective, not performance proof, and The Group's count is unaudited. Reported Group outcomes [contemporaneous record · strong] UC's IVP III–VIII observations range from 1.07x/1.5% to 6.80x/93.6% net to one LP. UC report, IVP III–VIII [researcher inference · strong] The rows show vintage dispersion, but combining their paid-in and distributions is not a valid aggregate TVPI or IRR because each has its own timing. Fund table and arithmetic boundary

[researcher inference · strong] Report company-count survival, capital-weighted returns, ownership, reserves, gross/net cash flows, DPI/TVPI/IRR, and ex-top-one/top-vintage views separately. The record supports no 80/20 company computation because only 10 funded cases were reconstructed and the career denominator and cash flows remain absent. Viability/economics model · coverage denominator

6. Vehicle and incentive fit

[researcher inference · moderate] Vehicle design changes the feasible strategy: source of money, legal authority, check size, reserve capacity, partner time, holding horizon, liquidity obligation, and who receives the economics. It does not prove patience or skill. Applied demonstrates that a strong company can outlive a fund's expected clock; the IVA split and Redpoint formation show that organizational continuity cannot be inferred from a brand. Applied liquidity postmortem · Redpoint formation

[researcher inference · insufficient public record] No public evidence establishes Dennis's fee, carry, clawback, personal co-investment allocation, universal reserve policy, or succession economics. [researcher inference · moderate] A modern reader should compare an opportunity with the actual partnership agreement, mandate, side letters, decision rights, key-person provisions, reserve model, and liquidity alternatives—not with the generic label “venture fund.” Vehicle gaps and retrieval boundary

7. Team attribution and work with founders

[investor-stated · strong] Venture was collective in Dennis's own framing: entrepreneurs create companies, while investors finance, question, support, recruit, govern, and sometimes replace. Raw transcript 2, pp. 6–10 [researcher inference · moderate] The analytical unit is therefore a role ledger, not a hero label. Material-case responsibility matrix

Conflicts, dissent, and intervention

[researcher inference · moderate] ROLM's disclosed pre-existing position and recusal is the clearest conflict-control story; LSI is the clearest exception because Dennis admits overriding the spirit of his consensus norm when he expected the allocation to vanish, and the good outcome cannot validate the governance breach ex post. Recognition and an unnamed rescue show remembered CEO replacement after accumulated evidence; Applied shows documented rights used against company preference. ROLM decision lab · LSI decision lab · Recognition and Applied postmortems

[researcher inference · moderate] The transferable process stores pre-discussion views, the sponsor and objector, conflict and recusal, legal decision right, action, board owner, later intervention, and cash-flow result. This makes learning possible without stealing credit from founders or laundering a lucky override into policy. Roles and decision-rights model

8. Views on the craft

Capital and people are complements

[investor-stated · strong] Dennis described entrepreneurs as the motive force and capital as fuel or lubricant, while his 1979 statement made management supply a binding growth constraint. Prepared statement, p. 1061 · Stanford ETL/CHM talk synthesis [researcher inference · moderate] The investor's job is therefore to identify the scarce complement—people, technical truth, customer access, governance, or time—not automatically enlarge the round. Capital-plus-complements model

Company building precedes transactions, but investors still have clocks

[investor-stated · moderate] Dennis criticized a finance-first, transaction-oriented craft, admired Arthur Rock's quiet committee work at Intel as company building, favored building to last, and described IPOs and distributions as financing or liquidity events rather than company endpoints. Raw transcript 2, pp. 29–31 [contemporaneous record · strong] Applied documents that a fund can seek liquidity while an operating company prefers continued privacy. Applied registration record [researcher inference · moderate] The doctrine is conditional patience: define which actor needs liquidity, by when, and through what rights before calling a long hold virtuous. Company-building and patience model

Optimism needs disconfirming evidence

[investor-stated · moderate] Dennis called optimism necessary but repeatedly credited luck and disappointment. Raw transcript 2, pp. 6–10 and 17–18 [researcher inference · moderate] The best version of his craft is hopeful participation under falsifiable tests; the worst is intuitive confidence protected by survivor stories. Reference, Collagen's first indication, and public positions held to zero are the controls. Case-specific failure postmortems

Capital cycles change conduct, not just supply

[researcher inference · moderate] Across 1974, 1979, 2000, 2005, and 2008/09 records, Dennis alternately welcomed useful new capital and warned that excess money could inflate price, accelerate decisions, enlarge rounds, weaken selection, and push firms toward sales; these are mechanism claims, not a proven timing signal. Written intellectual chronology · spoken intellectual chronology [researcher inference · moderate] The reader must measure fund/partner capacity, price, round size, follow-on burden, and exit liquidity rather than infer “more capital is good” or “boom means immediate crash.” Capital-cycle model

9. Signature bets and decision process

10. Mental models and maxims

The canonical inventory reconciles duplicate note-level rules into ten models. Each model includes provenance, context, applications, at least one exception, and a current translation; the two evidence-thin models disclose their own attempted-route shortfalls.

1. Underwrite the chain from technology to security economics.
2. Test role-relevant competence; reject proxies and hand off beyond competence.
3. Use networks for leads and syndication, then audit sponsor and funnel.
4. Map capital plus scarce complements.
5. Match vehicle capacity, partner time, duration, and liquidity to the opportunity.
6. Make roles, conflicts, dissent, and decision rights legible.
7. Treat plans as hypotheses and boards as update mechanisms.
8. Build companies rather than transactions, with explicit patience boundaries.
9. Separate company viability from investment economics and audit outliers.
10. Scale selection capacity through capital cycles without timing pretense.

11. Evolution and contradictions

Three contradictions should remain productive rather than be harmonized away:

1. Consensus versus urgency — [researcher inference · moderate]: Dennis preferred reasoned consensus but overrode it at LSI; outcome cannot resolve process quality. LSI decision lab
2. Openness versus familiarity — [researcher inference · moderate]: Dennis described Silicon Valley financing as open to good ideas, while the Leza record exposes a different gate in venture hiring and a homophily risk. SFGate source analysis
3. Patient company building versus fiduciary liquidity — [researcher inference · moderate]: the build-to-last ideal collides with distributions, personal needs, fund lives, and Applied's demand rights. Applied decision lab · company-building and patience model

12. Failures and limits

[researcher inference · strong] The record does not demonstrate a consistent postmortem practice, a changed checklist after each miss, a formal reserve rule, or a Dennis-authored valuation method. [researcher inference · moderate] His honesty about luck, people errors, Apple, Reference, and public losses improves the evidence base but does not itself prove learning or performance. Postmortems and missing controls · retrieval ledger

13. Historical operating environment

What a contemporary Dennis could use

• [researcher inference · moderate] Stanford and Peninsula engineering networks; brokers seeking help with technical issuers; employer research autonomy. Historical environment
• [researcher inference · moderate] Family/trust capital for a small personal check and an insurer balance sheet for larger corporate opportunities. Ampex access record · vehicle record
• [researcher inference · moderate] Early electronics, instrumentation, defense-adjacent and computing demand; smaller rounds and fewer specialized competitors. Historical environment
• [researcher inference · moderate] Repeat personal syndication before a deep institutional venture market. Group operating record
• [researcher inference · moderate] An American Express anchor and insurer credibility for IVA fundraising. Vehicle record
• [researcher inference · moderate] Later tax, pension, public-market, and limited-partnership changes that enlarged the capital pool. Historical environment · policy chronology

What cannot be carried forward as personal skill

[researcher inference · strong] Inherited capital, trustee approval, employer capital, privileged geography, low early valuations, institutional scarcity, policy changes, founders' execution, co-investor work, and favorable public markets are conditions or other actors. Dennis's own policy claims rely partly on temporal association; the official chronology does not prove that he or one reform caused venture growth. Historical environment · 1979 statement and evidentiary boundary

Denominator and hindsight controls

[researcher inference · strong] The public record is survivor-selected: 33 candidate names were searched, 26 met inclusion logic, only 11 entered decision labs, and only 10 were funded. No lab satisfies a full decision schema; 15 included names remain unresolved. This does not mean the missing companies failed. It means sector, stage, hit rate, ownership, reserve behavior, and career return cannot be inferred from the visible winners. Coverage accounting

14. Modern VC translation (as of 2026-08-02)

All transfer claims below are [researcher inference · moderate] unless marked otherwise. Current premises are bounded to sources already processed in the Reid record. NVCA/PitchBook reports more than $400 billion raised by U.S. startups in the first half of 2026 with capital heavily concentrated in AI and $100-million-plus rounds; its 2026 Yearbook reports AI at 65.4% of 2025 deal value and 487 mega-deals at 67% of value. These are association/market aggregates, not a bubble verdict or a Dennis-model efficacy test. NVCA Q2 2026 Venture Monitor · NVCA 2026 Yearbook

[researcher inference · strong] Two current legal/tax controls constrain historical analogy. [contemporaneous record · strong] The eCFR current through 30 July 2026 requires ERISA plan fiduciaries to evaluate an investment's portfolio role, alternatives, diversification, liquidity/current return, projected return, and funding objectives. 29 CFR §2550.404a-1(b) [researcher inference · strong] Dennis's 1977 cost-basis proposal is not a blanket prudence rule. 1977 proposal [contemporaneous record · strong] IRS Topic 427 distinguishes statutory and nonstatutory options and notes AMT and holding-period consequences. IRS Topic 427, reviewed 2026-04-28 [researcher inference · moderate] Employee options therefore carry exercise, AMT, and holding-period consequences and should not be treated as cash-equivalent compensation. Capital-plus-complements model

15. Comparison with completed peers

The eligible peer set as of 2026-08-02 is limited to completed Georges Doriot, Laurance Rockefeller, and Jock Whitney, in that order. Comparisons below are [researcher inference · moderate] unless marked provisional. Every row includes underlying evidence; incompatible owners, periods, legal forms, and return methods prohibit a ranking.

16. The strongest case against greatness

[researcher inference · provisional] The strongest case against Dennis as a personally great venture investor is that a celebrated pioneer narrative may combine inherited access, an early technical niche, corporate and institutional capital, partner-led successes, unusually favorable vintages, and retrospective memories without enough data to demonstrate repeatable personal outperformance. Profile edge audit · portfolio forensics

1. Access before skill — [researcher inference · moderate]. Trust capital, trustee approval, Stanford geography, warm referrals, employer autonomy, and insurer credibility shaped both opportunity and financing. Ampex access record · vehicle boundaries
2. Team misattribution — [researcher inference · strong]. Draper/Wythes/Bancroft at Measurex, McMurtry at ROLM, Marquardt at Seagate, Sequoia/Corrigan at LSI, and all partners at Exabyte performed material work. Better attribution strengthens the institution while weakening a Dennis-only edge. Material-case responsibility matrix
3. Incomplete denominator — [researcher inference · strong]. Twenty-five Group companies have no ledger; the selected career list is not a portfolio; full cash flows and ownership are unavailable. Searched denominator
4. Unaudited and conflicting economics — [researcher inference · strong]. Ampex, Recognition, IVA, and Seagate figures conflict or lack cash-flow definitions; UC starts at IVP III and represents one LP. Portfolio forensics
5. Fund dispersion — [researcher inference · strong]. IVP VI/VII are exceptional in the UC table while VIII is near capital return, contradicting a uniformly exceptional firm story. Comparable IVP III–VIII observations
6. Failure in both timing directions — [researcher inference · moderate]. Reference was too early; Apple was passed as too expensive; Dennis warned about Internet valuation but reported holding unnamed public positions to zero. Case-specific failure postmortems
7. People judgment is vulnerable to bias — [researcher inference · moderate]. Dennis called people his largest errors; the rejected body-size heuristic and Leza dispute show how “frame of reference” can encode invalid proxies or closed access. People-judgment evidence and limits
8. Intuition may not scale — [researcher inference · moderate]. Marquardt described Dennis's fast judgment as hard to explain and transfer; IVA's split and later organizational discontinuities weaken the idea of a durable Dennis system. Marquardt, printed pp. 23–25 · role and succession evidence
9. Company building can conceal investor tradeoffs — [researcher inference · moderate]. Collagen required repeated capital, while Applied shows patience colliding with LP liquidity and legal rights. Collagen decision lab · Applied decision lab
10. Era and luck — [researcher inference · moderate]. Dennis credited fortune, while early electronics demand, cheap rounds, limited competition, maturing public markets, and policy changes supplied potential tailwinds. Historical environment · Ampex decision lab

[researcher inference · moderate] The favorable thesis strengthens only if original memos, votes, role ledgers, board records, and full cash flows repeatedly connect Dennis-specific actions to better ordinary-case outcomes after access, partners, price, vintage, and outliers are controlled. The skeptical thesis strengthens if the missing record shows generic technical diligence, partner-heavy execution, selection by familiarity, weak ex-outlier economics, or governance assistance that did not change outcomes. Ranked retrieval ledger

17. Analyst study guide, glossary, and go deeper

Recommended reading path

1. Read the profile for chronology, causality, vehicles, and people map.
2. Read investments before using any company anecdote; the labs preserve ex-ante information and missing fields.
3. Compare Dennis's direct spoken and written records; repeated late-life stories are not independent corroboration.
4. Apply the mental models only with their exceptions and evidence status.
5. Use the source map to distinguish full, partial, blocked, and unprocessed sources.

Decision-lab path

• Technical/customer/price: Exabyte → Reference Technology → Apple → January 2000 valuation warning.
• People and governance: Seagate → LSI → Recognition → unnamed rescue → rejected body-size heuristic.
• Networks and attribution: Ampex → The Group → Measurex → ROLM.
• Vehicle and liquidity: personal/corporate boundary → IVA formation → IVP split → Applied.
• Portfolio evidence: Group claims → Seagate conflicts → UC IVP III–VIII dispersion → missing denominator.

Glossary

Reusable sourcing checklist

• [researcher inference · moderate] Record referral origin, source owner, sponsor, date, and legal/ethical use of information. Network and funnel model
• [researcher inference · moderate] Define intake unit, duplicate policy, period, stage, and complete pass/invest denominator. Searched denominator
• [researcher inference · moderate] Compare referred and non-referred opportunities on the same evidence standard. Network and funnel model
• [researcher inference · moderate] Track who never enters the funnel and audit familiarity/geography/credential filters. People-judgment model · network and funnel model
• [researcher inference · moderate] Treat syndicate participation as separate from sourcing, selection, board work, and proceeds. Roles and decision-rights model

Reusable diligence and IC checklist

• [researcher inference · moderate] Separate technical efficacy, present customer need, differentiation, timing, and security economics. Technology-to-economics model
• [researcher inference · moderate] Name the role and job-relevant evidence for each key person; reject appearance/familiarity proxies. People-judgment model
• [researcher inference · moderate] State what is believed ex ante and what evidence would falsify it. Plans-as-hypotheses model
• [researcher inference · moderate] Record vehicle, conflicts, recusal, sponsor, pre-discussion votes, dissent, final right, and urgency exception. Roles and decision-rights model
• [researcher inference · moderate] Identify the binding noncapital complement and a capacity-budgeted owner. Capital-plus-complements model
• [researcher inference · moderate] Model initial/follow-on capital, ownership/dilution, downside terms, reserves, and exit/liquidity paths. Vehicle and capacity model

Reusable portfolio and governance checklist

• [researcher inference · moderate] Report company-count and capital-weighted outcomes; gross/net; DPI/TVPI/IRR; top-one/top-vintage sensitivity. Viability-versus-economics model
• [researcher inference · moderate] At each board cycle, compare plan with evidence and choose support, recruit, pivot, replace, stop, or continue. Plans-as-hypotheses model
• [researcher inference · moderate] Require changed evidence before rescue capital; preserve the original thesis and sunk cost separately. Plans-as-hypotheses model
• [researcher inference · moderate] Reconcile company duration with fund, LP, founder, employee, and personal liquidity clocks. Company-building and patience model
• [researcher inference · moderate] Audit partner board load, domain competence, reserve capacity, and succession before scaling fund size. Capital-cycle model · vehicle and capacity model

Ranked sources

1. Dennis two-session Berkeley oral history — broadest direct chronology, philosophy, cases, and admissions; retrospective.
2. Raw Something Ventured transcript 2 — less-polished direct testimony, decision exceptions, failures, and competence boundary.
3. 2005 CHM Dennis–Johnson video — auditable audiovisual discussion and denominators, with collective-speaker boundaries.
4. 1979 Dennis prepared statement — contemporaneous authored policy and people/capital argument.
5. 1977 JCT testimony summary — official ex-ante policy position, but staff paraphrase.
6. UC private-equity report — comparable net-to-one-LP IVP III–VIII observations.
7. Marquardt oral history — partner attribution, Seagate, communication, and transferability critique.
8. Applied Medical registration, withdrawal, and opinion — vehicle duration, rights, liquidity, and legal boundaries.
9. Almanac January 2000 interview — ex-ante adoption/value/timing distinction.
10. Annotated source map — family de-duplication, access states, exact gaps, and retrieval priorities.

Go deeper: decisive missing evidence

[researcher inference · strong] Retrieve Ampex subscription/trust/sale records; The Group's company/member ledger; Fireman's/American Express committee and Recognition files; IVA's agreement, LP list, fees/carry and cash flows; IVP I–II and complete Dennis-era LP schedules; and company memos/minutes assigning source, votes, board actions, follow-ons, stops, and exits. Recover the 1981 authored article, full Done Deals chapter, WSJ interview, MBA paper, Dennis Applied declaration, archive video, and original speech/board files. Those records—not another selected-company list—can adjudicate edge, transferability, bias, fund economics, and learning. Ranked retrieval ledger · annotated source map

Pareto 80/20 — what drove the record

[researcher inference · strong] No defensible lifetime 80/20 percentage can be calculated from the public record. The surviving figures mix Dennis's personal and family-trust money, the personal-capital syndicate called The Group, Fireman's Fund/American Express corporate capital, Institutional Venture Associates (IVA), multiple Institutional Venture Partners (IVP) funds, paper values, partnership distributions, and one limited partner's fund-level cash flows. The denominator is incomplete and the measures are not commensurable. The rankings below are therefore directional answers to three different questions, not a synthetic return table. Dennis oral history, printed pp. 13–60 · UC methodology, p. 7

Economic contribution — ranked named-deal evidence, not comparable returns

Ranking rule. [researcher inference · strong] Eligibility requires a named Dennis-associated case with a reported dollar outcome. Order is lexicographic: (1) distributed/realized cash or reported corporate profit precedes paper value; then (2) larger nominal reported scale precedes smaller scale within that evidence tier. That yields Seagate's reported partnership distribution, Recognition's reported corporate profit, and Ampex's reported paper value, in that order. This is an economic-evidence priority, not a ranking of comparable return contribution: the values span decades and different owners, and none supplies compatible dates, dilution, fees, carry, taxes, or complete cash flows. Seagate witnesses · Recognition and Ampex account

Sensitivity. [researcher inference · strong] If corporate-account profit is excluded, the eligible order is Seagate then Ampex; if paper value is excluded, it is Seagate then Recognition. Neither variant makes these returns commensurable or admits a whole-fund aggregate into a named-deal ranking. Ranked cases and controlling measures

Unranked fund-level controls

Dominant-outlier tests. [researcher inference · strong] At deal level, exclusion of Seagate cannot be calculated because no compatible cash-flow ledger exists. At the only comparable public fund-vintage level, UC paid $92m and received $305.403m across IVP III–VIII, or 3.32x on simple cash totals. Excluding IVP VII—40.1% of those observed distributions—leaves $74m paid and $182.936m distributed, or 2.47x. Excluding both VI and VII leaves $62m and $113.138m, or 1.82x. These are researcher calculations from one LP's fully distributed rows, not a reported aggregate TVPI or IRR, Dennis economics, or proof that any named deal drove a vintage. [contemporaneous record · strong] The UC report warns that timing, fees, carry, expenses, secondaries, and valuation methods can make its figures differ from other investors' results. [researcher inference · strong] The observable IVP LP record is concentrated, but it does not support a literal “20% of deals produced 80% of returns” claim. UC table, IVP III–VII p. 1 · IVP VIII p. 2 · methodology p. 7

Historical importance — separate from return contribution

Ranking rule. [researcher inference · strong] Cases are ordered first by whether they mark a change in Dennis's capital-formation or governance regime, then chronologically within that qualifying set: personal/trust capital, corporate capital, pooled partnership formation, first-fund startup practice, and finally the fund-duration conflict. The order is historical judgment, not economic contribution; each row carries its claim-local evidence. Historical cases and evidence

Decision-learning value — ranked for replayability

Ranking rule. [researcher inference · strong] Cases are ordered by replayability: an observable decision tension plus a specific action plus post-decision evidence ranks ahead of a story missing one of those elements; ties are broken by field completeness. The row citations define the evidence ceiling, and the order does not claim economic importance. Decision-learning cases and evidence

Evidence rules and searched denominator

Scope and attribution rules

• Vehicles are never collapsed. [researcher inference · strong] Dennis personal, family trust, Peggy Dennis personal, The Group, Fireman's Fund/American Express, IVA, each numbered IVP partnership, its management entity, and later public holdings are separate capital accounts until a source proves otherwise.
• A name enters the discovery ledger. [researcher inference · strong] Inclusion requires a direct witness, institutional biography, court/SEC record, or public ownership filing connecting Dennis to an investment, board, management entity, or disclosed beneficial holding. A filing that merely names him among shared managers proves legal control/beneficial-ownership reporting, not sourcing, approval, or operating work.
• Follow-ons, distributions, successor securities, and duplicate filings are one company lineage. [researcher inference · strong] Collagen's Cohesion distribution remains visible because it created a separate public holding; it does not become a second original sourcing claim.
• Outcome measures stay separate. [researcher inference · strong] Paper value, beneficial ownership, company IPO/acquisition, partnership distribution, gross fund proceeds, and one LP's net cash flows cannot be added or converted into a Dennis multiple without basis, dates, dilution, fees, carry, and allocation.
• Support uses two axes. [researcher inference · strong] The first says what kind of evidence exists—investor-stated, contemporaneous record, documented behavior, retrospective witness, or researcher inference; the second says strong, moderate, provisional, or insufficient public record.
• No selection-list laundering. [researcher inference · strong] The Computer History Museum biography supplies a selected name set, not a portfolio ledger, and IVP's firm-wide promotional totals or later partner wins are excluded from Dennis performance. CHM profile

Reproducible search protocol

• Time boundary. [researcher inference · strong] The case search covered Dennis-associated activity described from his 1952 Ampex financing through the 2013 Applied registration withdrawal; retrieval/context sources were checked through 2 August 2026. Later sources could corroborate an older event but did not turn a retrospective account into contemporaneous evidence.
• Repositories and routes. [researcher inference · strong] Searches covered the Berkeley Venture Capitalists oral histories, Stanford's Dennis transcripts, Computer History Museum profiles/video/oral histories, SEC EDGAR issuer and beneficial-ownership filings, UC and CalSTRS private-equity disclosures, the Online Archive of California Fireman's finding aid, the cited state-court opinion, and targeted press/archive searches. The repeatable routing set was: Dennis name variant + company, Dennis name variant + vehicle alias, exact company name + IVP/IVA/director/investment, and SEC issuer/CIK searches. Berkeley Dennis oral history · Stanford transcript 2 · CHM profile · UC 2020 disclosure · Fireman's finding aid
• Name and vehicle variants. [researcher inference · strong] Queries used Reid Weaver Dennis, Reid W. Dennis, Reid Dennis, R. W. Dennis, and Dennis, Reid, paired with Institutional Venture Associates, IVA, Institutional Venture Partners, IVP, The Group, Fireman's Fund, Firemen's Fund, American Express, and AMEXCO; company-name spelling and successor lineages were also checked.
• Inclusion, deduplication, and depth. [researcher inference · strong] A named candidate entered only when a direct witness, institutional biography, court/SEC record, or public ownership filing connected Dennis to an investment, board, management entity, or disclosed holding. Follow-ons, successor securities, and duplicate filings were collapsed to one lineage. Searched means a query/repository route was executed; researched means the candidate received an evidence-boundary and structured field screen. All 33 named candidates reached both stages: 26 included associations are routed in the ledger and seven excluded shared-control issuers are documented as controls. Sufficiently reconstructed means enough decision fields survived for a bounded no-hindsight lab, not that the case is complete.

Mechanical reconciliation

Overlap and partition rule. [researcher inference · strong] 33 publicly known = 26 included + 7 excluded, with no overlap. Searched and researched are audit stages applied to all 33, not additive buckets. Within the 26 included associations, 11 sufficiently reconstructed + 15 unresolved = 26, again without overlap. Apple sits inside the 11 reconstructed cases but outside the 25 funded/held count. Cohesion remains one successor-security lineage within Collagen, not a second original investment. Mechanical reconciliation table

Outside the named funnel. Dennis and Johnson's claimed roughly 25 Group companies are a separate unreconciled public claim, not part of the 33 because no company list survives; they remember about eighteen successes, two capital returns, and roughly three substantial losses. [retrospective witness · provisional] CHM video, 00:35:07–00:39:15 At least five unnamed unresolved leads—Portland, Boulder, uncertain “Mini Store,” unnamed public stocks, and an unnamed management rescue—also sit outside the 33-name funnel. [researcher inference · provisional] 1983 report, ¶¶7–9 · CHM, 01:06:39 · raw transcript 2, printed pp. 12–13 and 17–18/PDF pp. 7–10

Pre-candidate exclusions. [researcher inference · strong] Later IVP XII/XVII/XVIII LP rows, IVP's promotional counts of 400 companies and 135+ IPOs, co-speaker companies, operating employers, duplicate filings, and repeated follow-ons are excluded before deal-candidate counting; they therefore do not alter 33, 26, 11, or 15. Dennis on GP/adviser boundary, printed pp. 47–48/PDF pp. 53–54 · CalSTRS later-fund rows, p. 5 · IVP promotional homepage

Full-schema reconstruction: 0 / 25 funded/held cases. [researcher inference · strong] No funded case has a complete contemporaneous memo, legal vehicle, security, ownership/dilution, approvals, board chronology, follow-ons, exit, and investor cash-flow schedule. The deep ledger is bounded reconstruction, not a complete file.

Coverage fraction: not computable. [researcher inference · strong] 11 / 26 is analytical depth inside the selected included set, not career coverage; 0 / 25 is full-schema completeness among included funded/held cases, not a hit rate. The true denominator across personal, Group, corporate, IVA, and IVP I–XI activity is absent, and the unnamed Group portfolio alone prevents a compatible career fraction. The 33-name audit funnel is reproducible but selection-biased toward remembered wins, board stories, public issuers, and filing controls. Mechanical reconciliation table

Historical context — what could have been known then

• 1950s corporate and personal experimentation. [retrospective witness · moderate] Ampex predated Dennis's institutional investing and Recognition used an insurer's balance sheet. [researcher inference · strong] The public record does not supply a contemporary private-company pricing index or success base rate, so later outcome fame cannot substitute for the evidence available at approval. Dennis, Ampex and Recognition, printed pp. 13–21/PDF pp. 19–27
• 1973–1974 fundraising shock. [retrospective witness · moderate] Jamieson recalls the oil shock and market decline during IVA's raise; Dennis recalls American Express offering a conditional $5m anchor and the completed fund reaching $19m through six insurers and the Ford Foundation. [researcher inference · moderate] That setting supports cautious fundraising and diversified institutional capital, not a claim that the fund faced modern fundraising terms. Jamieson, printed pp. 39–41/PDF pp. 51–53 · Dennis, printed pp. 30–40
• Caution after close. [retrospective witness · moderate] McMurtry recalls IVA waiting about fifteen months after the close before making its first investment. [researcher inference · moderate] This is evidence of pacing under uncertainty; without the pipeline and cash ledger it cannot establish superior discipline. McMurtry, printed pp. 42–43/PDF pp. 52–53
• Policy changed the opportunity set, but causality is bounded. [contemporaneous record · strong] The Revenue Act of 1978 changed capital-gains taxation; the Department of Labor's 1979 final rule under 29 CFR 2550.404a-1 said an investment's riskiness is not per se prudent or imprudent and that prudence depends on its role in the plan's overall portfolio (effective 23 July 1979); and the 1981 tax act changed rates again. [researcher inference · strong] These are contemporaneous institutional facts, not proof that any single reform caused pension allocations, an IVP deal, or a return. Revenue Act of 1978 · 1979 DOL final rule, 44 FR 37221–23 · Economic Recovery Tax Act of 1981
• 1980 promise versus opportunity. [retrospective witness · moderate] Dennis recalls telling first-fund LPs that only one or two raw startups would be attempted, then backing six or seven because the observed opportunities changed. [researcher inference · strong] That is a contemporaneous-governance question reconstructed retrospectively: whether the partnership agreement permitted the change, whether LPs were notified, and how reserves changed are unknown. Dennis, printed pp. 46–48/PDF pp. 52–54
• 2000s capital abundance. [investor-stated · moderate] In 2008 Dennis and Johnson described excess capital, higher prices, later rounds, fewer IPOs, and build-to-sell behavior. [researcher inference · strong] These observations help interpret the late portfolio; they do not retroactively establish the 1950s–1980s base rate. Dennis–Johnson, printed pp. 9–12/PDF pp. 10–13

Vehicles, economics, and incentive boundaries

Structural comparison. [researcher inference · strong] IVA/IVP had pooled LP capital, GP decision rights, finite fund expectations, management entities, and shared partner control. The evidence does not support generic claims about fee levels, carry, governance quality, reserve formulae, or “more patient” old capital. The clearest supported contrast is within Dennis's own record: personal/trust and corporate capital had different owners and duration constraints, while Applied shows how a ten-year fund expectation can collide with a twenty-four-year holding. OCBJ report of Dennis remarks · Applied S-1/A

People, capital, and influence map

Institutions and recurring roles

Material-case responsibility matrix

Comprehensive scoped deal ledger

This is comprehensive only for the narrow 25 funded/held names plus one Apple pass defined above. It is not Dennis's full career portfolio. Every missing decision field is written as unknown or not publicly established; neither means zero. Every outcome is kept behind its vehicle boundary.

