Reid Hoffman

1. Snapshot and testable edge thesis

This is a testable process thesis, not proof of alpha. The bounded screen is 164 relationships; only 81 received deep disposition, 71 were included, four meet strict reconstruction, and none supplies complete Hoffman-level cash flows. The counts come from the documented ledger; their interpretation is analyst work. [researcher inference · strong] count reconciliation

2. Core philosophy

Hoffman’s earliest explicit investment screen asks for mass distribution, a distinctive idea, and capital-efficient evidence; his later founder screen adds adaptability, references, and a coherent “theory of the game.” [investor-stated · strong] Rule of Three (numbered screen); YC founder lecture (10:02–15:06; 34:16–34:27). His scale doctrine is conditional: buy relative speed under uncertainty only when market structure rewards critical scale and survival capital can carry the inefficiency. [investor-stated · strong] CHM Blitzscaling (06:56–08:31; 18:20–18:25).

The synthesis is a loop: write a causal theory; test founders’ learning; map product and distribution together; finance the next uncertainty; use networks and boards for information/capability; revise the whole company when a fatal premise breaks; and distinguish risk-bearing actors before claiming value or return. [researcher inference · strong] Game On (“Why Games Are Serious Strategy”); Myths About Failure (thesis/failure passages); vehicle map.

The philosophy is not “networks and speed always win.” Hoffman calls network effects a bias rather than a universal gate, makes speed relative, and later adds harm, responsibility, runway, and scarce-capital constraints. [investor-stated · strong] 20VC (00:46–03:05); Bear Market Blitzscaling (capital-efficiency sections).

3. Sourcing

The public record shows relationship-led sourcing: PayPal/LinkedIn networks, founders and colleagues, references, prior operating contact, and Greylock’s institutional platform. Friendster led to the Facebook conflict route; James Slavet introduced Jeff Weiner; prior relationships appear in Nauto, Coda and Aurora accounts. Airbnb instead shows repeated trusted referrals and reference checks before meeting, not a long prior founder relationship. [retrospective witness · moderate] Part 2 note (14:29–16:08); Transformational Moments (02:01–15:43); Airbnb Reflections (referral/reference-check paragraphs); written theses.

The repeatable practice is to treat the network as an information channel, not as selection proof: log introducer, prior relationship, conflicts, missing cold channels, reference diversity, sponsor, IC roles, and who benefits. [researcher inference · strong] Airbnb shows a warm route can still produce disagreement; Neeva and SocialNet show access cannot solve distribution or product failure. Airbnb process; Neeva postmortem.

Repeatability is constrained by access and homophily. LinkedIn/PayPal founder status, Greylock’s brand, Microsoft governance, and a public media platform are institutional advantages unavailable to most analysts. No channel-by-channel funnel or demographic conversion table survives, so sourcing alpha is unproven. [researcher inference · insufficient public record] coverage boundary.

4. Picking

Founder/team. Test a founder’s capacity to learn faster than the problem changes: ask for a changed-mind example, a failure diagnosis, complementary-team evidence, integrity references, and the theory that organizes current action. [investor-stated · strong] How to Be a Great Founder (10:02–15:06; 34:16–34:27); How to Decide (decision-window/accountability sections).

Product/market/distribution. Identify the first user’s value, how the next user changes it, the channel/cold start, retained use, trust/safety costs, capture, and incumbent response. Early-stage metrics can be sparse, but causal interpretation cannot be. [researcher inference · strong] LinkedIn pitch (network/strategy slides); From Seed to Series A (18:03–21:14).

Competition. Seek markets where incumbents are structurally prevented from serving the new job, rather than merely large or slow. SpaceX is Hoffman’s retrospective changed-mind case; applying the rule to Neeva is analyst work because differentiation did not overcome defaults and acquisition economics. [researcher inference · moderate] Choosing Bad Competition (SpaceX passage); Neeva lab.

Diligence and rights. Public evidence is strongest for references, product use, partner debate, and board selection; it is thin on IC votes, original alternatives, check/security, ownership targets, and kill thresholds. A relationship, famous founder, narrative TAM, or later valuation is not a decision rule. [researcher inference · strong] people map.

5. Deal and portfolio mechanics

No complete public record establishes a universal Hoffman price, target ownership, reserve ratio, concentration limit, or sell discipline. The visible evidence instead supports stage-specific financing: the 2004 LinkedIn founder deck contains staged financing scenarios; Hoffman’s 2013 annotations add an A/B fallback; explicit Plan A/B/Z belongs to the 2012 Start-up of You/Permanent Beta framework and is restated in the later Startup Pivot essay. Other applications include Covid runway insurance, Blitzcapital for a critical-scale race, patient capital for science/commercialization, and SPAC sponsor/PIPE mechanics for public transactions. [investor-stated · strong] 2004 deck (financing slides); 2013 annotations (Plan A/B commentary); Permanent Beta (risk/permanent-beta passages); Startup Pivot (Plan A/B/Z framework); Calculating Risk (runway/insurance sections); Sunny Day Fund (investment proposal).

The practical rule is to price the next uncertainty: state security, entry ownership, dilution, reserves, next milestone, downside runway, follow-on trigger, and stop condition. [researcher inference · strong] The 71-row ledger shows why aggregate round sizes, IPOs, acquisitions, and issuer marks cannot stand in for Hoffman checks, proceeds, or returns. deal ledger.

Exit doctrine is optional rather than fixed: M&A and IPO are different capability/capital paths, while actual security-holder allocation and sale decisions remain mostly unobserved. [investor-stated · moderate] M&A or IPO (choice framework); uncertainty ledger.

6. Vehicle and incentive fit

[researcher inference · strong] Capital duration, fundraising pressure, decision rights, reserve capacity, liquidity, regulation, geographic authority, stakeholder objective, succession, and key-person exposure determine whether the same verbal thesis produces the same behavior. Blending these vehicles would make both performance and incentives incoherent. [researcher inference · insufficient public record] No complete public territory mandate or geography-specific decision-right schedule was recovered for Hoffman’s personal, Greylock, Discovery, Reinvent, Reprogrammed, foundation, or impact roles.

7. Team attribution and work with founders

The founder-service doctrine is support plus accountability: help recruit, connect, frame decisions and fill board capability gaps without seizing operating ownership. [investor-stated · strong] Board Part I (board-role/capability sections). Public intervention-outcome logs are sparse, so “value add” remains less proven than the doctrine. [researcher inference · insufficient public record]

8. Views on the craft

• Power law and outliers: Hoffman directly argues for funding more experiments; the bounded visible economic record is dominated by LinkedIn founder equity, but that is a researcher reading of a selected corpus, not a located Hoffman portfolio-construction rule. [investor-stated · moderate] Stimulus 2.0 (experiment argument). [researcher inference · insufficient public record] A Hoffman-specific power-law allocation view, audited hit rate, loss ratio and complete personal/fund denominator were not found. economic boundary.
• Access versus selection: relationships create information and access; learning velocity, product/distribution mechanism and independent partnership judgment are intended selection controls. Their incremental efficacy is unmeasured. [researcher inference · moderate] Airbnb Reflections (referral/partnership-debate paragraphs); founder screen (10:02–15:06; 34:16–34:27).
• Professional VC: repeat capital, board/recruiting networks and institutional memory can improve company building; compensation, carry, ordinary outcomes and conflicts must still be audited. [investor-stated · moderate] Professional VC (claimed advantages).
• Governance: early boards combine help and control; public boards emphasize fiduciary control. Stage, independence, conflicts, information rights and removal rules matter. [investor-stated · strong] Board Part II (00:34–04:43).
• Fund size: Discovery was reported as a $20 million pool with $25,000–$500,000 checks, while current Greylock 18 is a $1.5 billion early-stage fund with a stated 1–2-new-investments-per-partner annual model. [contemporaneous record · strong] Discovery (fund/check paragraphs); Greylock 18 (opening/model paragraphs). [researcher inference · insufficient public record] The record does not disclose Hoffman’s allocation, fund fees/carry, reserves, distributions or a causal fund-size/performance relation.
• Exits: Hoffman’s public doctrine compares M&A and IPO by capability, mission, talent and stakeholder paths rather than treating either label as success. [investor-stated · strong] M&A or IPO (¶¶12–18, 33–48). [researcher inference · strong] A listing, merger headline or company consideration is not a realized security-holder return; no complete sell/distribution discipline is public. outcome audit.
• Great investor: write a theory, find adaptive people, use networks to test it, preserve dissent, finance the next uncertainty, and revise on evidence. [researcher inference · strong] Game On (“Why Games Are Serious Strategy”); Airbnb case; decision-learning ranking. The record supports this vocabulary more strongly than personal alpha.

9. Signature bets and decision process

Use the no-hindsight cases before reading outcomes:

1. Aurora 2018: technical/team advantage, safety milestones, capital intensity, terms, and later vehicle separation.
2. Joby 2021: certification probability, manufacturing, runway, public-vehicle incentives, and dilution.
3. Hippo 2021: digital acquisition versus underwriting quality, sponsor/redemption mechanics, and adverse repricing.
4. Neeva 2021–2023: aligned-payer thesis, distribution/default friction, shutdown, and salvage.

[researcher inference · strong] The cases show one method cannot ignore vehicle or feedback latency: software distribution, autonomous safety, aerospace certification, insurance loss ratios, and public sponsor incentives require different milestones. Each worksheet freezes invest/pass, kill criteria, ownership/terms, and next evidence before the actual path.

10. Mental models and maxims

The reconciled field guide contains 14 supportable models; slogans and variants are merged rather than counted twice:

1. Write the theory of the game and expose it to reality.
2. Select learning velocity and complementary teams.
3. Map the network product and distribution loop.
4. Exploit constrained incumbents, not merely weak ones.
5. Buy speed only when critical scale pays.
6. Plan A/B/Z and finance across states.
7. Trade decision speed against reversibility, harm, and inaction.
8. Pivot the whole system when the fatal premise breaks.
9. Build boards as help, control, and succession systems.
10. Assemble ecosystems for missing capabilities.
11. Fence actor, vehicle, objective, and conflict.
12. Audit outcomes with denominators and cash flows.
13. Use alliances and network intelligence as reciprocal information systems.
14. Preserve exit optionality; do not confuse listing with return.

[researcher inference · strong] “Permanent beta,” Rule of Three, blitzscaling variants, network slogans and superagency are provenance or shorthand inside these models, not additional rules.

11. Evolution and contradictions

No public evidence proves a single continuous doctrine existed before it was written. Philosophy, PayPal and LinkedIn are dated antecedents; causal continuity is a researcher hypothesis, not biography fact. [researcher inference · strong]

12. Failures and limits

• SocialNet: retrospective accounts link delayed feedback and distribution weakness to later launch/learning doctrine; team records and loss economics are absent. [retrospective witness · strong] Candid transcript (SocialNet passage).
• Neeva: the consumer product failed despite differentiation; Snowflake’s $185.4m purchase is company consideration, not thesis validation or Hoffman proceeds. [contemporaneous record · strong] failure note.
• Hippo: early repricing and underwriting concern later coexist with improving issuer results; neither share price nor company earnings proves sponsor/personal return. [researcher inference · strong] Hippo case.
• Stripe pass / SpaceX changed mind: domain knowledge and incumbent assumptions can harden into blind spots. These are retrospective explanations, not contemporaneous memos. [retrospective witness · moderate] Stripe letter (“Reid’s response”); SpaceX passage.
• WeWork/Theranos: after-the-fact anti-cases bound blitzscaling but are not Hoffman portfolio losses. [researcher inference · strong] failure boundary.

The method is limited by winner-written doctrine, retrospective coherence, missing passes/ordinary cases, missing IC records, vehicle changes, no complete cash flows, and era dependence. The public record can support decision rules and falsifiers; it cannot establish audited alpha. [researcher inference · strong]

13. Historical operating environment

The earliest located company artifact is a 2004 consumer-internet fundraising deck that stages identity, distribution, trust, network proof and financing rather than assuming category certainty. It does not establish a market-wide cost or base-rate claim. [contemporaneous record · strong] LinkedIn deck (physical p.35 and financing slides).

In 2009 Hoffman argued that startup investment could counter the downturn; in 2010 reporting described a $20m Discovery seed pool with $25k–$500k checks. Those dated claims do not prove a market-wide 2010s capital-abundance thesis. [investor-stated · moderate] Stimulus 2.0 (startup-investment argument); Discovery (pool/check paragraphs).

The 2020–2021 Reinvent records show sponsor warrants, redemptions, PIPEs and public projections. A later SEC sample found average realized de-SPAC redemptions of 38% in 2020, 45% in 2021 and 85% in 2022; those retrospective market base rates are not Reinvent-specific performance. [contemporaneous record · strong] SEC 2024 final rule (p.402, Figure 2); Reinvent note.

Deep tech and AI differ from social software: certification, safety, science, manufacturing, energy, compute, privacy, competition law and public-interest harms lengthen feedback and raise irreversible costs. [researcher inference · strong] Aurora/Joby cases; AI chronology. Numerical VC hit-rate/base-rate comparisons are unavailable and are not invented.

14. Modern VC translation (as of 2026-08-27)

Every row is an analyst translation, not Hoffman doctrine.

[researcher inference · strong] Current-fund context is bounded: Greylock 18 is a $1.5 billion early-stage LP vehicle whose stated model is 1–2 new investments per partner per year. Greylock 18 (opening/model paragraphs); Form D (filer/3(c)(7)). [researcher inference · insufficient public record] The retained corpus does not establish a market-wide 2026 funding, valuation, loss, or exit base rate, so the translations below use company-specific evidence or say explicitly when the application is hypothetical.

Each row identifies its own grounded example or states that the 2026 application is hypothetical. None converts a modern company fact into Hoffman doctrine or return. [researcher inference · strong]

15. Comparison with completed peers

The closest defensible completed peers are David Sze (same firm and exact attribution controls), Ron Conway (relationship-led angel breadth), and Brad Feld (public doctrine, boards and vehicle experiments).

[researcher inference · moderate] The table is comparative researcher synthesis. Adjacent underlying sources establish the stated premises and public-record limits; they do not establish comparative alpha, rank, or causation.

[researcher inference · strong] The comparison suggests Hoffman’s distinctive public contribution is integration—network product, founder learning, conditional scale and public doctrine across roles—rather than a demonstrated advantage on ownership or realized fund economics. Underlying evidence is linked claim-locally above; completed peer packages provide navigation, not corroboration.

16. The strongest case against greatness

The skeptical thesis is that Hoffman’s visible reputation combines one extraordinary founder-equity event, privileged PayPal/LinkedIn access, Greylock’s institutional platform, partner and founder work, winner-selected writing, and later media coherence. LinkedIn’s gross merger schedule is the only large actor-linked economic anchor; without it, Facebook, Airbnb and Aurora cannot be ordered as personal returns. [researcher inference · strong] economic ranking.

Credit is easily overstated: Sze led Greylock’s LinkedIn/Facebook rounds; Slavet supported Airbnb; founders and teams built the companies; Reinvent, Reprogrammed, Aphorism, Microsoft and nonprofit capital have different owners and beneficiaries. [contemporaneous record · strong] people map. The 71 included deep-core relationships are neither a career nor performance denominator, and public profiles preferentially retain successes. [researcher inference · strong] coverage boundary.

Contradictions also matter: speed doctrine gained responsibility after criticism and anti-cases; Neeva challenges differentiated-product optimism; Stripe/SpaceX show expert priors can mislead; AI deployment advocacy overlaps investment and governance interests. [researcher inference · moderate] evolution; AI conflicts.

Evidence favoring the edge would be dated memos across winners, ordinary outcomes and failures showing the same mechanisms; a complete funnel; partner-role logs; stage/sector matched outcomes; and audited actor/vehicle cash flows. Evidence overturning it would show access/partner attribution explains selection, mechanisms were added after success, or ex-LinkedIn cash performance is ordinary. [researcher inference · strong]

17. Analyst study guide, glossary, and go deeper

Recommended path

1. Start with the final biography for chronology and causal boundaries.
2. Read WRITINGS, then TALKS to distinguish written doctrine from spoken application.
3. Work the Aurora, Joby, Hippo, and Neeva failure exercises before reading outcomes.
4. Use the mental-model field guide, then reread the skeptical case.

Glossary

Reusable IC checklists

• Sourcing: introducer; relationship/conflict; channels searched; missing populations; sponsor; decision-right owner.
• Diligence: theory/game; founder changed-mind evidence; reference diversity; product/distribution loop; constrained-incumbent proof; scenarios; fatal premise.
• IC: actor/vehicle/objective; security/ownership/dilution; reserve; relative-speed gate; harm/3S; dissent; kill/reopen milestone.
• Governance/follow-on: board capability matrix; help/control boundary; conflict/recusal; intervention log; next uncertainty; stop/pivot/leadership rule. Outcome: security/cost; realized proceeds; remaining position; company versus vehicle value; denominator; team attribution; counterfactual.