Deep-reconstructable cases

Other funded/held names — association is not deal leadership

Unnamed and excluded evidence

• Portland and Boulder deals. A 1983 report says IVP invested outside the Valley and that Boulder housing costs helped recruit a CEO; company, vehicle, check, security, ownership, board and result remain unknown. These are two references, not two proven unique deals. [contemporaneous record · provisional] Christian Science Monitor, ¶¶7–9
• “Mini Store” / likely MiniStor. Automatic captions render the storage failure unclearly. The plausible identity remains a researcher lead, not a ledger name, until a contemporaneous portfolio or company record links Dennis's vehicle. [retrospective witness · provisional] CHM video, 01:06:39–01:10:30
• Other unnamed losses/interventions. [retrospective witness · provisional] Dennis describes public stocks held to zero and an unnamed management-rescue story, but neither supplies issuers or legal vehicles. [researcher inference · strong] They enter failure-pattern analysis, not the 33-name funnel. raw transcript 2, printed pp. 12–13 and 17–18/PDF pp. 7–10
• Shared-control filing controls. [contemporaneous record · strong] Filings for Netflix, Thermage, ArcSight, Vonage, Argonaut, Turnstone, and Digital Impact name a different board partner or list Dennis among multiple IVP managers. [researcher inference · strong] Without separate Dennis deal evidence, those seven are excluded candidates: legal control does not prove source, vote, board work, or economics.

No-hindsight decision labs

[researcher inference · strong] Each lab stops at the decision date before revealing the remembered outcome. The questions are analyst-created exercises, not claimed historical IVP checklists. “Unknown” is decision-relevant: it marks the evidence that a real committee would have needed and the public record does not preserve. Scope and attribution rules

Decision lab 1: Ampex — access, inherited capital, and concentration

Decision-date freeze. [retrospective witness · moderate] Dennis had seen the Ampex recorder at Stanford and says a friend of his brother referred him into the private-company financing roughly a month before he joined Fireman's. The account does not provide a dated memo, then-current revenue, valuation, security, ownership, customer concentration, or liquidity plan. He remembers roughly $13,000 from a family trust and $2,000 of his own money; a $20,000 summary conflicts with the detailed body. His mother, as trustee, was a capital decision maker. Preface, PDF p. 4 · body, printed pp. 13–15/PDF pp. 19–21

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? What does prior product exposure establish, and what still needs independent diligence beyond a trusted referral?
2. What would kill the investment? Define customer, governance, valuation, and personal-concentration limits.
3. What ownership and terms would you require? Specify security, information rights, and a maximum personal/trust exposure.
4. What evidence unlocks the next check? Name operating milestones, customer proof, and trustee authorization.

Actual decision and outcome revealed. Dennis and the trust invested. He remembers a paper value of about $800,000–$900,000 roughly three-and-a-half years later; the preface says more than $1m. No sale record supports a realized multiple. [retrospective witness · provisional] [researcher inference · strong] The supported lesson is to distinguish access and paper appreciation from realized skill. Preface, PDF p. 4 · body, printed pp. 13–15/PDF pp. 19–21

Decision lab 2: Recognition Equipment — corporate scale and board responsibility

Decision-date freeze. [retrospective witness · moderate] Fireman's could write a much larger check than Dennis personally. The remembered opportunity was automated document processing, but the public account omits an approval memo, product evidence, security, ownership, downside case, reserve plan, and initial management assessment. Dennis reports roughly $600,000–$650,000 initially. Dennis, printed pp. 16–22/PDF pp. 22–28

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Is the automation need sufficiently urgent and proven to support an insurer-sized position?
2. What would kill the investment? Define accuracy, throughput, customer adoption, and management triggers.
3. What ownership and terms would you require? Specify board rights, information rights, staged funding, and executive-change authority.
4. What evidence unlocks the next check? Require customer acceptance, unit economics, working-capital data, and hiring milestones.

Actual decision and outcome revealed. Fireman's invested; Dennis served nineteen years on the board, sold enough after a tenfold rise to recover reported cost, and later participated in a CEO change. He says the broader corporate position ultimately produced about $15m of profit after a major price round trip. [retrospective witness · moderate] The corporate owner—not Dennis personally—held the economics. Dennis, printed pp. 16–21/PDF pp. 22–27

Decision lab 3: Measurex — syndication without credit laundering

Decision-date freeze. [retrospective witness · moderate] Bossen sought roughly $1.2m–$1.3m, above a typical Group round. Draper and Wythes opened the fundraising route; Bancroft/Bessemer became the largest original investor; other investors filled the round; Dennis and Peggy invested separately. Product evidence, pricing, securities, allocations, board rights, and reserves are not public. Dennis, printed pp. 24–27 · Bossen, printed pp. 19–20/PDF pp. 25–26

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Which founder, process-control customer, and technical claims must be independently verified?
2. What would kill the investment? Set financing-completion, prototype, customer-pilot, and founder-reference thresholds.
3. What ownership and terms would you require? Allocate pro rata, board, information, and follow-on rights among a crowded syndicate.
4. What evidence unlocks the next check? Require a fully subscribed round, working system, named industrial users, and installation economics.

Actual decision and outcome revealed. The network financed Measurex and Honeywell acquired it in 1997. Bossen remembers Peggy holding her separate investment through the sale for roughly 50x; no comparable Reid result survives. [retrospective witness · moderate] [researcher inference · strong] A strong company outcome does not collapse sourcing, capital, board work, and proceeds into one investor's credit. Bossen bio, PDF p. 5 · investment account, printed pp. 19–20/PDF pp. 25–26

Decision lab 4: ROLM — disclose, recuse, then decide

Decision-date freeze. [retrospective witness · moderate] McMurtry brought ROLM to IVA after already holding a personal/Palo Alto Investment position. He disclosed the conflict and did not join the approval decision. The public account establishes an experienced team but omits IVA's check, security, ownership, valuation, market memo, and reserve model. McMurtry, printed pp. 30–32/PDF pp. 40–42

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Can the merits be reproduced independently of the conflicted sponsor?
2. What would kill the investment? Define conflict-disclosure, customer, technology, and valuation failures.
3. What ownership and terms would you require? Specify clean allocation, independent approval, board seat, and related-party documentation.
4. What evidence unlocks the next check? Require customer references, a cap table, use of proceeds, and conflict-compliance record.

Actual decision and outcome revealed. Dennis, Jamieson, and Norris approved; McMurtry became the board representative. The company later went public, but public evidence does not reconstruct IVA's cash return. [retrospective witness · moderate] [researcher inference · strong] The learnable fact is the role separation preserved in McMurtry's later oral history, not a claimed multiple. McMurtry, printed pp. 30–32/PDF pp. 40–42

Decision lab 5: Seagate — sponsor persistence versus partnership concentration

Decision-date freeze. [retrospective witness · moderate] Marquardt backed a known disk-drive team and persisted after Dennis's initial reluctance. Oak/Glassmeyer offered IVP half of a reported 20%/$1m position; Marquardt says IVP's proposed share was $500,000. The initial security, dilution, product evidence, customer commitments, reserves, and exact legal vehicle remain unresolved. Glassmeyer, printed p. 42/PDF p. 47 · Marquardt, printed pp. 28–30/PDF pp. 33–35

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? How much weight belongs to team history versus disk-drive market and manufacturing risk?
2. What would kill the investment? Define product, yield, customer, financing, and concentration thresholds.
3. What ownership and terms would you require? Reconcile the proposed 10% stake, board right, anti-dilution, and follow-on capacity.
4. What evidence unlocks the next check? Require working drives, qualified customers, manufacturing yields, and cash-burn milestones.

Actual decision and outcome revealed. [retrospective witness · provisional] Dennis authorized the IVP investment; Marquardt led IVP's board work and recalls Glassmeyer as another director. Marquardt recalls roughly $63m distributed from IVP's $500,000. Dennis separately recalls a later $500,000 IPO purchase and an in-kind distribution worth about three times the fund's contributed capital. [researcher inference · strong] These unaudited narratives are not merged. Marquardt, printed pp. 28–30/PDF pp. 33–35 · Dennis, printed pp. 58–60

Decision lab 6: LSI Logic — scarcity, dissent, and excess cash in 1981

Decision-date freeze. Corrigan was an experienced semiconductor executive pursuing custom logic. Dennis recalls a scarce allocation, a younger partner's objection, and reported checks of $750,000 from Kleiner Perkins, $600,000 from his uncertain IVA/IVP vehicle, and $400,000–$500,000 from Sequoia. Corrigan's broader recollection says the startup raised $6m, gave up about 50%, began in a semiconductor downturn with excess fabrication capacity, and earned 18%–19% bank interest—about $100,000 monthly—while software was the critical path. [retrospective witness · moderate] Dennis raw transcript 2, pp. 25–29 · LSI oral-history panel, transcript pp. 15–16

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Does founder experience and foundry slack justify the price, ownership surrendered, and product uncertainty?
2. What would kill the investment? Define dissent-resolution, software completion, CMOS performance, customer, and burn triggers.
3. What ownership and terms would you require? Reconcile the $6m total, investor allocations, observer/board rights, and staged deployment.
4. What evidence unlocks the next check? Require a functioning design system, foundry agreement, customer design wins, and a written response to the partner objection.

Actual decision and outcome revealed. Dennis moved before the long weekend and later admitted violating his consensus ideal. Corrigan says Dennis had observer status, not a board seat. LSI reached an IPO in 1983; investor cash flows remain unavailable. [retrospective witness · moderate] [researcher inference · strong] A good outcome cannot validate every urgent override. Dennis raw transcript 2, pp. 25–29 · LSI panel, p. 36

Decision lab 7: Exabyte — application change as the thesis

Decision-date freeze. [retrospective witness · moderate] Dennis brought the partnership a Boulder team proposing to reuse consumer-video recording technology for computer data storage. He recalls initial partner disbelief followed by an all-partner trip. Public evidence does not give the check, security, ownership, product benchmark, customer commitments, board assignment, or reserve plan. Dennis, printed pp. 53–56/PDF pp. 59–62

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Is the technical reuse real, and does data storage solve a budgeted customer problem?
2. What would kill the investment? Set error-rate, durability, throughput, customer-validation, and cost thresholds.
3. What ownership and terms would you require? Specify a staged preferred round, board rights, reserves, and technical-information rights.
4. What evidence unlocks the next check? Require prototype endurance tests, customer trials, supplier commitments, and production-cost proof.

Actual decision and outcome revealed. All partners traveled to Boulder and approved the investment. Dennis remembers Exabyte among the successful first-fund startups, but the public record supplies no IVP cash flows. [retrospective witness · moderate] [researcher inference · moderate] The portable learning is to underwrite the new application independently of the inherited hardware. Dennis, printed pp. 53–56/PDF pp. 59–62

Decision lab 8: Collagen/Cohesion — finance the learning, not the original story

Decision-date freeze. [retrospective witness · provisional] The initial disclosed thesis was collagen as a burn-dressing material. Before revealing the later indication history, the public record supplies no original clinical evidence, adverse-event profile, regulatory route, check, security, ownership, burn, reserve plan, or contemporaneous platform thesis. Dennis, printed pp. 50–55/PDF pp. 56–61

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Is the biomaterial platform valuable independently of the burn indication?
2. What would kill the investment? Define safety, efficacy, regulatory, cash-runway, and alternative-use thresholds.
3. What ownership and terms would you require? Specify tranched preferred financing, board rights, protective provisions, and reserve limits.
4. What evidence unlocks the next check? Require controlled clinical data, regulator feedback, reproducible manufacturing, and paid demand for the new indication.

Actual decision and outcome revealed. IVP continued financing and Collagen shifted toward cosmetic use. Dennis called the outcome modestly positive. Later SEC filings show his 7.9% Collagen beneficial ownership at 1996 year-end and an 8.3% Cohesion holding received in Collagen's 1998 distribution; they do not show cost or realization. [contemporaneous record · strong] for holdings; [retrospective witness · moderate] for the pivot. Dennis, printed pp. 50–55 · Collagen 1997 13G · Cohesion 1999 13G

Decision lab 9: Applied Medical — company duration versus fund duration

Decision-date freeze. [documented behavior · strong] This October 1988 first-financing boundary comes from Applied's February 2013 filing—not a contemporaneous 1988 memo. The filing retrospectively records that the company was founded in 1987, IVP/IVM invested $1.5m in October 1988, and Peter Thomas joined the board that month. It does not preserve the product, regulatory, customer, valuation, security, ownership, sponsor, fund-life, or reserve evidence available at approval. Later financings and the exit conflict are withheld until the reveal; a true no-hindsight underwriting freeze cannot be reconstructed from the public file. Applied February 2013 S-1/A, prospectus pp. 45–46

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? What medical need, regulatory path, manufacturing advantage, and founder evidence justify the first and follow-on checks?
2. What would kill the investment? Define product, regulatory, customer, governance, and fund-duration triggers.
3. What ownership and terms would you require? Specify preferred rights, board representation, demand-registration rights, transfer rights, and a long-stop liquidity plan.
4. What evidence unlocks the next check? Require product clearance, hospital adoption, gross-margin proof, manufacturing milestones, and a revised exit path.

Actual decision and outcome revealed. The later filing says IVP added $4.2m through December 1990 but did not participate in Applied's final March 1992 venture financing. [researcher inference · moderate] The 1988-to-2012 chronology implies a roughly twenty-four-year hold; it is not a reported return period. In 2011 IVP invoked demand-registration rights, while the company and board said the timing/manner was not in Applied's best interest. The proposed 2012 offering involved 729,798 selling-holder shares, not primary capital. A June 2013 withdrawal said the registration never became effective and no securities were issued or sold under it. [documented behavior · strong] [investor-stated · moderate] The OCBJ report says Dennis wanted to retain his personal shares; that preference is distinct from IVP IV's LP-liquidity position. OCBJ report of Dennis's personal-share rationale The 2017 appellate opinion records allegations about Thomas, Dennis, and IVM IV; it time-barred fraud claims and revived contract/conversion claims without deciding their merits or establishing misconduct by Dennis. February 2013 S-1/A, financing history · October 2012 S-1/A · withdrawal · court opinion

Decision lab 10: Reference Technology — timing failure

Decision-date freeze. [retrospective witness · moderate] Reference Technology proposed adapting Sony's 14-inch videodisc technology to digital storage. Dennis remembers a good team and good underlying technology. [researcher inference · moderate] From that later description, an analyst can reconstruct the large physical format, uncertain customer readiness, and new-category adoption as risks that should have been tested; the record does not prove they were identified at decision date. No dated memo, market test, vehicle, check, security, ownership, milestones, board record, or follow-on plan survives. The later timing diagnosis and competing formats are withheld until the reveal. Dennis, printed pp. 52–53/PDF pp. 58–59

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Can a 14-inch optical-storage product solve an urgent need despite its physical burden?
2. What would kill the investment? Define paid-customer, form-factor, competing-medium, access-speed, cost-curve, and runway thresholds.
3. What ownership and terms would you require? Specify small option-preserving tranches, milestone pricing, board information, and hard reserve caps.
4. What evidence unlocks the next check? Require paid pilots, data-density/reliability tests, integration proof, and a financing plan matched to adoption.

Actual decision and outcome revealed. The firm invested. Dennis later estimated it was about five years too early: the cumbersome discs found too little market before IBM's 8-inch and Seagate's 5-inch alternatives supplied a more practical path. Investor loss, follow-ons, stop decision and salvage are unknown. [retrospective witness · moderate] [researcher inference · moderate] The reconstructed failure mechanism is market timing plus form factor, not bad people or nonexistent technology. Dennis, printed pp. 52–53/PDF pp. 58–59 · raw transcript 2, pp. 17–18

Decision lab 11: Apple — valuation discipline or category error?

Decision-date freeze. [retrospective witness · provisional] Dennis remembers a roughly $78m post-money price that looked extreme against Apple's then revenue and his own prior experience. The transcript is uncertain about the referring investor and supplies no date, term sheet, revenue definition, security, allocation, rights, growth evidence, or committee record. raw transcript 2, pp. 39–40

Analyst worksheet — answer before reading the reveal.

1. Would you invest then? Which growth, product, channel, and market evidence could justify a price outside the investor's historical range?
2. What would kill the investment? Define valuation, governance, customer concentration, execution, and downside thresholds.
3. What ownership and terms would you require? Specify allocation, preferred protections, information/board rights, and follow-on capacity.
4. What evidence unlocks the next check? Require verified revenue/growth, unit economics, channel proof, cap table, and a dated term sheet.

Actual decision and outcome revealed. Dennis passed and later recounted Apple as a miss; there is no Dennis anti-portfolio dollar calculation. [retrospective witness · provisional] [researcher inference · strong] The lesson is not “always pay up”; it is to make the valuation model, evidence boundary, and competence limit explicit before passing. raw transcript 2, pp. 39–40

Portfolio forensics

Comparable fund-vintage observations — one LP only

[contemporaneous record · strong] These rows are fully distributed observations reported by the University of California as of 30 June 2020, and UC defines vintage year as the year of first cash flow. Until 31 March 2017 Cambridge Associates prepared UC's periodic disclosure reports; beginning 31 December 2017 UC prepared them from internal recordkeeping, and UC warns that differences from prior reports can reflect that source-data change. UC's IRRs are net to its own account, calculated by a third-party service provider from UC-maintained partnership records, and may differ across investors because of cash-flow timing, carried interest, fees, expenses, distributed-stock sales, secondaries, and valuations. UC states that none of the information in the report was reviewed or approved by the funds' General Partners. The simple distributed/paid quotient above is therefore a researcher calculation, not an IRR. The 2010 and 2003 reports are earlier snapshots useful for chronology, not independent economic “votes.” UC 2020 table, III–VII p. 1 · VIII p. 2 · methodology and source switch, p. 7 · UC 2010 snapshot, pp. 1–3 · UC Regents 2003 snapshot, p. 1

[contemporaneous record · strong] CalSTRS reports IVP XII, XVII, and XVIII. [retrospective witness · moderate] Dennis described himself as an adviser—not GP—to XII, and the other vintages are later still. [researcher inference · strong] Those rows are excluded from Dennis-era economics rather than used to smooth the highly dispersed III–VIII record. Dennis, printed pp. 47–48/PDF pp. 53–54 · CalSTRS FYE 2024, p. 5

Funnel, stage, sector, geography, and check size

• Funnel. [investor-stated · provisional] Dennis said a fund invested in roughly 1.5%–2.5% of opportunities seen, but did not define seen, the time period, funds, duplicates, screens, or funded numerator. [researcher inference · strong] It is process-color, not a computable denominator. CHM video, 01:15:31–01:17:05
• Stage. [researcher inference · strong] The observed set ranges from a private-company personal/trust financing (Ampex), early startups (Measurex, ROLM, Seagate, LSI, Exabyte), repeated biomedical financing (Collagen), later-stage/public work, and distributed public shares (Microsoft). A single “seed investor” label would erase real vehicle differences. Ampex, printed pp. 13–15/PDF pp. 19–21 · first-fund cases, printed pp. 46–60/PDF pp. 52–66 · Microsoft, raw transcript 2, printed p. 40/PDF p. 40
• Sector. [researcher inference · moderate] Electronics, computing, storage, semiconductors, process control, and biomedical devices dominate the selected public record. That may reflect Dennis's electrical-engineering experience and IVP specialization, but it also reflects CHM selection and public-company survival. Sector weights cannot be calculated from 25 names. CHM selected profile · Dennis on specialization, printed pp. 47–57/PDF pp. 53–63
• Geography. [researcher inference · moderate] Silicon Valley dominates the selected record, with explicit Boulder and other out-of-Valley activity. [contemporaneous record · moderate] A 1983 report says Boulder housing was 10%–20% cheaper than Silicon Valley in a CEO-recruiting example, but omits the company and investment terms. [researcher inference · strong] The anecdote supports one talent-cost comparison, not a regional return edge. Christian Science Monitor, ¶¶7–9
• Check size. [retrospective witness · moderate] Dennis's recollections range from $15,000 of Ampex personal/trust capital to Group pools of roughly $80,000–$100,000 and Fireman's checks generally above $250,000. [documented behavior · strong] A later filing retrospectively records Applied financing in the millions. [researcher inference · strong] Dates, stages, follow-ons, inflation, and vehicle ownership differ, so no median or trend is supportable. Dennis, printed pp. 13–28/PDF pp. 19–34 · Applied February 2013 S-1/A, prospectus pp. 45–46

Ownership, concentration, reserves, governance, and exits

• Ownership. [retrospective witness · moderate] LSI's founder recalls the $6m syndicate taking about 50% collectively, while Seagate witnesses imply IVP took half of Oak's proposed 20% block. [contemporaneous record · strong] Dated filings separately show Dennis's later 7.8%–8.1% Collagen and 8.0%–8.3% Cohesion beneficial holdings. [researcher inference · strong] These are three different entities and dates, not a universal target. LSI panel, transcript p. 15 · Glassmeyer, printed p. 42/PDF p. 47 · Collagen 1998 13G, 7.8% · Collagen 1999 amendment, 8.1% · Cohesion 2000 amendment, 8.0% · Cohesion 1999 13G, 8.3%
• Concentration. [researcher inference · strong] Seagate is the only named case with a reported distribution large enough to test fund dominance, yet its witness cash-flow accounts cannot be reconciled. The UC fund rows show vintage dispersion and sensitivity to VI/VII, not company concentration. The bounded result is deal-level concentration unknown; vintage-level observed concentration high. Seagate witnesses · UC table, IVP III–VII p. 1 · IVP VIII p. 2
• Reserves. [researcher inference · moderate] Dennis recalls repeated Collagen financing, Reference as too early, and a first IVP fund that exceeded its promised startup workload; a later Applied filing says IVP skipped the final 1992 venture round even though the stake remained visible in 2012. No reserve ledger shows initial/follow-on ratios, loss escalation, recycling, or write-offs. Collagen and Reference, printed pp. 50–55/PDF pp. 56–61 · first-fund promise, printed pp. 46–48/PDF pp. 52–54 · Applied February 2013 S-1/A, prospectus pp. 45–46
• Governance. [researcher inference · strong] The visible modes include a long board tenure and CEO replacement (Recognition), sponsor recusal and a different approval group (ROLM), observer status (LSI), all-partner field diligence (Exabyte), repeated biomedical oversight (Collagen), and contractual demand rights against company preference (Applied). “Founder friendly” or “control investor” alone describes none of them. Recognition, printed pp. 16–21/PDF pp. 22–27 · ROLM, printed pp. 30–32/PDF pp. 40–42 · LSI panel, transcript p. 36 · Exabyte, Dennis printed pp. 53–56/PDF pp. 59–62 · Collagen, Dennis printed pp. 50–55/PDF pp. 56–61 · Applied S-1/A
• Exits/liquidity. [researcher inference · strong] Visible paths include partial public sale to recover basis, IPO and in-kind distribution, strategic acquisition, a corporate distribution into a successor holding, a personal sale for family liquidity, holdings retained to zero, and an attempted registration that never became effective. An IPO is not automatically a realized investor return. Recognition and Seagate, printed pp. 16–21 and 58–60/PDF pp. 22–27 and 64–66 · Measurex, Bossen printed pp. 19–20/PDF pp. 25–26 · Cohesion 13G · Microsoft/public losses, raw transcript 2, printed pp. 17–18 and 40/PDF pp. 17–18 and 40 · Applied withdrawal

Reported Group outcomes remain separate

[retrospective witness · provisional] Dennis and Johnson recall about twenty-five Group companies, around eighteen successes, two that returned capital, and roughly three large losses. [researcher inference · strong] The counts do not perfectly reconcile, names are absent, “success” is undefined, member allocations differ, and neither value-weighted nor cash-flow-weighted performance can be computed. The recollection supports frequent positive company outcomes in memory, not audited personal returns or a literal hit rate. CHM video, 00:35:07–00:39:15

People and capital versus personal “alpha”

[researcher inference · strong] The supported mechanism is collective: family/trust access at Ampex; employer balance-sheet scale at Recognition; syndicate sourcing at Measurex; partner disclosure and approval at ROLM; Marquardt sourcing and board work at Seagate; Sequoia access and Corrigan execution at LSI; all-partner diligence at Exabyte; specialist biomedical partners and repeated capital at Collagen; and LP rights at Applied. Dennis made consequential choices—authorizing Seagate, overriding LSI dissent, serving long on Recognition, and explaining Applied liquidity—but public evidence cannot separate his selection effect from co-investor access, founder execution, customer demand, market cycles, and vehicle duration. It would be equally wrong to erase him from those decisions or assign him every IVP outcome. Dennis oral history, printed pp. 13–60/PDF pp. 19–66 · Measurex, Bossen printed pp. 19–20/PDF pp. 25–26 · McMurtry on ROLM, printed pp. 30–32/PDF pp. 40–42 · Marquardt on Seagate, printed pp. 28–30/PDF pp. 33–35 · LSI decision account, raw transcript 2, printed pp. 25–29/PDF pp. 25–29 · LSI founder/operator account, transcript pp. 15 and 36 · Applied S-1/A

Failure, miss, and anti-portfolio postmortems

Uncertainty and retrieval ledger

Priority rule. [researcher inference · strong] Retrievals are ordered first by their expected ability to change the career denominator or economic conclusion, then person/deal attribution, then a single-case decision lab; ties favor records that could fill multiple missing schema fields. This is an analyst research queue, not a probability-of-retrieval estimate or a ranking of historical importance. Unresolved-record table

Saturation result. [researcher inference · strong] The final three searches—SEC records for IVP I–II plus Dennis, UC records for IVP I–II, and CHM records for a Dennis Group portfolio/loss ledger—yielded no new material deal or cash-flow evidence. A duplicate 2017 UC table repeated the IVP III–VII values and was not promoted into an independent source family. The absence is a bounded retrieval result, not evidence the underlying records never existed. Source-map attempted-route ledger

[researcher inference · strong] The evidence most likely to change the named-deal economic-evidence ranking is a compatible cash-flow schedule for Seagate, Recognition, and Ampex; the first IVA fund belongs in the separate fund-level controls. The evidence most likely to change personal attribution is a sponsor/approval/board ledger for Measurex, ROLM, Seagate, LSI, Exabyte, Collagen, and Applied. Until then, the narrow conclusion is: Dennis helped create several successive venture-capital institutions and made identifiable consequential decisions, but the selected public record cannot produce a career return, strict hit rate, complete denominator, or auditable personal 80/20 result. Dennis oral history · McMurtry oral history · UC methodology, p. 7

Coverage and evidence base

Literal priority-source reconciliation

At the WRITINGS cutoff, the source map contained 66 priority rows: 29 ★★★ and 37 ★★. Every priority row was routed once below by its primary analytical destination, and the counts reconcile exactly: 14 WRITINGS + 17 TALKS + 12 DEALS + 23 PROFILE/SYNTH = 66. Routing says which phase owned deep analysis at that cutoff; it does not claim that a gated body was read.

The ★ “Institutional Venture veterans” feature is outside the literal priority denominator but is included in the written corpus as a supplementary, eligible direct-profile route because its registration barrier could be tested cheaply. It received a bounded note rather than being silently discarded. Low-priority speech reports and event fragments remain with TALKS; aviation-only, derivative, posthumous, or third-party hiring narratives remain contextual rather than inflating the direct written denominator.

Written-record denominator

The written record contains 15 distinct artifacts or publication families: five authored/signed candidates or reported artifacts, plus ten published direct-response, interview, handout, or official-testimony records. The first group is an inventory denominator, not a claim that every blocked artifact or Dennis-only wording has been independently authenticated. The access accounting is complete:

The authored/signed candidate sub-denominator is 5 = 2 full + 0 materially partial + 3 blocked. The two full items are the 1979 prepared statement and 2002 Perkins introduction. The three blocked candidates are the 1981 article, untitled MBA paper, and Applied Medical declaration. The 1977 committee print is deliberately outside this candidate sub-denominator because it is staff paraphrase, not Dennis prose.

Exact access boundaries for the blocked record

Researcher inference · moderate: no owned blog, memoir, public LP letters, personal newsletter, published memo series, or standalone papers collection surfaced. The 1981 article is the only known major authored work likely recoverable through an authenticated Internet Archive loan or institutional ScienceDirect access; no other major openly retrievable authored item remains unprocessed as of 2026-08-02. Source-map attempted-route and direct-voice audit

What the surviving written record actually establishes

The public written record does not expose a complete Reid Dennis investment system. Its strongest contemporaneous authored item is policy advocacy, not an investment memorandum: in 1979 Dennis argues that venture-company formation depends on capital and people, and that employee equity helps young firms recruit when they cannot match incumbent salaries or security. Investor-stated · strong for the two-input diagnosis; investor-stated · moderate for his tax remedy; researcher inference · insufficient public record for whether the remedy improved company or fund outcomes. Prepared statement, printed pp. 1060–1066

The best ex-ante investment judgment appears in a reported interview rather than an authored essay. In January 2000 Dennis simultaneously expected broad Internet adoption and rejected then-current Internet-stock prices as disconnected from value. He refused a precise correction date. Investor-stated · strong for the price/value distinction and timing humility; researcher inference · moderate for the reusable two-stage underwriting rule; researcher inference · insufficient public record for any corresponding IVP trade or return. Almanac, body ¶¶4 and 13–19

The retrospective corpus describes a small, relationship-based syndication mechanism and a preference for more viable companies rather than accepting a few extreme winners as the governing portfolio design. Yet the reported 60–65% success rate and roughly eighteen Group successes have no company ledger, stable definition, ownership record, cash-flow series, or benchmark. Retrospective witness · provisional for the remembered counts; researcher inference · insufficient public record for superior selection skill or a non-power-law return distribution. Inc., body ¶¶4–5 and 13–14 · Institutional Investor, body ¶4

Finally, the record contains material counterevidence about access. Richard Leza alleged that Dennis described him as outside the industry's familiar image; Dennis said he did not remember the meeting and emphasized scarce positions and unusually credentialed applicants. Retrospective witness · provisional for Leza's allegation; investor-stated · moderate for Dennis's response; researcher inference · strong that neither is an adjudication, admission, disproof, or exoneration. SFGate, body ¶¶3–5 and 11–14

Overall boundary — researcher inference · moderate: Dennis's written record supports a capital-plus-people model, employee ownership as recruiting compensation, separation of technology adoption from entry price, timing humility, small-network syndication, and a preference for viable-company batting average. It does not establish his full sourcing screen, pricing method, ownership targets, portfolio construction, reserve policy, board operating system, exit rules, or audited performance. 1979 statement, printed pp. 1060–1066 · Almanac, body ¶¶11–17 · Institutional Investor, body ¶4 · Inc., body ¶¶13–14

Source-by-source analysis

JCT/CRS — summary of Dennis's 1977 pension-investment testimony

Context, authorship, and access. Dennis testified on May 11, 1977 before a Senate panel considering the Tax-Exempt Private Pension Investment Act. The accessible Joint Committee on Taxation/CRS print, issued August 2, supplies one staff-written paragraph at printed p. 8. It is a complete official summary but a partial Dennis record: his prepared text, evidence, exhibits, questions, and wording are absent. JCT/CRS summary, printed p. 8/PDF p. 5 · source note

Argument and reasoning. Contemporaneous record · strong: staff report that Dennis supported changing a five-percent concentration rule from “any class” of security to voting or equity securities, wanted the bill's two-percent leeway measured at asset cost rather than fluctuating market value, and stressed that section 3 permitted rather than compelled investment. JCT/CRS summary, printed p. 8/PDF p. 5

Evidence, assumptions, and operational rule. The staff paragraph preserves no example or empirical support. Researcher inference · moderate: the proposals share a design logic—measure the economic exposure that matters, stabilize the denominator, and open fiduciary capacity without imposing a quota. That is an analyst reconstruction, not Dennis's phrasing. Cost basis may stabilize compliance but can also conceal appreciated concentration, so it cannot be imported as a modern prudence rule without portfolio-level analysis. JCT/CRS summary, printed p. 8/PDF p. 5

Biography, decision relevance, and contradiction. The official hearing notice places Dennis on a panel with David Morgenthaler, Richard Hanschen, and Stewart Greenfield; it does not make him sole architect of the proposal or later law. This is an ex-ante coalition position before the 1978 tax act and 1979 final investment-duties rule. Researcher inference · strong: later stories that compress ERISA, Dennis's advocacy, and the prudence rule into a single 1978 event are historically unsafe. Congressional Record, 1977-05-09, printed pp. 14000–14001 · 1979 final rule, 44 FR 37221–22

Dennis/NVCA — “Emerging Innovative Companies, Capital Formation and the Need for Incentives” (1979)

Context, authorship, and access. The Senate print identifies Dennis as NVCA chair and an IVA general partner and prints his full prepared statement at pp. 1057–1067. It is direct authored testimony in an association-policy role, not an IVA fund letter or sole-person policy project. The body and exhibits are fully visible through the public Google Play Books reader. Introduction and title page, printed pp. 1056–1057 · source note

Central argument and reasoning chain. Dennis begins with a supply claim: the 1978 capital-gains reduction revived independent venture fundraising. He then says capital is only one input. Emerging companies sit between founder/family resources and a possible public offering, face higher financing needs and underwriting thresholds, and require qualified managers and key employees. Because they cannot match established-company cash pay or job security, he presents prospective employee ownership as risk compensation and recommends restoring qualified stock options. Investor-stated · strong for the people constraint; investor-stated · moderate for the capital-gains causal story and policy remedy. Prepared statement, printed pp. 1058–1062

Evidence and examples. Exhibit I reports independent venture firms raising $20 million in 1977, $215 million in 1978, and $69 million through May 15, 1979, with twenty other groups seeking $250–$300 million. The exhibit names NVCA and Capital Publishing Corporation as sources and combines known data with estimates described as reliable. Contemporaneous record · provisional: the association-supplied series is an advocacy dataset without a fund list, methodology, counterfactual, or coverage of all informal, corporate, SBIC, and personal capital. Exhibit I, printed p. 1063

Exhibit II compares then-current and proposed option treatment through one stylized grant at 5, exercise at 20, and sale at 30, under simplified tax-rate assumptions. It projects 7 of government revenue under the proposal against 2.5 under then-current law. Contemporaneous record · provisional: the association-supplied exhibit is a constructed example, not an aggregate revenue or employee-welfare estimate. Exhibit II, printed pp. 1064–1066

Assumptions, rule, and operational implications. The argument assumes more risk capital creates worthwhile company formation, qualified employees value illiquid equity enough to offset lower certainty, and favorable option taxation improves recruitment and ownership. It does not model dilution, exercise financing, AMT, concentration of worker wealth, failed-company options, retention, or unequal grant allocation. Researcher inference · moderate: underwrite the scarce complement to money—management, technical, and operating capacity—and test whether contingent equity credibly compensates each recruit for cash, liquidity, tax, dilution, and company-specific risk. Prepared statement, printed pp. 1060–1066

Biography, actual decisions, and evolution. Investor-stated · moderate: Dennis welcomes new venture firms even though they increase competition, saying the incumbent market had become too selective for the economy's good. Researcher inference · moderate: that sits usefully beside his 2002 Perkins introduction recalling 1974, where he describes scarce good deals immediately after IVA closed its first fund: supply of funds and supply of fundable projects can move differently. Researcher inference · insufficient public record: the statement names no Dennis investment, hiring decision, board action, reserve decision, or exit, so its firm-level application is unproven. Prepared statement, printed pp. 1059–1061 · Perkins introduction, PDF p. 23/printed p. vii

Contradictions and limit. Researcher inference · moderate: Dennis acknowledges fundraising that began before the 1978 bill was signed, weakening a simple post-enactment causal story. Contemporaneous record · strong for Senator Gaylord Nelson later quoting the personnel passage; researcher inference · strong that quotation shows legislative uptake but not Dennis authorship of S. 2239 or causal influence on enactment. Prepared statement, printed p. 1059 · Congressional Record, 1980-02-06, printed pp. 2108–2109

SFGate — “Roadblocks for Latino Venture Capitalist” (1998)

Context, authorship, and access. Laura Castaneda's full reported profile examines Richard Leza's difficulty entering venture capital. Retrospective witness · provisional: it presents Leza's account of an early-1980s conversation and Dennis's 1998 response; the alleged event is remembered about fifteen years later without notes, witnesses, correspondence, or hiring records. It is not Dennis-authored. SFGate, body ¶¶1–14 · source note

Claims and reasoning. Retrospective witness · provisional: Leza says Dennis told him he would struggle because he did not talk or look like incumbent venture capitalists. Investor-stated · moderate: Dennis says he does not remember the meeting and describes venture hiring as many unusually qualified applicants competing for very few positions. His response supplies a scarcity explanation but no decision criteria, applicant denominator, or evidence separating competitive selection from familiarity-based exclusion. SFGate, body ¶¶3–5 and 11–12

Evidence and counterevidence. The article reports Leza's Stanford MBA, engineering degree, 4.0 GPA, and large-company and startup experience. Those facts make Dennis's visible “good, average education” explanation incomplete, but the full candidacy and firms' evaluations are absent. Leza's broader explanation is homophily: investors recruit people with familiar backgrounds and search narrowly. Other named examples illustrate the claim but do not create a representative hiring study. Retrospective witness · moderate for Leza's mechanism; researcher inference · insufficient public record for prevalence or Dennis-specific intent. SFGate, body ¶¶1–2 and 11–14

Decision implication and boundary. Researcher inference · moderate: relationship-based selection can lower coordination costs while shrinking the observed opportunity set. A defensible process must measure who enters the referral funnel, who is screened, the reasons for rejection, and whether similarly qualified candidates receive comparable evaluation. The article does not prove a discriminatory act; Dennis's non-recollection does not disprove one. SFGate, body ¶¶6 and 10–14

Evolution and contradiction. Two years later Dennis told the Almanac that a good startup idea could obtain funding regardless of family or school. That statement concerns entrepreneurs seeking financing; the 1998 article concerns hiring investment professionals. Researcher inference · strong: the records create a tension between openness rhetoric and a possible network gate, not a literal same-market reversal. SFGate, body ¶¶11–14 · Almanac, body ¶¶5–8

Almanac — Internet adoption, crowded entry, and valuation (2000)

Context, authorship, and access. Contemporaneous record · strong: Barbara Wood's full January 5, 2000 profile records Dennis during the dot-com boom, before the later collapse, making the warning genuinely ex ante. It is a short edited interview without the question list or transcript. Almanac, full article · source note

Central argument and reasoning. Dennis separates the technology from its securities. He expects large remaining growth in Internet connectivity while calling Internet-stock prices irrationally disconnected from value. His reasoning is analogical: Cleveland once led automobiles, crowded startup funding resembles musical chairs, and speculative enthusiasm recalls tulip bulbs. Investor-stated · strong for the price/value distinction; investor-stated · moderate for the adoption and historical forecasts; the article supplies no comparative valuation or adoption dataset. Almanac, body ¶¶4 and 9–19

Picking and timing rules. Dennis expects late or unnecessary dot-com companies to fail but names no company or necessity test. He declines to time the correction closely, offering an eighteen-to-thirty-six-month range while saying he sees no near-term blow-up. Researcher inference · moderate: first test whether the technology creates durable utility; then independently test whether the particular company is differentiated and whether entry price permits a return. Recognizing a dislocation is not a market-timing edge. Almanac, body ¶¶11–17

Evidence, application, and limits. The profile names Ampex, Collagen, Seagate, and Excite as companies in which Dennis invested but gives no vehicle, date, round, role, ownership, governance action, or outcome. It supplies no evidence that Dennis or IVP reduced exposure, changed terms, or sold positions before the correction. Researcher inference · strong: an accurate warning without an action and cash-flow record is intellectual evidence, not investment-performance evidence. Almanac, body ¶¶2–3 and 11–19

Access rhetoric and contradiction. Investor-stated · provisional: Dennis calls Silicon Valley uniquely attractive and says family and school do not matter if an entrepreneur has a good idea. Researcher inference · strong: the claim has no applicant or funding denominator and cannot erase the SFGate hiring record. Researcher inference · moderate: it also should not be universalized beyond the 2000 Peninsula market. Almanac, body ¶¶5–8 · SFGate, body ¶¶3–14

Dennis — introduction to the Tom Perkins oral history (2002)

Context, authorship, and access. Contemporaneous record · strong: Dennis signed this one-page introduction in Menlo Park on January 6, 2002 as IVP Partner Emeritus. Researcher inference · moderate: it is a commemorative tribute selected to honor Perkins, so favorable tone and hindsight are structural features. Dennis introduction, PDF p. 23/printed p. vii · source note

Biography and market context. Dennis dates his first meaningful Perkins encounter to a May 1974 University of Washington seminar. He says that one month earlier Burge Jamieson, Burt McMurtry, and Dennis closed a $19 million IVA fund and then faced rising interest rates, a falling stock market, and scarce good entrepreneurs and deals. Retrospective witness · moderate for both the remembered chronology and market characterization nearly twenty-eight years later. Dennis introduction, PDF p. 23/printed p. vii, paragraph 1

Judgment and reasoning. Dennis praises Perkins for decomposing venture capital like an engineering problem, attending to detail, thinking analytically, setting high goals, and seeking industry leadership. Investor-stated · moderate: these are qualities Dennis admired in a peer, not direct proof of Dennis's own process or of Perkins's outcomes. His claim that Perkins built the “#1 firm” supplies no date, metric, comparison, or return denominator. Dennis introduction, PDF p. 23/printed p. vii, paragraphs 2 and 4

Example and operational rule. Retrospective witness · moderate: Dennis recalls Kleiner Perkins joining IVA in Collagen and Perkins serving as a constructive director; FDA approval risk inverted normal staging because the necessary capital had to be committed before the decisive risk could be reduced. The explicit maxim—invest a little to reduce risk before committing more—is Perkins's speech as reported by Dennis. Researcher inference · moderate: Dennis's choice to preserve the exchange shows that risk-retirement sequencing was memorable; it does not transfer authorship of the rule or prove Dennis applied it consistently. Dennis introduction, PDF p. 23/printed p. vii, paragraph 3

Actual-decision and access limits. Researcher inference · insufficient public record: the example identifies two firms and a board role but no date, check, ownership, financing sequence, milestone, vote, or outcome. Investor-stated · provisional: Dennis calls the FDA highly politicized without evidence. Researcher inference · moderate: the useful diagnostic is narrow—distinguish risk that staged spending can reduce from regulatory or technical risk requiring substantial precommitment, then price and size that irreducible exposure explicitly. Dennis introduction, PDF p. 23/printed p. vii

Institutional Investor — “Venture Capital's Early Pioneers” (2010)

Context, authorship, and access. Udayan Gupta's full trade-press history combines a short Dennis recollection with the editor's industry narrative. Researcher inference · moderate: because Gupta also edited Done Deals, overlap may derive from one interview family; repetition would show persistence, not independent corroboration. Institutional Investor, full article · Done Deals route · source note

Mechanism and evidence. Dennis recalls Group members bringing opportunities, evaluating them together, contributing about $25,000–$30,000 each, and calling friends to fill typical $250,000–$300,000 rounds. He reports roughly twenty-five investments over ten years, eighteen of them highly successful. Retrospective witness · provisional: the mechanism is concrete, but the result claim lacks names, dates, membership by deal, ownership, follow-ons, loss handling, success definition, and cash proceeds. Institutional Investor, body ¶4

Decision rule and limits. Researcher inference · moderate: small personal checks, repeated relationships, and referrals allowed a local network to assemble financings larger than any member's check. The article does not show whether this improved selection, governance, or returns, and it gives no reserve or concentration rule. Group screening, individual ownership, friendship syndication, and a formal fund must remain separate legal and analytical categories. Institutional Investor, body ¶4

Material chronology defects. The article says Dennis launched IVP in 1974 with $18 million and that “they” made a first investment in 1952. Better-specified direct testimony distinguishes a reported $19 million IVA close in 1974 from IVP's formation in 1980, and places Dennis's personal Ampex investment in 1952 before the Group's later period. Researcher inference · strong: the article cannot safely establish vehicle, fund size, founder roster, collective start date, or deal attribution. Institutional Investor, body ¶4 · Dennis oral history, printed pp. 27, 31, and 47

Policy contradiction. Gupta attributes institutional entry to “ERISA in 1978.” ERISA was enacted in 1974, while the relevant final investment-duties regulation was published in 1979. Contemporaneous record · strong for the official dates; researcher inference · moderate that later narrative compression overstates a single-event explanation. Institutional Investor, body ¶¶5–6 · 44 FR 37221–22

Inc. — “Then and Now: Venture Capital” (2011)

Context, authorship, and access. Eric Markowitz's full interview/profile accompanied Something Ventured, a documentary celebrating industry pioneers. Researcher inference · moderate: the source is direct edited voice, but selection, promotional, nostalgia, and survivor bias are unusually salient. Inc., full article · source note

Syndication mechanism and people thesis. Dennis describes roughly five working professionals inviting a Peninsula entrepreneur to lunch, pooling about $100,000, and phoning prior co-investors to complete a round. He contrasts early participants' constructive interest in helping people build companies with later entrants whom he sees as motivated chiefly by wealth. Retrospective witness · moderate for the remembered process; investor-stated · moderate for his normative contrast; researcher inference · insufficient public record for the unnamed actors, deal allocation, diligence, board work, or any survey of motives. Inc., body ¶¶4–8

Portfolio claim and reasoning. Dennis reports roughly 60–65% of early companies successful and another 10–15% returning capital, then rejects a philosophy satisfied with two large wins in ten. Retrospective witness · provisional for the remembered results; investor-stated · moderate for the higher-batting-average preference. Neither establishes a less-skewed return distribution: the cohort, vehicle, years, success definition, capital weighting, follow-ons, losses, and realized cash flows are unknown. Inc., body ¶¶13–14

Operational implication and failure mode. Researcher inference · moderate: know people through direct interaction, syndicate with repeat collaborators, and seek a larger fraction of viable enterprises instead of accepting failure as costless. The failure mode is optimizing an easy-to-remember company count while ignoring ownership, magnitude, time, dilution, and opportunity cost. A venture fund can have many viable companies and weak returns, or a low hit rate and exceptional returns. Inc., body ¶¶4–5 and 13–14

Evolution and contradictions. The 2011 story turns a 1950s apprenticeship into a critique of an industrialized asset class. Its ERISA date repeats the same 1978 compression as the 2010 article, while the legal record supports 1974 enactment and a 1979 final prudence rule. The “constructive pioneers” frame also omits failed pioneers and modern investors who provide operating help. Researcher inference · moderate: motive is a diligence question, not a performance proxy or moral halo. Inc., body ¶¶6–10 and 13–14 · 44 FR 37221–22

Cross-source synthesis

Capital is necessary, but the scarce complement changes

In 1977 Dennis seeks legal room for pension fiduciaries to invest rather than a mandate that they do so. In 1979 he says capital supply has recovered but company formation still depends on capable people willing to accept startup risk. In his 2002 recollection of 1974, a newly closed $19 million IVA fund faced scarce good entrepreneurs and deals. Researcher inference · strong: across these records, money is a system input whose usefulness depends on fiduciary discretion, fundable projects, and operating talent; more capital alone is never a complete company-formation thesis. 1977 JCT summary, printed p. 8/PDF p. 5 · 1979 statement, printed pp. 1060–1062 · 2002 introduction, PDF p. 23/printed p. vii

The causal chain is conditional: policy can expand eligible capital; capital can finance more candidates; company formation still requires differentiated projects and recruits; employee equity may help recruit them; execution and exits determine whether the capital creates returns. The sources support the first four links as Dennis's argument or memory, not the last link as a demonstrated outcome. Researcher inference · moderate. 1979 statement, printed pp. 1058–1066

Technology truth and investment truth are different propositions

The 2000 profile is the corpus's clearest foundational investment insight: broad Internet adoption can be real while particular companies are unnecessary and traded prices are irrational. Investor-stated · strong for the separation; researcher inference · moderate for a three-part underwriting sequence—technology utility, company differentiation, and price/terms. Neither a correct technology forecast nor an eventual market correction proves that Dennis selected the right private company or timed an exit. Almanac, body ¶¶4 and 11–19

This distinction also disciplines the 1979 policy argument. A larger venture-capital supply and useful employee option regime may improve financing capacity without making every funded project worthwhile or every option valuable. Researcher inference · moderate: analyze system capacity, company quality, and security economics separately. 1979 statement, printed pp. 1058–1066

Relationships are both financing infrastructure and a possible gate

The 2010 and 2011 retrospectives describe a repeated network: members source opportunities, meet founders directly, pool small personal checks, and call prior co-investors. That network plausibly reduces search and trust costs and expands financing capacity. Retrospective witness · moderate for the remembered mechanism; researcher inference · insufficient public record for its return advantage. Institutional Investor, body ¶4 · Inc., body ¶¶4–5

The SFGate record supplies the necessary adverse mechanism: familiar networks can also narrow who gets considered for investor roles. Dennis's 2000 openness claim applies to entrepreneur funding rather than VC employment, so it does not resolve the tension. Researcher inference · moderate: use relationships as evidence channels, not eligibility rules; measure reach, conversion, and rejection reasons. SFGate, body ¶¶3–14 · Almanac, body ¶¶5–8

Batting average is a preference, not a return proof

Dennis twice remembers unusually high early-company success and rejects a portfolio design content with two big wins out of ten. Researcher inference · moderate: the recurrence supports a stable preference for viable-company frequency. It does not establish the reported rate, its economic merit, or a portable process. Retrospective witness · provisional for the remembered counts; investor-stated · moderate for the stated preference; researcher inference · strong that company count, capital-weighted loss rate, fund TVPI, DPI, IRR, and partner attribution are different measures. Institutional Investor, body ¶4 · Inc., body ¶¶13–14

Researcher inference · strong: the strongest test would reconcile The Group's approximately twenty-five companies, checks and follow-ons by member, write-offs, distributions, holding periods, and market values. Until then, the claim is useful as a portfolio question—what failure distribution are we underwriting?—not as evidence that Dennis beat a power-law strategy. Institutional Investor, body ¶4 · Inc., body ¶¶13–14

Risk should be sequenced, but some risk requires precommitment

The Collagen anecdote preserves a useful tension. Perkins's stated norm is to invest incrementally and use each check to retire risk; FDA approval forced the parties to commit all required capital before the decisive risk could be reduced. Researcher inference · moderate: a staged-financing model must distinguish milestones that genuinely create information from unavoidable fixed-cost or regulatory gates. Otherwise “staging” becomes either false comfort or chronic underfunding. Dennis introduction, PDF p. 23/printed p. vii

Rhetoric is much richer than documented action

Researcher inference · moderate: the written corpus supports macro policy, people, valuation, network, and risk-sequencing ideas more strongly than actual fund operations. The evidence by function is deliberately uneven. 1979 statement, printed pp. 1058–1066 · Almanac, body ¶¶11–19 · Inc., body ¶¶4–14 · source-map denominator audit

Intellectual chronology, evolution, and contradictions

Three tensions should remain unresolved rather than averaged away:

1. More capital versus too much undisciplined capital. Dennis welcomes increased fund supply in 1979, warns about crowded Internet financing in 2000, and criticizes wealth-motivated entrants in 2011. Researcher inference · moderate: the consistent variable may be productive intermediation rather than capital quantity, but no source supplies a threshold. 1979 statement, printed pp. 1059–1061 · Almanac, body ¶¶11–19 · Inc., body ¶¶7–8
2. Trusted relationships versus open access. — Researcher inference · moderate. Group accounts celebrate repeat networks; SFGate documents a credible exclusion mechanism; the Almanac offers broad entrepreneur-openness rhetoric. These sources cover different gates and cannot adjudicate each other. Inc., body ¶¶4–5 · SFGate, body ¶¶3–14 · Almanac, body ¶¶5–8
3. High viability versus venture return skew. — Researcher inference · insufficient public record. Dennis prefers more company successes, but the public record does not show whether that preference generated better net returns or simply reflects a different company, era, or success definition. Inc., body ¶¶13–14

Decision playbook reconstructed from the written record

This is an analyst reconstruction, not a Dennis-authored checklist. Each rule preserves its evidence grade, diagnostic questions, and failure boundary.