Best sources

Primary, direct, and official evidence — ranked

1. 2004 LinkedIn company deck — earliest located founder/company-planning artifact; it is not a Hoffman investment memo. The 2013 annotated publication is retrospective commentary.
2. How to Be a Great Founder — most compact direct founder/theory/reference screen (10:02–15:06; 34:16–34:27); it is doctrine, not an outcome test.
3. Airbnb Reflections — best named direct partnership-process account, with retrospective and self-attribution limits.
4. Game On — clearest direct intellectual mechanism for uncertainty and updating; examples do not establish investing efficacy.
5. HBR Blitzscaling — direct/coauthored conditional-speed doctrine (pp.4–8), not portfolio evidence.
6. Aurora S-4 — strongest official transaction/recusal record (“Background of the Business Combination — Hoffman’s role”); it does not reveal Greylock or personal returns.
7. Snowflake FY2024 10-K — strongest official Neeva outcome/economics boundary (business-combinations note); purchase consideration is not Hoffman proceeds.
8. SEC 2024 SPAC rule — strongest official market/base-rate control (p.402, Figure 2), not Reinvent-specific performance.

Independent secondary and analysis — ranked

1. Lee and Kim — strongest peer-reviewed large-sample challenge to early scaling (abstract; §§4.5–4.6); association and sample design do not adjudicate Hoffman's portfolio.
2. O’Reilly — strongest independent practitioner critique of monopoly, responsibility, and boundary conditions; it is argument, not matched performance evidence.
3. Wired 2012 profile — strongest independent narrative inventory for early angel attribution (“Hoffman’s fortunes as an investor”); amounts and holdings remain reported rather than transaction-ledger proof.

Pareto 80/20 — what drove the record

The public record does not support a literal return Pareto. None of the 71 included company/capital-relationship families has complete Hoffman-level cash flows; the 14-company set below is about 19.7% of that bounded denominator and is an evidence-weighted explanatory set, not a literal return share or permission to invent percentages. The set is LinkedIn, Friendster, Facebook, Airbnb, Blockstream, Xapo, Aurora, Joby, Hippo, Neeva, Change.org, OpenAI, Inflection AI, and Manas AI. Together they explain the transition from founder equity to personal angels, Greylock partnership capital, cryptocurrency, multi-vehicle/SPAC structures, failure, impact, foundation capital, and co-founded AI companies. [researcher inference · strong] Denominator/vehicle audit

Ranking 1 — economic contribution to the visible record

This ranks publicly observable economic significance, not personal net return. LinkedIn is the sole measured actor-linked anchor. The remaining cases are unordered directional tiers because round size, company value, fund holdings and sponsor securities are not Hoffman checks or proceeds. Every row states the observable measure, deserved attribution, support, mechanism, denominator caveat and ledger/case route.

The honest conclusion is concentration of visible significance, not measured economics. LinkedIn is the only large actor-linked gross transaction schedule; without it, Airbnb, Facebook and Aurora are plausible major contributors but cannot be ordered as returns. [researcher inference · strong]

Ranking 2 — historical importance

Historical rank is ordinal by demonstrated effect on Hoffman’s career model, not economics. Each row is one of 71 included records; all unlisted core and outer-screen candidates remain in the denominator and prevent success-selected generalization.

Ranking 3 — decision-learning value

Learning rank is ordinal by evidence completeness and transferability, not by outcome quality. Each mechanism is an analyst comparison; deserved actor credit and missing-field boundary remain explicit.

Evidence boundary and searched denominator

Fixed high-trust core

The deep core counts a relationship once when named in the pre-query processed direct-note/source bodies, original LinkedIn deck, current/legacy Greylock profile/master portfolio, Discovery reporting/issuer records, Wired’s 2012 profile, or Reinvent/SEC filings. A separate exact-name query layer screened—but did not deep-research—83 additional families. Queries combined "Reid Hoffman" with invest, investor, funding, portfolio, board, co-founder, Greylock, Discovery Fund, Reinvent, Reprogrammed, and exact surfaced names; the stopping rule was the terminal 2026-08-27 screen. Per-name routes for the outer 83 were not preserved or revalidated, so none enters evidence, rankings, reconstruction or performance analysis. The exact names are in the denominator note. [researcher inference · strong]

Career coverage is not computable. TechCrunch’s “60-plus” host characterization and Aurora’s “100-plus” issuer characterization are non-enumerated, incompatible claims. Even the 81/164 deep-research ratio is selection-biased and is not coverage of a career. TechCrunch (editorial introduction); Aurora 2023 proxy (director biography).

Known / unknown / disputed at decision date

• Known publicly: attributed actor/vehicle for the strongest cases; named round/transaction in filings or issuer records; direct thesis for LinkedIn, Airbnb, Blockstream, Change.org, Aurora, Coda, Nauto, Neeva, OpenAI, Inflection and Manas; sponsor terms for Aurora/Joby/Hippo. [researcher inference · strong]
• Unknown publicly: complete opportunity set, IC vote/dissent for most cases, check and security for most angels, ownership/dilution/reserves, follow-on policy, board rights, internal milestones/stop dates, transfers, distributions, fees/carry, taxes and net return. [researcher inference · strong]
• Disputed or source-sensitive: Facebook check amount; Inflection co-founder versus regulator chronology; OpenAI donation/investment timing and amount; TrialPay versus the live-profile TrialPlay typo; public-success narratives versus adverse/failed-company evidence. [researcher inference · strong]

Why the four-case reconstruction threshold is strict

“Material case evidence” is not enough. A case is reconstructed only when public evidence covers actor/vehicle, decision-era thesis/process, material security/transaction terms, milestone/governance path, and an observable outcome or current state. Aurora, Joby, Hippo and Neeva clear that bar; all still fail complete cash-flow reconstruction. Airbnb has unusually good decision-process evidence but lacks terms and lifecycle economics, so it remains unresolved under the same rule. [researcher inference · strong]

Ten exclusions and controls

Historical context and base rates

• 2002–2009, social-network formation: LinkedIn, Friendster, Facebook and the early web angels were made before mature social-platform playbooks. Identity, viral distribution, network density and trust were competing hypotheses; later ubiquity must not be projected backward. [researcher inference · moderate] LinkedIn deck (physical p.35); CHM Social Computing note (chronology/takeaways).
• 2009–2012, post-crisis seed institutionalization: Greylock reportedly placed a $20 million Discovery Fund inside Greylock XIII with reported $25,000–$500,000 checks. This converted a broad angel practice into a fund/partnership context but leaves fee, carry, reserve and loss data unavailable. [contemporaneous record · moderate] TechCrunch (fund/check paragraphs).
• 2014–2019, platform-to-deep-tech expansion: Blockstream/Xapo, autonomous systems, Coda, Entrepreneur First, Change.org and OpenAI widen the opportunity set and capital forms. The shift raises longer-duration technical, regulatory and governance risks than the early consumer-web portfolio. [researcher inference · moderate] Blockstream thesis (vehicle/thesis paragraphs); Aurora case; AI/impact case.
• 2020–2021, SPAC regime: Reinvent entered an unprecedented SPAC issuance wave in which the SEC highlighted sponsor compensation, conflicts, redemptions, PIPE terms and dilution. The base-rate comparison is therefore not a conventional Series A cohort. [contemporaneous record · strong] SEC 2021 statement; SEC 2020 guidance.
• The SEC’s later 2024 rulemaking sample reports average realized de-SPAC redemptions of 38% in 2020, 45% in 2021 and 85% in 2022. This is a market-period comparator, not Reinvent-specific performance and not interchangeable with other samples. [contemporaneous record · strong] SEC final rule (physical p.402, Figure 2).
• 2022–2026, AI formation and concentration: Inflection, OpenAI, Manas AI, Sanmai and related AI bets mix company formation, affiliated capital, boards and strategic counterparties. Technical progress and aggregate funding are not yet a realized-return base rate. [researcher inference · moderate] AI/impact case; Sanmai ledger row (#47).

No public source provides a complete seed/Series A cohort, loss ratio, write-off rate, DPI, TVPI, PME, reserve ratio or opportunity-cost ledger for Hoffman’s personal, Greylock, Discovery, Reinvent, Reprogrammed or foundation capital. Outcome frequency and hit rate therefore remain unmeasured. [researcher inference · strong]

Vehicle economics, incentives, and governance

Key incentive conclusion: the record is not one portfolio. A personal seed check, a Greylock preferred share, a Discovery micro-check, a SPAC sponsor warrant, a Reprogrammed affiliate stake, a foundation investment, fund backing, microlending support and a donation have different owners, horizons, downside, liquidity, governance and beneficiaries. Any blended “Hoffman return” would be economically incoherent. [researcher inference · strong]

Dimension-matched current-fund comparison. Greylock 18 is current LP-backed, $1.5b and early-stage; the firm describes a selective 1–2-new-investments-per-partner-per-year model. Discovery was reported as a $20m pool with $25k–$500k seed checks inside Greylock XIII. Personal angels use individual capital, while Reinvent SPACs used sponsor securities and finite search/de-SPAC mechanics. These differ in capital source, duration, approval rights, follow-on/reserve capacity, liquidity and succession exposure. Public evidence does not disclose Greylock/Discovery fees, carry, reserves, geographic mandates or key-person clauses, nor Hoffman’s personal/SPAC pecuniary allocations. [researcher inference · strong] Greylock 18 (opening/model paragraphs); Form D (filer/3(c)(7)); Discovery (fund/check paragraphs); SEC SPAC guidance (sponsor/PIPE/redemption/conflict headings).

People, capital, and influence map

Unknown by default: sourcing beyond named evidence, diligence owner, IC vote, negotiator, director/observer term, recruiting, reserve/follow-on decision, exit recommendation and narrator conflict. Silence is not attributed behavior. [researcher inference · strong]

<a id="detailed-deal-ledger--71-included-deep-core-families"></a>

Detailed deal ledger — 71 included deep-core families

Every row is one company family. Follow-ons, acquisitions, SPACs and affiliated vehicles remain separate events inside the row. Unless explicitly stated in a row, chronology/stage, capital vehicle, contemporaneous thesis, decision team, board/governance, follow-on/dilution, company status/outcome, retrospective account, and actor-level economics are not publicly established or not reconstructed. Check, ownership, reserves, proceeds and net return are never inferred from aggregate company/round values.

Ex-ante signature cases and no-hindsight worksheets

Aurora — 2018 Greylock preferred round, then separated later vehicles

Decision-date record. [contemporaneous record · strong] The S-1 fixes February 2018 terms: 29,948,750 Series A shares at $2.775672, including Greylock’s 12,272,330 shares for $34,063,962.77. Hoffman’s contemporaneous thesis emphasized the founders and a safe path to scale; the original IC memo, alternatives and vote remain private. Aurora S-1 (“Series A Convertible Preferred Stock Financing”); thesis note

Actual path, separately sourced. Greylock and Reinvent Capital participated again in 2019; a distinct Reinvent de-SPAC took Aurora public in 2021. The S-4 records Hoffman’s recusal from both sides’ transaction decisions. This reconstructs financing/governance, not personal return. [contemporaneous record · strong] Aurora announcement (opening investor paragraphs); S-4 (“Background…Hoffman’s role”)

Joby — 2021 public-market scale bet

Decision-date record. [contemporaneous record · strong] The transaction projected roughly $4.6b equity value and $1.6b gross cash and linked sponsor vesting/lockups to long-term price thresholds; company forecasts and a huge TAM were forward-looking. Internal sponsor alternatives, certification diligence and downside cases are not public. Joby Rule 425 (“Transaction Overview,” alignment and risk sections)

Actual path, separately sourced. As of 2026-07-31, Joby reported Stage 4 at 100% company/77% FAA and Stage 5 at 20%/10%, $2.264b cash/cash equivalents/short-term investments, and a $245.4m Q2 net loss; first passenger targets remain forward-looking. Reuters independently reported the first TIA certification-test aircraft flight while noting certification remained required. These are operating milestones, not sponsor return. [contemporaneous record · strong] Joby Q2 2026 exhibit (certification and liquidity sections); Reuters report (2026-03-11)

Hippo — 2021 de-SPAC adverse lab

Decision-date record. [contemporaneous record · strong] The filed structure exposed enterprise value, PIPE, redemptions, sponsor shares and performance vesting, while company projections supplied the operating case. Internal underwriting alternatives and sponsor downside analysis are absent. Hippo S-4/A (background, ownership and sponsor-summary sections)

Actual path. Forbes reported an early >50% share decline. By Q2 2026 Hippo reported $10.1m GAAP net income, $21m adjusted net income, 50.4% net loss ratio and 95.8% combined ratio; progress does not validate the original valuation or reveal sponsor return. [contemporaneous record · moderate] Forbes (performance/underwriting discussion); Hippo Q2 2026 (financial highlights)

Neeva — consumer-search failure and strategic salvage

Decision-date record. Ramaswamy stated the subscription/ad-free model; Hoffman added a “magic, surprise, and delight” and aligned-payer screen. Known: model, incumbent lock-ins and team. Unknown: original alternative set, IC dissent/vote, security/ownership, board rights, follow-ons, internal milestones and stop dates. Disputed/untested: whether paid product advantage could overcome default distribution. [contemporaneous record · strong] Full failure note

Actual path. Consumer search shut down; Snowflake acquired all equity interests for $185.4m cash. That is strategic salvage, not validation of the consumer thesis or a disclosed Greylock return. [contemporaneous record · strong] Snowflake 10-K (business-combinations note); postmortem

Portfolio-level forensics over the 71 included core

• Stage/check: the observable ledger is early-stage heavy, but missing stages and unlike structures make a numeric stage histogram misleading. Exact actor/vehicle checks are sparse beyond filed Aurora/Reinvent transactions; median check, entry ownership, dilution and reserve ratio are not computable. [researcher inference · strong] Ledger; Aurora case.
• Sector/time: the early social/consumer web cluster expands into crypto, autonomy, aerospace, insurance, AI, biotech and litigation capital. That is opportunity-set evolution, not evidence later sectors outperformed. [researcher inference · moderate] Wired (“Hoffman’s fortunes”); AI case; Reinvent case.
• Geography: evidence is US/Silicon-Valley heavy but includes UK/EU company formation and frontier-market payments; domicile, customer market and capital-source geography differ, so no complete allocation statistic is supported. [researcher inference · moderate] Entrepreneur First row (#19); Segovia/Taptap rows (#48/#55).
• Concentration/reserves: the 14-case set is explanatory only. Firm/personal/foundation/sponsor cash flows, reserve budgets and opportunity costs are absent, so DPI, TVPI, PME, hit rate and concentration are unmeasured. [researcher inference · strong] Denominator audit.
• Governance/exits: observer, director, chair, sponsor and cofounder roles recur but do not prove ownership. Company acquisitions/public listings are visible; liquidation preferences, investor proceeds and carry usually are not, and current/legacy profiles are success-selected. [researcher inference · strong] People map; outcome boundaries.
• Current LP-fund comparison: Greylock 18’s $1.5b, selective early-stage LP model differs dimensionally from Discovery’s reported $20m seed pool, personal angels and finite-duration SPAC sponsors; fees/carry, reserve authority, geography and key-person terms remain unknown. [researcher inference · strong] Greylock 18 (opening/model paragraphs); SEC Form D (filer/3(c)(7)); Discovery (fund/check paragraphs).

Failures, misses, passes, and process change

Neeva — full failure lab

• Initial belief: subscription search aligned user/customer incentives; product quality/privacy plus incumbent lock-ins created a wedge. [investor-stated · strong] Failure note
• Failure mechanism: product/model differentiation did not solve consumer switching/default distribution and acquisition economics at sufficient scale. [researcher inference · moderate] Failure note
• Stop/outcome: consumer product closed; Snowflake acquired all equity interests. Investor liquidation outcomes remain unknown. [contemporaneous record · strong] Failure note
• Process change: future incumbent challenges should require explicit distribution/default tests and paid-cohort thresholds, not only structural differentiation. This is a research prescription; public evidence does not show Greylock formally adopted it. [researcher inference · moderate] Failure note

Hippo — structure and underwriting anti-case

• Early adverse evidence showed market repricing and underwriting concern; later operating results improved. The lesson is to separate company performance, sponsor incentives and investor cash flows across time. [researcher inference · strong] Hippo case
• No public evidence establishes a Reinvent/Hoffman stop, sale, write-down, follow-on decision or realized return. [researcher inference · strong] Hippo case

SpaceX — initial pass, later invest

• Hoffman says he initially rejected SpaceX because he mapped auto/transport competition through incumbent strength, then later invested after recognizing an incumbent-misalignment opportunity. [retrospective witness · moderate] Choosing Bad Competition (“Elon Musk and Tesla”/SpaceX passage)
• Decision-date memo, timing, vehicle, check and actual belief-change evidence are absent. The retrospective lesson is to ask why incumbents underperform—not assume large incumbents are either unbeatable or inept. [researcher inference · moderate]

Stripe — documented 2012 pass

• In a 2023 first-person letter, Hoffman says he passed in 2012 despite admiring the Collisons because PayPal experience made him overweight regulatory friction, fraud/chargebacks and payments scale risk. He says he missed how YC/ecosystem volume could compound. [retrospective witness · strong] The Generalist (“Reid’s response,” paragraphs beginning “In 2012…” through “Sometimes knowing too much…”)
• This is strong for his later explanation, not contemporaneous proof of the 2012 deliberation. Price/valuation also appears in later reporting, but the direct letter’s process lesson is narrower: domain expertise can harden a once-correct model into a blind spot. [researcher inference · strong]
• Public evidence does not establish a formal Greylock process change, later opportunity, or quantified opportunity cost. [researcher inference · strong]

Anti-cases and failure-denominator limits

• Theranos and WeWork writings are after-the-fact doctrine tests, not portfolio failures. SocialNet is an operating failure, not investment loss. Convoy, Nanosolar, Tome and other adverse outcomes may affect the record, but public actor-level write-offs/recoveries and decision postmortems are incomplete. [researcher inference · strong] Theranos note; WeWork note
• Because firm profiles omit many quiet losses and the outer 83 were screened only, neither failure rate nor miss rate is computable. [researcher inference · strong]

Uncertainty and retrieval ledger

What would change the conclusions

• Audited actor/vehicle cash flows could replace the directional economic ranking with DPI/IRR/MOIC and a genuine concentration test.
• Original memos/IC records could materially reassign sourcing, dissent and decision credit and test whether later narratives match decision-date beliefs.
• Complete cap tables and distribution records could show that apparently large company outcomes were diluted, sold early, donated, held by a fund/foundation, or economically immaterial to Hoffman.
• A comprehensive failed/pass ledger could reverse the success-selected picture and permit stage/sector/base-rate comparison.