1. Map the system before underwriting the security. — Researcher inference · moderate. Historical rule: capital is one input alongside fundable projects and qualified people. Ask which complement is actually scarce, who controls it, and what evidence shows money can unlock it. Failure mode: assuming a larger round fixes a talent, demand, regulation, or execution bottleneck. 1979 statement, printed pp. 1060–1062

2. Create capacity without confusing permission with allocation. — Researcher inference · moderate. Historical rule: Dennis's 1977 position sought permissive fiduciary room, not compulsory venture exposure. Ask whether an institutional rule expands a choice or silently becomes a quota, and what portfolio analysis must precede use. Failure mode: treating asset-class eligibility as evidence that a particular fund is prudent. JCT/CRS summary, printed p. 8/PDF p. 5

3. Use a stable denominator, then separately monitor current exposure. — Researcher inference · moderate. Historical rule: Dennis preferred cost for a statutory leeway calculation. Ask what denominator best matches the legal or operating decision and how appreciation, impairment, follow-ons, and distributions alter actual risk. Failure mode: allowing stable measurement to hide current concentration. JCT/CRS summary, printed p. 8/PDF p. 5

4. Underwrite employee equity as risky compensation, not free upside. — Researcher inference · moderate. Historical rule: young firms can use prospective ownership to offset lower salary and security. Ask what cash discount the recruit bears, which option regime applies, when exercise and sale occur, whether AMT or a holding-period failure changes treatment, and whether the grant is broad or selectively allocated. Failure mode: counting headline option value as realized after-tax compensation. 1979 statement, printed pp. 1061–1066 · IRS Topic 427, reviewed 2026-04-28

5. Welcome financing capacity only if selection capacity keeps pace. — Researcher inference · moderate. Historical rule: Dennis welcomed new venture firms after believing incumbents had become overly selective. Ask whether new capital adds sourcing, domain judgment, governance, and follow-on capacity or only bidding pressure. Failure mode: equating fund count with useful company formation. 1979 statement, printed p. 1060

6. Separate technology adoption from company quality and entry price. — Investor-stated · strong for the distinction; researcher inference · moderate for the rule. Ask whether the underlying technology creates durable utility, whether this company captures it, and whether price and terms leave a return. Failure mode: using a correct secular trend to justify every participant or valuation. Almanac, body ¶¶4 and 11–17

7. Demand a necessity and differentiation test in crowded categories. — Researcher inference · moderate. Historical rule: late or unnecessary dot-coms can fail even when the Internet grows. Ask which user problem disappears without the product, why the company is not redundant, and what evidence shows timing advantage. Failure mode: financing category membership rather than a differentiated wedge. Almanac, body ¶¶11–14

8. Do not convert valuation conviction into false timing precision. — Investor-stated · strong for timing humility; researcher inference · moderate for the rule. Ask what evidence would change the value judgment, what liquidity runway can survive delayed correction, and which actions do not require predicting the exact turn. Failure mode: being directionally right but financially unable to wait. Almanac, body ¶¶15–17

9. Use repeated relationships to assemble information and capital. — Researcher inference · moderate. Historical rule: direct founder meetings, member referrals, and repeat co-investors can lower coordination costs. Ask who sourced the evidence, which participant independently challenged it, and who is responsible after financing. Failure mode: letting trust substitute for diligence or leaving decision rights implicit. Institutional Investor, body ¶4 · Inc., body ¶¶4–5

10. Audit the network's blind spots and entry funnel. — Researcher inference · moderate. Ask who never receives a referral, whether comparable candidates are judged by comparable criteria, and how sourcing expands beyond familiar backgrounds. Failure mode: interpreting homophily as merit because the excluded opportunity set remains unseen. SFGate, body ¶¶3–14

11. Keep person, syndicate, corporation, partnership, and fund attribution separate. — Researcher inference · strong. Ask who supplied capital, sourced, approved, negotiated, served on the board, authorized follow-ons, and influenced exit. Failure mode: assigning Dennis's 1952 personal investment to The Group or collapsing 1974 IVA into 1980 IVP. Institutional Investor, body ¶4 · Dennis oral history, printed pp. 27, 31, and 47

12. Stage capital only against milestones that retire real risk. — Retrospective witness · moderate for Perkins's reported statement; researcher inference · moderate for use. Ask what will be learned before the next check, whether the result changes invest/stop/terms, and whether management can reach it with the tranche. Failure mode: cosmetic staging that delays an inevitable full commitment. The maxim belongs to Perkins as quoted by Dennis, not to Dennis. Dennis introduction, PDF p. 23/printed p. vii

13. Name irreducible precommitment risk before investing. — Researcher inference · moderate. If regulation, manufacturing, infrastructure, or a pivotal trial requires capital before information arrives, ask for total exposure, financing contingency, survival runway, and explicit stop conditions. Failure mode: funding half of a binary program and discovering that the “option” has no standalone value. Dennis introduction, PDF p. 23/printed p. vii

14. Track viability and economics on separate scorecards. — Researcher inference · strong. Ask for company outcomes and capital-weighted cash flows, ownership, dilution, time, and opportunity cost. Failure mode: calling a portfolio superior because many companies survive while ignoring whether the fund returned capital competitively. Inc., body ¶¶13–14 · Institutional Investor, body ¶4

15. Diligence motivation, but never use it as a performance proxy. — Researcher inference · moderate. Ask why each investor and founder wants the work, what behavior that predicts under stress, and what incentives contradict the story. Failure mode: turning “constructive” motives into a halo for competence or treating financial ambition as proof of poor governance. Inc., body ¶¶7–10

16. Separate a stated belief, an action, and an outcome. — Researcher inference · strong. Ask what Dennis said before the event, what vehicle actually did, who made the decision, and what cash-flow result followed. Failure mode: treating the 2000 warning as a successful trade, the 1979 testimony as policy causation, or a remembered success count as audited performance. Almanac, body ¶¶11–19 · 1979 statement, printed pp. 1058–1066 · Inc., body ¶¶13–14

Modern VC translation (as of 2026-08-02)

Contemporaneous record · moderate: the transfer test must start with changed premises. NVCA's Q2 2026 page says U.S. startups raised more than $400 billion in the first half, with the overwhelming majority of capital flowing to AI companies and $100 million-plus rounds; investment, fundraising, and exits remained concentrated among relatively few companies and funds. Its 2026 Yearbook reports 65.4% of 2025 deal value in AI and 67% of total value in 487 mega-deals. Researcher inference · strong: these are current association/PitchBook market observations, not evidence that a Dennis rule produces returns. NVCA Q2 2026 Venture Monitor, summary and bullets · NVCA 2026 Yearbook, “Two Markets” and mega-deal sections

Researcher inference · strong: current pension prudence cannot be reduced to Dennis's 1977 cost-basis proposal. Contemporaneous record · strong: the eCFR, current through July 30, 2026, requires ERISA plan fiduciaries with investment duties to give appropriate consideration to an investment's portfolio role, risk and opportunity relative to reasonably available alternatives, diversification, liquidity/current return, projected return, and funding objectives. 29 CFR §2550.404a-1(b), current as of 2026-07-30

Researcher inference · strong: the 16 historical playbook rules map to 14 modern translations: relationship use and funnel audit are tested together, as are staged financing and irreducible precommitment risk. Every rule is therefore transferred once without pretending the paired diagnostics operate independently. Reconstructed playbook

Five writings to read first

Ranking method — researcher inference · moderate: prioritize directness, contemporaneity, decision usefulness, adversarial value, access completeness, and independence from the later oral-history corpus. This is a reading order, not a quality or performance ranking. Corpus accounting and access boundaries

1. “Emerging Innovative Companies, Capital Formation and the Need for Incentives” (1979). The only full, substantial contemporaneous Dennis-authored statement; it exposes his capital-plus-people causal chain, source data, tax example, assumptions, and advocacy limits. Prepared statement, printed pp. 1057–1067 · note
2. Almanac Internet/valuation profile (2000). The strongest ex-ante investment judgment: a correct adoption thesis can coexist with irrational security prices, crowded entry, and uncertain correction timing. Full article · note
3. “Roadblocks for Latino Venture Capitalist” (1998). Essential adversarial reading on the possible exclusion cost of a relationship-driven industry; preserves both Leza's allegation and Dennis's non-recollection without resolving them. Full article · note
4. Dennis introduction to the Tom Perkins oral history (2002). The other full signed writing; supplies IVA chronology, admired analytical traits, a concrete regulatory-risk example, and an attribution test because the staged-financing maxim belongs to Perkins. Introduction, PDF p. 23/printed p. vii · note
5. “Then and Now: Venture Capital” (2011). The clearest compact account of Group syndication, constructive motivation, and Dennis's high-batting-average preference—best read with its missing denominator and nostalgia bias in view. Full article · note

The 2010 Institutional Investor article is useful as a chronology-error exercise but ranks below these five because its Dennis passage is short, likely overlaps Gupta's Done Deals work, and compresses 1952 personal capital, 1974 IVA, 1980 IVP, and ERISA dates. Full article

Unresolved written record and retrieval agenda

Planning label — researcher inference · moderate: this ordering reflects expected ability to change the synthesis, not certainty that the records exist in the named repository. Source-map route audit

1. Retrieve the 1981 article body. Use authenticated Internet Archive lending, institutional ScienceDirect, or a library scan of Technology in Society 3(1–2):107–111. It could confirm, narrow, or reverse the 1979 capital-gains, employee-option, and technology-policy argument. DOI and route note
2. Recover the May 11, 1977 underlying testimony. Seek the committee hearing transcript, Dennis statement, exhibits, marked bill text, and panel correspondence. The staff summary cannot reveal his reasoning or coalition authorship. JCT/CRS route and note
3. Locate the Stanford MBA paper. Search 1950–1952 GSB course archives, registrar or student-paper holdings, yearbooks, Dennis family files, and likely period terms such as “new enterprise,” “risk capital,” and “small business.” Identify the coauthor before assigning any idea. Discovery note
4. Borrow Done Deals through a lawful route. Reconcile the full chapter against Gupta's 2010 article and Dennis's 2009 oral history; obtain interview date, recording, questions, and edits if preserved. Controlled-lending note
5. Retrieve the Dennis Applied Medical declaration and exhibits. Request San Mateo County Superior Court case CIV519758, the summary-judgment filing, IVM IV partnership agreement, and any attached allocation documents. Preserve the difference between stock ownership and sale-proceeds sharing and between procedural result and merits. Declaration route note
6. Complete the blocked interview set. Use lawful subscriptions or library databases for the WSJ interview and Private Funds CFO feature; locate the complete Autumn 2011 Gentry Wealth issue through the publisher, reporter, libraries, or Dennis papers. Gentry route note
7. Recover the Stanford ETL handout. Ask Stanford eCorner/ETL and Internet Archive for the original PDF payload, then determine authorship before treating it as writing; process the spoken event separately in TALKS. Handout route note
8. Search for the absent operating record. Original investment memoranda, partnership letters, board memoranda, annual LP communications, portfolio reviews, pass logs, and exit memoranda would test whether the public principles affected real decisions. No public corpus of these records has surfaced. Source map, direct-voice and archive assessment

Researcher inference · moderate: until those records are recovered, the sharpest defensible conclusion is limited: Dennis articulated how policy, capital, people, ownership incentives, relationships, valuation, and staged risk interact, but the written corpus cannot establish a complete investing process or its returns. 1979 statement, printed pp. 1058–1066 · Almanac, body ¶¶11–19 · Institutional Investor, body ¶4 · source-map denominator audit

Coverage and evidence base

Literal TALKS reconciliation

TALKS began with 17 priority source-map rows representing 14 independent evidence families: the Berkeley catalog/transcript pair, Dennis–Johnson joint oral history, two duplicate Stanford Something Ventured transcripts, CHM event/video pair, and ten colleague or counterevidence oral histories. Every priority family was acquired and reviewed in full. The two Stanford transcript objects are alternate renderings of one 2008-09-25 interview, not two testimonies; the CHM page and video are one 2005 appearance; and catalog metadata is not independent corroboration of its attached body.

Saturation work expanded the deliverable to 26 exact-schema notes: four core direct-voice families, ten original priority witness families, and twelve supplementary appearance or control families. Access reconciles exactly:

At the TALKS cutoff, the source map had 106 rows, 106 unique retrieval URLs, and 91 underlying families, with a 31 ★★★ / 48 ★★ / 27 ★ priority split. Of those rows, 63 had been reviewed to their stated boundary and 43 remained queued; later DEALS additions are intentionally outside this historical phase count. An [x] means reviewed to the stated boundary, not that a missing speech, restricted film, or gated recording was silently treated as available.

Evidence labels carry two separate judgments. The first is attribution—investor-stated, contemporaneous record, documented behavior, retrospective witness, or researcher inference. The second is support—strong, moderate, provisional, or insufficient public record. Repeated Dennis recollections show consistency, not independent verification; repeated praise from members of the same professional network may show a field norm or shared retrospective narrative rather than a distinctive edge. Source-map methodology and boundaries

Dated appearance and appearance-route inventory

Priority witness and supplementary attribution controls

Each row carries both axes locally; grades describe what the cited witness remembers, not independent verification of complete legal attribution or economics.

What the direct spoken record actually establishes

Dennis's origin story is not merit-only. Navy electronics work, Stanford engineering and business training, Fireman's Fund autonomy, a dense San Francisco referral network, and inherited assets he could risk in Ampex all preceded the repeatable process he later described. Retrospective witness · moderate for this combined opportunity set; researcher inference · strong that the first Ampex purchase was family/personal capital rather than fund capital. Berkeley, printed pp. 1–15/PDF pp. 7–21

His distinctive claimed advantage was translation: ask technical employees questions, connect a device to a customer need, and combine that understanding with financial and management judgment. The source record supports that he used this language repeatedly and that referrals followed his technical fluency. It does not show a blinded comparison, false-positive rate, or causal return advantage. Investor-stated · moderate. Berkeley, printed pp. 13 and 49–56/PDF pp. 19 and 55–62 · CHM, 00:35:07–00:39:15

Dennis's people-centered philosophy does not mean passive founder loyalty. He says financed plans often become obsolete by the first board meeting, boards learn how teams actually work, and long support can culminate in replacing a CEO. Recognition Equipment supplies the long apprenticeship; Collagen, NBI, and an unnamed recurring year-end rescue case supply adaptation and failure evidence. Investor-stated · moderate for the operating ideal; retrospective witness · moderate for the cases. Dennis–Johnson, printed pp. 5–7/PDF pp. 6–8 · raw interview, transcript 2 pp. 7–10 and 27–28

The spoken record also contains meaningful negative evidence. Dennis calls Ampex fortunate, admits he did not clearly recognize the dot-com bubble, held some public stocks to zero, invested too early in Reference Technology, passed on Apple because its price/revenue relationship exceeded his experience, and retired when younger partners understood newer markets better. Retrospective witness · strong that he made these admissions; researcher inference · moderate that his strongest reusable lesson is competence-boundary awareness rather than infallible intuition. raw interview, transcript 2 pp. 17–18 and 39–42

No spoken source establishes a complete personal portfolio, stable ownership denominator, fund-by-fund cash-flow series, repeatable scoring rubric, reserve model, or audited hit rate. The record is richest on narrative mechanism and role memory, weaker on ex-ante documents, and weakest on person-level economics. Researcher inference · strong. Source-map candidate denominators and attribution boundary

Intellectual chronology and changes over time

1952–1965: personal access becomes a relationship syndicate

The remembered sequence is personal Ampex capital, employer-supported field research, Fireman's Fund corporate investing, and then The Group's informal personal syndication. The Group was not a fund: one member screened, five members deliberated, individual checks were pooled, recurring followers completed the round, and one member took board responsibility. Retrospective witness · moderate. The model depended on wealth, other full-time jobs, trusted referrals, and geographic proximity as much as judgment. Berkeley, printed pp. 13–28/PDF pp. 19–34

The later people-centered doctrine grows out of this small-network setting, but so does its access problem. A fast lunch process among known financiers could lower coordination cost while restricting who reached the table. The corpus supplies no applicant funnel, demographic denominator, or rejected-company sample. Researcher inference · moderate. Berkeley oral-history note · source-map historical context

1960–1974: corporate apprenticeship becomes an institutional partnership

Recognition Equipment taught Dennis that a large order did not guarantee durable profitability, partial sale could return cost, and long board service could end in management change. These were Fireman's Fund acts and capital. IVA later used Dennis's insurer credibility and an American Express conditional anchor to attract other institutions; McMurtry, Jamieson, Norris, LPs, and outside counsel all belong in the formation account. Retrospective witness · moderate. Berkeley, printed pp. 16–22 and 30–40 · McMurtry, printed pp. 41–42 · Jamieson, printed pp. 37–41

1977–1981: company building becomes policy advocacy

Dennis later described capital-gains and pension-regulation work as necessary to enlarge the capital base. Independent witnesses support substantial work but allocate origins and tasks across Heizer, Hanschen, Bancroft, Morgenthaler, Golder, Lea, Pratt, association staff, legislators, regulators, and Dennis. The official former-chair table places Morgenthaler in 1977–1978, Dennis in 1978–1979, and Golder in 1979–1980. Retrospective witness · moderate for Dennis's major coalition role; contemporaneous record · strong for association-chair chronology; researcher inference · strong against sole credit. Morgenthaler, printed pp. 101–106/PDF pp. 105–110 · Pratt, printed pp. 19–23/PDF pp. 24–28 · NVCA former-chair table

The spoken chronology repeatedly compresses a 1978 tax change, a 1979 Labor rule, and a second 1981 tax statute. Researcher inference · strong: official law controls dates; oral history controls Dennis's remembered causal model. 1978 Revenue Act · 1979 final investment-duties rule · 1981 Economic Recovery Tax Act

1980s: stated workload limits meet an exceptional opportunity wave

Dennis says the 1980 vehicle told LPs it expected only one or two raw startups because company creation consumed partner time, then made six, six or seven, or seven or eight as opportunities appeared. The varying denominator prevents an exact success rate, but the invariant ex-ante point is useful: the partnership recognized a workload constraint and then made a deliberate exception. Retrospective witness · moderate for the premise; retrospective witness · provisional for count and results. CHM, 00:50:54–00:55:18 · Berkeley, printed pp. 46–48 · raw interview, transcript 2 pp. 18–20

The cases do not reduce to one formula: Seagate relied on a proven entrepreneurial team; LSI involved a scarce allocation and disputed internal timing; Stratus required bridge-like urgency before a fund close; Exabyte repurposed consumer-video technology for data storage. Researcher inference · moderate: Dennis's practice was case-based judgment bounded by people, time, technical translation, and capital availability—not a fixed scorecard. CHM, 00:51:39–00:55:18

1999–2009: succession, specialization, and capital-cycle skepticism

By 1999 Dennis publicly said deeper technical knowledge was necessary because venture had become more competitive; named partners moved to Redpoint while old IVP/Brentwood partnerships continued portfolio obligations. Investor-stated · strong for Dennis's diagnosis and contemporaneous record · strong for named formation-time roles; researcher inference · strong that this is organizational discontinuity, not a timeless founder-centered IVP. Almanac, paragraphs 3–12

In 2005 and 2008 he warned that excess fund capital raised prices, accelerated decisions, enlarged rounds, and pushed investors toward financial engineering or sales when IPO liquidity weakened. In the private 2008/09 cycle panel, Lynn recalls three senior investors contesting claims that venture capital was dying by pointing to historical recurrence. Investor-stated · moderate for Dennis's dated market diagnosis; retrospective witness · moderate for Lynn's collective panel memory. CHM, 01:17:33–01:25:56 · Dennis–Johnson, printed pp. 9–12 · Lynn, PDF p. 4

2012: patient capital reaches a legal and liquidity boundary

Applied Medical is the sharpest late counterexample to an unlimited build-to-last story. Dennis reportedly still admired the company and wanted personally to retain shares, but IVP IV exercised demand-registration rights because a ten-year fund had held the company for twenty-four years and some LPs needed liquidity; Applied's board and executives publicly opposed the IPO. Investor-stated · moderate for Dennis's missing-video rationale; documented behavior · strong for the filing-described demand right; researcher inference · strong that company duration, fund duration, and personal conviction are separate variables. OCBJ, paragraphs 3–9 · Fortune video transcription · Applied Medical S-1/A, “Reasons for this Offering”

Recurring frameworks, stories, and their limits

People judgment is central—and least codified

Dennis repeatedly says his largest mistakes involved people and that accumulated experience provides the comparison base for assessing them. McMurtry, Jamieson, and Bancroft remember him as optimistic or people-oriented; Marquardt regards Dennis's style as intuitive and chooses McMurtry because McMurtry's method looked more learnable. Investor-stated · moderate for Dennis's preference; retrospective witness · moderate for the reputation; researcher inference · strong that no transferable Dennis rubric survives. Dennis–Johnson, printed pp. 5–7 · McMurtry note · Jamieson note · Bancroft note · Marquardt, printed pp. 20–21

The raw interview also records Dennis adopting an acquaintance's rule against substantially overweight CEOs. Investor-stated · strong that he recounted the proxy; researcher inference · strong that body size is a stigmatizing, unsupported substitute for diligence. This counterexample is crucial: accumulated pattern recognition can encode prejudice as easily as expertise. raw interview, transcript 2 p. 28

Plans are hypotheses; boards update them

The first board meeting may reveal that the financed plan is already obsolete. Dennis's preferred response is sustained attention, team observation, advice, and—when necessary—leadership change or an end to financing. Investor-stated · moderate. Recognition Equipment demonstrates long involvement and CEO replacement; Collagen demonstrates a failed initial use and costly pivot; Reference Technology demonstrates being right about direction but wrong about timing. Berkeley, printed pp. 16–21 and 49–55 · raw interview, transcript 2 pp. 7–10

Technical truth, market timing, and entry economics remain separate

Engineering fluency helps interrogate a product, but Reference Technology was too early, Collagen's original clinical use failed, and Apple was passed because the remembered price/revenue relationship exceeded Dennis's experience. Exabyte worked through application change. Researcher inference · moderate: ask separately whether the technology works, a customer problem exists now, the team can translate one into the other, and the security's price/terms permit a return. The cases support the questions, not universal answers. Berkeley, printed pp. 50–56 · raw interview, transcript 2 pp. 39–40

A high batting average is a preference, not performance evidence

Dennis consistently recalls roughly twenty-five Group investments and approximately eighteen large successes and rejects a portfolio philosophy satisfied with one or two winners in ten. The repetition is almost entirely Dennis-derived and lacks names, ownership, follow-ons, cash flows, time weighting, or a stable definition of success. Investor-stated · moderate for the preference; retrospective witness · provisional for the denominator; researcher inference · insufficient public record for superior returns or a non-power-law distribution. Berkeley, printed pp. 27–28 · CHM, 00:37:41–00:39:15

Proximity was a mechanism, not an invariant law

Lea remembers Dennis describing a roughly twenty-minute investing radius, consistent with The Group's other jobs and the value of frequent access. Yet the 1983 interview reports Dennis's two latest deals in Portland and Boulder, and Exabyte later took the partnership to Boulder. Retrospective witness · moderate for the remembered heuristic; investor-stated · moderate for the 1983 exceptions; researcher inference · moderate that proximity was period- and case-dependent rather than categorical. Lea, printed pp. 137–139 · CSM, paragraphs 7–9 · Berkeley, Exabyte at printed pp. 53–56

Build to last is an ideal constrained by vehicle obligations

Dennis contrasts company building with transactions, praises Arthur Rock's long Intel engagement, and treats IPOs as financing and liquidity events rather than company endpoints. Pratt independently remembers a similar contrast. But Applied Medical shows LP liquidity and contractual rights overriding management's desired timing. Investor-stated · moderate for the ideal; retrospective witness · provisional for Pratt's characterization; documented behavior · strong for the counterexample. Dennis–Johnson, printed pp. 9–12 · Pratt, printed p. 43 · Applied Medical S-1/A, “Reasons for this Offering”

Decision evidence and safe attribution

The governing rule is simple: personal capital, spouse capital, employer capital, informal syndicate capital, IVA capital, each IVP fund, and shared GP control are different. Sourcing, approval, board work, follow-ons, exit, and economics may belong to different people. Researcher inference · strong. Source-map vehicle and attribution boundary

Contradictions, counterevidence, and unresolved claims

The preservation and resolution rules below are researcher inference · strong; the underlying factual components retain the source-specific grades and claim-local links in each row. Source-map chronology and attribution controls

Reconstructed operating playbook

This is an analyst reconstruction, not a Dennis-authored checklist. The global grade is researcher inference · moderate; each step must be tested against contemporaneous deal records before use as historical fact. Source-map attribution and denominator controls

1. Name the capital before evaluating the story. Identify personal, employer, syndicate, partnership, fund, and co-investor capital; identify who can approve, who owns the board role, and whose liquidity clock applies. Berkeley note · Bossen note · Applied note
2. Use referrals as leads, not truth. Dennis's network helped surface technical founders quickly, but familiarity can narrow access and amplify shared blind spots. Record the full funnel and reasons for rejection. raw-interview note · Lea note
3. Translate technology into a dated customer problem. Test technical feasibility, current use, customer budget, distribution, and whether timing—not merely direction—is right. Collagen, Reference Technology, and Exabyte show why these are separate gates. Berkeley note · raw-interview note
4. Evaluate people through job-relevant evidence. Direct work, references, conflict behavior, learning speed, and operating history are defensible inputs. Appearance, body size, social familiarity, and confidence are not. Dennis–Johnson note · raw-interview counterexample
5. Match company needs to vehicle capacity. Check initial capital, reserves, partner time, ownership, duration, and exit flexibility. The 1980 startup exception and Applied Medical show opposite ends of this constraint. CHM note · Applied note
6. Make dissent explicit. Record the objection, evidence that could resolve it, decision owner, timing pressure, and follow-up review. Do not use LSI's good outcome to validate every override. raw-interview LSI account
7. Treat the financed plan as a hypothesis. At each board cycle, compare customer, product, team, cash, and market evidence with the underwriting case; decide whether to support, pivot, recruit, replace, stop, or sell. Dennis–Johnson note · Berkeley note
8. Separate company success from investment return. Track security, ownership, dilution, follow-ons, distributions, time, fees, carry, and opportunity cost. A high company batting average is not itself a high return. CHM denominator note · source-map candidate denominators
9. Plan liquidity before it becomes coercive. Model fund term, extensions, LP needs, secondary options, IPO rights, management preference, and in-kind distribution. “Patient” must have a legal and economic definition. Applied note · Applied Medical S-1/A
10. Keep a loss and pass ledger. Reference Technology, public stocks held to zero, Apple, Collagen's failed first use, and prolonged disk-drive exposure are more useful for calibration than selected winners alone. raw-interview note · CHM note

Five records to read or watch first

The ranking criterion is decision density, directness, auditability, and counterevidence—not fame or positivity. The ranking itself is researcher inference · moderate. Corpus reconciliation

1. CHM: “Innovation Is Nothing New”. The only complete public audiovisual appearance in the corpus; full history, cases, failures, Q&A, and exact timestamps permit speaker and tone checks.
2. Berkeley oral history. The broadest direct record across biography, capital sources, vehicles, policy, cases, boards, and exits; pair it with the note because publication rights constrain quotation and figures remain retrospective.
3. Raw Something Ventured interview. Less polished than the documentary and unusually candid about LSI, Apple, public-stock losses, retirement, and the body-size proxy; count the two transcripts once.
4. Dennis–Johnson Venture Voices. A compact direct comparator recorded on 2008-09-22, preserving speaker-specific views on people, markets, boards, capital scale, and building to last.
5. David Marquardt oral history. The best skeptical internal witness on Dennis's intuitive style, IVA communication, the split, and Seagate sourcing; it prevents the direct corpus from becoming self-validating.

For policy attribution, read Morgenthaler and Pratt immediately after those five. For fund-life and founder/LP conflict, add the Applied Medical route.

Unresolved spoken record and retrieval agenda

The priority order is researcher inference · moderate, based on how much each missing record could change attribution, chronology, decision reconstruction, or economics. The access states and attempted routes are documented in the source-map coverage assessment.

1. Berkeley audiovisual originals: inspect the five cataloged videodiscs, session logs, unedited audio, release forms, and correction correspondence for the 2009 oral history. Berkeley note and catalog boundary
2. Stanford raw interview media: obtain the companion moving-image objects for all 66 clips and reconcile names, amounts, requested cuts, and transcript variants against audio. Stanford raw-interview note
3. Early Stanford appearances: locate the spring 1970 panel program/recording and the 1979 small-business-class schedule, handout, recording, and student notes; keep Johnson/Draper and Marquardt attribution separate. Byers note · Marquardt note
4. 2005 ETL: recover the dead handout and biography from Stanford storage or a valid web archive and search event audio/video, course records, dinner notes, and attendee archives. Stanford ETL route note
5. Public event originals: request the 2007 IBF acceptance speech, 2009 GTS panel, 2010 SFAero program, and records of the unnamed 1992 Westin venture forum; the Computer Museum memo establishes only planned attendance and does not establish Museum sponsorship. IBF note · GTS note · SFAero note · 1992 board memo
6. Private cycle panel: search Morgenthaler/Canvas calendars, invitations, internal notes, and recordings to date the 2008/09 Kramlich–Dennis–Morgenthaler appearance and allocate its examples. Lynn cycle-panel note
7. Applied Medical video: recover the WR Hambrecht file and surrounding page, then reconcile the complete remarks with IVP IV documents, Applied filings, board records, and final transaction outcome. Applied video-route note · filing reconstruction
8. Decision artifacts: obtain The Group ledger, Fireman's Fund committee minutes, IVA/IVP investment memoranda, board minutes, partner votes, and capitalization records for Recognition, Measurex, ROLM, Seagate, LSI, Exabyte, Collagen, Reference Technology, Apple, and Applied. Ranked retrieval ledger
9. Performance denominator: acquire fund-by-fund LP statements and cash flows, including IVP I–II and Dennis-era vehicles, before reusing any claimed hit rate, multiple, market share, or “three times the fund” story. Comparable LP observations and limits
10. Saturation refresh: repeat exact-title, archive-catalog, event-host, transcript, and video searches when collections change. The present three broad passes plus final exact-title/video check found no additional Dennis speech body beyond duplicates and bounded routes; that is not proof that no other recording exists. Source-map attempted-route ledger

1. Underwrite the chain from technology to security economics

• Rule: Do not invest in “a technology.” Test a five-link chain: technical efficacy → present customer problem and willingness to pay → differentiated company → market timing and adoption path → entry security economics. A strong link never cures a broken one.
• Evidence status: [investor-stated · strong] for Dennis's January 2000 separation of Internet adoption from public-company value and his refusal to time the correction; [retrospective witness · moderate] for Exabyte, Reference, Collagen, B2B-customer, and Apple memories; [researcher inference · moderate] for integrating them into one chain. Almanac, body ¶¶4 and 11–19 · Dennis oral history, printed pp. 50–56
• Provenance and origin: Dennis's engineering degree and securities-analysis apprenticeship preceded Ampex; he said technical founders were routed to him because he could interpret the “black box.” The clearest ex-ante written formulation came much later, when he expected continuing Internet utility but called prevailing prices irrational and some entrants late or unnecessary. Dennis oral history, printed pp. 13–18 · Almanac, ¶¶11–19
• Mechanism and reasoning: Technical literacy can expose what must be true; customer evidence tests whether anyone needs it now; competitive analysis tests whether a company captures the value; timing tests whether the form factor, infrastructure, regulation, and budget exist; price, preferences, ownership, dilution, follow-ons, and exit time determine whether company success becomes investor return. Reference failed at timing/form factor, while the Apple pass shows that price discipline can also reject an extraordinary company. Reference and Apple postmortems
• Historical and vehicle context: The model is most grounded in Dennis's electronics, computing, storage, process-control, and biomedical record from the 1950s–1990s. Fireman's corporate balance sheet, IVA/IVP partnership capital, and personal holdings faced different approval and liquidity conditions; Dennis's public-market warning cannot be silently converted into an IVP private-company trade. Sector and vehicle boundary · public-position admission, raw transcript 2 pp. 17–18
• When to use: Use at first screen, before price approval, at every follow-on, when a platform pivots uses, when a category is fashionable, and whenever an analyst is tempted to use technical inevitability as a complete thesis.
• How to apply it: Write one falsifiable test per link. Record independent efficacy evidence; a named customer workflow, budget and switching behavior; alternatives and the reason this company remains necessary; adoption prerequisites and time-to-proof; and security-level ownership/return under slower growth, down round, extra follow-on, and delayed exit. Give each broken link a kill, re-price, milestone, or pass action.
• Examples:
  • Exabyte — retrospective account; supporting case, [retrospective witness · moderate]: Dennis sourced a team redirecting video-hardware knowledge to data storage; all partners flew to Boulder and approved. The application-change logic is visible, but initial customer evidence, price, ownership, and return are not. Dennis oral history, printed pp. 53–56/PDF pp. 59–62
  • Reference Technology — retrospective account; negative case, [retrospective witness · moderate]: Dennis remembered good people and technology but a large 14-inch optical-storage form factor about five years too early. Technical possibility did not create a current market. Dennis, printed pp. 52–53/PDF pp. 58–59
  • Collagen — retrospective account; exception/pivot case, [retrospective witness · moderate]: the emotionally compelling burn-dressing use failed; a cosmetic use emerged after investors repeatedly financed the company. A platform can survive a broken first link only when new evidence supports a different chain. Raw transcript 2, pp. 8–10
  • Apple and the Internet warning — researcher-applied illustration; opposite-direction control, [researcher inference · moderate]: the Apple pass and later public-loss admission are retrospective accounts, while the January 2000 adoption/value distinction is a contemporaneous articulation rather than evidence of a portfolio application. Comparing them is the analyst's control: correct concern is not correct action. Raw transcript 2, pp. 17–18 and 39–40 · Almanac, ¶¶11–19
• Evidence: Useful citation placements span Dennis's Berkeley oral history, a separate raw documentary interview, and the contemporaneous January 2000 article. The model has more than three cases and an exception, but only Exabyte offers a reasonably direct positive Dennis-source/partnership-decision example; no complete memo or cash-flow record proves efficacy. Decision labs 7, 8, 10 and 11
• Limits, failure modes, and exception ledger: Analysts can overfit a neat chain to retrospective stories, demand present revenue from markets that require creation, or use “valuation discipline” to rationalize every missed winner. B2B preference is not universal; Google-like consumer exceptions were conceded. No term sheet verifies Apple, no clinical file fixes Collagen's decision sequence, and no Exabyte/Reference memo exposes the actual tests. Treat each case as bounded evidence, not a scoring formula. Raw transcript 2, pp. 37–40
• Evolution and contradictions: The underlying skill appears at Ampex, becomes a repeated storage/computing screen, and is articulated at security level in 2000. Contradictions run both ways: Reference was too early, Apple looked too expensive, Collagen's first use was wrong, and public holdings went to zero despite the valuation warning. The record shows humility, not a stable predictive batting record.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: separate technical efficacy, paid customer need, differentiation, timing, and security economics. What fails / changed assumptions: frontier or market-creating products may lack paid history, and current AI infrastructure/platform dependencies differ from Dennis-era hardware adoption. Observable signals: retained paid workflow, independent customer value, durable distribution, infrastructure cost, ownership/dilution, and delayed-exit return. Likely misuse: treating model benchmarks or NVCA's reported AI/mega-deal concentration as company proof or a bubble clock. [researcher-applied illustration]: compare an AI application with retained paid workflow against one whose use disappears when a platform bundles the feature, then stress both securities under slower adoption. NVCA 2026 Yearbook · NVCA Q2 2026 Venture Monitor

2. Test role-relevant competence; reject proxies and hand off beyond competence

• Rule: Define the actual role, test observable learning and execution evidence, and update the judgment. Do not substitute familiarity, pedigree, appearance, age, body size, or confidence for competence; an investor must also hand off when current markets exceed the team's knowledge.
• Evidence status: [investor-stated · moderate] for Dennis's preference for relevant operating experience, small teams, trust, and his retirement competence boundary; [investor-stated · strong] for the body-size statement as a rejected counterexample; [researcher inference · moderate] for the falsifiable, bias-audited procedure. Raw transcript 2, pp. 26–32, 37–42
• Provenance and origin: Dennis described accumulated experience as a “frame of reference,” preferred people over markets, and recommended operating-company experience to would-be venture investors. He also said people were his largest errors, conceded exceptional inexperienced founders, and retired when younger partners understood new markets better. These admissions turn a person-first maxim into a revisable competence test rather than intuition worship. CHM video, 01:06:29–01:10:44 · raw transcript 2, pp. 37–42
• Mechanism and reasoning: In uncertain companies, people must learn before the plan is known. Relevant prior work, honest explanation, reference behavior, complementary teams, and evidence of adapting after error can predict that capacity. Familiarity and visible identity may reduce perceived friction without predicting the work, so they create hidden false negatives and correlated teams. Handing off protects the institution when the evaluator's experience no longer maps to the market. Leza access account and Dennis response, ¶¶3–14
• Historical and vehicle context: Dennis formed his early frame in electronics and securities analysis, then evaluated small teams through The Group, corporate boards, and partnerships. His “experienced manufacturing founder” preference belongs to that context, not a universal age rule. Later IVP specialization and Redpoint formation show institutional competence can migrate away from a founder. Redpoint formation · Dennis on specialization, printed pp. 47–57
• When to use: Use for founder and executive assessment, investor hiring, board succession, sector expansion, deal-sponsor selection, and any pattern-recognition claim based chiefly on resemblance to past winners.
• How to apply it: State the role's next twelve-to-eighteen-month jobs; list evidence of comparable learning, execution, integrity, and team complementarity; seek disconfirming references or work samples; record what would change the score; compare referred and non-referred candidates under the same standard; and run a competence review for the evaluating partner before approval.
• Examples:
  • Seagate — retrospective account; supporting but team-attributed, [retrospective witness · moderate]: Marquardt emphasized the experienced disk-drive founders and persisted; Dennis eventually authorized IVP. Team experience was relevant, but source and board work were Marquardt's and economics conflict. Marquardt, printed pp. 28–30
  • LSI Logic — retrospective account; supporting/exception, [retrospective witness · moderate]: Wilf Corrigan's semiconductor experience and vision attracted a multi-firm syndicate, yet Dennis overrode a younger partner rather than resolving dissent through a legible evidence standard. A good outcome cannot validate the override. Raw transcript 2, pp. 25–29
  • Body-size heuristic — retrospective account; rejected case, [researcher inference · strong]: Dennis repeated an acquaintance's rule against substantially overweight CEOs. It is a stigmatizing, unsupported proxy and direct evidence that accumulated “frame of reference” can encode bias. Raw transcript 2, p. 28, clip 43
  • Leza and retirement — researcher-applied illustration; access/competence controls, [researcher inference · moderate]: Leza alleged a familiarity-based gate, while Dennis offered a scarcity/credential explanation; later Dennis voluntarily ceded decisions to partners with more current knowledge. One warns about external false negatives, the other about internal evaluator decay. SFGate, ¶¶1–14 · Raw transcript 2, pp. 41–42
• Evidence: Direct Dennis testimony supplies the positive preference, explicit error admission, invalid proxy, exceptions, and handoff. Marquardt and Leza provide independent skeptical perspectives; issuer/firm records show later roles belonged to other partners. No complete hiring or founder-evaluation denominator tests predictive accuracy. Netflix prospectus, principal stockholders
• Limits, failure modes, and exception ledger: Job evidence can be backward-looking, references socially correlated, and a scorecard false precision. Operating experience may overfit yesterday; inexperienced founders can succeed; a sector expert may miss a cross-category innovation. The Leza meeting is disputed and has no IVP hiring ledger, so it demonstrates a credible access mechanism, not measured Dennis discrimination. Appearance-based heuristics are not “limits” to tune—they are excluded.
• Evolution and contradictions: Dennis moved from personal technical pattern recognition to larger specialist teams and finally a competence-based handoff. Yet his people-first doctrine coexists with acknowledged people mistakes, a biased heuristic, LSI's governance exception, and an access dispute. The modern rule is stricter than the historical behavior because the corpus itself supplies the counterevidence.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: define the role, test job-relevant work and learning evidence, audit false negatives, and hand off beyond evaluator competence. What fails / changed assumptions: prior titles can become obsolete quickly, references can be socially correlated, and expertise at one AI layer does not establish competence at another. Observable signals: a predefined job, work sample, disconfirming reference, comparable-candidate review, exclusion log, and named competent evaluator. Likely misuse: relabeling pedigree, affiliation, appearance, or category fluency as pattern recognition. [researcher-applied illustration]: require different evidence for a model researcher, infrastructure operator, workflow seller, and regulated-deployment leader, and record which partner can judge each. NVCA concentration is context, not a hiring base rate. NVCA Q2 2026 · NVCA 2026 Yearbook

3. Use networks for leads and syndication, then audit sponsor and funnel

• Rule: Let trusted networks route specialized leads, information, talent, and co-investors; never treat the referral as proof. Record source, sponsor, independent diligence, board owner, pass reason, and who never reached the funnel.
• Evidence status: [retrospective witness · moderate] for Dennis's early brokerage referrals, The Group mechanics, and witness-described Measurex and ROLM roles; [researcher inference · moderate] for the funnel-audit extension. Dennis oral history, printed pp. 13–28 · Inc., ¶¶4–5 · Measurex and ROLM responsibility rows
• Provenance and origin: Ampex arrived through a family-connected friend; brokers sent technical companies; The Group pooled member leads and called prior co-investors; IVA fundraising used Dennis's insurer credibility. The network was an information and financing infrastructure before it was a fund. Raw transcript 2, pp. 1–6 and 24
• Mechanism and reasoning: Repeated relationships lower search and coordination cost, make references richer, and assemble rounds larger than one investor can fund. Named sponsorship creates accountability after sourcing. The same mechanism can create homophily, overconfidence, geographic correlation, and stolen credit; independent diligence and funnel denominators separate access from selection.
• Historical and vehicle context: In the 1950s, roughly five part-time professionals could pool $80,000–$100,000 and call friends to fill a $250,000–$350,000 round. This is not a modern fund acceptance rate or proof of a legal pool. IVA later converted institutional relationships into committed partnership capital. Geographic closeness helped board work but was not invariant: Exabyte required travel to Boulder. Institutional Investor, ¶4 · Exabyte, printed pp. 53–56
• When to use: Use in thesis sourcing, founder referrals, co-investor assembly, executive recruiting, customer/reference diligence, and portfolio adjacency—especially when someone calls a network “proprietary.”
• How to apply it: Capture first source and every handoff; name the sponsor and board owner; record whether capital, diligence, or reputation came from each participant; obtain independent customer/technical evidence; define the intake unit and duplicates; track rejects and no-access populations; and compare referred versus non-referred outcomes without using confidential portfolio data.
• Examples:
  • The Group — retrospective account; supporting mechanism, [retrospective witness · provisional]: one member sourced, five screened, personal checks pooled, friends filled the round, and a member boarded. The claimed outcomes lack a ledger. Inc., ¶¶4–5 and 13–14
  • Measurex — retrospective account; attribution control, [retrospective witness · strong]: Draper sourced Bossen through Chicago, Wythes helped fundraise, Bancroft/Bessemer supplied major capital, and Reid/Peggy held separate positions. A network success cannot become a Dennis-only deal. Draper, printed pp. 49–50/PDF pp. 58–59 · Bossen, 02-00:01:26
  • ROLM — retrospective account; sponsor/conflict case, [retrospective witness · moderate]: McMurtry brought a position he already partly owned, disclosed and recused; Dennis, Jamieson, and Norris approved; McMurtry boarded. The network opened the door while procedure allocated responsibility. McMurtry, printed pp. 30–32
  • Leza / out-of-region deals — researcher-applied illustration; exceptions, [researcher inference · moderate]: relationship familiarity may close the investor-employment funnel; Boulder and other out-of-Valley cases disprove a strict local-radius law. SFGate, ¶¶3–14 · Christian Science Monitor, ¶¶7–9
• Evidence: Dennis, Inc., and Institutional Investor are overlapping retrospective narratives for The Group, not three independent confirmations. Independent support comes from Draper/Bossen and McMurtry role accounts. No Group member/deal ledger, complete IVP funnel, or referred-versus-cold outcome series exists. Source-family de-duplication
• Limits, failure modes, and exception ledger: A dense network can amplify shared error, exclude unfamiliar people, leak confidential information, or turn a prestigious referral into waived diligence. Proximity can waste coverage if the actual work is remote-compatible. Dennis's 1.5%–2.5% funding claim lacks intake definition; The Group's twenty-five-company claim lacks roster. Do not calculate acceptance or network alpha from either. CHM video, 01:15:31–01:17:05
• Evolution and contradictions: Warm access at Ampex became a local personal syndicate, then insurer-backed LP fundraising, then specialized multi-firm deal networks. Scale increased but attribution and communication weakened: Marquardt recalled poor partner communication, and later firm lineages split. The relationship mechanism persisted; its legibility did not automatically scale. Marquardt, printed pp. 23–25
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: use networks to route specialized evidence and syndicate capacity, then preserve independent diligence and a complete funnel. What fails / changed assumptions: co-investors may share the same AI thesis, confidential information cannot be repurposed, and referred/non-referred cohorts may be incomparable. Observable signals: source, sponsor, independent evidence, rejection reason, duplicate policy, stage/price, and no-access denominator. Likely misuse: calling prestigious or repeated inbound proprietary selection alpha. [researcher-applied illustration]: compare referred and non-referred AI-infrastructure opportunities on identical technical, customer, price, and outcome fields. Treat NVCA concentration only as a prompt to test correlation. NVCA Q2 2026

4. Map capital plus scarce complements

• Rule: Before adding capital, map the full company-building system and identify the binding scarce complement—qualified people, technical proof, customer access, distribution, regulatory permission, governance, or time. Assign an owner and measurable release condition; money alone is not the intervention.
• Evidence status: [investor-stated · strong] that capital alone was insufficient and qualified managers/key employees were scarce; [investor-stated · moderate] that entrepreneurs are the motive force; [researcher inference · provisional] for a general bottleneck map. This model is below the normal case-evidence target because no Dennis memo shows the map used before a named approval. 1979 statement, printed pp. 1060–1062
• Provenance and origin: In May 1979 Dennis argued contemporaneously that young firms needed a steady supply of managers and key employees and used employee equity to compensate for lower cash and security. In later talks he described entrepreneurs as the motive force and capital as fuel or lubricant. The generalized complement map is an analyst reconstruction, not his phrase. Prepared statement, p. 1061 · Stanford ETL talk synthesis
• Mechanism and reasoning: More money raises the chance of success only if cash can acquire or unlock the actual constraint. If the bottleneck is regulatory evidence, a scarce operator, customer trust, or board decision, a larger round may merely increase burn and price. Employee options may broaden upside and offset risk, but exercise timing, applicable tax, and statutory holding periods can materially change the employee's outcome. IRS Topic 427, reviewed 2026-04-28
• Historical and vehicle context: Dennis spoke for NVCA while identified as an IVA partner, not from a company memo. Startups then faced cash-pay and job-security disadvantages, and the policy objective was to restore qualified stock options. Recognition, LSI, Collagen, and Exabyte illustrate possible complement problems, but public files do not prove Dennis diagnosed them this way ex ante. Prepared statement, printed pp. 1056–1062
• When to use: Use when a company asks for a larger round, misses milestones despite adequate cash, cites hiring as the reason for financing, relies on a regulatory or distribution gate, or claims investor “value add.”
• How to apply it: Draw the causal system; rank constraints; name the evidence, accountable operator, investor capacity, authority, cost, and date for the top constraint; compare capital-only and noncapital interventions; model employee-equity value under failure/flat/down-round/secondary/exit scenarios; and decline to promise help that lacks a named capable owner.
• Examples:
  • 1979 employee-equity proposal — researcher-applied illustration; direct but non-deal case, [researcher inference · moderate]: Dennis identified manager/key-employee recruitment as a complement and options as risk compensation. The attached tax example was stylized NVCA advocacy, not a company outcome study. Prepared statement and Exhibit II, pp. 1061–1066
  • Recognition — retrospective account; plausible operating complement, [retrospective witness · moderate]: Dennis remembered long board work and eventual CEO replacement after management weakness. This shows that capital did not eliminate leadership risk, but missing board books prevent an ex-ante bottleneck attribution. Dennis, printed pp. 16–21
  • Collagen — retrospective account; exception, [retrospective witness · moderate]: capital had to support a long clinical/regulatory path and a pivot after the first use failed. Repeated funding may have bought learning or escalated commitment; financing and clinical minutes are absent. Raw transcript 2, pp. 8–10
  • Exabyte / LSI — researcher-applied illustration; candidate complements, [researcher inference · provisional]: application insight, specialized technical teams, syndicate capacity, and speed appear, but no memo identifies the binding constraint or tests whether investor help released it. Exabyte, Dennis pp. 53–56 · LSI, raw transcript 2 pp. 25–29
• Evidence: The direct 1979 authored source and later direct talks support “capital plus people,” while Recognition/Collagen/Exabyte/LSI provide bounded illustrations from different source families. They do not meet the normal standard of three grounded decisions showing Dennis used a complement map ex ante and achieved a linked result.
• Limits, failure modes, and exception ledger: Model-specific attempted-route ledger: the 1981 Dennis article body was sought through DOI/ScienceDirect and an access-restricted Internet Archive volume but remained blocked; the MBA paper was sought through Stanford catalogs, web/archive, exact-phrase, obituary, and degree routes but title/coauthor/body remain absent; the 1977 testimony was checked in the JCT summary, Congressional Record notice, GovInfo, Congress, and hearing routes but the full Dennis text was not recovered. Fireman's/OAC was searched for Recognition files but resolved to an unrelated ten-item maritime collection; SEC/issuer filings, CHM/Berkeley/Stanford oral-history archives, company searches, and note saturation did not produce Recognition, LSI, Collagen, and Exabyte investment memoranda, hiring plans, board records, and employee-option evidence, or original IVP portfolio reviews and LP communications. Collagen clinical/financing minutes and Exabyte customer diligence are specific subsets of that missing record. Blocked written-record routes · 1977 and 1979 source analysis · Fireman's dead end
• Evolution and contradictions: Dennis welcomed new capital in 1979 because it could widen company formation, yet later warned that excess capital raises price and weakens discipline. Employee options can recruit scarce people, but exercise may create tax consequences before sale and failure to meet statutory holding periods can alter tax treatment. The model therefore asks what capital unlocks, not whether capital is generally scarce or abundant. IRS Topic 427, reviewed 2026-04-28
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: identify the binding noncapital constraint, a capable owner, authority, capacity, and a release condition before enlarging a round. What fails / changed assumptions: the inferred bottleneck may be wrong or uncontrollable, and extra capital can raise burn without buying evidence; employee option outcomes also depend on exercise timing, AMT exposure, and statutory holding-period treatment. Observable signals: named constraint/owner, decision authority, capacity budget, milestone, release condition, and employee-option outcomes after applicable tax. Likely misuse: quoting headline grant value as cash-equivalent after-tax value or prescribing more money for integration, approval, distribution, or talent the investor cannot unlock. [researcher-applied illustration]: before financing more compute, test whether customer approval or integration is actually binding and fund only the authorized release plan. NVCA Q2 2026 · IRS Topic 427, reviewed 2026-04-28

5. Match vehicle capacity, partner time, duration, and liquidity to the opportunity

• Rule: Underwrite the investor as rigorously as the company: legal capital, mandate, decision rights, reserves, partner skill/time, holding duration, liquidity obligations, and succession must fit the work from first check through realization.
• Evidence status: [retrospective witness · moderate] for Dennis's successive capital forms; [contemporaneous record · strong] for Applied's financing, registration-right, and withdrawal sequence; [researcher inference · moderate] for vehicle fit as a unified rule; [researcher inference · insufficient public record] for exact fees, carry, reserves, and most partnership terms. Vehicle chronology · Applied case record
• Provenance and origin: Dennis moved from trust/personal capital to The Group, Fireman's/American Express, IVA, and successive IVP funds. He described a personal/corporate check gap, used insurer credibility to raise IVA, and later distinguished his personal wish to hold Applied from a fund's LP-liquidity duty. Dennis oral history, printed pp. 21–48 · Applied account and filing
• Mechanism and reasoning: Vehicle determines whose capital bears risk, who may approve and govern, check/follow-on capacity, incentive and conflict structure, time available for learning, and when cash must return. Partner board load and current domain knowledge determine whether promised help exists. A company can be excellent while its security, fund, or manager is a poor fit.
• Historical and vehicle context: Personal capital allowed small discretionary bets; The Group enlarged checks without a permanent pool; corporate capital enabled Recognition-scale investing but belonged to the employer; IVA introduced institutional LPs and dedicated partners; IVP added fund vintages, shared control, and succession. These were different systems, not one continuous Dennis portfolio. Vehicle/economics boundaries
• When to use: Use before every fund commitment and company investment, before increasing fund size, when adding board seats or sectors, at each follow-on, and whenever company duration may exceed the fund or key person's authority.
• How to apply it: Name the exact legal owner/manager and mandate; map decision/veto/conflict rules; model initial and reserve checks, ownership/dilution, partner hours, board load, skill coverage, key-person departure, delayed exit, and LP liquidity; list primary, acquisition, secondary, tender, distribution, and continuation alternatives; then state what mismatch causes a pass or different vehicle.
• Examples:
  • Ampex / Fireman's boundary — retrospective account; supporting origin, [retrospective witness · moderate]: Dennis's personal/trust check required trustee approval, while larger corporate opportunities used Fireman's authority. Size and ownership changed the feasible decision. Dennis, printed pp. 13–22
  • IVA formation — retrospective account; capacity case, [retrospective witness · moderate]: a conditional $5 million American Express anchor and institutional LPs helped close a reported $19 million pool and support a dedicated team; terms and return remain unaudited. Dennis, printed pp. 35–48
  • First IVP fund — retrospective account; workload exception, [retrospective witness · provisional]: Dennis remembered promising only one or two startups and making six to eight after an opportunity wave. The varying count and missing capacity record prevent a success claim, but show strategy can outrun planned partner time. Raw transcript 2, pp. 18–20
  • Applied Medical — researcher-applied illustration; duration/liquidity exception, [researcher inference · moderate]: filings document that IVP IV invested in 1988, followed through 1990, skipped the last 1992 round, later invoked registration rights, and withdrew without a sale; they do not show that Dennis or IVP applied this synthesized vehicle-fit rule. Dennis's separate personal-versus-LP account makes the sequence a bounded analyst test. Applied S-1/A · withdrawal
• Evidence: The Berkeley and raw transcripts provide direct vehicle testimony; filings independently document Applied's financing/right/withdrawal sequence; UC supplies net-to-one-LP vintage observations. No partnership agreement or full manager economics verifies the claimed causal advantage. UC IVP III–VIII
• Limits, failure modes, and exception ledger: A long-duration label does not prove patience; stable family/corporate capital can conceal concentration, weak incentives, or poor stops. A finite fund can still support a long company through planned liquidity alternatives. Dennis's fee/carry, co-investment, reserve ratios, side letters, key-person rules, and IVP I–II LP results remain missing. Do not infer fund terms from witness praise or later IVP vehicles.
• Evolution and contradictions: Each new vehicle solved a constraint—small checks, corporate ownership, dedicated capital, partner specialization—while adding another: trustee/employer authority, LP liquidity, partner economics, communication, or succession. The 1980 IVA split and 1999 Redpoint formation show organizational discontinuity; Dennis's retirement shows a responsible competence handoff, not continuous personal authority. Marquardt, printed pp. 23–25 · Redpoint formation
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: match the actual mandate, decision rights, reserves, partner capacity, duration, and liquidity to the company. What fails / changed assumptions: historical vehicle labels do not establish current legal authority, side-letter obligations, key-person continuity, or the company's horizon. Observable signals: governing documents, portfolio role, alternatives, diversification, liquidity/current return, projected return, funding objectives, board load, reserve model, and delayed-exit cases. Likely misuse: treating “patient capital” or the historical “ERISA opened VC” story as permission to skip current prudence. [researcher-applied illustration]: stress a delayed-exit company against the fund's actual agreement, reserves, partner load, and LP obligations, then use another vehicle or pass if the fit fails. 29 CFR §2550.404a-1(b), current through 2026-07-30

6. Make roles, conflicts, dissent, and decision rights legible

• Rule: Before outcome is known, record who sourced, sponsored, diligenced, objected, recused, approved, negotiated, boarded, followed on, and exited; identify the legal vehicle and final decision right. Learning requires attribution without hero laundering.
• Evidence status: [retrospective witness · moderate] for role separation at ROLM, Seagate, LSI, Exabyte, and Measurex; [contemporaneous record · strong] for named later issuer holdings and board/control roles; [researcher inference · strong] that historical attribution must keep roles separate; [researcher inference · moderate] for the proposed IC ledger. Material-case responsibility matrix · later issuer filing boundary
• Provenance and origin: Dennis preferred small partnerships, reasoned objections, and no automatic veto, but admitted that LSI violated his consensus ideal. ROLM supplies a disclosure/recusal pattern; the other deal records repeatedly separate sponsor, capital, board, and result. The formal ledger is a synthesis from attribution failures, not a recovered IVP form. Raw transcript 2, pp. 27–29 · ROLM, McMurtry printed pp. 30–32
• Mechanism and reasoning: Pre-discussion views reduce conformity and hindsight editing. Conflict disclosure and recusal protect the decision object. Naming the final right distinguishes persuasion from authority. Later action/outcome fields reveal whether a thesis was implemented and whether claimed help occurred. Without this chain, a firm's brand absorbs partner and founder work and lucky outcomes validate opaque process.
• Historical and vehicle context: The Group decided informally, Fireman's used corporate approval, IVA/IVP involved partners and legal entities, and issuer filings later show shared managing-director power. A “Dennis investment” can mean personal capital, a corporate security, a partnership vote, a board role, shared GP control, or no Dennis action at all. Netflix prospectus · ArcSight S-1
• When to use: Use for every IC decision, conflict, co-investment, partner override, board allocation, follow-on, management change, exit, performance attribution, and historical case study.
• How to apply it: Timestamp independent pre-meeting views; record source and sponsor; list information owners and evidence; disclose conflicts and recusal; preserve dissent and response; name the governing document and final right; log security/vehicle and board actor; then attach later operating action, ownership, cash flow, and counterfactual without rewriting the original belief.
• Examples:
  • ROLM — retrospective account; supporting conflict case, [retrospective witness · moderate]: McMurtry disclosed a prior position and recused; Dennis, Jamieson, and Norris approved; McMurtry boarded. Formal documents are missing, but the remembered roles are unusually legible. McMurtry, printed pp. 30–32
  • LSI Logic — retrospective account; governance exception, [retrospective witness · moderate]: Sequoia brought Corrigan, several firms supplied capital, a younger partner objected, and Dennis accelerated the decision. The good outcome is hindsight; the vehicle and formal vote remain unresolved. Raw transcript 2, pp. 25–29
  • Seagate / Exabyte — researcher-applied illustration; sponsor contrast, [researcher inference · strong]: Marquardt sourced and persisted on Seagate while Dennis authorized; Dennis sourced Exabyte while all partners traveled and approved. “IVP backed” obscures different work. Marquardt, printed pp. 28–30 · Dennis, printed pp. 53–56
  • Measurex / later issuer filings — researcher-applied illustration; anti-hero controls, [researcher inference · strong]: Draper/Wythes/Bancroft/Bossen/Peggy and Reid occupy separate roles; Netflix and Thermage filings identify other IVP directors. Founder status cannot assign every later firm action to Dennis. Bossen, 02-00:01:26 · Thermage Form 3
• Evidence: Independent source families include Dennis, McMurtry, Marquardt, Draper/Bossen, SEC filings, and company records. They strongly establish attribution risk and several role maps, but no complete IVP IC minute book proves a standard decision procedure or Dennis's formal authority across funds.
• Limits, failure modes, and exception ledger: A ledger can become bureaucratic theater, suppress informal contribution, or manufacture false precision. Legal authority does not equal intellectual causality; dissent can be tactical; recusal can be incomplete; a successful override can encourage outcome bias. Formal IC rules, partner economics, LSI allocation deadline, ROLM vote, Seagate board status, and most exit actors remain missing.
• Evolution and contradictions: Informal personal syndication evolved into corporate committees and LP partnerships, but communication did not automatically improve. ROLM shows disciplined recusal; LSI shows senior urgency overriding the norm; Marquardt describes Dennis's intuitive speed as hard to transfer; later filings show distributed control. Legibility is the remedy the historical organization did not consistently document publicly.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: preserve source, sponsor, evidence owners, conflicts, independent views, dissent, authority, vehicle, board actor, and later outcome separately. What fails / changed assumptions: structured records can become performative, suppress informal contribution, or expose confidential material if access is poorly controlled. Observable signals: timestamped pre-discussion views, conflict/recusal record, final-right citation, security terms, board capacity, and later operating/cash-flow entries. Likely misuse: scorecard theater, rushed consensus, or assigning every outcome to the firm. [researcher-applied illustration]: log a concentrated 2026 AI round's source, sponsor, independent votes, dissent, final security, board owner, and later cash flows as separate access-controlled fields. NVCA Q2 2026