Until those records exist, the durable conclusion is methodological: Hoffman’s public record is unusually rich in narrative and unusually fragmented in economic ownership. The correct unit is the actor × vehicle × transaction, not the famous person × company name. [researcher inference · strong]

Coverage and evidence base

Literal priority-source reconciliation

The source map contains 45 priority rows in “Direct writings, books, and named investment theses.” Those rows resolve into 42 atomic named works plus three discovery records: the Reid Hoffman writing index, books index, and Greylock legacy blog index. Indexes are not essays, books, or independent evidence; they receive no source note and no content analysis.

The 33 full bodies are “Stimulus 2.0,” “My Rule of Three,” the annotated LinkedIn Series B pitch, the CEO-succession essay, the Blockstream thesis, “Managing Talent in a Networked Age, Part I,” Coda, Nauto, Aurora Series A, “In Praise of Professional Venture Capital,” “Is Blitzscaling Over?”, “VC Investing During Covid,” “Calculating Risk,” “Myths About Failure,” “The Startup Pivot,” Board Part I, Airbnb Reflections, “Congrats, Aurora,” the Inflection launch, “Bear Market Blitzscaling,” Impromptu, Change.org, Theranos, “Responsible Blitzscaling,” WeWork, “M&A or IPO?”, “Re-inventing the SPAC,” “Choosing Bad Competition,” “How to Decide,” “Sunny Day Fund,” “Blitzcapital,” “Building a better board of directors,” and “How We Wrote Blitzscaling.” The five partial records are The Alliance, official Chapter 1, pp. 1–20, Board Part II's three-paragraph introduction, the title/subtitle/date record for “Blitzscaling During Covid,” Entrepreneur First's official opening excerpt, and Callisto's official opening excerpt; the two written records remain partial while their audio bodies were separately acquired and timestamped in TALKS, and the two missing post bodies are not reconstructed from contemporaneous reporting. The exact inaccessible book bodies are Reid Hoffman and Greg Beato, Superagency (288 pages); Reid Hoffman with June Cohen and Deron Triff, Masters of Scale (304 pages); Reid Hoffman and Chris Yeh, Blitzscaling (336 pages); and Reid Hoffman and Ben Casnocha, The Startup of You, revised 2022 body. Their official publisher descriptions, guide/edition material, authors, dates, and access barriers are recorded, but this document makes no claim to have read those four books.

This is a coherent high-priority slice, not a lifetime-writing denominator. The final index-level triage count is 279 dated records, spanning 2012-10-01 through 2025-06-15 on the live writing index; all 279 were treated as discovery metadata, not as 279 accessed content bodies. The ledger below assigns every zero-based index row once: 27 selected + 45 derivative/repost + 51 routed to TALKS + 17 metadata/dead/inaccessible/duplicate + 139 topical/political/promotional/general leadership = 279. A reproducible rule governs this run: retain distinct ex-ante allocation or vehicle doctrine, named-deal theses, failure/anti-cases, and governance decisions, plus the minimum dated texts needed to test material evolution in those rules. Derivative summaries, reposts, generic scaling/leadership essays without a distinct decision rule, audio/video beyond exposed written introductions, index pages, non-Hoffman posts unless required for role formation or firsthand intellectual-history evidence, and political, promotional, cultural, or topical commentary were excluded or routed to TALKS. An archive record and its canonical full body count once; a derivative that merely repeats a processed book/post does not become an independent work or corroborating source. The 27 selected index rows plus 15 deduplicated external/book works produce the 42 atomic-work denominator above. The ledger also records the eight-candidate audit: Building a better board and How We Wrote Blitzscaling were retained after body-level inspection; six overlapping or out-of-scope candidates were named and excluded or routed. Jeff Markowitz’s Part II was read as an attribution check but excluded because it is not Hoffman-authored; it is not independent validation of Part I.

Writing-index triage ledger

The row IDs are zero-based positions in the live index extraction inspected 2026-08-27. Every row is assigned exactly once after title, date, route/format, and apparent-authorship inspection; canonical bodies were opened for retained candidates.

The eight-candidate boundary audit retained Building a better board of directors for its distinct capability-matrix and executive-information rule and How We Wrote Blitzscaling for its distinct intellectual-history evidence. Six were excluded or routed: Competitive Strategy is an edited oral capsule for TALKS and overlaps Choosing Bad Competition/WeWork; Scalable Culture is downstream hiring/culture operations without a distinct decision rule; Competitive Edge overlaps Choosing Bad Competition; Beyond Startups repeats application boundaries already covered by Is Blitzscaling Over?, Bear Market Blitzscaling, and Responsible Blitzscaling; Founder to Leader is a co-authored delegation/succession excerpt overlapping Professional CEO; and Microsoft Board is a role announcement for which SEC filings are superior.

Evidence and attribution rules

[researcher inference · strong] Direct prose establishes what an author argued at a date; it does not establish consistent practice, sole decision credit, vehicle ownership, realized economics, or causal skill. First-person retrospectives are separated from decision-era records. Named Greylock theses establish firm investment and public sponsorship only to the extent stated; founder equity, personal angels, Greylock funds, Reinvent sponsorship, Reprogrammed/family-affiliated capital, philanthropy, and board-only roles remain separate throughout.

[researcher inference · strong] Co-authorship is substantive evidence, not a footnote. The Startup of You is with Ben Casnocha; The Alliance with Casnocha and Chris Yeh; Blitzscaling with Yeh; Masters of Scale with June Cohen and Deron Triff; and Superagency with Greg Beato. Mustafa Suleyman—not Hoffman—authored the Inflection launch, while “How We Wrote Blitzscaling” is explicitly Chris Yeh's guest account despite its Hoffman-profile host/byline. Quotes below are verified against accessed originals, contain no more than 25 words, and use a pinpoint locator. Per-note metadata records access dates from 2026-08-24 through 2026-08-27.

Source-by-source analysis

1. Stimulus 2.0 — widen the experiment pool during crisis

Context and argument. [investor-stated · strong] In March 2009, while identified as LinkedIn CEO/founder and before joining Greylock, Hoffman argued that recession did not eliminate good ideas. He proposed small-business credit, immigration reform, and as much as $100 million of public matching for qualifying venture and angel investments, accepting higher failures to increase the chance of funding eventual winners (proposals 1–3). source note

Reasoning, assumptions, and implication. [investor-stated · strong] The causal chain is more entrepreneurial experiments → more products and services → more durable jobs. The useful decision rule is countercyclical search: investor liquidity preference can move faster than the quality of invention. [researcher inference · strong] But a stable “good idea” proportion, net job creation, and program returns are asserted rather than demonstrated. Public matching also creates adverse-selection and governance risks because private managers keep normal economics plus part of the public upside while taxpayers share loss (proposal 3).

Biography and boundary. [contemporaneous record · moderate] TechCrunch’s more-than-60-venture characterization is [single-source], appears again in its Rule of Three introduction, and provides no list, vehicle split, date boundary, or outcome ledger. It is neither independent corroboration nor a personal portfolio denominator (editor’s note).

2. My Rule of Three — an early consumer-internet screen

Context and argument. [investor-stated · strong] One month later, Hoffman compressed late personal-angel practice into three questions: can the product reach millions, is it categorically distinct, and can it scale with capital efficiency? “In consumer Internet, think ‘distribution, distribution, distribution’” (criterion 1; 7 words). source note

Reasoning and operating rule. [investor-stated · strong] Differentiation earns attention, distribution converts attention into mass adoption, and financeability preserves the option to learn. Capital efficiency includes a software cost curve and the ability to reach evidence before the next financing. [researcher inference · moderate] “Millions” is a dated Web 2.0 consumer threshold, not a portable enterprise, biotech, or autonomous-systems metric. A modern screen must substitute the relevant unit—seats, workflows, fleets, safety miles, validated targets, or regulated approvals (criteria 1–3).

Evidence and evolution. [contemporaneous record · moderate] The same [single-source] 60-plus characterization has no failures, passes, or vehicle map (introduction). [investor-stated · strong] The 2022 bear-market essay retains financeability but permits speed-over-efficiency when time creates a durable lead. [researcher inference · strong] The two positions reconcile only if competition, learning evidence, and a spending stop rule are specified ex ante.

3. LinkedIn’s Series B Pitch — networks plus staged proof

Context and argument. [retrospective witness · strong] Hoffman’s 2013 annotations surround LinkedIn’s 2004 deck and explicitly recall a company without substantial organic growth, revenue, or an obvious category lead. [investor-stated · strong] The case was that a professional-identity network could become a trusted people-search and recruiting layer, then a broader platform (Introduction; Slides 2–10). source note

Method. [investor-stated · strong] The annotations recommend a clear investment thesis, proactive risk disclosure, bottom-up product evidence, and a financing sequence in which each round proves the next uncertainty. “Begin and end with the investment thesis” (Slide 37, Advice; 7 words). Analogy should orient the audience, not substitute for reasoning; team roles and risks should be explicit; one committed champion can matter more than diffuse approval (Slides 12, 19, 21–22, 34, 37).

Decision and attribution. [contemporaneous record · strong] The underlying deck is ex-ante founder fundraising evidence, while the advice and six-part thesis are hindsight. LinkedIn’s later S-1 attributes Greylock’s position and board representation to David Sze and says Hoffman lacked voting or dispositive power over Greylock’s shares (LinkedIn S-1, principal-stockholder footnotes). The work therefore cannot be converted into a Hoffman-led Greylock deal. It also names an early team rather than supporting sole-founder execution credit (Slide 34).

4. The Startup of You — career as a revisable venture (body inaccessible)

Access and argument boundary. [contemporaneous record · moderate] Penguin Random House identifies Hoffman and Ben Casnocha, retains the original February 14, 2012 product date, and markets a revised edition; Casnocha’s dated note says the revision appeared in May 2022 with post-pandemic material (publisher description · Casnocha edition note, ¶¶1–2). [researcher inference · strong] Neither full body was acquired. source note

Verified theme map. [contemporaneous record · provisional] The publisher synopsis describes adaptability, differentiated assets, professional networks, intelligent risk, engineered serendipity, and network intelligence (publisher description). [researcher inference · provisional] Together these imply a career portfolio of hypotheses: build assets, seek distributed information, take bounded risks, and revise. That inference is deliberately weaker than a chapter analysis; no edition-level textual evolution is claimed.

Counterweight. [researcher inference · moderate] A startup analogy can increase agency while individualizing structural exposure to health, caregiving, immigration, discrimination, geography, and macro shocks. Whether the authors address those limits cannot be judged from the available publisher synopsis.

5. The Alliance — explicit reciprocity in an unstable labor market

Access and central argument. [contemporaneous record · strong] The official 20-page Chapter 1 excerpt is the only book-body partial in this phase. [investor-stated · strong] Hoffman, Ben Casnocha, and Chris Yeh argue that at-will employment combined with indefinite loyalty language creates reciprocal distrust; their alternative is “a mutually beneficial deal, with explicit terms, between independent players” (Chapter 1, p. 8; 10 words). source note

Reasoning and examples. [investor-stated · strong] A company should invest in an employee’s market value while the employee invests in the company’s mission; both state what they can keep. A team analogy clarifies performance and transition better than a literal family, while compassion, trust, and mutual benefit should remain (pp. 8–12). The chapter uses Disney/Pixar and Amazon Web Services as stories about suppressing or enabling entrepreneurial employees (pp. 14–18).

Implication and limit. [researcher inference · strong] A decision-grade alliance needs a mission, horizon, reciprocal contributions, checkpoints, transition rule, confidentiality boundary, and remedy. The “independent players” frame understates employer power over pay, information, evaluation, visa status, and termination. Chapters 2–8, implementation appendices, and conclusion were not accessed, so tours-of-duty mechanics are not reconstructed from derivative summaries (official Chapter 1, pp. 8–18).

6. Managing Talent in a Networked Age, Part I — portfolio application, not corroboration

Context and argument. [investor-stated · strong] Hoffman applies the co-authored Alliance framework to Greylock portfolio advice: replace assumed permanence with explicit tours of duty whose mission and horizon advance company and employee, and discuss entrepreneurial ambition honestly (¶¶1–7). source note

Decision relevance. [contemporaneous record · strong] He says he and Greylock talent partner Jeff Markowitz used these principles with portfolio companies (¶2). That connects network philosophy to post-investment talent work but supplies no named decision, cohort, retention result, or dispute outcome. Part II is Markowitz-authored and therefore excluded from the Hoffman corpus; it illustrates claimed Greylock practice without independently validating Part I.

Contradiction. [researcher inference · moderate] Encouraging external networking can increase learning and honest retention or accelerate departures, conflicts, and information leakage. Treat the alliance as a governance contract, not a cultural slogan, and test company and employee outcomes separately (¶¶1–7).

7. Our Investment in Nauto — a data flywheel with an immediate wedge

Context and thesis. [investor-stated · strong] In a July 2017 named Greylock thesis, Hoffman says driving will become networked. Nauto’s dual-camera AI platform was intended to deliver immediate value to fleets, drivers, and insurers while accumulating data for autonomy and human-machine adaptation (body ¶¶1–6). [contemporaneous record · strong] Greylock co-led the Series B and Hoffman joined the board. source note

Reasoning and decision evidence. [contemporaneous record · strong] The founder case rests partly on a relationship with Stefan Heck dating to 1988 and on systems-thinking reputation. [researcher inference · strong] The scalable wedge pattern is portable: solve a current paid problem while generating evidence for a larger system. But one million reported miles [single-source] is an exposure count, not proof of safety, causal model improvement, lawful data advantage, or defensibility (body ¶¶4–6).

Evolution and limit. [researcher inference · strong] The network metaphor has migrated from people to vehicles, but physical reliability, hardware margin, liability, insurer incentives, privacy, and regulation make the underwriting qualitatively different. The post supplies no price, ownership, burn, error rates, or independent fleet cohort (body ¶¶1–6).

8. Our Investment in Coda — networked productivity and experiential diligence

Context and thesis. [investor-stated · strong] Hoffman’s October 2017 Greylock announcement argues that fragmented documents, spreadsheets, applications, and communication should converge into one programmable collaborative surface. [contemporaneous record · strong] Greylock led Coda’s Series A, Hoffman joined the board, and the firm used the beta internally (body ¶¶1–6). source note

Reasoning and implication. [investor-stated · strong] Live data, application logic, and collaboration could create a shared system and perhaps network effects as teams build and reuse workflows. Shishir Mehrotra’s and Alex DeNeui’s operating experience and Greylock’s own use serve as founder and product evidence. [researcher inference · strong] Internal use is valuable only if compared with external retention, switching, workflow depth, permissions, and competition; otherwise familiarity can become confirmation (body ¶¶3–6).

Boundary. [researcher inference · strong] The post establishes a Greylock investment, public sponsor, and board role—not personal capital, sole approval, entry ownership, or return. It gives no customer count, retention, pricing, margin, competitive win/loss, price, or downside case, and “network effects” may mean collaboration value rather than durable cross-user defensibility (body ¶¶3–6).

9. Our Investment in Aurora — partnership as the scale architecture

Context and thesis. [investor-stated · strong] In February 2018, Hoffman argued that autonomous vehicles could form a fleet-learning network, while safety and manufacturing made partnership central. Aurora would supply a full autonomy stack across manufacturers instead of building every vehicle. [contemporaneous record · strong] Greylock co-led the Series A with Index, and Hoffman joined the board (body ¶¶2–9). source note

Reasoning and record. [contemporaneous record · strong] The founder case combines Chris Urmson’s prior relationship, technical-team pedigree, and partnership orientation. Aurora’s S-1 independently records the February 2018 Greylock financing and the transfer mechanics of Hoffman's director grant to Greylock-related entities (Aurora S-1, “Certain Relationships” and director sections). This is formal financing/governance corroboration, not a personal-return calculation.

Operational implication and tension. [researcher inference · strong] Partner-embedded scale needs precommitted technical safety, simulation-to-road transfer, data rights, partner concentration, integration, capital, and regulatory milestones. Partnership can distribute faster or create dependency and bargaining constraints. The essay acknowledges a long replacement horizon but provides no intervention-rate, safety, burn, valuation, or OEM concentration evidence (body ¶¶2–9).

10. Blitzscaling — publisher-level access, independently testable doctrine

Access and argument boundary. [contemporaneous record · moderate] Penguin Random House identifies Hoffman and Chris Yeh, October 9, 2018, and 336 pages. Its description presents a rapid startup-to-scaleup method whose business model, hiring, founder role, and culture change by stage; [researcher inference · strong] the body was not acquired (publisher description). source note

What can be supported. [researcher inference · strong] Detailed doctrine comes from the separately accessible 2016 HBR text and 2022 update, not reconstructed book chapters. Tim O’Reilly argues that the method can mistake survivor stories and subsidized market power for sustainable business, while conceding the case where market size, durable advantage, distribution, gross margin, and competition really make speed decisive (O’Reilly, “The case for blitzscaling” and “Why I’m skeptical”).

Empirical counterweight. [contemporaneous record · moderate] Lee and Kim report early scaling associated with higher failure and no systematic successful-exit benefit, with a stronger platform association; experimentation via A/B testing attenuates some failure association. They disclose survival, job-posting measurement, and omitted-variable limits. [researcher inference · moderate] This is evidence against a universal timing rule, not a causal verdict on each case (abstract; §§4.5–4.6; Tables 5–6).

11. Airbnb Reflections — updating after a miss and deciding through disagreement

Context and argument. [retrospective witness · strong] Published at Airbnb’s 2020 public debut, Hoffman recounts failing to pursue an early “couches” introduction, updating after repeated trusted referrals, conducting reference checks before meeting, and moving quickly. He calls Airbnb his first Greylock investment; David Sze remained skeptical, James Slavet supported it, and the partnership proceeded (“The Introduction,” “The Meeting,” and “Greylock’s Investment”). [retrospective witness · moderate] Greylock's current Hoffman profile separately says, “Reid was a board observer for Airbnb,” rather than describing a director seat (role-summary paragraph; accessed 2026-08-27). source note

Decision rule. [retrospective witness · strong] Independent, trusted repetition should raise diligence priority, not lower the diligence standard. Hoffman describes a preference for asymmetric cases: “could be zero, but if they succeed, could be huge” (“Greylock’s Investment”; 10 words excerpted). Founder design practice and crisis adaptation are treated as evidence of product and organizational learning.

Attribution and bias. [researcher inference · strong] The account properly preserves Slavet support, Sze skepticism, referrer access, and partnership approval; Greylock's current profile supplies the separate observer-status boundary (role-summary paragraph beginning “Reid was a board observer for Airbnb”). Publication at a winner milestone creates hindsight and selection bias; there is no original memo, price, ownership, reserve plan, losing-marketplace denominator, or Hoffman/firm realized-return series (“No Strangers” through “Think Different”).

12. Masters of Scale — a theme map, not an analyzed book body

Access and corpus. [contemporaneous record · moderate] Penguin Random House names Reid Hoffman with June Cohen and Deron Triff, a September 7, 2021 date, and 304 pages. The publisher says the book synthesizes more than 100 interviews [single-source] plus unaired material; the official reader’s guide surfaces idea selection, “squirmy no,” 100 loved users, culture, fire triage, learning, customer feedback, Plan B, and value/profit alignment (description · reader guide, questions 2–10). source note

What follows—and what does not. [contemporaneous record · provisional] The accessible map complicates scale-at-all-costs: love before reach, continuous learning, selective fire-fighting, and fallback plans are explicit guide topics (publisher description and reader's guide). [researcher inference · strong] The more-than-100 count describes an editorial interview pool, not an investment denominator or independent sample. The body is inaccessible, and book, podcast, and publisher guide are related products rather than corroborating sources.

Selection concern. [researcher inference · strong] A corpus centered on successful leaders is exposed to outcome selection and retrospective coherence. The later phase must acquire the body, enumerate guests and failed cases, and separate guest testimony, Hoffman argument, and Cohen/Triff editorial synthesis (publisher description and reader's guide).

13. Congrats, Aurora — a retrospective four-gate risk review

Context and argument. [retrospective witness · strong] At Aurora’s November 2021 public debut, Hoffman revisits Greylock’s Series A, later technology, acquisitions, partnerships, and a Reinvent-sponsored transaction. He groups the original uncertainty into technology, regulation, business model, and scale (“A Long-Term Partnership” and “Critiques”). source note

Use and boundary. [retrospective witness · strong] Four separate gates are a useful long-duration review structure. Simulation, sensors, OEM/trucking ties, and the Uber ATG acquisition are presented as adaptation evidence, but the essay provides no independent safety or commercial threshold. Aurora’s S-1 controls financing and securities facts (Aurora S-1).

Vehicle fence. [researcher inference · strong] Greylock investment, Hoffman director work, and Reinvent sponsorship have different capital, fiduciaries, dates, and economics. A listing and announced valuation are market events—not return, realized proceeds, or proof of mass autonomy. A passenger-to-trucking emphasis and acquisitions may be intelligent updating or evidence the original path was incomplete; the retrospective cannot decide alone (“Long-Term Partnership” and “Critiques of Autonomous Driving” · Aurora S-1, “Certain Relationships”).

14. Inflection launch — formation evidence, not Hoffman-authored doctrine

Context and attribution. [contemporaneous record · strong] Mustafa Suleyman’s March 2022 announcement says he and Hoffman were co-founding Inflection AI, names Karén Simonyan as co-founder/chief scientist, and says the company was incubated at Greylock (¶¶1–3). source note

Evidence boundary. [contemporaneous record · strong] Suleyman—not Hoffman—presents a consumer natural-language-interface thesis. The post establishes an explicit public co-founder label and launch context, not Hoffman's selection reasoning, corporate title, check, ownership, or day-to-day work. The CMA’s narrower transaction chronology foregrounds Suleyman and Simonyan and identifies Hoffman as a board member; that scope difference does not negate the launch label (CMA full decision).

Governance consequence. [contemporaneous record · strong] Microsoft’s 2024 proxy says Reprogrammed and Greylock each held less than 10% [single-source], Hoffman could be deemed a beneficial owner through relationships, and he recused from Microsoft board review of the licensing/hiring transaction (proxy, related-person transaction disclosure). [researcher inference · strong] Those percentages are filing disclosures, not a personal cash-return calculation, and the transaction does not validate the original consumer-company thesis.

15. Bear Market Blitzscaling — speed becomes a conditional variable

Context and argument. [investor-stated · strong] In December 2022, Hoffman restates blitzscaling as speed over efficiency under uncertainty but says scarce capital lowers the affordable inefficiency. The relevant comparison is competitive: spend for speed only when delay destroys an advantage; otherwise preserve runway, revenue proof, and the next financing milestone (“What Is Blitzscaling?” and “The Investment Thesis”). source note

Operating method. [investor-stated · strong] Begin with defense—cash, fewer simultaneous risks, revenue/profitability path—then use competitor weakness for selective offense. A-plan/B-plan/Z-plan logic extends to founders and careers; first-time founders without a network face different opportunity sets from repeat founders (“Defense and Offense” and career sections).

Contradiction and evidence. [retrospective witness · moderate] PayPal and LinkedIn are survivor examples, a bias Hoffman acknowledges. O’Reilly’s critique and Lee/Kim’s study show why “competition” is not enough: product-market learning, capital transmission, unit economics, externalities, and experiment quality require explicit thresholds (O’Reilly · Lee and Kim, abstract and §§4.5–4.6).

16. Impromptu — agency through iterative AI deployment

Context and central argument. [investor-stated · strong] Hoffman calls the 247-page GPT-4 collaboration a dated travelogue rather than a definitive account. Across education, creativity, justice, journalism, social media, work, hallucinations, and “Homo Techne,” he argues that conversational AI can amplify human agency when people remain active users and judges (full book PDF, Introduction, Contents, and Conclusion). source note

Reasoning and risk. [investor-stated · strong] The book distinguishes statistical prediction from consciousness and discusses toxicity, fabricated facts, and imperfect guardrails. Its bias is toward deployment, participation, scrutiny, and rapid correction rather than waiting for zero risk. “For Homo techne, utopia is a direction, not a destination; a process, not an outcome” (physical PDF p. 232 / printed p. 225; 15 words).

Interest and implication. [investor-stated · strong] Hoffman discloses early OpenAI funding and Microsoft directorship while describing GPT-4 access. [researcher inference · strong] The text extends distribution logic from connecting people to distributing cognitive capability, but it is an interested demonstration, not independent model evaluation or an investment memo. A decision translation must separately underwrite model capability, user benefit, error harm, distribution, concentration, governance, and role conflicts (Introduction and Conclusion).

17. Superagency — a later collective-agency claim (body inaccessible)

Access and thesis boundary. [contemporaneous record · moderate] Simon & Schuster identifies Hoffman and Greg Beato, Authors Equity, January 28, 2025, and 288 pages. The publisher synopsis markets “superagency” as widespread empowerment when millions access AI, with examples in education, medicine, and personal navigation; [researcher inference · strong] the body was not acquired (“About the Book,” “About the Authors,” and “Product Details”). source note

Provisional evolution. [researcher inference · provisional] Publisher material suggests a shift from Impromptu’s individual co-pilot demonstration to a social diffusion claim in which users benefit from one another’s capabilities. This cannot establish the book’s reasoning, evidence, critics, or falsifiers. The page foregrounds upside and promotional endorsements; it is not independent evaluation of disinformation, labor displacement, concentration, safety, or unequal access (publisher “About The Book”).

Attribution and current-record warning. [researcher inference · strong] The co-authorship prevents Hoffman-only credit. Its live marketing biography also cannot control time-sensitive board status or precise investment roles; those require filings. No book-body quotation or unseen argument is used here (publisher author and publication metadata).

18. If, Why, and How Founders Should Hire a “Professional” CEO — succession as network rewiring

Context and argument. [investor-stated · strong] In January 2013, Hoffman argued that founder permanence and automatic “adult supervision” are both inadequate rules. The founding CEO should compare passion and capability with the actual growth-stage job; a successor must own product as well as process, arrive early enough to shape the company, build trust with founders, and receive unambiguous authority (“How do I know when to replace myself?” and “How do we make the transition?”). source note

Evidence and operating method. [retrospective witness · strong] Hoffman says he initiated the LinkedIn search, David Sze challenged whether a COO would suffice, Dan Nye built organizational systems, James Slavet introduced Jeff Weiner, and Weiner combined scaled management with product ownership. “My mistake was thinking you could divorce product strategy from the CEO role” (same section; 13 words). Hoffman describes physically stepping away so internal decisions would route through Weiner, then returning as a candid but non-undermining adviser.

Boundary and contradiction. [researcher inference · strong] “Founder” as a state of mind can foster ownership but must not overwrite legal formation, equity, title, or historical credit. LinkedIn is a selected winner, the founder-control studies cited do not establish that surrendering control caused more value, and the essay provides no failed-transition denominator or contemporaneous board scorecard (“How to Tell When Your Company Needs a New CEO” and “Founder as a State of Mind”).

19. Why I Invested in Blockstream — a personal investment with public-good priority

Context and argument. [investor-stated · strong] In the November 2014 thesis, Hoffman treats Bitcoin as an open protocol and shared ledger whose network could support programmable asset exchange. Blockstream's sidechains were intended to add functionality without fragmenting the common network, conditional on core-developer and community adoption (¶¶1–14). source note

Vehicle and objective. [contemporaneous record · strong] Hoffman explicitly says he joined Blockstream's first round as an individual after Greylock discussion and prioritized the open ecosystem/public good; the same essay separately says he led Greylock's Xapo investment (¶¶10, 15–16). No check, ownership, security, voting right, or return is disclosed. Personal Blockstream capital and Greylock Xapo capital therefore remain separate.

Decision implication and risk. [researcher inference · strong] A protocol investment memo should name the capital owner, public-good covenant, change-governance process, value-capture path, and adoption/safety thresholds. The essay overstates fraud elimination and exchange assurance by underweighting custody, keys, implementation, oracles, governance, and regulation. Ecosystem value can also accrue away from the company (¶¶10, 15–16).

20. In Praise of Professional Venture Capital — an interested institutional defense

Context and argument. [investor-stated · strong] Hoffman's December 2019 essay defends professional VC as a flawed institution whose deploy-and-distribute incentive can align capital with company value, broaden allocation beyond inherited family networks, and add repeated board practice. It contrasts that model with politically constrained government allocation and concentrated family wealth (opening; “Option 1”; “Option 2”). source note

Portable version. [researcher inference · strong] The useful claim is narrower: prefer an allocator when access is contestable, incentives are disclosed, decision rights are legible, and selection, intervention, performance, and harms can be audited. Cash, network access, and board help should be separately valued; a public-company winner list does not show marginal VC causality (“The Efficiency of Professional Venture Capital” and “Accountability”).

Contradiction. [researcher inference · strong] The essay understates bootstrapping, customers, grants, bank and revenue-based credit, angels, corporate capital, and public funds. Professional VC can reproduce its own school, geography, race, gender, and network gates; exit alignment does not automatically align investors with founders, workers, customers, or the public (“The Problem with Alternative Sources of Capital” and “Why Venture Capital is Better”).

21. Is Blitzscaling Over? — relative speed and an explicit anti-case

Context and argument. [investor-stated · strong] Hoffman and Chris Yeh's May 2020 essay preserves blitzscaling only where a winner-take-most structure rewards early scale. Covid can reduce absolute growth while a survivor gains relative share; WeWork is named as a misapplication because its spending did not create that market structure (¶¶3–5). source note

Ex-ante test. [contemporaneous record · strong] The authors predicted that Airbnb's two-sided network would strengthen relative to hospitality competitors after the pandemic (¶¶6–16). “What matters is that you win the race, not whether or not you set a new world record” (¶4; 18 words). This is a dated, testable market-share prediction, not a disclosed investment or proof that the general method works.

Boundary. [researcher inference · strong] Relative share can improve while customers, workers, suppliers, or shareholders lose. The underwriting must specify the share metric, persistence mechanism, cash and unit-economics limit, stakeholder harm, and stop rule; “network effects” cannot be accepted as a label (¶¶3–16).

22. Blitzscaling During Covid — written denominator preserved, audio separately completed

Access result. [contemporaneous record · strong] Hoffman's archive identifies a June 18, 2020 item, “Blitzscaling During Covid: Growth Strategies Amid Turbulence,” associated with Hoffman and Chris Yeh (official landing page · dated archive entry). source note

Hard boundary. [researcher inference · strong] The live page exposes only title/subtitle; its Greylock legacy slug now resolves to the firm homepage, and no stable written transcript was reached. TALKS lawfully acquired the official RSS audio and completed timestamped analysis in the oral-source note. WRITINGS remains partial and infers no defense, offense, runway, growth, or company claim from metadata or adjacent 2020/2022 essays.

23. VC Investing During Covid — diligence and runway adapt to a missing channel

Context and argument. [investor-stated · strong] In July 2020, Hoffman said Greylock continued investing while spending more remote time with founders, adding calls with non-founder team members, widening geography, and adjusting sector hypotheses. He reported funding companies for 36 months rather than the traditional 18–24 months [single-source] to protect building and early-customer validation (¶¶1–7). source note

Attribution and method. [contemporaneous record · strong] Saam Motamedi supplies several Greylock-practice examples and Chris Yeh moderates the underlying conversation; this is not Hoffman-only action. The reusable rule is to recreate missing information channels—more time, broader team contact, longer financing against the dominant risk—rather than silently waive diligence (¶¶1–7).

Counterweight. [researcher inference · strong] Remote access may widen sourcing while privileging founders who perform well on video and references. More runway may reduce financing risk or increase dilution, price error, burn tolerance, and the cost of a wrong thesis. The post names no cohort, partner leads, checks, outcomes, or matched pre-pandemic control (¶¶1–7).

24. Calculating Risk — decision windows plus severity, scope, and systematicity

Context and argument. [investor-stated · strong] Hoffman and Chris Yeh argue that crisis creates less-contested opportunity only for companies that can survive it. Decisions should default to a short window, with delay purchased only by questions capable of changing the choice; high-impact risk should be assessed by severity, scope, and systematicity (“The Upside of Risky Times,” “Defining Decision Windows,” and “Managing The Three S’s”). source note

Decision evidence. [retrospective witness · moderate] Hoffman credits David Sze with urging a pre-crisis LinkedIn financing, says he opposed an immediate layoff after the 2008 shock, and concludes that pausing before choosing offense or defense would have been wiser. This preserves Sze's financing credit and distinguishes Hoffman from CEO Dan Nye; it is hindsight without minutes, cash cases, or a control company (“Reducing People Risk”).

Operational rule. [researcher inference · strong] Make a reversible default, name decision-changing questions, price information in time and cash, and escalate mitigation as severity/scope/systematicity rise. “Reducing risks to zero is both impossible and counterproductive” (Conclusion; 9 words), but the phrase is safe only with explicit harm and irreversibility limits.

25. Myths About Failure — test the thesis, not the founder's identity

Context and argument. [investor-stated · strong] Hoffman rejects failure as an objective: write the market/customer/product/competition/revenue thesis, attack its most dangerous assumptions, update confidence, and pivot or fold decisively. Integrity, peer recommendations, and learning from a prior failure can remain founder-selection evidence (“Don’t Fail Fast – Learn Fast”; “Know When to Pivot and When to Fold”; “The Psychology of Failure”). source note

Anti-case and attribution. [retrospective witness · moderate] Hoffman attributes SocialNet's failure to delayed launch, weak product/GTM integration, “knower” hiring, and prescriptive investors. In PayPal, he describes challenging the Palm thesis as a director and credits Max Levchin with the email-payments alternative and Levchin, Peter Thiel, and the team with the change. This is director/operator learning, never PayPal co-founder evidence (“What You Can Learn From My Failure” and “A Winning Pivot”).