7. Treat plans as hypotheses and boards as update mechanisms

• Rule: Finance an explicit hypothesis, not a fixed forecast. Use the board to compare plan with evidence and choose support, recruit, pivot, replace, continue, or stop; require changed evidence before new capital.
• Evidence status: [investor-stated · moderate] for plans changing quickly and boards learning team dynamics; [retrospective witness · moderate] for Recognition, Collagen, and an unnamed rescue; [researcher inference · moderate] for the full update cadence and stop logic. Raw transcript 2, pp. 7–10 and 27–28
• Provenance and origin: Dennis described board work as continuous updating after the financed plan changes. Recognition supplied a nineteen-year apprenticeship and management replacement. Collagen preserved a failed use and pivot. An unnamed rescue showed Dennis listening for months before independently agreeing that a CEO change was needed. Dennis oral history, printed pp. 16–21 and 50–55
• Mechanism and reasoning: Early plans compress uncertain technology, people, customers, and regulation. A board adds value only if it collects decision-relevant evidence, names accountable owners, changes resources or leadership when warranted, and stops when option value no longer exceeds capital/time. Repeated financing without a new hypothesis is sunk-cost escalation; changing management without diagnosis is scapegoating.
• Historical and vehicle context: Dennis's visible modes range from corporate board tenure to sponsor recusal, observer status, all-partner travel, repeated biomedical financing, and contractual liquidity action. This is not one universal “active board” style. The investor's legal right and partner capacity vary by Fireman's, IVA/IVP, and company. Governance modes
• When to use: Use at underwriting, board formation, budget approval, missed milestone, pivot, executive hire/change, rescue round, regulatory delay, and exit/liquidity conflict.
• How to apply it: Preserve the initial causal plan; define the highest-risk belief, evidence cadence, owner and threshold; keep a board decision log; on each cycle compare forecast with fact; distinguish support/recruit/pivot/replace/stop; attach capital to the new information path; and record why contrary evidence did not trigger action.
• Examples:
  • Recognition Equipment — retrospective account; supporting governance case, [retrospective witness · moderate]: Dennis remembered partial de-risking, a later price reversal, long board work, an interim committee, and CEO replacement. The action sequence is visible only retrospectively. Dennis, printed pp. 16–21
  • Collagen — retrospective account; pivot/financing case, [retrospective witness · moderate]: the burn treatment failed, cosmetic use emerged, and investors repeatedly financed the company to a remembered modest result. The case supports updating but cannot distinguish disciplined learning from escalation. Raw transcript 2, pp. 8–10
  • Unnamed management rescue — retrospective account; diagnosis case, [retrospective witness · provisional]: Dennis replaced a departing board member, observed for three or four months, and reached the same CEO conclusion. The issuer, intervention, capital, and outcome are unknown; it cannot be merged with NBI. Raw transcript 2, pp. 7 and 12–13
  • Reference / Applied — researcher-applied illustration; stop and objective exceptions, [researcher inference · moderate]: Reference's good-technology/team story lacked timely adoption, while Applied's successful company triggered a fund-liquidity action. Board updating must include market stops and stakeholder-specific objectives, not only operating support. Reference, Dennis pp. 52–53 · Applied withdrawal
• Evidence: The Berkeley oral history and raw documentary transcript are separate artifacts but share Dennis as narrator and therefore are not independent corroboration. This single-narrator limit governs Recognition, Collagen, and the unnamed-rescue account. Applied filings independently show a rights action, not that Dennis used the board-update model. The named positive cases remain retrospective and lack board materials, so causal efficacy is unproved.
• Limits, failure modes, and exception ledger: Active boards can micromanage, create founder dependency, confuse noisy data with structural failure, or use management replacement to hide an investor's bad thesis. Pivot tolerance can eliminate falsifiability; milestone finance can underfund; rescue rounds can protect sunk cost. Required missing controls: Recognition board books and committee/CEO-change record; Collagen clinical evidence, financing minutes, and pivot/stop decisions; identity and complete file for the unnamed rescue; and contemporaneous, case-specific stop criteria across all three. None surfaced in Fireman's/OAC, SEC/issuer, Berkeley/Stanford/CHM, litigation, or company searches. Ranked retrieval ledger
• Evolution and contradictions: Recognition taught long board participation; later stories frame plans as provisional and people problems as observable over time. Collagen shows a useful pivot but weak stop evidence; Reference shows patience cannot fix premature timing; Applied shows a board/fund action can serve liquidity rather than company strategy. No public postmortem shows a firm-wide checklist change.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: preserve the initial hypothesis, define evidence thresholds, and pre-commit support, recruit, pivot, replace, continue, or stop actions. What fails / changed assumptions: AI or regulated milestones can be gamed, technical progress may not yield deployment or approval, and a board may lack authority or capacity. Observable signals: independent technical evidence, deployment/retention, security/compliance, burn/runway, named owner, cadence, threshold, and financing consequence. Likely misuse: endless pivots or follow-ons that merely protect sunk cost. [researcher-applied illustration]: connect the next check to a verified deployment/retention threshold with explicit support, pivot, and stop actions rather than a generic runway extension. NVCA 2026 Yearbook

8. Build companies rather than transactions, with explicit patience boundaries

• Rule: Optimize for a durable operating company and useful work, not a financing headline or fast transaction; at entry, define whose patience is promised, the resources supporting it, the liquidity clocks, and the conditions under which sale, distribution, secondary, or stop is legitimate.
• Evidence status: [investor-stated · moderate] for Dennis's company-building/transaction distinction and build-to-last ideal; [retrospective witness · moderate] for Recognition, Seagate, Microsoft, and board examples; [contemporaneous record · strong] for Applied's registration/withdrawal boundary; [researcher inference · moderate] for conditional patience. Raw transcript 2, pp. 29–31 and 40 · Applied case record
• Provenance and origin: Dennis said venture investors influence rather than create companies, contrasted operating consequences with investment-banking transactions, admired Arthur Rock's quiet Intel committee work, and criticized later pressure to build to sell. Pratt independently remembered Dennis as non-transactional, though without a named case. Raw transcript 2, pp. 6–7 and 29–31 · Pratt, printed p. 43/PDF p. 48
• Mechanism and reasoning: Company value compounds through product, customers, people, governance, and reinvestment after a financing closes. A patient investor can avoid premature sale and support that compounding. But patience consumes reserves, partner time, and LP duration; a strategic sale may be rational, and a long hold may be loss aversion. Naming stakeholder clocks prevents “build to last” from erasing fiduciary or personal needs.
• Historical and vehicle context: Personal holdings, a corporate balance sheet, and LP funds had different clocks. Dennis remembered recovering Recognition cost through a partial sale, receiving Seagate distributions, selling a small Microsoft distribution for family medical needs, and preferring to hold Applied personally while the fund sought LP liquidity. These are not one exit doctrine. Recognition/Seagate, Dennis printed pp. 16–21 and 58–60 · Microsoft, raw transcript 2 p. 40
• When to use: Use at entry, board/financing decisions, strategic offers, IPOs, tenders, distributions, secondary requests, fund extensions, continuation structures, and any claim of patient capital.
• How to apply it: Separate company financing from investor/employee/founder liquidity; list stakeholder horizons and rights; budget reserves and board time; define operating milestones and opportunity cost; model primary, acquisition, IPO, secondary, tender, distribution, and shutdown; and explain why the chosen route serves the authorized vehicle rather than a personal preference.
• Examples:
  • Recognition — retrospective account; conditional patience/de-risking, [retrospective witness · moderate]: Fireman's reportedly sold enough after appreciation to recover cost, then remained through a reversal, management change, and eventual corporate profit. No ledger tests whether the partial sale or long hold was optimal. Dennis, printed pp. 16–21
  • Seagate — retrospective account; company versus fund event, [retrospective witness · provisional]: Dennis described an IPO purchase and distribution while Marquardt described one check and a large distribution. Both treat the public event as a fund-liquidity path, but economics conflict. Marquardt, printed pp. 28–30
  • Microsoft — retrospective account; personal exception, [retrospective witness · moderate]: Dennis sold a small distributed position for a family medical need. It is a personal liquidity decision, not evidence that the fund exited badly; the lost upside is hindsight. Raw transcript 2, p. 40
  • Applied — researcher-applied illustration; hard boundary, [researcher inference · moderate]: filings document the company's opposition to IVP's demand-registration route and a withdrawal that sold no securities; they do not document application of the synthesized patience rule. Dennis's separate personal-versus-LP account makes the conflict an analyst test of actor-specific patience. Applied S-1/A · withdrawal
• Evidence: Direct Dennis testimony supports the ideal and several personal/fund distinctions; Pratt offers independent but thin character evidence; SEC and court records control Applied's documented sequence. No complete hold/sell ledger or matched long-hold outcome set proves patience added return.
• Limits, failure modes, and exception ledger: “Company first” can excuse over-capitalization, delayed stops, weak distributions, or disregard of LP mandate. A quick acquisition can maximize company/stakeholder value; a long hold can reflect no market. IPO, filing, mark, and company sale are not automatically fund cash. Applied's later lawsuit preserved allegations and procedural outcomes, not a merits finding against Dennis. Applied appellate opinion
• Evolution and contradictions: Dennis's career moved from personally patient concentrated holdings to institutional funds with explicit liquidity pressure. His late ideal grew more critical of build-to-sell behavior, while Applied made the counterconstraint concrete. The correct evolution is from unconditional rhetoric to actor-specific patience, not a claim that Dennis always followed one holding rule.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: separate company financing, employee/founder liquidity, fund liquidity, and company continuity, with actor-specific patience limits. What fails / changed assumptions: concentrated large rounds and uneven exits can exhaust reserves or duration, while a long hold may reflect loss aversion rather than company building. Observable signals: stakeholder clocks, consent/rights, reserve and board capacity, realizable liquidity paths, and delayed-exit effects on ownership and LP cash. Likely misuse: branding every long hold as patient or every sale as betrayal. [researcher-applied illustration]: maintain distinct primary, employee-secondary, fund-secondary/distribution, and company-sale scenarios and identify whose authorized objective each serves. NVCA Q2 2026

9. Separate company viability from investment economics and audit outliers

• Rule: A viable company count and an attractive investment return are different claims. Report the complete denominator, capital weights, ownership/dilution, follow-ons, time, gross/net cash flows, DPI/TVPI/IRR, and results excluding dominant companies and vintages.
• Evidence status: [investor-stated · moderate] for Dennis's preference for a higher batting average; [retrospective witness · provisional] for Group hit counts and named return claims; [contemporaneous record · strong] for UC's net-to-one-LP IVP III–VIII rows; [researcher inference · strong] for separation, but model efficacy remains unresolved because no complete Dennis-career ledger exists. Inc., ¶¶13–14 · UC report
• Provenance and origin: Dennis reported roughly 60%–65% early successes plus another 10%–15% returning capital and rejected a two-winners-in-ten objective. Other appearances report about twenty-five Group companies, eighteen successes, two capital returns, and roughly three losses. The numbers express preference and nostalgia; definitions and arithmetic are unstable. Institutional Investor, ¶4 · CHM video, 00:35:07–00:39:15
• Mechanism and reasoning: Company-count viability measures selection/support frequency; return measures the price, ownership, dilution, reserves, magnitude, timing, fees/carry, and realization of those outcomes. One dominant winner can rescue a low hit rate, while many survivors can tie up capital at poor returns. Ex-outlier and capital-weighted views expose the difference without claiming one portfolio shape is universally best.
• Historical and vehicle context: The Group invested member capital deal by deal; Recognition belonged to Fireman's; IVA and each IVP fund had different pools and clocks. Combining their counts creates a false fund. UC observes one LP in IVP III–VIII as of a date, not Dennis's personal economics, IVA, IVP I–II, or a lifetime aggregate. Reported Group outcomes · Comparable vintages
• When to use: Use in fund underwriting, portfolio reviews, reserve allocation, performance claims, pioneer comparisons, track-record attribution, and any “hit rate,” “X-times fund,” or selected-company narrative.
• How to apply it: Freeze the company/vehicle/time denominator; define success before classification; reconcile paid-in, follow-on, ownership/dilution, realized/unrealized value and dates; calculate metrics only within commensurable cash flows; show loss/break-even/viable/outlier counts and weights; remove top one/top vintage; and attribute management-company, fund, partner, and personal economics separately.
• Examples:
  • The Group — retrospective account; count without economics, [retrospective witness · provisional]: roughly twenty-five, eighteen successes, two capital returns, and three large losses do not fully reconcile and have no company ledger or value weights. Dennis, printed pp. 27–28
  • Seagate — retrospective account; outlier with conflicting cash flows, [retrospective witness · provisional]: Marquardt's one-$500,000-to-$63-million memory conflicts with Dennis's two-$500,000 and roughly-three-times-fund account. Preserve both; calculate neither as audited. Marquardt, printed pp. 28–30 · Dennis, printed pp. 58–60
  • IVP III–VIII — researcher-applied illustration; supporting dispersion control, [researcher inference · strong]: UC reports III 2.15x/13.7%, IV 2.34x/9.9%, V 3.00x/27.5%, VI 5.82x/63.1%, VII 6.80x/93.6%, VIII 1.07x/1.5%, net to that LP as of 2020-06-30. The rows are contemporaneous records; using their dispersion as an analyst control is not evidence Dennis applied this model, and different timing prohibits a synthesized aggregate IRR. UC report, pp. 1–2
  • Collagen / Applied — researcher-applied illustration; viability exceptions, [researcher inference · moderate]: Dennis remembered Collagen as only modestly positive after a pivot; Applied was a successful-looking long-lived company whose fund liquidity route failed. Survival and operational quality do not settle fund return. Collagen, raw transcript 2 pp. 8–10 · Applied withdrawal
• Evidence: The UC table is the strongest comparable performance evidence but covers one LP and six funds. Dennis interviews supply denominator-like counts and conflicting outlier memories; filings supply ownership snapshots, not proceeds. No full cash-flow/control series tests whether the higher-batting preference produced superior risk-adjusted returns.
• Limits, failure modes, and exception ledger: Hit rates depend on definitions and horizon; ex-outlier analysis can understate a deliberately power-law strategy; IRR can be gamed by timing; TVPI mixes marks and cash; one-LP data may differ by fees/timing. Model-specific attempted-route ledger: searches sought personal/trust records, The Group's member/company ledger, Fireman's and AMEXCO investment/committee ledgers, IVA partnership/capital-account/audited statements, and IVP I–XI fund schedules; no complete set or stable success definition surfaced. IVP I–II LP statements were specifically sought through UC performance archives and web/LP routes without result. Final saturation searches for SEC filings naming Dennis with IVP I–II, UC records for IVP I–II, and CHM records for a Group portfolio/loss ledger produced no new material evidence; a duplicate 2017 UC table was not treated as an independent source. Retrieval ledger · source-map saturation record
• Evolution and contradictions: Dennis's early count-based pride persists into late interviews even as institutional funds create more measurable but dispersed vintages. The stated dislike of two-winner portfolios does not show lower concentration: Seagate is a remembered outlier, and IVP VI/VII dominate simple pooled arithmetic. The model preserves the preference while withholding efficacy.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: separate company-count viability from capital-weighted, gross/net, realized/unrealized, and ex-outlier economics. What fails / changed assumptions: incomplete denominators, incomparable marks/cash-flow dates, and intentional power-law construction can make simple counts or mechanical outlier removal misleading. Observable signals: ownership/dilution, reserves, DPI/TVPI/IRR, realization date, top-one and top-vintage sensitivity, and a stable success definition. Likely misuse: substituting survival, unicorn count, or marked value for realized fund return. [researcher-applied illustration]: show survival beside DPI/TVPI and results excluding the dominant AI company or vintage without combining differently timed rows into a synthetic IRR. NVCA's 487 mega-deals and 67% value share illustrate concentration, not fund efficacy. NVCA 2026 Yearbook

10. Scale selection capacity through capital cycles without timing pretense

• Rule: When capital supply expands, scale domain judgment, diligence quality, governance time, reserve discipline, and exit analysis before scaling fund size, price, round size, or decision speed. Diagnose cycle pressure without pretending to call its reversal date.
• Evidence status: [investor-stated · moderate] for Dennis's capital-supply mechanisms and timing humility; [contemporaneous record · strong] for his authored 1979 welcome to new capital; [researcher inference · moderate] for the measurable capacity rule; no causal estimate shows the model improved returns. 1979 prepared statement, printed pp. 1058–1060 · Almanac, ¶¶11–19
• Provenance and origin: In 1974 Dennis remembered scarce deals and entrepreneurs after IVA closed; in 1979 he welcomed new firms because excessive selectivity left capable entrepreneurs unfunded; around 2000–2009 he warned that excess capital raised prices, compressed decisions, enlarged rounds, changed motives, and encouraged building to sell. These are different cycle states, not one anti-capital doctrine. Dennis Perkins introduction, PDF p. 23 · spoken chronology
• Mechanism and reasoning: More capital can fund neglected good companies and attract skilled investors. Beyond selection/governance capacity it can reduce underwriting time, bid up entry price, overfund before evidence, overload boards and reserves, correlate portfolios, and create exit dependence. A dislocation may persist; the actionable variables are capacity and price, not a precise crash date.
• Historical and vehicle context: Dennis's comments span a newly raised $19 million IVA fund, post-1978 industry fundraising, first-IVP startup workload, dot-com public valuation, and later institutionalized venture. Fund size, security type, regulation, and market liquidity differ. His 2008/09 cycle-panel remarks were collective and must not be assigned solely to him. Cycle-panel attribution boundary
• When to use: Use in fund-size decisions, hiring and partner promotion, annual planning, hot-sector underwriting, mega-rounds, reserve reviews, compressed allocations, IPO/M&A assumptions, and public/private valuation comparisons.
• How to apply it: Track capital raised/deployed by stage/sector, deals and boards per partner, specialist coverage, diligence time and exception rate, entry price/preferences, round size versus milestone need, reserve burden, concentration/correlation, and realizable exit paths. Predefine actions—slow, syndicate, re-price, narrow, hire, or pass—without attaching them to a forecast correction date.
• Examples:
  • 1974 IVA — retrospective account; scarcity case, [retrospective witness · moderate]: Dennis remembered raising $19 million into high rates, falling markets, and scarce deals/entrepreneurs, then needing deal flow. A full-purse investor can lower standards if deployment becomes the objective. Dennis introduction, PDF p. 23
  • 1979 advocacy — researcher-applied illustration; contemporaneous articulation and productive-capital control, [researcher inference · moderate]: Dennis welcomed new entrants despite competition because existing firms had become too selective. The authored statement articulates a position; it does not document portfolio application of the synthesized capacity rule, and its estimated fundraising data cannot establish tax causality. Prepared statement, pp. 1058–1063
  • First IVP fund / LSI — retrospective account; capacity and speed exceptions, [retrospective witness · moderate]: the fund reportedly exceeded its planned startup count, while scarcity pushed Dennis to act before a long weekend over partner reluctance. Outcomes do not prove workload or rushed governance was sound. Raw transcript 2, pp. 18–20 and 25–29
  • January 2000 — researcher-applied illustration; contemporaneous articulation and timing-humility control, [researcher inference · strong]: Dennis called prices disconnected from value and predicted failures among late/unnecessary entrants but declined to claim a near-term break. The article documents a contemporaneous view, not a portfolio application; later public-loss admissions prevent turning the warning into a successful trade. Almanac, ¶¶11–19 · Raw transcript 2, pp. 17–18
• Evidence: The 1979 authored statement supplies an ex-ante pro-capital position; the 2000 article supplies an ex-ante valuation/timing distinction; direct later talks supply excess-capital mechanisms. None identifies a complete portfolio action or provides a causal capacity/return study.
• Limits, failure modes, and exception ledger: Capacity metrics can encourage box-checking and miss exceptional urgent opportunities; high prices can accompany durable growth; larger rounds can fund irreducible fixed cost; low deal pace can reflect weak sourcing rather than discipline. Dennis's forecasts lack named company/security actions, and his policy fundraising series had estimated components. Never label current concentration a bubble or predict its end from this record.
• Evolution and contradictions: Dennis moved from capital scarcity, to welcoming competition, to warning about excess supply and transaction incentives, while consistently resisting precise timing. The apparent reversal is conditional: capital is useful when selection and company-building capacity scale with it. The contradiction remains that his own first fund reportedly exceeded planned startup workload and he held some public positions to zero.
• Modern VC translation (as of 2026-08-02): [researcher inference · moderate] What transfers: scale selection, diligence, governance, reserves, and exit analysis before scaling dollars or speed; do not time a reversal. What fails / changed assumptions: large rounds may reflect real fixed costs or durable growth, aggregate funding does not measure one fund's capacity, and metrics can become box-checking. Observable signals: deals and boards per partner, specialist coverage, diligence time/exceptions, price/preferences, round-to-milestone need, reserve burden, concentration, and realizable exits. Likely misuse: calling AI a bubble or forecasting an immediate correction from funding concentration alone. [researcher-applied illustration]: use NVCA's reported first-half-2026 capital and AI/mega-round concentration to trigger a named fund-capacity and price audit, slowing or syndicating only when measured load or return fails. NVCA Q2 2026 Venture Monitor · NVCA 2026 Yearbook

Evidence gaps and candidate models

[researcher inference · strong] The ten models above are the full reconciled set supported at synthesis. Note-level rules were merged when they shared a mechanism; isolated ideas remain below rather than being promoted into artificial models. Philosophy model index · Writings playbook · Talks playbook

Highest-value adjudication remains unchanged: complete vehicle/fund ledgers and cash flows; company memos and board minutes; pass and loss files; hiring/funnel records; and the blocked authored/interview corpus. Until those arrive, a model's clarity must not be mistaken for Dennis's authorship, consistent application, or demonstrated return efficacy.

What the evidence supports

The strongest defensible thesis is narrower than a pioneer legend. Dennis combined engineering literacy, securities-analysis training, access to personal and family-trust capital, an insurer's balance sheet, and a dense Peninsula referral network. Across later vehicles, the record shows him translating technical claims for financial decision makers, assembling syndicates, authorizing some investments, serving on boards, and helping convert an informal network into institutionally funded partnerships. [researcher inference · strong] for this bundle; [researcher inference · insufficient public record] for repeatable personal return alpha. Dennis oral history, printed pp. 13–57/PDF pp. 19–63 · Testable edge thesis

The direct record supports several useful habits, but not a recovered universal checklist. Dennis distinguished technical adoption from security value, treated plans as hypotheses that boards must update, emphasized role-relevant people judgment, and recognized that fund and company clocks can diverge. His best ex-ante public judgment is the January 2000 separation of broad Internet adoption from listed-company prices and his refusal to predict when the gap would close. [investor-stated · strong] for that dated distinction; [researcher inference · moderate] for the generalized underwriting sequence. Almanac, body ¶¶4 and 11–19 · Written-record boundary

The operating evidence is collective. At Measurex, Draper and Wythes routed the founder, Bancroft supplied the largest remembered block, and Peggy Dennis's reported holding is separate from Reid's. At ROLM, McMurtry sourced, disclosed his prior position, recused, and joined the board while Dennis, Jamieson, and Norris approved. At Seagate, Marquardt sourced, persisted, and led IVP's board work; Oak/Glassmeyer supplied half the reported opportunity; Dennis authorized IVP's participation. At LSI, Sequoia brought Corrigan and Dennis made the urgent approval decision but held observer rather than board status. [retrospective witness · moderate] for these role reconstructions; [researcher inference · strong] that no lone-picker account survives them. Bossen, printed pp. 19–20/PDF pp. 25–26 · McMurtry, printed pp. 30–32/PDF pp. 40–42 · Marquardt, printed pp. 28–30/PDF pp. 33–35 · LSI panel, transcript p. 36

The policy record supports coalition participation, not sole-person causation. The 1977 JCT staff summary preserves Dennis's proposal for a permissive pension-investment rule and a cost-basis denominator; enacted tax and pension changes followed through different statutes, agencies, dates, and named advocates. Morgenthaler, Pratt, and Bancroft remember Dennis as important within the coalition, while the official records do not allocate causal shares. [contemporaneous record · strong] for Dennis's summarized proposal and enacted chronology; [retrospective witness · moderate] for coalition contribution; [researcher inference · insufficient public record] for sole authorship or causal effect. JCT summary, printed p. 8/PDF p. 5 · 1979 DOL final rule, 44 FR 37221–23 · Morgenthaler note

The economic evidence permits only bounded conclusions. Reported Ampex, Recognition, first-IVA-fund, and Seagate figures use different owners, dates, measures, and incomplete cash-flow sets. The only comparable public series is UC's net-to-one-LP IVP III–VIII table: researcher calculations from the disclosed paid-in and distributed cash yield 3.32x across those six rows, 2.47x excluding VII, and 1.82x excluding VI and VII. Those are simple cash ratios, not reported fund TVPI, personal economics, or proof that a named company drove a vintage. [contemporaneous record · strong] for UC's rows; [researcher inference · strong] for the arithmetic and boundary. UC IVP III–VII rows, p. 1 · IVP VIII row, p. 2 · methodology, p. 7

[researcher inference · strong] Accordingly, what survives is an institutional and process contribution: Dennis helped connect engineering evidence to capital, built successive vehicles, made identifiable consequential decisions, and articulated several useful boundaries. The public record does not establish a complete career return, a stable hit rate, a Dennis-only contribution to IVP performance, or a literal 80/20 result. Investment evidence boundary · Strongest skeptical case

What could falsify the thesis

[researcher inference · strong] The favorable thesis would weaken if original opportunity logs showed that family, Stanford, broker, employer, and co-investor access—not Dennis's technical translation—selected nearly all attractive opportunities; if contemporaneous memoranda showed generic rather than discriminating technical work; if partners supplied the decisive judgment in cases later attached to Dennis; or if matched ordinary cases showed no relation between his board interventions and company or investor outcomes. It would fail as a return thesis if complete, commensurable, net cash flows showed ordinary ex-outlier performance after price, ownership, sector, vintage, partner work, and access were controlled. Predictions and falsifiers · Ranked retrieval ledger

The strongest skeptical explanation is inherited and institutional access plus favorable geography and era plus team execution plus survivor memory. Dennis entered with trust capital and elite educational access; Fireman's Fund supplied a balance sheet and unusual autonomy; electronics and storage companies clustered nearby; later LP capital and policy changes expanded capacity; famous outcomes survived while the Group ledger, most losses, and complete IVP cash flows did not. [researcher inference · provisional] because the missing records could either strengthen or weaken this account. Ampex access record · Historical context · Source-map highest-value retrievals

[researcher inference · strong] The record already contains disconfirming observations. Reference Technology was remembered as roughly five years early; Apple was passed at a remembered price Dennis considered beyond his frame; Collagen's first use failed and required repeated financing; Dennis said people mistakes were his largest errors, repeated a stigmatizing body-size heuristic that this package rejects, and reported holding unnamed public securities to zero; LSI succeeded after an admitted consensus exception; IVP fund observations are sharply dispersed; Applied shows company-building patience colliding with LP liquidity and contractual rights. Failure and anti-portfolio table · People-judgment model · Applied filing

[researcher inference · strong] The discriminating evidence is concrete: contemporaneous source and pass logs; technical and customer diligence; term sheets and votes; dissent and recusal records; board minutes tied to a diagnosed constraint; ownership, dilution, reserves, and complete cash flows; ordinary and failed cases; and matched Dennis-led versus partner-led outcomes. Until those records surface, process and institution building pass, while personal alpha remains unproved. Predictions and falsifiers · Ranked retrieval ledger

Attribution and denominator audit

[researcher inference · strong] The package correctly separates personal and family-trust capital, Peggy Dennis's separate holding, The Group members' personal capital, Fireman's Fund/American Express corporate capital, the 1974 IVA partnership, numbered IVP funds and management entities, and later public holdings. It does not convert a management-control filing into sourcing, a board role into ownership, an IPO into realized proceeds, a company outcome into a fund return, or a firm result into Dennis economics. Vehicle analysis · People and capital map

[researcher inference · strong] Seven material role cross-checks were reopened against the underlying records. Material-case responsibility matrix

[researcher inference · strong] The named-deal funnel also reconciles. There are 33 named candidates: 26 included associations, comprising 25 funded/held lineages plus the Apple pass, and seven filing-based exclusions. All 33 were searched and researched; 11 of the 26 included cases have bounded decision labs and 15 remain unresolved. None of the 25 funded/held cases has every full-schema field. The unnamed roughly 25-company Group claim is outside the named funnel because no member-company ledger survives. Thus 11/26 is analytical depth and 0/25 is full-schema completeness, not career coverage or hit rate; career coverage remains not computable. Mechanical reconciliation · Group boundary