Falsifiability. [researcher inference · strong] “Learn fast” becomes retrospective cover unless hypotheses, disconfirming thresholds, decision owner, and review date precede the result. SocialNet and PayPal are selected opposite-outcome cases, not a comparative base rate (“Don't Fail Fast – Learn Fast” through “A Winning Pivot”).

26. The Startup Pivot — belief revision becomes resource reallocation

Context and argument. [investor-stated · strong] A pivot is a material rewrite of an explicit investment thesis; loss of belief can come from data or another rational argument. Initial product-market fit may not survive scale, and a true pivot may require product, GTM, model, technical platform, organization, and team to change together (“Investment Theses Are Critical”; “What Leads to Pivots?”; “What Changes Does Pivoting Entail?”). source note

Method and example. [investor-stated · strong] A/B/Z planning preserves an incremental alternative and a survival/restart option, while remaining cash bounds the number of reinventions. [retrospective witness · strong] The PayPal sequence names Thiel and Levchin as starting founders, Hoffman and Scott Banister as directors, and later decisions across Thiel, Levchin, Luke Nosek, Hoffman, and team. “Hard does not equal valuable” (“Why You Might Have to Burn Boats”; 5 words).

Contradiction. [researcher inference · strong] Burning boats can focus a team or destroy option value. Precommitted confidence thresholds, reversibility, runway, stakeholder harm, and learning latency must decide; the survivor PayPal account does not reveal the failed paths of companies that exhausted their tool chest (“How Do I Know When It's Time to Pivot?” and “Conclusion”).

27. Building a Great Startup Board, Part 1 — supportive catalyst, not company-builder credit

Context and argument. [investor-stated · strong] Hoffman separates oversight from management: a board should test the CEO, material commitments, opportunity, and risk without taking the operating wheel. Its role changes from concentrated upside/founder support toward executive recruiting, asset protection, and public-company reinvention as the company matures (“What Does A Board of Directors Do?” and stage section). source note

Attribution and evidence. [investor-stated · strong] “Companies succeed because of the entrepreneurs and executives, not the investors” (“Trying to take all the credit”; 11 words). [retrospective witness · moderate] SocialNet's prescriptive board and Hoffman's independent-director role at Confinity/PayPal are presented as the learning path. Those accounts do not allocate later success without minutes and other participants.

Tension. [researcher inference · strong] Replacing an early founder may signal a bad investment while fiduciary duty can require leadership change later; stage, evidence, and asset value are boundary conditions. Board seats and advice volume are not value-add proof: record what the board knew, what decision changed, counterfactual, owner, and result (“5. How Do Board Members Know When Change Is Necessary?”).

28. Building a Great Startup Board, Part 2 — selection topic, missing written body

Access result. [contemporaneous record · strong] The January 2021 page introduces director selection, future board diversity, and examples of effective directors as Part II topics (official-page introduction). source note

Hard boundary. [researcher inference · strong] The page exposes three introductory paragraphs but no written transcript or stable body for its embedded audio. TALKS completed timestamped analysis from the public SoundCloud/YouTube route in the oral-source note. WRITINGS remains partial and reconstructs no candidate scorecard, diversity mechanism, named-director example, conflict rule, or appointment outcome from the introduction or Part I.

29. My New Investment in Change.org — mission, network, and vehicle are separate tests

Context and argument. [investor-stated · strong] In May 2017, Hoffman framed Change.org as a civic marketplace in which petition creators, signers, feedback, replication, and persistence could turn participation into collective action. He argued that durable impact requires financial autonomy, diversified revenue, meaningful engagement, and outcomes rather than user count alone (¶¶1–16, 23–35). source note

Vehicle and objective. [investor-stated · strong] Hoffman explicitly separated the round from Greylock: the capital was personal, and he pledged realized equity appreciation to nonprofits (¶¶17–22). He described a round of more than $30 million, his largest personal impact investment to date, and platform figures near 200 million users and 21,000 petition “victories”; all are [self-reported] [single-source], with no disclosed instrument, ownership, donation mechanism, causal-outcome audit, or realized economics (¶¶4, 7–10, 20–22).

Decision implication and limit. [researcher inference · strong] Impact underwriting should separate participation, attributable social outcome, mission durability, revenue incentives, financial return, and the recipient of realized economics. A pledge aligns rhetoric with mission only when basis, realization, taxes, recipients, enforcement, and governance are inspectable; the essay supplies none. Named directors also prevent converting a public sponsor into sole governance credit (¶¶22, 28–35).

30. Our Investment in Entrepreneur First — backing a company-generating institution (partial)

Access and context. [contemporaneous record · strong] The recoverable Hoffman-authored official excerpt places Entrepreneur First inside Greylock's early-stage software mission; the rest of the primary post was not recovered. Same-day reporting identifies Greylock as lead of a $12.4 million financing [single-source] and Hoffman as a new board member, and separates EF's operating company from the fund investing in participant companies (TechCrunch ¶¶1–7). source note

Argument and reasoning boundary. [investor-stated · strong] In the preserved same-day interview, Hoffman called EF's pre-team, pre-idea model contrary to conventional wisdom and described an institution that selects technical talent, forms cofounder teams, and may create entrepreneurial networks across cities (TechCrunch ¶¶12–18). [researcher inference · strong] The investable unit moves upstream from one startup to a repeat company-formation system, but the operating-company financing is not a look-through claim on every EF fund or graduate.

Attribution and test. [contemporaneous record · moderate] Mosaic says Toby Coppel introduced the opportunity and that Mosaic and Greylock collaborated, so public Hoffman sponsorship does not erase sourcing or co-investor work (Mosaic account). [researcher inference · strong] The thesis requires cohort denominators for admission, team formation, company formation, shutdown, follow-on, exit, participant welfare, and realized economics by city and vintage. Missing body, price, ownership, fund rights, and cohort outcomes prevent a complete memo.

31. Our Investment in Callisto — nonprofit support, not venture-return evidence (partial)

Access and central argument. [investor-stated · strong] The recoverable official excerpt frames Callisto as reporting infrastructure for sexual coercion, harassment, and assault affecting women and women founders in technology. A same-day report preserves Hoffman's claim that the platform changes the reporting equation by giving survivors options, information, support, and control over disclosure timing and method (VentureBeat ¶11). source note

Vehicle and theory-of-change boundary. [contemporaneous record · strong] VentureBeat reports donations from Greylock and four other venture firms to a nonprofit, not a personal Hoffman check or conventional equity investment (¶¶1–10). [researcher inference · strong] The title's “investment” cannot be used as return evidence without a legal instrument; no amount, ownership, security, board right, restriction, or return objective is exposed.

Decision implication and limit. [researcher inference · moderate] Survivor-controlled information may change reporting incentives, but a launch thesis cannot establish safety, adoption, repeat-offender identification, false-report handling, data governance, or cultural change. The inaccessible primary remainder prevents reconstruction of Hoffman's diligence, objections, success measures, and governance conditions; contemporaneous journalism preserves a fragment, not the missing memo.

32. Theranos: Blitzscaling, Blitzfailing, or Blitzfrauding? — speed stops at deception and serious harm

Context and argument. [investor-stated · strong] In August 2018, after allegations and enforcement were public, Hoffman distinguished an imperfect product exposed for learning from concealed unsafe performance or misrepresented reality. Speed never excuses fraud, and verification should intensify with the severity and scope of potential harm (¶¶5–24). source note

Governance rule. [investor-stated · strong] Founders must distinguish current fact, future aspiration, plan, and risk; culture and boards should protect dissent while founders retain honesty obligations and directors retain fiduciary accountability (¶¶22–35). [researcher inference · strong] A pre-scale gate therefore needs claim substantiation, a harm taxonomy, independent verification proportional to irreversibility, protected reporting, and an escalation owner.

Attribution and hindsight limit. [contemporaneous record · strong] The essay neither claims a Hoffman investment nor treats Elizabeth Holmes's prior class appearance as one; its August 2018 legal descriptions are preserved as contemporaneous allegations, settlement, and pending indictment rather than backfilled outcomes (¶¶2–4). [researcher inference · strong] An ex-post anti-case does not show that Hoffman or the doctrine detected Theranos before exposure, and the claim that experienced VCs avert major landmines lacks an invested-and-rejected fraud denominator.

33. Responsible Blitzscaling — responsibility and velocity are simultaneous constraints

Context and argument. [investor-stated · strong] In November 2018, Hoffman and Chris Yeh argued that a blitzscaling company can acquire institutional power before culture, governance, and regulator relationships mature. Their response is not to reject scale categorically but to pair first-scaler velocity with a moral compass: “The art lies in marrying responsibility and velocity” (¶¶1–6; 8 words). source note

Operating implication. [researcher inference · strong] Responsibility cannot remain a value statement: name affected stakeholders, harm severity and reversibility, cultural capacity, regulatory readiness, accountable owner, independent checks, and a dated stop rule before the next growth stage. The authors acknowledge winners and losers while emphasizing innovation, consumer value, ambitious projects, and jobs (¶¶6–9).

Evidence boundary. [contemporaneous record · strong] The ten-paragraph text is fully accessible but explicitly adapted from the co-authored Blitzscaling book; it is partial book-text access, not independent corroboration of the unavailable book body (¶10). [researcher inference · strong] Selected favorable companies supply no measurable readiness threshold, matched failure cohort, or distribution of stakeholder costs.

34. WeWork: Blitzscaling or Blitzflailing? — an explicit screen applied after exposure

Context and screen. [investor-stated · strong] After WeWork's failed 2019 offering attempt, Hoffman distinguished hypergrowth from blitzscaling. His screen requires a large market, effective distribution, a credible path to high margins, and network effects or another durable moat, bounded by product-market fit and operational scalability (¶¶3–22). source note

Anti-case. [investor-stated · strong] He found market size, distribution, and product appeal, but inadequate future margins, lock-in, and operational scalability; subsidy is defensible only as a temporary route to sustainable capture (¶¶23–43). [researcher inference · strong] The decision-grade version precommits unit-economics, retention, moat, operating-capacity, and stop thresholds rather than labeling an exposed failure after the fact.

Quantitative and causal limit. [contemporaneous record · moderate] The essay's capital, expense, revenue, occupancy, and margin figures are source-linked but remain [single-source] secondary calculations here, pending primary S-1 reconciliation (¶¶24, 31–34). The ex-post diagnosis does not show the framework rejecting WeWork or a matched look-alike before large financings, and one correct anti-case does not validate the screen across a full denominator.

35. M&A or IPO? — strategic value should preserve exit optionality

Context and argument. [investor-stated · strong] In July 2020, Hoffman argued that founders should normally build toward an enduring public company rather than try to sell, because acquisition depends on a specific buyer's strategy. The common route to IPO and M&A optionality is a scarce strategic asset—network, data, technology, or capability—and either exit should extend value creation rather than serve as an end itself (¶¶1–20, 38–48). source note

Cases and evidence. [retrospective witness · moderate] Hoffman recounts a PayPal negotiation sequence from $550 million through $1.5 billion [self-reported] [single-source], then argues that eBay support improved PayPal's path; the latter is an unobserved counterfactual (¶¶21–25). He presents LinkedIn's IPO as marketing and its Microsoft sale as a capability/culture choice, from the interested position of LinkedIn cofounder and Microsoft director (¶¶29–46).

Decision implication and limit. [researcher inference · strong] Exit diligence should separate mission, strategic asset, financing need, buyer-specific synergy, public-market readiness, fiduciary duties, dilution, and shareholder liquidity. Hoffman's stated decade-long Greylock public-company orientation has no disclosed complete outcome denominator, and strategic value may be regulated, non-exclusive, costly, or hard to monetize (¶¶3, 11–20).

36. Re-inventing the SPAC — “venture capital at scale” depends on the contract

Context and vehicle. [contemporaneous record · strong] In October 2020, Hoffman identified Reinvent Technology Partners, sponsored with Mark Pincus and Michael Thompson, as a SPAC vehicle distinct from Greylock and personal angels. [investor-stated · strong] He proposed a faster, more price-certain public route plus patient capital and operator partnership after listing (¶¶1–8). source note

Mechanism and missing terms. [investor-stated · moderate] The proposed alignment rests on concentrated ownership and continued engagement by a “long-term financial co-founder” (¶¶7–9; 3 words). [researcher inference · strong] Concentration is not alignment without price, sponsor promote, warrants, vesting, lockup, redemptions, dilution, control, downside exposure, and exit constraints; none is disclosed in the essay.

Evidence boundary. [investor-stated · moderate] Hoffman acknowledges historical SPAC underperformance and cites a favorable operator-led subset without exposing its sample, period, fees, selection method, or benchmark (¶7). [researcher inference · strong] This is an ex-ante vehicle thesis, not evidence that any Reinvent transaction produced durable company or shareholder outcomes; filings and post-listing comparisons must test “venture capital at scale.”

37. Choosing Bad Competition — underwrite the incumbent's structural constraint

Context and taxonomy. [investor-stated · strong] Hoffman prefers a valuable market with structurally bad competition to a low-value market with no rivals. He divides bad competition into lazy incumbents using stale methods, overlooked industries whose value has changed, and incumbents trapped in inherited patterns; all reduce to failure to recognize a consequential change (opening and “The Three Types of Bad Competition”). source note

Cases and belief revision. [retrospective witness · strong] LinkedIn illustrates job boards preserving classified-ad logic while public professional identity opened passive-candidate discovery. Hoffman also records initially advising Elon Musk against SpaceX, later concluding that defense contractors optimized for contracts rather than product, and later investing (“How LinkedIn Changed the Game” and “Elon Musk and SpaceX”). That is useful changed-mind evidence, not a contemporaneous rejection memo.

Decision implication and attribution. [researcher inference · strong] A memo should name the valuable change vector, the incumbent's rational incentive or organizational constraint, evidence it cannot adapt, likely retaliation, capital/regulatory burden, and a falsifier. “Legacy players are slow” is circular after a startup wins. The later SpaceX vehicle, date, price, rights, and economics are undisclosed, while the Aurora/Nuro/Nauto references establish neither sole Greylock credit nor outcomes (“Elon Musk and Tesla”).

38. How to Decide — speed, accuracy, and visible accountability

Context and argument. [investor-stated · strong] In March 2021, Hoffman defined entrepreneurial decision quality as both accuracy and speed: waiting for complete information or comfort is usually wrong under startup uncertainty. During blitzscaling, consensus may become impractical, but ownership should remain explicit through RACI or DACI and feedback loops (“What is Good Decision-Making?” and “Making Decisions While Blitzscaling”). source note

Selection and learning. [investor-stated · strong] Greylock's repeat-founder test asks whether experience produced an inspectable decision calculus and improvement across hiring, strategy, and sales. [retrospective witness · strong] Hoffman says he decided too slowly and poorly at SocialNet, accepted responsibility, and tried to improve at PayPal (same sections). [researcher inference · strong] Ask for the information available then, owner, action, update trigger, and postmortem rather than treating founder experience as a credential.

Arithmetic and boundary. [researcher inference · moderate] The opening illustration in which a 0.5% improvement across 1,000 independent decisions compounds to roughly 147 times is arithmetic, not empirical evidence that startup choices are independent, identically measurable, or multiplicative (opening). The essay does not classify reversible, regulated, safety-critical, or existential choices; speed must yield to the decision-window, harm, and irreversibility gates developed elsewhere in the corpus.

39. Sunny Day Fund — patient climate capital across the commercialization gap

Context and argument. [investor-stated · strong] In December 2015, Hoffman argued for a public-private climate-energy pipeline: governments fund broad basic and applied research, while patient private investors finance technologies that have left the laboratory but remain too risky and slow for traditional capital. A broad portfolio should still require credible paths to commercialization and eventual scale (¶¶1–14). source note

Role, resources, and vehicle boundary. [contemporaneous record · strong] Hoffman says he participated in the Breakthrough Energy Coalition with Bill Gates and other business leaders and separately identifies a Columbia energy-policy advisory-board role (¶¶3–7). Neither fact establishes a binding personal commitment, investment entity, fund interest, company security, ownership, decision right, or realized return. [investor-stated · moderate] Capital alone is insufficient; networks, expertise, iteration, commercialization, and deployment must bridge the gap (¶¶14–21).

Decision implication and contradiction. [researcher inference · strong] A climate memo should fence research grants, coalition pledges, fund commitments, direct company investments, adviser roles, and follow-ons, then stage technical, safety, manufacturing, cost, regulatory, permitting, offtake, and infrastructure evidence. “Eventual rapid scaling” sits in tension with the essay's long-duration premise: even validated technology can remain bottlenecked by physical deployment. The text supplies no company cohort, allocation amounts, milestones, or comparative return evidence.

40. Blitzcapital — finance critical-scale races, not status

Context and allocation rule. [investor-stated · strong] In May 2017, Hoffman answered a bootstrapping critique by placing capitalization on a continuum. Customer-funded growth is preferable when it can reach scale; all-out blitzcapital is justified only when a winner-take-most race makes time to critical mass a durable competitive variable (¶¶1–6 and decision questions). source note

Cases and decision test. [retrospective witness · moderate] Hoffman uses Rocket Internet's European challenge to Airbnb as the case for raising and accelerating, while Google and six geographically diverse bootstrap examples illustrate unusually profitable economics or weak focused competition (¶¶4–10). The closing test asks about actual and potential rivals, both sides' access to capital, whether a network product needs critical mass, and whether deliberate product-market fit and revenue can work instead (¶11 and decision questions).

Boundary and modern rule. [researcher inference · strong] “Blitzcapital” is a financing regime, not a Hoffman vehicle. A decision-grade version must define critical scale, the mechanism making first scale durable, competitor capital, staged milestones, dilution, burn, governance pressure, and a stop-financing rule. Airbnb/Rocket Internet and the bootstrap list are selected narratives; the claim that financed successes outnumber bootstrappers lacks a defined universe, failures, ownership, or survivorship adjustment (¶¶8–10).

41. Building a better board of directors — fill capability gaps, not prestige slots

Context and argument. [investor-stated · strong] In September 2017, Hoffman argued that directors should fill mission-critical skill, knowledge, and network gaps the company cannot readily hire. Board composition is team design: each member needs a defined contribution, enough working time, and capabilities matched to company stage, rather than brand-name ornamentation (¶¶2–9). source note

Information and succession. [investor-stated · strong] At public-company scale, directors should form careful relationships with key executives to evaluate the CEO and successors without undermining the CEO (¶¶10–13). [retrospective witness · moderate] Change.org and Nauto illustrate mission/business-model and automotive/technology gaps, while the early Confinity/PayPal story names Peter Thiel, Max Levchin, Scott Banister, and Hoffman; these are relationship examples, not sole appointment, company outcome, or PayPal co-founder evidence (¶¶3–10).

Decision implication and limit. [researcher inference · strong] The portable method is a dated board-capability matrix covering current risks, independence, conflicts, availability, executive-information protocol, succession, appointment owner, and refresh trigger. Direct executive relationships can improve information or create shadow reporting, fragmented authority, confidentiality risk, and CEO undermining. The selected cases supply no minutes, appointment votes, removals, weak-board denominator, diversity outcome, or causal effect.

42. How We Wrote Blitzscaling — a revision process, not an efficacy test

Authorship and chronology. [retrospective witness · strong] Although the October 2018 page sits on Hoffman's profile, its body explicitly identifies Chris Yeh as guest author. Yeh dates the concept to spring 2015, when Hoffman proposed shifting attention from startups to “scaleups,” then credits framework formation across Hoffman, Yeh, Allen Blue, John Lilly, the Stanford CS183C course, and unnamed readers (¶¶1–10, 15–19). source note

Revision method. [retrospective witness · strong] Yeh says the first Stanford playbook failed to establish the who, what, where, when, and why, prompting the pair to discard a book's worth of material and rewrite. The later manuscript expanded beyond high technology and Silicon Valley, added responsibility, and used peer draft feedback as network intelligence (¶¶10–20). [researcher inference · moderate] A modern IC analogue archives the initial thesis, names the test audience and disconfirming feedback, preserves discarded versions, and distinguishes reframing from validation.

Evidence boundary. [researcher inference · strong] A class and friendly draft network are feedback mechanisms, not controlled tests of investment or company outcomes. The discarded manuscript, dissenting feedback, and claim-level changes are unavailable. This witness account supports intellectual evolution and distributed credit; it does not establish a Hoffman investment, vehicle, board decision, portfolio return, or the efficacy of capital-intensive speed.

Cross-source synthesis

Network philosophy becomes an underwriting mechanism

[researcher inference · strong] The through-line is not merely “networks are good.” The 2004 LinkedIn deck stages proof that identity and trust can produce a professional network; the 2009 Rule of Three turns mass distribution and categorical distinction into a screen; Coda, Nauto, Aurora, Blockstream, Change.org, Entrepreneur First, and later AI texts move the same logic into collaborative documents, vehicle-data flywheels, partner ecosystems, open protocols, civic participation, company formation, and distributed cognition (LinkedIn pitch · Rule of Three · Change.org · Impromptu).

[researcher inference · strong] The repeatable test is a mechanism map: who joins, what contribution becomes more valuable with participation, what prevents multi-homing or imitation, how value is captured, and who bears fraud, privacy, safety, or concentration costs. “Network effects” without this chain is a metaphor, not evidence (LinkedIn pitch · Blockstream).

Selection is learning velocity inside a financeable opportunity

[investor-stated · strong] The early screen combines large distribution, distinctive product, and capital-efficient evidence. Later work adds founder learning, integrity, theory of the game, and reference networks; the named deep-tech theses add technical-team credibility, an immediate paid wedge, and partnerships for capabilities the startup cannot own (Rule of Three · Nauto · Aurora thesis).

[researcher inference · strong] The founder screen should therefore test learning with a pre-mortem and changed-mind example, not charisma; test distribution with cohort behavior, not total users; and test capital efficiency against the next uncertainty, not a generic software benchmark. “Bad competition” requires a structural incumbent constraint rather than an ex-post loser label, while deep-tech and regulated markets require milestones in safety, science, manufacturing, reimbursement, and approval rather than “millions of users” (Rule of Three · Choosing Bad Competition · Aurora thesis).

Capital is optionality, but vehicle and objective control the claim

[investor-stated · strong] The 2009 stimulus essay argues for more experiments during capital withdrawal; Sunny Day Fund allocates patient private capital at a public-research commercialization gap; Blitzcapital makes abundant financing conditional on a critical-scale race; the professional-VC defense values repeat allocation and board networks; Covid practice lengthens runway; Calculating Risk treats pre-emptive financing as insurance (Stimulus 2.0 · Sunny Day Fund · Blitzcapital · Calculating Risk).

[researcher inference · strong] Optionality is valuable only if capital owner, mandate, price, dilution, reserve, decision rights, public-good constraint, and stop rule are known. Blockstream is explicitly personal while Xapo is Greylock; Change.org is personal impact capital with a donation pledge; Callisto is reported Greylock nonprofit support; Reinvent is a jointly sponsored public-market vehicle. Aurora later crosses Greylock financing, a director grant transfer, and Reinvent sponsorship. These are not one portfolio or one return stream (Blockstream · Change.org · Reinvent · Aurora S-1).

Post-investment practice is thesis maintenance, not control theater

[investor-stated · strong] The governing loop is consistent across failure, pivot, CEO, board, and decision texts: write the thesis; attack the fatal assumptions; time-box decision-changing questions; keep ownership visible; refactor company capabilities when belief changes; support the CEO without seizing execution; fill dated board-capability gaps; change leadership when stage and fiduciary evidence require it (Myths About Failure · Startup Pivot · How to Decide · Better Board · Board Part I).

[researcher inference · strong] Value-add claims need an intervention ledger: date, role, information available, advice, decision owner, dissent, chosen action, counterfactual, and outcome. SocialNet, PayPal, LinkedIn, Airbnb, and Aurora are valuable cases but remain predominantly first-person or winner-selected without that denominator (Myths About Failure · Airbnb Reflections · Congrats, Aurora).

Blitzscaling narrows from slogan to conditional choice

[investor-stated · strong] The mature doctrine is relative speed under uncertainty, not growth at any cost: winner-take-most structure, durable scale advantage, survival capital, operational scalability, responsibility, and a time-limited reason to tolerate inefficiency. Theranos puts deception and severe harm outside the permission set; Responsible Blitzscaling makes stakeholder capacity simultaneous with speed; WeWork supplies an ex-post margin/moat/operations anti-case; the 2022 update lowers acceptable inefficiency when capital is scarce (Theranos · Responsible Blitzscaling · WeWork · Bear Market Blitzscaling).

[retrospective witness · strong] Chris Yeh's intellectual history shows that this conditionality was partly constructed through revision: a Stanford teaching test exposed missing premises, a first manuscript was discarded, and responsibility was added as scope widened (How We Wrote Blitzscaling, ¶¶10–20). [researcher inference · strong] Revision is evidence of learning behavior, not proof that the resulting doctrine predicts company or investment outcomes.

[contemporaneous record · moderate] O'Reilly's critique says survivor narrative, monopoly rents, and subsidized share can masquerade as durable business; Lee and Kim report early scaling associated with more failure and no systematic successful-exit benefit, subject to observational and measurement limits (O’Reilly · Lee and Kim, abstract and §§4.5–4.6). [researcher inference · strong] The synthesis is a gate, not a compromise: scale only after mechanism, learning rate, financing, harms, and a stopping condition are explicit.

The central contradictions are productive only when measured

• [researcher inference · strong] Founder support versus fiduciary correction: founder agency matters, yet board duty can require a CEO or strategy change. Stage and pre-agreed evidence decide (Board Part I).
• [researcher inference · strong] Openness versus capture: Blockstream and AI “agency” rhetoric values public benefit, while venture structures need private value capture. Governance and surplus allocation decide (Blockstream · Impromptu).
• [researcher inference · strong] Network access versus network exclusion: trusted introductions accelerate diligence but can reproduce homogeneity; sourcing conversion by channel and demographic must be measured (Airbnb Reflections).
• [researcher inference · strong] Speed versus learning and harm: fast action matters only when feedback arrives before irreversible capital, safety, regulatory, labor, or social damage (Calculating Risk · Is Blitzscaling Over?).
• [researcher inference · strong] Retrospective coherence versus ex-ante record: LinkedIn annotations, Airbnb/Aurora retrospectives, and PayPal stories clarify method but cannot replace original memos, failure cohorts, and return ledgers (LinkedIn pitch · Myths About Failure).
• [researcher inference · strong] Mission versus economic ownership: Change.org's donation pledge, Callisto's nonprofit support, Blockstream's public-good priority, Greylock fund capital, and Reinvent sponsorship have different beneficiaries, contracts, and return objectives; “impact” or “investment” cannot collapse those differences (Change.org · Reinvent).
• [researcher inference · strong] Patient capital versus rapid scale: Sunny Day Fund accepts long technical and deployment horizons, while Blitzcapital finances time-sensitive races to critical mass. The market's feedback latency, physical bottlenecks, capital terms, and durable scale mechanism decide; neither “patient” nor “blitz” is an asset-class default (Sunny Day Fund · Blitzcapital).

Intellectual chronology, evolution, and contradiction

[researcher inference · strong] The philosophy-to-investment path is: philosophical concern with identity and agency → professional identity as trusted network → distribution and network effects as selection → network intelligence and alliance as operating systems → blitzscaling as competitive acceleration → protocols, autonomous systems, and AI as networks of machines/capabilities. This is an interpretation of dated works, not proof that every decision followed the path.

[researcher inference · strong] The main evolution is increased conditionality. The 2009 screen is concise and consumer-Web-specific; Sunny Day Fund and Blitzcapital split patient commercialization from critical-scale races; 2017–2018 theses add technical founders, current wedges, partners, a board-capability matrix, and documented framework revision; 2020–2022 writing adds runway, relative competition, decision windows, thesis falsification, organizational refactoring, and stakeholder risk. The persistent weakness is evidence: public writing offers far more doctrine and winner retrospectives than rejected cases, board records, full economic outcomes, or systematic falsification.

Decision playbook and modern VC translation

A Reid Hoffman-derived decision playbook

1. Fence the role, vehicle, and objective. [researcher inference · strong] Record founder/operator equity, personal or foundation capital, Greylock fund, nonprofit support, Reinvent sponsorship, Reprogrammed/family-affiliated holding, incubator/co-founder role, and board-only duty separately. Name who approved, who owns economics, and whether return or mission is primary (Blockstream · Change.org · Reinvent).
2. Write the thesis and fatal assumptions. [investor-stated · strong] State customer, market, product, distribution, competition, revenue, financing, regulation, and team hypotheses; rank what can kill the company (Myths About Failure).
3. Map the network mechanism. [researcher inference · strong] Identify participant sides, value contribution, feedback loop, cold start, multi-homing, trust/safety, capture, degradation, and attributable outcome. Reject “network effect” without cohort or structural evidence (LinkedIn pitch · Change.org · Entrepreneur First).
4. Test founder and team learning. [investor-stated · strong] Ask for a changed-mind example, evidence of learning from a prior failure, integrity references, and the proposed learning loop. Distinguish “learner” from experienced “knower” without dismissing relevant expertise (Myths About Failure · How to Decide).
5. Stage proof by financing. [investor-stated · strong] Each round should retire the next uncertainty. Use A for the current thesis, B for a nearby alternative, Z for survival/restart; predefine runway and financing failure. Fund speed only when critical scale and competition make time decisive; otherwise preserve evidence discipline and capital efficiency (LinkedIn pitch · Startup Pivot · Blitzcapital).
6. Price information and irreversibility. [investor-stated · strong] Default to a decision window; extend it only for questions likely to change the action. Escalate caution with severity, scope, and systematicity (Calculating Risk).
7. Specify the board compact and capability matrix. [investor-stated · strong] The board challenges thesis, CEO, material commitments, and risk; management operates. Define each director's missing capability, time, independence, conflicts, executive-information protocol, succession role, and refresh trigger; record dissent and intervention as company stage changes (Better Board · Board Part I).
8. Pivot the system, not the slide. [investor-stated · strong] When belief changes materially, refactor product, GTM, business model, platform, team, and incentives together; preserve options until commitment has more value than reversibility (Startup Pivot · Choosing Bad Competition).
9. Blitzscale through a responsibility gate. [researcher inference · strong] Require winner-take-most structure, durable first-scale advantage, adequate capital, measurable relative progress, operational capacity, tolerable and reversible stakeholder harm, verified high-severity claims, and a dated stop rule. Otherwise fast growth is not the method (Is Blitzscaling Over? · Responsible Blitzscaling · Theranos · WeWork).
10. Audit outcomes and attribution. [researcher inference · strong] Track write-offs, shutdowns, pivots, follow-ons, dilution, realized/unrealized value, intervention outcomes, externalities, mission outcomes, and rejected deals by vehicle. Do not infer return from valuation, listing, acquisition, board seat, donation, public credit, or an exit counterfactual (Change.org · M&A or IPO · Reinvent).

Modern VC translation (as of 2026-08-27)

[researcher inference · strong] For AI, the old “distribution, distribution, distribution” screen becomes capability × workflow distribution × trust × cost decline. Underwrite model dependence, compute and data rights, evals, error severity, user retention after novelty, human override, vendor concentration, security, and regulatory exposure. A consumer-interface thesis such as Inflection's must be tested separately from the later personnel/license transaction; Hoffman's co-founder/board/investor and Microsoft-director roles require a conflict ledger and recusal evidence (Inflection launch · Microsoft 2024 proxy). The proxy's less-than-10% Reprogrammed and Greylock positions are [single-source] filing disclosures, not personal proceeds.

[researcher inference · strong] For autonomy, biotech, climate, and other deep tech, replace viral growth with validated safety/science/manufacturing/cost/regulatory/permitting/offtake milestones. Separate an immediate commercial wedge from the long-duration platform and distinguish public research, coalition commitment, fund, direct security, and adviser role. Partnership can accelerate distribution while creating data, integration, concentration, and bargaining dependencies. Simulation miles, targets screened, coalition participation, partnerships, or capital raised are exposure/activity measures until linked to validated outcomes (Sunny Day Fund · Aurora thesis).

[researcher inference · strong] For governance, archive the ex-ante memo and every material board intervention. Founder support is not immunity; “supportive catalyst” requires independent directors, conflict review, decision ownership, and a trigger for escalation or leadership change. Current records control live biography: Microsoft's June 2026 8-K says Hoffman would not stand for re-election, remained through the annual meeting, and reported no disagreement—not whatever an older marketing bio implies (Microsoft 2026 8-K, Item 5.02).

[researcher inference · strong] For portfolio construction, no public complete denominator exists. A modern evaluation must keep the Greylock profile union, personal/foundation angels, nonprofit support, three Reinvent SPACs, Reprogrammed holdings, founder/operating equity, incubated companies, and board-only roles in separate tables. The 60+ description is a [single-source] TechCrunch host characterization and the 100+ description is a [single-source] issuer characterization; neither is an auditable career portfolio. Without invested capital, ownership, losses, proceeds, and dates, no performance ranking is warranted.

Best writings and unresolved record

Ranked top five for decision usefulness

1. LinkedIn’s Series B Pitch to Greylock. [researcher inference · strong] Best bridge between a contemporaneous deck and later annotation: explicit thesis, risk disclosure, staged financing, and a hard attribution correction via the S-1. Its hindsight layer is visible rather than hidden (source · note).
2. The Future of the Bitcoin Ecosystem / Why I Invested in Blockstream. [researcher inference · strong] Rarest vehicle-clean decision-era thesis: personal Blockstream capital, separate Greylock Xapo capital, public-good priority, founder reasoning, and an adoption dependency (source · note).
3. Myths About Failure. [researcher inference · strong] Strongest failure/anti-case text because it names SocialNet errors, turns “fail fast” into falsifiable thesis learning, and preserves Levchin/Thiel/team credit in PayPal (source · note).
4. Building a Great Startup Board, Part 1. [researcher inference · strong] Best governance and attribution discipline: oversight versus management, stage change, supportive catalyst, founder credit, and the CEO-replacement tension (source · note).
5. Bear Market Blitzscaling. [researcher inference · strong] Best mature boundary-setting of the signature doctrine: relative speed, capital scarcity, defense/offense, revenue proof, and survivor-bias acknowledgment, with independent critique and empirical counterweight (source · note).