[researcher inference · strong] The source and phase denominators are internally exact. The map has 119 rows, 119 unique row-level retrieval URLs, and 99 defined source families; 107 rows are processed to their stated boundary and 12 remaining rows are low-priority ★ routes. All 31 ★★★ and all 61 ★★ rows are processed. WRITINGS routes its cutoff set of 66 priority rows exactly once across 14 WRITINGS, 17 TALKS, 12 DEALS, and 23 PROFILE/SYNTH destinations; its 15-artifact corpus has seven accessible full/partial records analyzed and eight exact blocked boundaries. TALKS reconciles 26 notes: 20 full-body and six partial/metadata routes, with no unprocessed priority family. Source coverage assessment · Writings reconciliation · Talks reconciliation

Transfer limits for modern VCs

The most defensible transfers are procedures, not slogans: separate technical efficacy, current customer need, differentiation, and security economics; test people on role-relevant evidence and reject familiarity or appearance proxies; distinguish source, sponsor, vote, board work, ownership, and exit; write down conflicts and dissent; connect the next check to information gain; match vehicle duration, reserves, partner capacity, and liquidity to the company; and report company viability separately from capital-weighted outcomes. These are [researcher inference · moderate], not recovered Dennis rules with demonstrated efficacy. Mental-model field guide · Reusable checklists

[researcher inference · moderate] Historical conditions do not transfer mechanically. Dennis's trust capital, insurer balance sheet, small early rounds, nearby electronics cluster, limited competition, and successive corporate/partnership vehicles differ from a modern institutional fund. A remembered proximity advantage can become a closed funnel; patient capital can become escalation; consensus can be bypassed by scarcity; company building can collide with LP duties; a high company-success count can conceal weak capital-weighted performance; and category growth cannot justify any security price. Historical context · Vehicle-fit analysis

[contemporaneous record · strong] The current factual premises were refreshed as of 2026-08-02. NVCA reports extreme 2025–H1 2026 AI and mega-round concentration. [researcher inference · moderate] That concentration is a capacity and price-audit signal, not a company proof or crash clock. [contemporaneous record · strong] Current 29 CFR §2550.404a-1(b) requires ERISA plan fiduciaries with investment duties to analyze portfolio role, alternatives, diversification, liquidity/current return, projected return, and funding objectives. [researcher inference · strong] It does not turn Dennis's 1977 cost-basis proposal into a modern blanket rule. [contemporaneous record · strong] IRS Topic 427 supports option-regime, exercise, AMT, and holding-period consequences, not an uncited claim about preferred-stock mechanics. NVCA Q2 2026 Venture Monitor · NVCA 2026 Yearbook · 29 CFR §2550.404a-1(b) · IRS Topic 427

[researcher inference · moderate] The practical transfer test is therefore observable: name the decision owner and legal vehicle; freeze the evidence available then; state the customer, price, ownership, and milestone assumptions; record dissent and conflicts; budget partner and reserve capacity; define the evidence that releases the next check; and show gross/net, realized/unrealized, and ex-outlier economics only within a compatible denominator. Without those fields, technical fluency, network quality, patience, and company building are stories rather than testable mechanisms. Reusable diligence and IC checklist · Decision-lab protocol

Unresolved evidence

• Ampex subscription, trust-authority, security, ownership, sale, tax, and realized-proceeds records remain unavailable; the $13,000/$15,000/$20,000 inputs and $800,000–$900,000/above-$1m values remain conflicting recollections. Ampex lab
• No Group member/company ledger reconciles the roughly 25 companies, reported successes, capital-return cases, losses, member allocations, board sponsors, or cash flows. Group evidence boundary
• Fireman's Fund/American Express committee minutes, the AMEXCO charter or licensing file, Recognition board/accounting records, and the corporate venture ledger remain missing. The public Fireman's finding aid is a bounded maritime-collection dead end. Fireman's finding aid · Vehicle boundary
• The IVA agreement, LP schedule, capital calls, fees, carry, reserves, extensions, company schedule, and cash flows are unavailable; the reported $19m and $180m–$200m accounts remain unaudited, and witness chronology differs. IVA retrieval priority
• No public IVP I–II LP statement or complete Dennis-era fund series surfaced. UC III–VIII represents one LP; later CalSTRS funds do not fill the Dennis-era gap. Comparable LP observations
• Seagate's two reported check/distribution accounts, LSI's legal vehicle and allocation, Collagen's clinical and financing decisions, Reference's full loss mechanics, Apple's pass terms, and Applied's final disposition remain unreconciled. Deep case ledger · Failure table
• The 1981 Dennis article, Done Deals chapter, MBA paper, Stanford ETL handout, full WSJ and Gentry interviews, Applied declaration, original event recordings, and Berkeley/Stanford audiovisual originals remain bounded by exact legal, institutional, or access routes. Written access boundaries · Spoken retrieval agenda

[researcher inference · strong] The review corrected source-map priority parity, the Financiers and LSI page locators, unsupported family chronology, ambiguous missing-field dashes, Markdown-breaking evidence-label separators, claim-local links, note metadata keys, legal and tax scope, direct-looking phrases without note parity, and overbroad employee-option inferences. The remaining items above are explicit archival, denominator, attribution, or outcome limits after bounded public-route work; none is a silently waived accessible priority source or unfinished phase requirement. No REVIEW continuation is required on the present record. Source coverage · Uncertainty ledger

Twenty-point acceptance evidence

[researcher inference · strong] After the recorded corrections, no acceptance item requires a public-record continuation. Remaining uncertainty is preserved in the ranked retrieval ledgers and the substantive boundaries above. Source coverage · Uncertainty ledger

Central-claim citation audit trail

Ordinary-claim citation audit trail

The sample was fixed at two claims from each pre-review core document before results were recorded; it was not selected after checking.

Citation and link QA

[researcher inference · strong] Mechanical reconciliation returns 119 source rows and 119 unique row-level retrieval URLs: 107 checked and 12 unchecked ★ routes, with 31/31 ★★★ and 61/61 ★★ processed. All 41 notes have the exact metadata keys and six required analytical H2 sections. All eight reader-facing roots open directly into a substantive H2; investments.md opens with the required Pareto section. The package contains 11 four-question decision labs, nine failure/anti-portfolio rows, 17 philosophy sections, and 10 twelve-component mental models. Source coverage · Decision labs · Mental models

[researcher inference · strong] The final sweep checked 743 local links, including 574 heading anchors, with no missing target; it also checked 186 unique external HTTP routes, of which 151 returned 2xx, 31 returned access-controlled 401/403, four ended in a transport timeout or HTTP/2 stream error, and none returned 404. No Markdown table-width defect, raw compound evidence-label separator, or Git whitespace error remained. Authentication, automation controls, and transport failures were treated as access states, not automatic evidence that a reader-facing route was dead; material routes were manually reopened where required. Source access boundaries · Written access boundaries

[researcher inference · strong] The quotation audit checked direct-looking phrases against notes or underlying audio/text and paraphrased phrases that lacked note parity. Economic claims retain owner, period, reported-versus-calculated status, and cash-flow boundary. No unsupported literal 80/20 result, audited personal return, firm-wide hit rate, sole-policy-causation claim, or hidden accessible-source waiver survived the review. Economic evidence rules · Spoken-record evidence policy

Reader conclusion

[researcher inference · strong] Read Dennis as an important technical translator, syndicator, board participant, and builder of successive venture-capital institutions—not as a demonstrated lone picker with an auditable personal track record. The best lessons are disciplined separations: technology from security price, company viability from investor return, source from approval and board work, personal from corporate and fund capital, patient company building from LP duration, and historical policy advocacy from current legal duty. The same evidence requires skepticism: access and era mattered, partners and operators supplied decisive work, famous economics are incomplete and conflicting, fund observations are dispersed, and the career denominator is missing. The package is valuable because it preserves those limits rather than converting pioneering status into proof of alpha. Testable edge thesis · Strongest skeptical case

Direct voice, oral history, and first-person records

• ★★ | Reid Dennis oral-history catalog record | https://digicoll.lib.berkeley.edu/record/218448 | 2009 | archive catalog; reliability A for metadata | Canonical metadata, rights route, and five-videodisc holdings record for the two-session Berkeley interview; processed with the transcript in TALKS.
• ★★★ | Reid Dennis: Pioneer Venture Capitalist oral-history transcript | https://digicoll.lib.berkeley.edu/record/218448/files/dennis_reid.pdf | interviews 2009-01-13 and 2009-02-10; published 2009 | full transcript; reliability A | Core direct account of family access, Ampex, Fireman's Fund, The Group, IVA, IVP, policy, deals, and failures; processed in TALKS.
• ★★★ | Reid Dennis and Pitch Johnson, Venture Voices oral history | https://digicoll.lib.berkeley.edu/record/218377/files/dennis_johnson_donated.pdf | 2008-09-22 interview; 2009 copyright | full transcript; reliability A | Joint testimony and a high-priority cross-check for industry chronology and first-fund cases; processed in TALKS.
• ★★ | Tom Perkins oral history with Reid Dennis introduction | https://digicoll.lib.berkeley.edu/record/218111/files/klienerperkins00perkrich.pdf | 2002 | transcript and signed introduction; reliability A | Dennis's own short assessment of Perkins, Collagen, and early venture relationships; signed introduction processed in WRITINGS.
• ★★★ | Emerging Innovative Companies, Capital Formation and the Need for Incentives | https://play.google.com/books/reader?id=_7XYivWt-6sC&hl=en_US&pg=GBS.PA1057 | 1979-05-22 | prepared Senate statement; reliability A; access full | Dennis speaking as NVCA chair on capital supply, management recruitment, and qualified stock options; printed pp. 1057–1067 processed in WRITINGS.
• ★★★ | Venture capital, technology and taxes | https://doi.org/10.1016/0160-791X(81)90016-6 | 1981 | authored journal article; reliability A for metadata; body restricted | Technology in Society 3(1–2):107–111; DOI, ScienceDirect, and controlled-lending routes attempted, with no body claim inferred.
• ★★ | Stanford MBA paper on startup-financing challenges, title unknown | https://www.legacy.com/us/obituaries/sfgate/name/reid-dennis-obituary?id=54696176 | date unknown; written while completing MBA; reported 2024 | coauthored student paper; reliability C until located | Family obituary establishes the artifact but supplies no title, coauthor, course, repository ID, or text; Stanford and catalog routes produced no copy.
• ★★★ | Done Deals, Reid Dennis chapter beginning p. 179 | https://archive.org/details/donedealsventure00gupt/page/178/mode/2up | 2000 | controlled-digital-lending book; reliability B | Interview-based chapter is a near-contemporaneous Dennis route; authenticated loan required and no unauthorized mirror was used.
• ★★ | Venture capitalist Reid Dennis: “We've just scratched the surface of the Internet” | https://www.almanacnews.com/morgue/2000/2000_01_05.yearbw.html | 2000-01-05 | interview profile; reliability B; access full | Ex-ante separation of Internet adoption from listed-company value, crowded-entry risk, access rhetoric, and timing humility processed in WRITINGS.
• ★ | Reid Dennis acceptance-speech notes | https://willprice.blogspot.com/2007/06/reid-dennis.html | 2007-06-08 | attendee notes; reliability C | Contemporary paraphrase of Dennis's IBF speech; processed to the absent-recording boundary in TALKS.
• ★ | Global Technology Symposium highlights with Reid Dennis | https://techcrunch.com/2009/04/03/highlights-from-this-years-global-technology-symposium/ | 2009-04-03 | event report; reliability B | Contemporary collective-panel recap and event metadata processed to the absent-transcript boundary in TALKS; no statement is assigned to Dennis alone.
• ★★★ | Something Ventured Reid Dennis interview, transcript one | https://stacks.stanford.edu/file/vf476gy6652/Reid_Dennis_Transcript.pdf | interview 2008-09-25; transcript object circa 2009 | raw interview transcript; reliability A | Long direct interview that overlaps the second Stanford transcript and was processed as one evidence family in TALKS.
• ★★ | Something Ventured Reid Dennis interview, transcript-one record | https://purl.stanford.edu/vf476gy6652 | interview 2008-09-25; transcript object circa 2009 | repository metadata and rights record; reliability A for metadata | Canonical Stanford control for transcript one; processed with its PDF and collapsed into the same interview family.
• ★★ | Something Ventured Reid Dennis interview, transcript two | https://stacks.stanford.edu/file/mj877gg1809/Reid_Dennis_Transcript_2.pdf | interview 2008-09-25; file dated 2009-08-27 | raw interview transcript; reliability A | Overlapping rendering includes Dennis's retirement and limits-of-competence reflections; processed with transcript one in TALKS.
• ★★ | Something Ventured Reid Dennis interview, transcript-two record | https://purl.stanford.edu/mj877gg1809 | interview 2008-09-25; file dated 2009-08-27 | repository metadata and rights record; reliability A for metadata | Canonical Stanford control for transcript two; processed with its PDF and collapsed into the same interview family.
• ★★ | Something Ventured collection record | https://purl.stanford.edu/mb678nw9491 | circa 2011 | repository collection metadata; reliability A for metadata | Controls collection provenance and routes companion media; processed without counting a new Dennis appearance.
• ★ | Something Ventured Reid Dennis moving-image component, early history | https://purl.stanford.edu/th350dw5024 | interview 2008-09-25; repository object circa 2009 | moving-image metadata; reliability A for metadata | Component record for early interview material; metadata reviewed, media not replayed, and collapsed into the raw-interview family.
• ★ | Something Ventured Reid Dennis moving-image component, NBI/Collagen/Fireman's Fund | https://purl.stanford.edu/hj715nt1591 | interview 2008-09-25; repository object circa 2009 | moving-image metadata; reliability A for metadata | Component record for the named segment; metadata reviewed, media not replayed, and collapsed into the raw-interview family.
• ★ | Something Ventured Reid Dennis moving-image component, Carter/Doriot/East–West | https://purl.stanford.edu/tb017pn8590 | interview 2008-09-25; repository object circa 2009 | moving-image metadata; reliability A for metadata | Component record for the named segment; metadata reviewed, media not replayed, and collapsed into the raw-interview family.
• ★★ | Something Ventured official study guide | https://zeitgeistfilms.com/sitelets/somethingventured/pdf.php | 2011 | distributor study guide; reliability B | Official synopsis, participants, runtime, and closed-caption metadata for the edited documentary; processed without treating the edit as an independent interview.
• ★ | Something Ventured film listing | https://tv.apple.com/us/movie/something-ventured/umc.cmc.4h745fuvxxm12zvdzp2411pe | 2011 film | commercial platform metadata; reliability C | Runtime, credits, and legal full-film route; processed to the rental boundary without implying full-film review.
• ★★ | Something Ventured Smithsonian finding aid | https://sirismm.si.edu/EADpdfs/NMAH.AC.1467.pdf | interview 2008-09-25; finding aid 2018 | archive authority; reliability A for metadata | Resolves the interview date and proves both Stanford transcripts are one Dennis interview family.
• ★★ | Innovation is Nothing New event record | https://computerhistory.org/events/innovation-nothing-new-100odd-years/ | 2005-04-27 | institutional event record; reliability A for metadata | Metadata for Dennis and Pitch Johnson's full recorded discussion; processed with the video in TALKS.
• ★★★ | Innovation is Nothing New full video | https://www.youtube.com/watch?v=gC6qJ7HuCX8 | recorded 2005-04-27; uploaded 2019-01-31 | full captioned event video; reliability A | Entire 1:39:48 program and captions reviewed with quoted passages checked against audio in TALKS.
• ★★★ | Then and Now: Venture Capital | https://www.inc.com/articles/201109/then-and-now-venture-capital.html | 2011-09-22 | interview; reliability B; access full | Direct retrospective remarks on The Group, motivation, and changing portfolio expectations processed with denominator and nostalgia controls.
• ★★ | The Pioneer | https://www.gwenbooks.com/pdf/gentry_press.pdf | Autumn 2011 | Gentry Wealth interview/profile; reliability C until retrieved | Public clipping exposes cover, masthead, editor's note, and contents metadata but omits the article body; the contents say the article begins at printed p. 20 and credit Kristine Carber.
• ★★ | Five Questions With Reid Dennis, a VC Investor Since 1952 | https://www.wsj.com/articles/BL-VCDB-2219 | 2009-06-24 | paywalled interview; reliability C until retrieved | Exact canonical route inspected; body remained unavailable and no title-derived claim was used.
• ★ | Stanford ETL archive | https://stanfordetl.wordpress.com/archive/ | 2005-01-12 event | archive index; reliability A for metadata | Confirms Venture Capital as a Career—Then and Now; processed to the dead biography/handout boundary in TALKS.
• ★ | Stanford ETL Reid Dennis biography snapshot | https://web.archive.org/web/20100706174327/http://etl.stanford.edu/handouts/0405_win_speakerbio/ReidDennis.html | 2010-07-06 snapshot of 2005 page | archive snapshot; reliability C until retrieved | Exact snapshot timed out during retrieval on 2026-08-01; no body claim was inferred.
• ★★ | Stanford ETL Reid Dennis handout snapshot | https://web.archive.org/web/20120406220815/http://etl.stanford.edu/handouts/0405_win/ReidDennis_1_12_2005.pdf | 2012-04-06 snapshot of 2005 handout | archive snapshot; reliability C until retrieved | Live and archived payloads failed; authorship and contents remain unknown and no talk-title inference was used.
• ★ | San Francisco Aeronautical Society 2010 Reid Dennis honoree page | https://sfaero.org/honorees/2010-reid-dennis/ | 2010 | institutional award biography; reliability B | Award metadata and aviation chronology processed to the absent program/speech/recording boundary in TALKS.
• ★★ | The Spoils of Success | https://www.csmonitor.com/1983/0513/051337.html | 1983-05-13 | contemporaneous reported interview; reliability B | Direct fragments on out-of-Valley deals, capital abundance, and competition; processed in TALKS.
• ★ | Automatic Pilots | https://www.forbes.com/asap/2000/0529/148.html | 2000-05-29 | reported lifestyle interview; reliability B | Direct fragments on piloting, retirement, and generational time pressure; processed without transferring aviation risk tolerance into investing claims.
• ★★ | Stakeholder Explains Push for Applied IPO | https://www.ocbj.com/finance/stakeholder-explains-push-applied-ipo/ | 2012-11-16 | contemporaneous report quoting a missing WR Hambrecht video; reliability B | Preserves Dennis's IVP IV fund-life and LP-liquidity rationale; processed to the unavailable-video boundary in TALKS.
• ★★ | The kicking-and-screaming IPO | https://fortune.com/2013/01/07/the-kicking-and-screaming-ipo/ | 2013-01-07 | contemporaneous report quoting the missing WR Hambrecht video and filing; reliability B | Preserves additional Dennis fragments and transaction context; processed without treating the article as the primary record for Applied's legal position.

Witnesses, colleagues, and team attribution

• ★★★ | Burt McMurtry oral history | https://digicoll.lib.berkeley.edu/record/218493/files/mcmurtry_burt.pdf | interviews 2009-06-09 and 2009-06-30; published 2009 | full transcript; reliability A | Co-founder's account of The Group, IVA fundraising, ROLM, and the 1980 partnership split; processed in TALKS.
• ★★★ | Burgess Jamieson oral history | https://digicoll.lib.berkeley.edu/nanna/record/103808/files/jamieson_burgess.pdf?registerDownload=1&version=1&withMetadata=0&withWatermark=0 | interviews 2009-07-21 and 2009-07-28; published 2010 | full transcript; reliability A | IVA colleague's chronology and disputed four-founder account, including Jim Norris; processed in TALKS.
• ★★★ | David Marquardt oral history | https://archive.computerhistory.org/resources/access/text/2021/05/102792229-05-01-acc.pdf | interviews 2011-10-11 and 2012-02-28; published 2021 | full transcript; reliability A | Skeptical witness on IVA's internal communication, Dennis's intuitive style, and the 1980 split; processed in TALKS.
• ★★ | T. Bondurant French oral history | https://archive.computerhistory.org/resources/access/text/2021/09/102792280-05-01-acc.pdf | interviews 2011-12-16 and 2012-05-08; published 2021 | full transcript; reliability A | Alternative IVA date and lineage account that tests retrospective chronology; processed in TALKS.
• ★★★ | David Morgenthaler oral history | https://archive.computerhistory.org/resources/access/text/2019/03/102781071-05-01-acc.pdf | interview 2010-04-14; published 2019 | full transcript; reliability A | Independent witness to Dennis's NVCA coalition work on capital-gains and pension policy; processed in TALKS.
• ★★ | Thom Weisel oral history | https://archive.computerhistory.org/resources/access/text/2020/09/102740478-05-01-acc.pdf | interview 2018-12-10; released 2020 | full transcript; reliability A | Former Fireman's Fund employee's account of Dennis and the insurer investment department; processed in TALKS.
• ★★★ | Stanley Pratt oral history | https://archive.computerhistory.org/resources/access/text/2021/05/102792221-05-01-acc.pdf | interview 2011-08-09; published 2021 | full transcript; reliability A | Credits Dennis and Stanley Golder within the tax-and-ERISA advocacy coalition; processed in TALKS.
• ★★★ | Pete Bancroft oral history | https://digicoll.lib.berkeley.edu/record/218541/files/bancroft_pete.pdf | interviews 2009–2010; published 2010 | full transcript; reliability A | Witness account of Dennis's Washington policy work and the shared West Coast investment network; processed in TALKS.
• ★★ | William Draper oral history | https://digicoll.lib.berkeley.edu/record/103600/files/draper_william.pdf | interviews 2008; published 2009 | full transcript; reliability A | Measurex sourcing and investor-network counterevidence; the transcript contains no Dennis mention and was processed as a negative control in TALKS.
• ★★★ | David A. Bossen oral history | https://digicoll.lib.berkeley.edu/record/103597/files/bossen_dave.pdf | interview 2010; published 2011 | canonical full transcript; reliability A | Printed p. 20/PDF p. 26 names the wider Measurex investor network and separates Peggy Dennis's personal position from Reid Dennis's investor role; processed in TALKS.
• ★★ | Dave Bossen Measurex oral-history mirror | https://ed-thelen.org/MeasurexBossenOral.html | interview 2010; published 2011 | authorized timestamped transcript mirror; reliability B | Duplicate of the canonical Berkeley transcript used only for audio timestamps and collapsed into the same evidence family.
• ★★ | Edward F. Glassmeyer oral history | https://digicoll.lib.berkeley.edu/record/103633/files/vcg-glassmeyer.pdf | interview 2008-10-17; published 2009 | full NVCA transcript; reliability A | Co-investor testimony says Oak bought 20% of Seagate for $1m, invited IVP to take half, and Marquardt joined the board; processed without restricted quotation.
• ★★ | LSI Logic oral-history panel | https://archive.computerhistory.org/resources/access/text/2012/04/102746194-05-01-acc.pdf | recorded 2011-11-30; published 2012 | full Computer History Museum transcript; reliability A | Founder/operator testimony cross-checks LSI's 1981 syndication, financing environment, Dennis's observer rather than board role, and the limits of person-level check attribution.
• ★★ | Charles Lea oral history | https://digicoll.lib.berkeley.edu/record/103633/files/vcg-lea.pdf | interviews 2008-10-10 and 2008-10-11; published 2009 | full transcript; reliability A | Witness evidence on NVCA coalition membership and Dennis's reported proximity heuristic; processed without restricted quotation in TALKS.
• ★★ | Brook Byers oral history | https://digicoll.lib.berkeley.edu/record/218191/files/byers_brook.pdf | interviews 2002–2005; published 2006 | full transcript; reliability A | Reconstructs a spring 1970 Stanford panel with Johnson, Draper, and Dennis while preventing retroactive IVP attribution; processed without restricted quotation in TALKS.
• ★ | Rebecca Lynn, Creating Your Own Canvas | https://ecorner.stanford.edu/wp-content/uploads/sites/2/2015/11/3600.pdf | 2015-11-11 | full Stanford eCorner talk transcript; reliability B | Retrospective witness to Dennis, Kramlich, and Morgenthaler's private 2008/09 venture-cycle panel; processed with collective attribution in TALKS.
• ★★ | The Next New Thing | https://computerhistory.org/stories/the-next-new-thing/ | accessed 2026-08-01 | institutional web feature; reliability B | Processed as historical context for The Group and WAVC; it supplies no member ledger, investment economics, or basis for formal-lineage claims beyond its institutional narrative.
• ★★ | Computer History Museum Reid Dennis profile | https://computerhistory.org/profile/reid-dennis/ | accessed 2026-08-01 | institutional biography; reliability B | Its selected personal, corporate, and IVP company names were screened into the DEALS discovery ledger as role leads, not treated as a complete portfolio or as proof of economics or sole attribution.

Books, scholarship, and historical interpretation

• ★★★ | Financiers of Innovation: The Development of the Venture Capital Industry | https://creativeclass.com/articles/Financiers%20of%20Innovation.pdf | 1997 | scholarly book manuscript; reliability B | Full manuscript screened: printed p. 135/PDF p. 135 names five Group members and describes AMEXCO, while printed p. 139/PDF p. 139 conflicts with direct witnesses on Bryan's given name, IVA's split date, and the successor name; used only with those conflicts preserved.
• ★★★ | Venture Capital: An American History | https://books.google.com/books?id=uSmWDwAAQBAJ | 2019 | limited-preview scholarly book; reliability B | Canonical limited-preview route screened; it supports completed-peer and institutional context but exposed no Dennis-specific body passage suitable for promotion, so no title- or snippet-derived Dennis claim is used.
• ★★ | Social Networks in Silicon Valley | https://web.mit.edu/ecastill/www/publications/CastillaExtract.pdf | 2000 | scholarly book chapter; reliability B | Full extract screened; printed p. 236/PDF p. 20 names Dennis as a charter Group member and gives a derivative IVA/Seagate account, useful as network context but not independent verification of deal economics or individual leadership.
• ★★ | Venture Capital-Financed Innovation and Technological Change in the USA | https://kenney.faculty.ucdavis.edu/wp-content/uploads/sites/332/2018/03/Venture-Capital-Financed-Innovation-and-Technological-Change-in-the-US.pdf | 1988 | research paper; reliability B | Full paper screened; Dennis appears only in the acknowledgments/interview list on PDF p. 1, so the paper is retained as fund-market context and a negative control rather than Dennis-specific evidence.
• ★★ | Financing the Next Silicon Valley | https://wustllawreview.org/wp-content/uploads/2021/10/87.4.1.pdf | 2010 | law-review article; reliability B | Full article screened; printed p. 741/PDF p. 26 names Dennis and four peers as angels but generalizes the SBIC solution, so it does not establish that Dennis or AMEXCO used an SBIC.
• ★★★ | Solving the problem of abundance: venture capital and the making of asset-driven inequalities | https://researchonline.lse.ac.uk/id/eprint/125890/1/Solving_the_problem_of_abundance_venture_capital_and_the_making_of_asset-driven_inequalities.pdf | 2024 | critical scholarly article; reliability B | Full article screened; printed pp. 14–15/PDF pp. 16–17 supplies an adversarial asset-abundance account and the six-insurer-plus-Ford/$19m IVA claim, but it relies on Dennis's 2009 oral history and is not independent verification of economics or policy causation.
• ★ | Understanding Silicon Valley | https://www.researchgate.net/publication/37702342_Understanding_Silicon_Valley_The_Anatomy_of_an_Entrepreneurial_Region | 2000 | hosted scholarly volume; reliability B | Provides regional institutions and network context for testing person-centered claims.
• ★ | The Power Law sample | https://cdn.penguin.co.uk/dam-assets/books/9780141988948/9780141988948-sample.pdf | 2022 | book sample; reliability B | Modern narrative treatment useful for locating claims, not for a portfolio or return denominator.
• ★★ | Venture Capital's Early Pioneers | https://www.institutionalinvestor.com/article/2btgdtwr5w5i99g1zmm0w/portfolio/venture-capitals-early-pioneers | 2010-02-06 | trade-press history; reliability B; access full | Dennis-derived Group mechanics and result claims processed; 1952, 1974/1980, fund-size, and ERISA compressions are explicitly corrected.
• ★ | The rise of risk capital, part 2 | https://steveblank.com/2009/10/29/the-secret-history-of-silicon-valley-12-the-rise-of-%E2%80%9Crisk-capital%E2%80%9D-part-2/ | 2009-10-29 | practitioner synthesis; reliability B | Useful overview with Ampex amount variants that should not override primary records.
• ★ | Bancroftiana 137 | https://digicoll.lib.berkeley.edu/record/255009/files/bancroftiana_137.pdf | 2010 | archive newsletter; reliability B | Describes the venture-capital oral-history project, provenance, and interview network.
• ★ | Institutional Venture Partners fund-size history in Entrepreneurial Finance | https://books.google.com/books?dq=%22institutional+Venture+Partners%22+founded+1980&id=qpBlEAAAQBAJ&pg=PA18 | 2022 edition | restricted book preview; reliability C | Secondary lead for IVP II–IV fund sizes pending legal and LP records.

Vehicles, portfolio, filings, and performance records

• ★★★ | University of California private-equity performance report | https://www.ucop.edu/investment-office/_files/updates/pe_irr_06-30-20.pdf | 2020-06-30 | official LP performance table; reliability A | Tables pp. 1–2 and methodology p. 7 supply the dated, net-to-UC cash-flow observations for IVP III–VIII used in DEALS; they are one LP's fund-level results, not Dennis's economics or a deal ledger.
• ★★★ | University of California private-equity methodology and fund table | https://www.ucop.edu/investment-office/_files/invinfo/ICC_12-31-10%20Master%20by%20Type.pdf | 2010-12-31 | official LP report; reliability A | Tables pp. 1–3 were processed as an earlier IVP III–VIII valuation snapshot and methodology control; interim values and UC cash-flow timing prevent person-level or deal-level attribution.
• ★★ | UC Regents private-equity performance snapshot | https://regents.universityofcalifornia.edu/regmeet/nov03/603spreadsheetnov.pdf | 2003-09-30 | official LP report; reliability A | Tables pp. 1–2 were processed as a historical IVP III–VIII snapshot; the report's early-fund and valuation caveats prevent treating the observation as audited lifetime performance.
• ★★ | CalSTRS private-equity performance report FYE 2024 | https://www.calstrs.com/files/5c3703f7c/CalSTRSPrivateEquityPerformanceReportFYE2024.pdf | 2024-06-30 | official LP report; reliability A | Methodology p. 1 and table p. 5 were processed only as excluded later-fund context for IVP XII, XVII, and XVIII, not as Dennis-era or person-level performance.
• ★ | CalSTRS current private-equity portfolio performance table | https://www.calstrs.com/private-equity-portfolio-performance-table | accessed 2026-08-01 | dynamic official LP table; reliability A | Drift-prone later-fund observations demonstrate why all performance figures need dated snapshots.
• ★★ | Reid W. Dennis SEC submission index | https://data.sec.gov/submissions/CIK0000903975.json | 1997–2008 filings | regulator data index; reliability A | All 15 indexed filings were screened, yielding eight disclosed issuer families; the set is a later public-disclosure denominator, not an early-career portfolio denominator.
• ★★ | Collagen Schedule 13G | https://www.sec.gov/Archives/edgar/data/903975/000095000597000127/0000950005-97-000127.txt | filed 1997-02-11; holdings as of 1996-12-31 | SEC ownership filing; reliability A | Reports Dennis's individual beneficial ownership of 697,043 shares, or 7.9%; processed without reproducing the personal identifier embedded in the legacy filing.
• ★★ | Collagen Schedule 13G | https://www.sec.gov/Archives/edgar/data/903975/000095000598000092/0000950005-98-000092.txt | filed 1998-02-10; holdings as of 1997-12-31 | SEC ownership filing; reliability A | Reports 700,043 shares, including exercisable options, or 7.8%, with separately disclaimed trustee shares; it does not establish original cost, vehicle, or realized return.
• ★★ | Collagen Schedule 13G amendment | https://www.sec.gov/Archives/edgar/data/903975/000095000599000220/0000950005-99-000220.txt | filed 1999-02-22; holdings as of 1998-12-31 | SEC ownership filing; reliability A | Reports 706,043 shares, including exercisable options, or 8.1%; later ownership is not evidence of the original investment thesis or cash flows.
• ★★ | Cohesion Technologies Schedule 13G | https://www.sec.gov/Archives/edgar/data/903975/000095000599000101/0000950005-99-000101.txt | filed 1999-02-10; holdings as of 1998-12-31 | SEC ownership filing; reliability A | Reports 696,043 shares, or 8.3%, received through Collagen's 1998 distribution; this documents the corporate lineage but not investment basis or realization.
• ★★ | Cohesion Technologies Schedule 13G amendment | https://www.sec.gov/Archives/edgar/data/903975/000095000500000044/0000950005-00-000044.txt | filed 2000-01-21; holdings as of 1999-12-31 | SEC ownership filing; reliability A | Reports 696,043 shares, or 8.0%, and separately identifies excluded options and trustee shares; economics remain unreconstructed.
• ★★★ | Netflix final IPO prospectus | https://www.sec.gov/Archives/edgar/data/1065280/000101287002002475/d424b4.htm | 2002-05-22 | SEC prospectus; reliability A | Principal-stockholder table and footnote 6 identify IVP VIII ownership, Tim Haley's board role, and Dennis as one of ten shared controllers; the filing does not credit Dennis with sourcing or board work.
• ★★ | Digital Impact Form 4 | https://www.sec.gov/Archives/edgar/data/903975/000120158003000003/primary_doc.xml | filed 2003-09-11 | SEC ownership filing; reliability A | Records a pro-rata IVP VIII/affiliate distribution and shared reporting-person control; it is distribution mechanics, not evidence that Dennis sourced or led the deal.
• ★★ | Turnstone Systems Form 4 | https://www.sec.gov/Archives/edgar/data/903975/000102691903000001/primary_doc.xml | filed 2003-10-30 | SEC ownership filing; reliability A | Identifies IVP VII and affiliated reporting persons, including Dennis; it establishes shared control rather than personal deal leadership or economics.
• ★★ | Argonaut Technologies Form 4 | https://www.sec.gov/Archives/edgar/data/903975/000100905604000001/primary_doc.xml | filed 2004-01-13 | SEC ownership filing; reliability A | Identifies IVP VI and affiliated reporting persons, including Dennis; it supplies no sourcing, board, cost, dilution, or realized-return record.
• ★★ | @Road Form 4 | https://www.sec.gov/Archives/edgar/data/903975/000090397504000001/xslF345X02/primary_doc.xml | filed 2004-07-28; earliest reported transaction 2004-07-27 | SEC ownership filing; reliability A | Defines Dennis's IVM VIII and X roles, shared voting and investment power, and pecuniary-interest disclaimers; it does not establish a Dennis-only deal.
• ★★ | Vonage Form 3 | https://www.sec.gov/Archives/edgar/data/903975/000133500506000046/edgar.xml | filed 2006-05-23 | SEC ownership filing; reliability A | Identifies IVP X entities and a group of affiliated reporting persons, including Dennis; personal sourcing, governance work, and economics are not established.
• ★★ | Thermage Form 3 | https://www.sec.gov/Archives/edgar/data/1026919/000118143106062776/xslF345X02/rrd134489.xml | filed 2006-11-09; signed 2006-11-04 | SEC ownership filing; reliability A | Assigns Samuel Colella's board role and documents IVP VII/shared fund control, countering a Dennis-led-deal inference.
• ★★ | ArcSight S-1 | https://www.sec.gov/Archives/edgar/data/1368582/000095013407019818/f28075orsv1.htm | 2007-09-11 | SEC registration filing; reliability A | Principal-stockholder and management tables show IVP X ownership and six managing directors sharing control; no Dennis board role appears.
• ★ | Yodlee S-1 | https://www.sec.gov/Archives/edgar/data/1161315/000119312514255452/d684206ds1.htm | 2014-06-30 | SEC registration filing; reliability A | Maps IVP X–XI entity hierarchy and partner-specific board attribution.
• ★ | comScore principal-stockholder filing copy | https://ir.comscore.com/static-files/387f7f04-77bc-4a88-8b6b-af7c287cf647 | SEC-era filing | issuer-hosted filing, partial access; reliability A for document | Retrieval route for IVP X holdings and shared-managing-director control.
• ★★★ | Applied Medical Corp. v. Thomas | https://law.justia.com/cases/california/court-of-appeal/2017/a145867m.html | 2017-05-10 | appellate opinion transcription; reliability A | The opinion's background and disposition were processed for IVP IV roles and the time-bar/contract/conversion rulings; party allegations are not converted into findings, and the court did not reach the stated Dennis liability theories.
• ★★ | Reid Dennis declaration supporting summary judgment in Applied Medical Corp. v. Thomas | https://cases.justia.com/california/court-of-appeal/2017-a145867m.pdf?ts=1494450041 | filed by 2014–2015; described 2017 | signed litigation declaration; reliability A for court characterization; body unavailable | Opinion p. 4 n.2 jointly characterizes Dennis and Thomas declarations; exact caption, date, and Dennis-only wording require trial docket CIV519758.
• ★★★ | Applied Medical Resources Corp. Form S-1/A | https://www.sec.gov/Archives/edgar/data/1534727/000119312512424444/d415166ds1a.htm | 2012-10-16 | SEC registration filing; reliability A | “Reasons for this Offering” records IVP IV's exercise of demand-registration rights and Applied's stated opposition; processed as company-stated transaction evidence, not a merits finding.
• ★★ | Applied Medical Resources Corp. Form S-1/A financing history | https://www.sec.gov/Archives/edgar/data/1534727/000119312513055761/d415166ds1a.htm | 2013-02-13 | SEC registration filing; reliability A | Prior-financing paragraphs record the October 1988 IVP/IVM round, IVP follow-ons through December 1990, nonparticipation in the March 1992 final venture round, Peter Thomas's board role, and shared IVM IV control; issuer accusations remain company-stated, not adjudicated facts.
• ★★ | Applied Medical request to withdraw Form S-1 | https://www.sec.gov/Archives/edgar/data/1534727/000119312513247925/d550189drw.htm | 2013-06-05 | SEC withdrawal request; reliability A | Confirms the registration never became effective, no securities were sold through it, and the company sought withdrawal after IVP IV, other sellers, and their underwriter stopped responding; the stated reasons are the issuer's account.
• ★★ | IVP official homepage | https://www.ivp.com/ | accessed 2026-08-01 | promotional firm summary; reliability B | Current firm-wide company and IPO counts were processed only as a survivor-selected promotional denominator boundary, not Dennis performance.
• ★★ | How IVP's Reid Dennis built a VC firm to last | https://fortune.com/2024/03/19/how-ivp-reid-dennis-built-a-vc-firm-to-last/ | 2024-03-19 | business feature, partial access; reliability B | The accessible succession and current-fund context was screened, then excluded from Dennis-era economics; the full article remained gated.
• ★ | Fundraising documents reveal IVP's returns | https://www.newcomer.co/p/scoop-fundraising-documents-reveal | 2026-07-23 | paywalled confidential-deck report; reliability C | Recent reported inception and fund-return claims to compare with independent LP tables.
• ★ | Institutional Venture veterans | https://www.privatefundscfo.com/venture-veterans-2/ | 2008-09-01 | registration-gated trade feature; reliability C | Title, deck, Jennifer Harris byline, and date inspected; body and any Dennis contribution remain unavailable, so no content claim is used.

Corporate, archive, and institutional records

• ★★★ | Fireman's Fund Insurance Companies records | https://oac.cdlib.org/findaid/ark%3A%2F13030%2Fc8pc37sv | 1882–1969 holdings | archive finding aid; reliability A for metadata | The full ten-item HDC 1780 / SAFR 24946 inventory was screened; it concerns the vessel Harvey Mills and exposes no Dennis, investment-committee, Recognition Equipment, AMEXCO, or venture file. This is a bounded negative archive result.
• ★ | Fireman's Fund heritage timeline, 1960s | https://www.ffic-heritageserver.com/time/1960.htm | accessed 2026-08-01 | corporate retrospective; reliability B | Dates the American Express acquisition and supplies corporate context requiring independent corroboration.
• ★★ | American Express to lower stake in Fireman's Fund | https://www.washingtonpost.com/archive/business/1986/02/27/american-express-to-lower-stake-in-fund/0ef4b7c4-70d4-49b5-9736-52c59c139b38/ | 1986-02-26 | contemporaneous business report; reliability B | Processed only for the 1968 acquisition and 1985–1988 divestment chronology and stated corporate-capital rationale; it supplies no Dennis deal economics or proof of AMEXCO's legal form.
• ★ | VIA funds history | https://www.viafunds.com/ | accessed 2026-08-01 | successor-firm history; reliability B | Maps a claimed American Express venture lineage that must not be conflated with IVA or IVP.
• ★ | Redpoint's formation | https://www.almanacnews.com/morgue/1999/1999_08_25.venture.html | 1999-08-25 | contemporaneous local report; reliability B | Dennis's direct explanation and successor-team counterevidence processed in TALKS, guarding against a seamless founder-centered firm narrative.
• ★ | Computer Museum board meeting memo, 9 October 1992 | https://tcm.computerhistory.org/GardnerHenrieFiles/CM%20Board%20Meeting%2010-9-92.pdf | 1992-10-09 | institutional board memo; reliability A for the plan | Says Dennis and John Doerr would attend an October venture forum; processed only as planned-attendance evidence because no program, recording, or proof of appearance surfaced.
• ★★ | Fearless: Stanford Engineering Venture Fund at twenty-five | https://engineering.stanford.edu/news/fearless-stanford-engineering-venture-fund-twenty-five | 2021-06-14 | university history; reliability B | Full university history screened: it records Dennis among three people approached about endowed chairs in 1984 and reports first- and second-fund annual returns, but the article supplies no cash-flow method or Dennis-level economics; used only for later institutional and vehicle context.
• ★★ | Stanford Magazine July 2024 | https://project-orion-production.s3.amazonaws.com/uploads/issue/189/StanfordMagazine_July2024.pdf | 2024-07 | alumni magazine PDF; reliability B | Full issue route screened; PDF p. 59 separately summarizes Dennis's Stanford degrees, Navy service, IVA/IVP chronology, civic roles, family, and death, with no investment economics or sole-deal attribution inferred.
• ★★ | Celebration of Life for Reid Dennis | https://www.legacy.com/us/obituaries/sfgate/name/reid-dennis-obituary?id=55393790 | 2024-06-20 | short family event notice; reliability B | Full notice screened: it gives the 1926-05-16 to 2024-03-14 lifespan and September 29 celebration, links to the separate March obituary, and supplies no independent investment evidence.
• ★★★ | Roadblocks for Latino Venture Capitalist | https://www.sfgate.com/news/article/Roadblocks-for-Latino-Venture-Capitalist-3006465.php | 1998-05-04 | contemporaneous reported article; reliability B; access full | Richard Leza's allegation and Dennis's non-recollection/labor-market response processed without treating either as adjudication or exoneration.
• ★★ | Alpha Girls excerpt on Mary Jane Elmore | https://www.sfchronicle.com/style/article/How-can-women-survive-Silicon-Valley-and-thrive-13867257.php | 2019-05-21 | adapted book excerpt; reliability B | Full accessible excerpt screened: it reports peer warnings before Dennis hired Elmore and says IVP had not previously had a woman partner, but remains a retrospective narrative excerpt and cannot establish a firm-wide gender denominator or adjudicate the Leza allegation.
• ★ | Smithsonian Reid Dennis transcript finding-aid component | https://sova.si.edu/components/NMAH.AC.1467 | 2008 interview collection | archive authority; reliability A for metadata | Processed with the full finding-aid PDF to resolve provenance and the 2008-09-25 date for the overlapping Stanford transcripts.

Policy, regulation, and industry formation

• ★★★ | JCT/CRS summary of testimony on the Tax-Exempt Private Pension Investment Act of 1977 | https://www.jct.gov/getattachment/6f1dd827-59d6-42f7-9951-4ca6d76956f1/jcs-47-77-4091.pdf | 1977-08-02; testimony 1977-05-11 | official staff summary; reliability A; partial as Dennis record | Printed p. 8/PDF p. 5 preserves staff paraphrase of Dennis's concentration, cost-basis, and permissive-rule proposals; underlying testimony remains unrecovered.
• ★★★ | Congressional Record notice of the May 1977 pension-investment hearing | https://www.congress.gov/95/crecb/1977/05/09/GPO-CRECB-1977-pt12-1-2.pdf | 1977-05-09 | official congressional record; reliability A | Printed pp. 14000–14001 names Dennis alongside other venture-capital witnesses; processed as coalition and appearance evidence, not proof of sole authorship or causation.
• ★★★ | Revenue Act of 1978 | https://www.govinfo.gov/content/pkg/STATUTE-92/pdf/STATUTE-92-Pg2763.pdf | 1978-11-06 | enacted federal statute; reliability A | Title IV, §§ 401–403 supplies the contemporaneous capital-gains changes and controls TALKS chronology.
• ★★★ | Economic Recovery Tax Act of 1981 | https://www.congress.gov/97/statute/STATUTE-95/STATUTE-95-Pg172.pdf | 1981-08-13 | enacted federal statute; reliability A | Section 101 controls the later rate-reduction date and corrects compressed oral-history chronology in TALKS.
• ★★ | General Explanation of the Economic Recovery Tax Act of 1981 | https://www.jct.gov/publications/1981/jcs-71-81/ | 1981-12-29 | official legislative explanation; reliability A | Official full explanation screened as the enacted-law control for the 1981 provisions; it supplies statutory context, not Dennis authorship or person-level causal credit.
• ★★★ | Investment Duties regulation, 44 Federal Register 37221 | https://archives.federalregister.gov/issue_slice/1979/6/26/37221-37229.pdf | 1979-06-26 | final federal regulation; reliability A | Primary record controlling the 1979 prudent-investor-rule date in TALKS; it does not allocate causal credit among advocates.
• ★★ | H.R. 10148, Venture Capital Tax Act | https://www.congress.gov/bill/95th-congress/house-bill/10148 | introduced 1977-11-29 | official congressional bill record; reliability A | Full record and CRS summary screened: Rep. Thomas Downey introduced the bill, it was referred to Ways and Means and never enacted, and its exchange/option/loss provisions are proposal context—not the enacted changes or evidence of Dennis authorship.
• ★★ | Congressional Record, 22 August 1978 | https://www.congress.gov/95/crecb/1978/08/22/GPO-CRECB-1978-pt20-5-1.pdf | 1978-08-22 | official debate record; reliability A | Full issue searched and the capital-gains debate screened; it identifies a wider legislative and industry coalition but contains no Dennis hit, so it is used against sole-person causal credit.
• ★★ | Congressional Record, 23 May 1978 | https://www.congress.gov/95/crecb/1978/05/23/GPO-CRECB-1978-pt11-6-1.pdf | 1978-05-23 | official debate record; reliability A | Full issue searched for Dennis and venture-tax terms; no Dennis-specific passage was located, so no appearance, testimony, or causal claim is inferred from this route.
• ★★ | Federal Reserve report on small-business capital | https://www.federalreserve.gov/boarddocs/rptcongress/sbc_rep.pdf | 2007 | government report; reliability A | Full report screened for retrospective small-business finance and limited-partnership context; it supplies no Dennis-specific evidence or basis for attributing tax or pension reforms to one advocate.
• ★★★ | NVCA Board of Directors Handbook | https://nvca.org/wp-content/uploads/2023/05/2023-2024-NVCA-Board-of-Directors-Handbook.pdf | 2023–2024 | association record; reliability A for role metadata | Former-chair table dates David Morgenthaler to 1977–1978, Reid Dennis to 1978–1979, and Stanley Golder to 1979–1980; processed as title chronology, not a contributor ranking.
• ★ | Capital Formation hearings catalog | https://books.google.com/books/about/Capital_Formation.html?id=KA6PJo8SisMC | 1978 | official hearing-volume catalog; reliability A for metadata | Retrieval route for Ed Zschau, industry witnesses, and contemporaneous policy arguments.

Coverage assessment

• Source-map protocol: ★★★ must process · ★★ should process · ★ if time; the live 31/61/27 split keeps the queue selective. Reliability A means primary/direct or authoritative record, B means institutional or quality secondary, and C means discovery lead, interested paraphrase, or uninspected restricted source. [x] means a later phase processed the item to the access boundary written on its row; it does not mean a gated body was read. Reported gains, multiples, hit rates, and IRRs remain reported unless a dated LP or audited record states their basis.
• Corpus shape: this map now contains 119 source rows and 119 unique row-level retrieval URLs, representing 99 underlying source families after additionally collapsing the repeated Collagen filings into one issuer family, the repeated Cohesion filings into another, and the Applied registration/withdrawal records into one transaction family. The pre-DEALS family rules still collapse the Berkeley Dennis catalog/PDF pair, the canonical Bossen transcript/mirror pair, the Stanford raw-interview PDFs, record pages, collection row, sampled moving-image records, and Smithsonian controls, the CHM event/video pair, the Stanford ETL index/snapshot set, and the study-guide/Apple routes to the edited film. Eleven primary SEC rows were added in DEALS: nine represent six newly mapped issuer families, while two expand the existing Applied family; the Glassmeyer and LSI records add two independent operator/co-investor witness families. The current check state is 107 processed / 12 unprocessed rows; by grade it is 31/31 ★★★, 61/61 ★★, and 15/27 ★. TALKS began with 17 priority rows representing 14 families and processed all fourteen; its 26 exact-schema notes comprise four core direct-voice families, ten priority witness families, and twelve supplementary appearance or control families. The full map spans direct testimony, witness oral histories, books, scholarship, LP performance tables, SEC filings, litigation, corporate archives, contemporaneous press, enacted law, regulation, and legislative records.
• Name and date variants searched: Reid Weaver Dennis, Reid W. Dennis, Reid Dennis, R. W. Dennis, and Dennis, Reid were paired with Institutional Venture Associates, IVA, Institutional Venture Partners, IVP, The Group, Fireman's Fund/Firemen's Fund, American Express, AMEXCO, Ampex, Recognition Equipment, Measurex, Seagate, Exabyte, colleague/company names, and interview/oral-history/speech/letter/archive/lawsuit/failure terms. Date variants covered 1926, 1944–1952, 1955/56–1965, 1960, 1968, 1972–1975, 1977–1981, the 1980s, 1998, 2000–2011, and 2024. Results for people who merely share the Dennis surname were excluded.
• Attempted-route ledger: direct voice was searched through Berkeley, Stanford/eCorner, CHM, Smithsonian, NVCA, YouTube, event archives, exact-title press, and Internet Archive; vehicles and deals through SEC/issuer filings, OAC, insurer/successor records, official LP reports, and company/litigation searches; policy through Congress, GovInfo, Federal Register, JCT, Federal Reserve, and NVCA; criticism through lawsuit, discrimination, gender, failure, write-off, regulatory, and criminality variants; books and archives through Google Books, Internet Archive, library catalogs, and exact Wayback snapshots. Three broad TALKS saturation passes—decade/speech/lecture, platform/event/podcast, then conference/forum/video combinations—plus a final exact-title/video check produced no additional Dennis speech body beyond duplicate families, the processed cycle-panel recollection, the processed Applied-video fragments, and an October 1992 board memo that says Dennis and John Doerr would attend a venture forum but does not prove they did. The last three DEALS saturation searches—SEC filings for IVP I–II with Dennis, UC records for IVP I–II, and CHM records for a Group portfolio or loss ledger—produced no new material deal or cash-flow evidence; a duplicate 2017 UC table was not counted as an independent family.
• Direct voice: the two-session 2009 Berkeley transcript is unusually rich; the 22 September 2008 Dennis–Johnson interview is a valuable joint comparator; the full 2005 CHM video supplies directly auditable audiovisual evidence; and the 25 September 2008 raw Something Ventured interview preserves less-polished testimony in two duplicate transcript renderings. TALKS also recovered bounded direct or edited fragments from 1983, 1999, 2000, and 2012; a 2007 attendee paraphrase; a 2009 collective-panel recap; and event metadata for 2005 and 2010. WRITINGS recovered the full 1979 prepared statement and 2002 signed Perkins introduction. The 1981 article, controlled Done Deals chapter, full Wall Street Journal interview, MBA paper, signed Applied Medical declaration, original speeches, correspondence, board memoranda, and any standalone papers remain retrieval targets. No owned blog, memoir, public LP-letter corpus, or standalone Dennis papers collection surfaced.
• Archive routes, rights, and barriers: Berkeley states that publication beyond fair use may require permission, the NVCA transcript reserves publication rights, and Stanford limits reproduction of its raw transcripts beyond research, teaching, private study, and fair use. The Berkeley catalog identifies five interview videodiscs, but only the transcript was publicly reviewable; Stanford exposes companion moving-image metadata for the raw documentary interview, not playable media. The edited documentary requires lawful rental, the original WR Hambrecht Applied Medical video is missing, the Stanford ETL biography/handout payloads fail, and the SFAero page contains honoree biography rather than an appearance body. Done Deals is controlled digital lending; trade press can be paywalled; CHM's older institutional files may require onsite access. The public Fireman's Fund finding aid was fully screened and proved to be a ten-item maritime collection rather than the venture files sought; it points to no investment-committee or Recognition records. The exact Stanford ETL biography snapshot and handout snapshot both timed out when retried on 2026-08-01; the live links return 404. No public IVA/IVP partnership agreements or IVP I–II LP statements surfaced. These are bounded access findings, not evidence that records do not exist.
• Candidate denominators: Dennis retrospectively reports about 25 Group investments but supplies no ledger. CHM and IVP publish selected-company or firm-wide promotional lists. Dennis's SEC CIK gives a complete 15-filing / eight-issuer public-disclosure set for 1997–2008—Collagen, Cohesion, Digital Impact, Turnstone, Argonaut, @Road, Vonage, and ArcSight—but neither an early-career denominator nor proof of sourcing; Thermage and Netflix appear through separate issuer filings. UC provides comparable, dated, net-to-one-LP observations only for IVP III–VIII; CalSTRS's disclosed IVP XII, XVII, and XVIII observations are later-fund context excluded from Dennis-era performance. A complete personal, Group, corporate, IVA, and IVP I–XI portfolio-and-cash-flow denominator remains unresolved.
• Vehicle and attribution boundary: at minimum, later phases must separate Dennis personal capital, The Group's member capital, Fireman's Fund/American Express corporate capital, the 1974 IVA partnership, and each IVP fund. Within funds, sourcing, investment-committee approval, board work, follow-on decisions, shared GP control, and economics may belong to different people.
• Chronology and identity conflicts: preserve $13,000/$15,000/$20,000 for Ampex; $1.2 million/$1.3 million for the remembered Measurex fundraising target (Dennis, printed p. 26/PDF p. 32 · Bossen, printed pp. 19–20/PDF pp. 25–26); Dennis's John Bryan versus Kenney and Florida's William Bryan; three senior IVA principals plus junior partner Jim Norris versus Jamieson's four-founder account; 1973 planning/1974 close versus a 1975 IVA recollection; seven LPs versus six-institution shorthand; direct-witness 1980/Technology Venture Investors versus scholarly 1976/Technology Venture Associates; The Group network overlap versus formal WAVC succession; 1978 versus 1981 capital-gains chronology; a proximity heuristic versus Dennis's 1983 out-of-Valley deals; reported Group/startup counts with shifting denominators; and the 2008-09-25 Smithsonian interview date versus later transcript-file dates. Dennis says no SBIC was needed, while Kenney and Florida call AMEXCO a venture subsidiary—not an SBIC—so AMEXCO regulatory status is an open-document question, not a source contradiction.
• Critical and negative coverage: direct testimony identifies Group losses, Collagen's failed first use, Reference Technology, and another uncertainly named storage failure. Marquardt questions process transferability; Draper's negative name search and Bossen's account prevent sole Dennis credit for Measurex; the LSI override complicates Dennis's consensus rule; LP tables show sharp fund dispersion; and the Applied Medical demand-registration episode qualifies unlimited-patience and founder-deference narratives. The Applied Medical opinion preserves allegations and procedural outcomes without converting them into a merits finding. Searches did not surface a substantiated personal criminal conviction or venture-related regulatory sanction, but that bounded result is not proof of universal absence.
• Historical context: inherited educational and trust access, an insurer balance sheet, inexpensive local startups, small early syndicates, defense and electronics demand, limited competition, tax changes, pension regulation, and a dense Stanford–San Francisco network are rival explanations to test alongside selection and governance skill.
• Completed-peer comparison: Doriot's ARD used a publicly traded institutional vehicle; Whitney deployed personal/family capital through a staffed firm; Rockefeller used inherited family capital and a family office; Dennis moved from trust-backed personal investing to an informal syndicate, corporate capital, then institutionally funded partnerships. This is a vehicle comparison, not a ranking or proof that any one person was “first.”
• Highest-value retrievals: (1) Ampex subscription, trust, ownership, and sale records; (2) The Group's member and company ledger; (3) Fireman's Fund/American Express committee minutes, AMEXCO charter, SBA license, and Recognition files; (4) IVA formation, LP, capital-account, fee/carry, valuation, and cash-flow records; (5) IVP I–II and complete Dennis-era LP statements; (6) board minutes assigning deal roles; (7) Berkeley's five interview videodiscs and Stanford's raw companion video; (8) original records for the 1970 and 1979 Stanford appearances, 2005 ETL lecture, 2007 IBF speech, 2008/09 cycle panel, 2009 GTS panel, 2010 SFAero award, 2012 Applied video, and planned 1992 forum; and (9) company files for losses, passes, recapitalizations, and exits.

Source notes