[researcher inference · strong] The closest exclusions are Blitzcapital, the cleanest financing-timing rule but success-selected and without a matched denominator; How to Decide, whose speed-plus-accountability rule rests on retrospective SocialNet/PayPal evidence; Building a better board, whose capability matrix lacks appointment/outcome records; WeWork, the clearest operating anti-case but written after public exposure; and Change.org, the cleanest mission/vehicle fence but without outcome or donation-accounting evidence (Blitzcapital · How to Decide · Better Board · WeWork · Change.org).

Unresolved evidence and access priorities

• [researcher inference · insufficient public record] Portfolio and economics: no complete deal/vehicle denominator; no consistent check, ownership, reserves, write-off, proceeds, or realized/unrealized return series. Company or press counts cannot substitute.
• [researcher inference · insufficient public record] Original decisions: Airbnb, Aurora, Coda, Nauto, Inflection, Blockstream, and other named theses need investment memos, partnership votes, price/ownership, dissent, follow-on rules, and rejected comparables.
• [researcher inference · insufficient public record] Interventions: board minutes and management/employee accounts are needed to test claimed help at SocialNet, PayPal, LinkedIn, Airbnb, Aurora, and other boards.
• [researcher inference · insufficient public record] Failure and selection cohorts: writings name SocialNet and some doctrinal anti-cases but do not enumerate missed deals, write-offs, founder replacements, governance failures, or false network-effect theses.
• [researcher inference · insufficient public record] Book bodies: Superagency, Masters of Scale, Blitzscaling, and revised The Startup of You were not acquired; The Alliance is Chapter 1 only. Publisher and guide material is not book-body evidence.
• [researcher inference · strong] Routed oral items: TALKS completed lawful audio acquisition and stable timestamps for Board Part II and “Blitzscaling During Covid”; their WRITINGS records remain partial because written metadata supports no unseen doctrine.
• [researcher inference · insufficient public record] Partial posts: full lawful bodies for Entrepreneur First and Callisto could materially change their selection, diligence, objection, and success-measure analyses; current conclusions stop at official excerpts and named contemporaneous fragments.
• [researcher inference · insufficient public record] Impact and public vehicles: Change.org needs instrument, ownership, donation enforcement, recipients, and realized outcomes; Callisto needs donor entity, restrictions, governance, and safety outcomes; Reinvent needs sponsor terms, dilution, redemptions, post-listing interventions, and shareholder results.
• [researcher inference · insufficient public record] Climate allocation and capital timing: Sunny Day Fund needs a legal coalition/vehicle map, commitments, selected-company cohort, milestones, and outcomes; Blitzcapital needs matched financed/bootstrapped cohorts, ownership/dilution, round decisions, critical-scale measures, and stop-financing evidence.
• [researcher inference · insufficient public record] Governance and intellectual revision: Building a better board needs capability matrices, appointment/removal records, minutes, conflicts, and outcomes; How We Wrote Blitzscaling needs the discarded manuscript, dissenting feedback, claim-level revisions, and later failure-driven updates.
• [researcher inference · insufficient public record] Current conflicts and outcomes: AI company formation, Microsoft/OpenAI relationships, later transactions, public-board roles, political activity, and philanthropy need dated role/recusal/economics records rather than narrative aggregation.

[researcher inference · strong] The written corpus supports a serious decision system—networks, explicit theses, learning, staged finance, conditional speed, and supportive governance. It does not yet support a complete portfolio, causal value-add claim, or investment-performance conclusion. Those are the decisive next evidence gates.

Coverage and evidence base

The spoken-corpus map begins with 73 raw routes: 20 priority rows seeded in the talks section of sources.md, two audio records routed from WRITINGS, and all 51 zero-based rows routed from Hoffman’s 279-row official writing index. Body-level triage retains 37 priority records and excludes 36: 34 retained records correspond to actual appearances or episodes, while three—Entrepreneurship and Loneliness, Talking Politics, and the Quartz fast-culture rebuttal—are explicit written route exceptions selected for distinct biographical or governance evidence. Companion essays for the two philosophy episodes and Game On establish access to those distinct episodes but are not treated as verbatim transcripts or independent corroboration.

The final denominator is one record per atomic appearance or justified route exception, not one per URL. Old Stanford PDFs that return HTML, current/legacy Greylock pages, RSS enclosures, SoundCloud routes, YouTube mirrors, and companion essays are access/version routes. They do not multiply the evidence. [researcher inference · strong] Repetition across appearances establishes consistency of public doctrine, not truth, attribution, or investing efficacy.

Raw-route ledger

The 20 seeded talk rows and two routed audio records—Building a Great Startup Board, Part 2 and Blitzscaling During Covid—are all retained. The official-index route is disjoint:

The 36 exclusions are named rather than silently discarded:

• Guest/subject/derivative capsules (17): row 49, Future Follows Fiction—guest/subject-led future-fiction route; 55, Human-Centered AI—guest/subject-led AI route; 64, Go Anywhere, Get Anything—guest-led marketplace capsule; 83, The Power of Ritual with YouTube and Coda’s Shishir Mehrotra—Mehrotra-led capsule; 110, Tory Burch: Combining Incredible Patience and Explosive Speed—Burch-led capsule; 123, The Power of Difference: Sallie Krawchek at Ellevest—Krawchek-led capsule; 139, How Marissa Mayer created Google’s school for young “superheroes”—Mayer-led capsule; 146, 5 Steps to Finding Your Next Big Idea from Spanx’s Sara Blakely—Blakely-led capsule; 149, The Right Way to Build an Online Community: 3 Rules From Investor and Flickr Cofounder Caterina Fake—Fake-led capsule; 152, What an Amazon explorer can teach you about finding your big idea—guest-led capsule; 154, How being values-driven rather than profit-driven helped Howard Schultz make Starbucks more profitable—guest-led capsule; 157, The price that bleeds your business, may save your business—guest-led capsule; 158, Don’t be a Unicorn. Be a Phoenix—guest-led capsule; 161, How to Build a Community That Thrives — with Entrepreneur/Investor Caterina Fake—Fake-led derivative; 168, Don’t Just Compete. Invent a New Game and Master It—guest-led episode; 177, How Code for America Makes Good Government Great—guest/organization-led capsule; 231, Networked Curiosity: How Oscar-winning producer Brian Grazer uses curiosity for career success—guest-led capsule. These are not counted as independent Hoffman appearances or corroboration.
• Topical, promotional, or general records (13): row 16, Humanity’s Hegelian Golden Braid speech—general intellectual speech; 17, Video: Reid Hoffman meets his AI twin—AI promotion/demonstration; 19, Bologna Business School Commencement Speech—general commencement; 62, Spiderman Ethics—topical ethics route; 77, Eric Lander’s great new podcast Brave New Planet—book/podcast promotion; 94, Entrepreneurship Now—general entrepreneurship; 117, Announcing the Blitzscaling Series, a new part of Greylock’s Greymatter Podcast—series promotion; 180, Why panels suck—general presentation advice; 207, Extra Sensory Production—topical production record; 220, Texting out an S.O.S.: Saving Lives With Smartphones and Big Data—topical promotion; 223, Changing the World, Silicon Valley Style—general regional commentary; 226, Compassionate Utilitarianism: Balancing the head and the heart in the pursuit of philanthropy—general philosophy; 229, Driving in the Networked Age—general network-era commentary.
• Covered elsewhere or not a distinct priority appearance (6): row 24, Talking AI with AI—AI route overlapped by retained direct AI appearances; 95, Building Companies in Times of Change—pandemic/company-building overlap; 128, How the Lessons of Role-Playing Games Help Build Great Teams and Companies—covered by retained Game On; 163, Rules to Live by — for Infinite Learners—career/learning overlap; 190, What’s the Secret of Silicon Valley? Hint: It’s Not Just Startups—general regional route; 211, Using Artificial Intelligence to Humanize Management and Set Information Free—AI/management overlap. This yields 37 selected + 36 excluded = 73 raw routes.

Dated source-by-source guide

1. Social Computing: From Message Boards to Social Networks — 2006

• [contemporaneous record · strong] In the CHM panel, Hoffman describes LinkedIn as a trusted professional network and discusses relationship context and reputation while the company is still independent (20:22–20:52; 44:21–45:39; 01:03:34–01:05:42). This is stronger historical evidence than later origin stories, but automatic captions constrain exact names and numbers.

2. Investing in Long-Term Ideas — 2011-06-30

• [investor-stated · strong] The mapped 1:48 passage in Stanford’s live printable transcript contrasts a one-to-three-month payoff with a three-to-five-year risk and argues that founders are better positioned to pursue the latter. It is an early horizon rule, not a full diligence system; the dead clip PDF is not cited as read.

3. Live Life in Permanent Beta — 2012-02-22

• [investor-stated · strong] The Stanford transcript joins intelligent risk, contrarian testing, network intelligence, and continual reinvention (paragraphs beginning “the model doesn’t work anymore,” “the network was the key thing,” and “being in permanent beta”). It generalizes operating experience into career advice; it does not validate an investment record.

4. Academy of Achievement oral history — 2014-09-13

• [retrospective witness · strong] Only the dated Q&A on the Academy page is used for the Apple–Fujitsu–SocialNet–PayPal–LinkedIn sequence and SocialNet failure lessons (questions on leaving academia for software, founding SocialNet, and lessons from its failure). The surrounding living biography is excluded because it says Thiel’s angel check was $500 million rather than $500,000 and later assigns Hoffman a Figma Series A/$50 million claim that conflicts with the deal record.

5. How to Be a Great Founder — 2014-11-04

• [investor-stated · strong] The Stanford/YC lecture is the corpus’s clearest founder-selection checklist: complementary teams (04:49–05:32), intelligent contrarianism (10:02–12:55), a decisive distribution edge (28:43–28:51), and reference-checked adaptability or an “infinite learning curve” (34:16–34:27). Creator captions give a stronger text route than automatic subtitles, but no candidate scorecard or base rate is disclosed.

6. Tim Ferriss Show #101 — 2015

• [investor-stated · moderate] The 30-page transcript connects PayPal, the Facebook introduction, network-as-mentor behavior, and professional-CEO selection (pp. 12–20). Michael McCullough’s contributions and the difference between an introduction, an investment, and firm ownership remain explicit.

7. Blitzscaling — 2016-04

• [investor-stated · strong] In the HBR interview, Hoffman defines the method as rapidly building for a large market and first-mover scale, then names product-market fit, competition, capital burn, and organizational transitions as conditions and costs (pp. 4–8). The 2016 wording is not retrofitted with later formulations.

8. Brian Chesky — Handcraft the Core Experience — 2017–2018

• [retrospective witness · strong] Chesky’s New York host work in the speaker-attributed transcript shows non-scalable learning before automation (pp. 2–4); Hoffman’s framing then examines how to preserve that learning at scale (pp. 8–10). Chesky and Airbnb own the operating work; Hoffman owns the narrative frame.

9. Ten Commandments for Startup Success — 2017-06-26

• [investor-stated · moderate] The Tim Ferriss transcript packages thesis, founder learning, distribution, networks, and ranked risks into memorable rules (“Commandment 2” through “Commandment 8”). The publisher warns the transcript may contain errors, so the source is not used for fragile wording or numbers.

10. Economic Club of New York conversation — 2018

• [investor-stated · strong] The event transcript defines speed over efficiency, uses PayPal as operating evidence, and distinguishes local network structure (pp. 6–10). Eric Schmidt’s questions and host biography are not recast as Hoffman testimony.

11. Blitzscaling: How Silicon Valley Learned to Grow — 2018

• [investor-stated · strong] The CHM event makes the mature doctrine conditional on uncertainty and relative competitive speed (06:56–08:24; 18:20–18:25) and acknowledges organizational inefficiency (08:20–08:31). Chris Yeh retains framework co-authorship despite not speaking in this event.

12. The Reid Hoffman Story (Part 2): Make everyone a hero — 2019-02-26

• [retrospective witness · strong] In the official ART19 feed record with GUID gid://art19-episode-locator/V0/asxj18qKrADugx8qjg2CB5mG46vms_ixtffmXdysEs4, Hoffman recalls PayPal's relative-speed edge on eBay (11:47–13:38), testing investing and entrepreneurship before calling Friendster his first pure angel investment (14:29–16:08), LinkedIn's conviction and acquisition-loop changes (20:29–29:03), and Greylock's experiment-friendly, entrepreneur-first culture (42:05–43:32). June Cohen hosts and structures the episode; teams and entrepreneurs retain operating credit. The current CMS URL serves Part 1's body, so the 2019-02-26 Part 2 RSS enclosure—not the misrouted text—was transcribed with the OpenAI API; timestamps refer to the 58:52 dynamic-ad rendition.

13. Quartz fast-culture rebuttal — 2019-03-19

• [investor-stated · strong] Hoffman and Yeh’s written rebuttal distinguishes quick from reckless growth and acknowledges cultural, regulatory, and governance harms (opening rebuttal; paragraphs beginning “These pressures are compounded” and “The art lies”). This is a route exception, not an appearance; the authors’ book-defense incentive and Tim O’Reilly’s criticism stay visible.

14. Strategy Session: Early Adopters — 2019-09-26

• [investor-stated · moderate] The official automatic transcript distinguishes transferable early-adopter attributes (Laura Douglas turn) and makes sequencing conditional on competition: improve product-market fit before scaling distribution when no rival is close, but iterate while scaling when relative speed is decisive (Dominik Eggert turn). The machine transcript is used only for clear, audio-checked turns.

15. Planning for Volatility — 2020-06-09

• [investor-stated · strong] In a contemporaneous Greylock portfolio town hall, Hoffman says volatility weakens apparently proven product-market-fit and capital assumptions while the investment horizon remains ten-plus years (00:59–03:27). David Thacker moderates; broad advice is not evidence that every portfolio company followed it.

16. Transformational Moments — 2020-06-11

• [retrospective witness · strong] Hoffman’s Jeff Weiner account credits Weiner as later-stage co-founder/co-architect, James Slavet for the introduction, and references plus an interim role for reducing succession risk (00:12–15:43). Weiner and the LinkedIn team retain operating credit.

17. Blitzscaling During Covid — 2020-06-18

• [investor-stated · strong] The official RSS episode, with Hoffman and Chris Yeh, re-parameterizes relative speed around changed product-market-scale fit, capital, and travel demand (00:22–02:30; 25:50–31:46). The private YouTube route and dead Greylock page were superseded by public audio transcribed with the OpenAI API.

18. The Virtual VC — 2020-07-16

• [investor-stated · strong] Hoffman and Saam Motamedi explain in the public episode that Greylock continued new and follow-on investing and redesigned remote diligence to include team members beyond founders (01:06–06:00). Motamedi owns the quoted video-diligence practice; firm claims are not Hoffman-only.

19. From Seed to Series A — 2020-08-06

• [investor-stated · strong] In the Hoffman/Guo episode, Sarah Guo rejects a rigid metric screen and Hoffman uses Airbnb’s tiny early numbers to illustrate behavior- and market-structure judgment (18:03–21:14). They also frame the VC as a long-term financing/board partner founders should diligence (25:23–34:00).

20. Talking Politics — 2020-08-13

• [investor-stated · strong] Hoffman’s prose route defines 3S as severity, scope by number of people affected, and systemic/cascade harm; uncertainty can increase the need to investigate. This written governance exception is not assigned oral timestamps, and organizational relevance is not smuggled into the definition of scope.

21. Building Ecosystems of Adaptability — 2020-08-27

• [investor-stated · strong] In Peter High’s interview, Hoffman argues that product and go-to-market should be co-designed and that shock conditions discount old product-market-fit evidence (05:06–11:40; 21:55–22:35). High retains interviewer credit.

22. Entrepreneurship and Loneliness — 2020-12-15

• [retrospective witness · strong] Hoffman’s first-person episode-derived essay contrasts SocialNet’s unfamiliar category and inexperienced team with the more network-supported LinkedIn path (“Facing Loneliness at SocialNet” and “Remember: You Are Not Alone”). It is retained for failure biography, not as a generic wellness appearance.

23. Reid Hoffman #AMA with 20VC — 2021-01-13

• [retrospective witness · strong] In the public audio, Hoffman recalls bringing Airbnb to Greylock, David Sze’s skepticism, and the partnership debate (09:13–11:46); he also names network effects as a bias rather than an absolute screen (00:46–03:05). The account does not imply unilateral approval.

24. Building a Great Startup Board, Part 2 — 2021-01-26

• [investor-stated · strong] Hoffman and Yeh distinguish management, investor, and independent director roles in the public video/audio route and make board-seat selection a working-relationship decision (00:34–04:43). The official page’s introduction remains a partial writing; the spoken body is complete through SoundCloud and audio-checked automatic captions.

25. Game On — 2021-05-18

• [investor-stated · strong] The companion essay moves from goals/strategy/opponents to incomplete information, probability, fast feedback, coalition dynamics, and Catan’s resources/trading/path changes (“Why Games Are Serious Strategy,” “What Kinds of Games Teach Strategy,” “Game #2: RuneQuest,” and “Game #5: The Settlers of Catan”). Chris Yeh retains interlocutor credit; the game model is intellectual biography, not performance proof.

26. People-First Capitalism — 2021-06-08

• [retrospective witness · strong] In the full Hoffman/Chesky transcript, Chesky owns the decisions to increase board cadence, sequence stakeholder actions, handle layoffs, and adapt travel before the IPO (Chesky turns beginning “The third thing is we have to act,” “So we had monthly all-hands,” “the hardest decision,” and the later “searches with dates” passage). Hoffman is board member/interviewer, not the operator.

27. The Philosopher-Entrepreneur — 2021-07-20

• [investor-stated · strong] The companion essay develops thesis → practice → observation → revision and explicitly applies it to investing (“Why Philosophy Matters,” “How Philosophy Works,” and “1. Aristotle & Human Nature”); “3. Wittgenstein & Language” supplies Hoffman’s retrospective Coda board/investor rationale, not causal proof of Coda’s product or outcome.

28. Philosopher Versus MBA — 2021-07-27

• [investor-stated · strong] The companion essay calls an MBA an adverse founder signal but makes behavior, network composition, and exceptions decisive (“Top Drawbacks of an MBA Program,” “Not Prioritizing Action,” and “The Network May Not be as Valuable as It Seems”); Chris Yeh’s countercase covers conventional-business understanding, analytical/finance/accounting skills, network, and brand (“The Case for Pursuing an MBA” and “Venture Capital and the MBAs of Silicon Valley”). Entrepreneurship itself remains learned through practice.

29. Taking on Tech Titans — 2021-08-17

• [investor-stated · strong] In the full transcript, Hoffman tests whether the target is an incumbent priority and whether the startup has a structural difference (opening Hoffman turns; responses beginning “Sridhar laid it out…” and “The thing I would add as an investor…”); Sridhar Ramaswamy owns Neeva’s subscription/no-tracking thesis (Yeh question beginning “So, Sridhar, you brought up Neeva”; Ramaswamy response beginning “So at a core level, my co-founder Vivek and I…”), and Chris Yeh moderates. Neeva’s later consumer-search shutdown makes this a failure test, not success validation (contemporaneous transcript; later outcome reported by TechCrunch).

30. The Knowledge Project #147 — 2022-09-20

• [investor-stated · strong] In the public-audio episode, Hoffman matches decision artifacts to depth/speed (12:11–13:37), distinguishes board help from governance across stages (24:24–29:11), and describes crisis-board questions and outside information (29:24–34:07). The publisher transcript remains members-only; timestamps derive from public audio transcribed with the OpenAI API.

31. Washington Post Live: The Path Forward — 2023-05-10

• [investor-stated · moderate] In the publisher transcript, Hoffman identifies bad-actor amplification, rejects a six-month pause because continued development can improve defenses, and proposes source/identity validation and deepfake safeguards (Christine Passariello risk/pause exchanges; Gregory Schuckman election question). Passariello raises OpenAI/Microsoft ties; Hoffman discloses his Microsoft board and Inflection/Pi co-founder roles.

32. Reid Hoffman Confronts the AI Critics — 2023-11-07

• [investor-stated · moderate] The official automatic transcript frames risk comparatively and favors iterative adoption plus governance (“How criticism of AI has restricted its use” and “Why Reid Hoffman didn’t sign the letter urging AI labs to pause development,” Berman questions and Hoffman responses). Critics do not speak in the episode; their original arguments and independent safety evidence remain necessary counterweights.

33. AI + Humanity — 2023

• [investor-stated · moderate] In the CHM event, Hoffman calls AI human amplification and acknowledges that it can elevate capability or amplify weakness and malicious use (08:58–13:28; 24:39–25:17). Automatic captions support timestamps only after speaker/audio checks.

34. Candid Fireside Chat — 2024

• [retrospective witness · moderate] The Stanford transcript revisits SocialNet’s nine months without user feedback and PayPal’s eBay-user pivot (paragraphs beginning “All of those aphorisms came from a failure” and “Two days after we launched”), then ABZ risk planning, LinkedIn, and current AI optimism. The polished resilience frame can compress earlier uncertainty, so contemporaneous records remain primary for chronology.

35. Possible Riffs 007 — 2024-07-31

• [investor-stated · strong] On the Possible page, Hoffman tests AI founders for learning velocity, a theory of the game, and commitment to a multi-year journey (turn beginning “Well, there’s two parts to this. One is general investing…”; paragraphs beginning “Now, the thing that I would add…” and “Because I’m trying to figure out…”). The page says the transcript was AI-generated and lightly edited; delicate wording is not treated as human-verbatim.

36. Video: Hoffman pitches LinkedIn to himself — 2024-10-03

• [researcher inference · moderate] The AI younger-self video reconstructs the early LinkedIn pitch, while current Hoffman says network growth should precede smaller revenue streams (01:00–03:02). It is a hindsight artifact checked against the separate 2004 deck, not contemporaneous Series B evidence.

37. In Good Company: Shaping the AI Era — 2026-02-25

• [investor-stated · strong] In the NBIM/Acast episode, Hoffman argues that incumbents should select bounded AI risks and deploy to learn (11:04–13:42), identifies his Manas AI co-founder/investor role (08:16–08:33), and discusses large-scale AI infrastructure conviction (23:17–24:00). Transaction terms and NBIM endorsement are not inferred.

Recurring frameworks and decision rules

1. Start with a theory of the game, then expose it to reality

[investor-stated · strong] Hoffman repeatedly asks decision-makers to name the game’s goal, other agents, rules, resources, uncertainty, and feedback loop. The model is explicit in Game On, becomes a founder test in How to Be a Great Founder (10:02–15:06), and appears as a current AI-investment screen in Possible Riffs. The intellectual mechanism is hypothesis → action → evidence → revision, not confidence in abstraction; the philosophy essay makes that loop explicit.

[researcher inference · strong] Diagnostic: What has to be true about users, competitors, distribution, capital, and time? Which observation would revise the theory? Boundary: games clarify structure but can understate regulation, path dependence, and genuinely unknown actors. No public base rate shows that teams using this vocabulary outperform.

2. Select learning velocity, not a fixed founder archetype

[investor-stated · strong] Complementary teams, intelligent contrarianism, distribution insight, references, and an “infinite learning curve” form the most stable picking doctrine in the 2014 founder lecture (04:49–05:32; 10:02–15:06; 28:43–28:51; 34:16–34:27). Possible Riffs adds current AI-specific commitment and theory-of-game tests. From Seed to Series A shows how sparse metrics increase—not eliminate—the need for behavioral and market evidence (18:03–21:14).

[researcher inference · strong] Diagnostic: What did the team learn faster than peers, how did its thesis change, and what do references say the founders do when a plan fails? Boundary: adaptability language is vulnerable to hindsight; a flexible storyteller can look like a learner without producing a repeatable update.

3. Treat product and distribution as one design problem

[investor-stated · strong] The 2006 CHM panel makes trust and relationship context part of LinkedIn’s product, while Early Adopters asks which early behavior transfers and Building Ecosystems of Adaptability says product and go-to-market should be co-designed (08:27–11:40). This is stronger than “build first, market later”: distribution is part of the causal product thesis.

[researcher inference · strong] Diagnostic: Which user behavior creates value for the next user, and which channel teaches the company fastest? Boundary: a channel advantage can mimic product-market fit until acquisition economics or platform rules change.

4. Buy speed only when market structure pays for it

[investor-stated · strong] The 2016 HBR interview conditions speed on large-market potential, product-market fit, competition, and first-mover scale (pp. 4–8). The 2018 CHM event emphasizes relative speed and temporary inefficiency (06:56–08:31; 18:20–18:25), while the Covid episode changes the parameters when capital and demand move (25:50–31:46). Chesky’s handcrafting account supplies the necessary opposite: intensive manual learning can be rational before scale (pp. 2–4).

[researcher inference · strong] Diagnostic: What concrete scale advantage is perishable, which competitor can capture it, and what capital/organizational debt is being purchased? Kill condition: no credible winner-take-most mechanism, weak product-market fit, or an inefficiency whose accumulated harm exceeds the speed advantage.

5. Plan across states, not around a single forecast

[investor-stated · strong] The 2012 Permanent Beta talk frames intelligent risk as a wager with alternatives. In June 2020, Planning for Volatility discounts previously “proven” assumptions (00:59–03:27), and Virtual VC adapts both portfolio capital and diligence (23:09–25:30). The 2022 Knowledge Project adds recurring board rituals that surface risks, opportunities, and what is top of mind (24:24–34:07).

[researcher inference · strong] Diagnostic: What survives the downside state, what option benefits from the upside state, and which signal triggers a replan? Boundary: broad scenarios are not useful unless tied to runway, hiring, product, and financing actions.

6. Make governance a help-and-control system

[investor-stated · strong] Board Part II distinguishes management, investor, and independent roles and makes selection deliberate (00:34–04:43). The Knowledge Project says early boards mix governance with company help, while public-company boards weight governance more heavily (24:24–29:11). People-First Capitalism offers Chesky’s crisis practice—weekly board cadence and explicit stakeholder sequencing—not a Hoffman-only board intervention.

[researcher inference · strong] Diagnostic: What unique question, outside information, and stage-specific skill does each director contribute? Boundary: a helpful director can still have conflicts; trust does not replace independence, removal rules, or control clarity.

7. Compare risk with the risk of inaction, then assign responsibility

[investor-stated · moderate] The Quartz rebuttal says velocity does not excuse irresponsibility. Talking Politics triages harm by severity, affected population, and systemic cascade. In AI, the CHM event acknowledges amplification of both capability and harm (24:39–25:17); the NBIM interview says incumbents should choose bounded risks rather than wait for zero uncertainty (11:04–13:42).

[researcher inference · moderate] Diagnostic: Who bears the downside, how broad and cascading can it become, what reversible deployment tests it, and what red line stops it? Boundary: “risk of inaction” can become an all-purpose acceleration argument unless decision rights and stop rules are specified.

Intellectual chronology, evolution, and contradictions

• 2006–2012 — network trust becomes adaptive strategy. [contemporaneous record · strong] LinkedIn is described as a trust/context system in the 2006 CHM panel; 2011 and 2012 talks add long horizons, intelligent risk, networks as information, and permanent beta. Sequence supports evolution; it does not prove that LinkedIn success caused the later investment method.
• 2014–2016 — picking and scaling become explicit. [investor-stated · strong] The founder lecture turns experience into a selection screen; HBR codifies when to buy speed. The two doctrines are complementary only if learning and distribution evidence precede scale.
• 2017–2019 — successful cases meet the responsibility critique. [researcher inference · moderate] Chesky’s handcrafting episode shows slow learning before fast growth; the Quartz rebuttal answers social-cost criticism but supplies few enforceable stop mechanisms.
• 2020 — shock turns doctrine into a live test. [contemporaneous record · strong] Planning, remote diligence, seed/Series A selection, board help, and crisis scaling were stated during—not after—the pandemic in Planning for Volatility, Virtual VC, From Seed, and Blitzscaling During Covid. That timing makes the rules testable against later portfolio outcomes.
• 2021–2022 — intellectual origins and governance receive more precision. [researcher inference · strong] The game and philosophy essays explain model formation; Board Part II and the Knowledge Project operationalize director selection and rituals.
• 2023–2026 — AI broadens the method and raises the conflict. [investor-stated · moderate] AI appearances retain iteration, networks, and bounded risk but add amplification and infrastructure conviction (CHM; Washington Post; NBIM). Portfolio exposure and public advocacy make counterevidence especially important.

Biography, deal rationales, and credit allocation

Contemporaneous versus retrospective checks

Decision playbook for a current analyst

1. Write the game before the memo. [researcher inference · strong] State actors, incentives, scarce resources, timing, uncertainty, and disconfirming evidence using the game framework. If the memo cannot say what would change the theory, it is a story, not a thesis.
2. Reference the learning loop. [investor-stated · strong] Ask for an instance where the founders changed a core view, not just worked harder; cross-check with former collaborators using the 2014 reference/adaptability test (34:16–34:27).
3. Model product and channel together. [researcher inference · strong] Identify the first user’s value, the next user’s value, the acquisition loop, and the point at which the loop fails. The historical basis is early-adopter transfer plus GTM co-design (08:27–11:40).
4. Charge speed to a named advantage. [researcher inference · strong] Every inefficiency must buy a measurable lead in density, learning, distribution, or standard-setting. If it buys only headline growth, the HBR conditions (pp. 4–8) are not met.
5. Use scenario-linked financing. [researcher inference · strong] Define downside runway, base milestones, upside acceleration, and explicit replan signals; the model comes from Planning for Volatility (00:59–06:10), not from pretending the median forecast is certain.
6. Diligence the investor and board as a product. [investor-stated · strong] Ask who will help, govern, recruit, challenge, and leave; use From Seed (25:23–34:00) and Board Part II (00:34–04:43). Founder friendliness without role clarity is not governance.
7. Pre-commit harm and stop rules. [researcher inference · moderate] Translate severity, affected population, and systemic cascade into escalation owners and kill conditions using 3S; do not let “risk of inaction” waive the control system.
8. Write the failure update. [researcher inference · strong] Compare the ex-ante thesis with the outcome, as Neeva permits (2021 thesis; 2023 shutdown). Preserve what was reasonable then, identify the broken premise, and change the next screen.

Listen to these five first

[researcher inference · strong] This ranking prioritizes decision-method density, source quality, distinct named cases, and complementary coverage; it is a reader route, not Hoffman’s ranking.

1. How to Be a Great Founder (2014). The best compact founder-picking source: complementary teams, defensible contrarianism, distribution, and reference-checked learning, with creator captions.
2. From Seed to Series A (2020). The most complete early-stage process discussion, with Sarah Guo’s metric-screen view, Hoffman’s Airbnb example, diligence, and investor-as-board-partner framing.
3. The Knowledge Project #147 (2022). The strongest decision-and-governance interview: artifact depth, board horizons, rituals, crisis questions, and CEO succession. Public audio is fully timestamped despite the members-only publisher transcript.
4. Reid Hoffman #AMA with 20VC (2021). The clearest named investment and partnership-debate account, plus the boundary between a network-effects bias and a hard screen.
5. Transformational Moments (2020). The best people-and-succession case because it credits Weiner, Slavet, references, team observation, and a reversible interim role rather than presenting founder omniscience.

Access failures and unresolved record

• The old Stanford clip/PDF URLs for Long-Term, Permanent Beta, and Candid return HTML/404 rather than PDF bodies. Live Stanford printable transcripts replaced them; the Long-Term analysis remains fenced to the mapped 1:48 passage.
• The Knowledge Project publisher transcript is members-only. The public Apple audio was acquired and transcribed with the OpenAI API, so the appearance is full-access; the paid text was not bypassed.
• The old Blitzscaling During Covid YouTube video is private and several Greylock legacy pages resolve to the firm homepage or 404. Official RSS enclosures, Apple audio, SoundCloud routes, and the NBIM/Acast enclosure supplied full public audio for the ten API-transcribed records.
• CHM exposed no useful manual captions for the three retained videos; YouTube automatic captions were used with speaker and audio checks. How to Be a Great Founder used the creator-provided English caption track. Board Part II used the public SoundCloud track plus an auto-captioned YouTube mirror.
• The Possible transcript is AI-generated and lightly edited by the publisher. Masters of Scale’s two retained automatic transcripts and Tim Ferriss’s error-warning transcript are not treated as flawless verbatim text. The Masters of Scale Part 2 CMS slug currently serves Part 1's body; the official ART19 RSS metadata and public audio establish the retained Part 2 record.
• The two philosophy records and Game On are distinct episodes, but only their official full companion essays were processed; no spoken claims or timestamps are assigned. Entrepreneurship and Loneliness, Talking Politics, and Quartz are written route exceptions, not oral appearances.
• No public Greylock partner-vote record, rejected-company denominator, board minutes, LP letter, full portfolio cash flow, or reliable pass ledger was found. The talks establish public decision rules and some named recollections; they cannot establish firm-wide hit rates, ownership, returns, or sole attribution.
• Highest-value next records are the Airbnb approval/board materials; contemporaneous LinkedIn CEO-search records; Greylock remote-investment performance; Neeva investment and shutdown analysis; and a named AI investment rejected or stopped for safety/governance reasons. Until those surface, claims about efficacy remain insufficient public record.

Write the theory of the game and expose it to reality

• Rule — Define the actors, goals, rules, resources, uncertainty, feedback, and disconfirming evidence before investing; revise the theory when observations disagree. Model type: causal-thesis and probabilistic-updating model. Falsifier: decisions and outcomes show no relation to the stated game. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for the game vocabulary and thesis/revision loop; [researcher inference · moderate] for portfolio efficacy because original memos and a complete outcome denominator are missing.
• Provenance and origin — LinkedIn’s underlying 2004 founder fundraising deck is the earliest located planning artifact; Hoffman’s 2013 annotated publication adds retrospective commentary. His 2021 Game On essay explicitly contrasts complete-information games with uncertainty, feedback, coalition, and probability; because the retained body is an edited companion essay, no source-verbatim quote was kept from it. Hoffman’s later compact formulation is “You need to figure out your theory of the game.” (10 words; “The Investment Thesis”). original deck (strategy/financing slides); 2013 annotations; Game On (“Why Games Are Serious Strategy”; “What Kinds of Games Teach Strategy”); Bear Market (“The Investment Thesis”).
• Mechanism and reasoning — Making assumptions explicit converts narrative disagreement into evidence: a signal changes a probability, a milestone changes financing, and a failed premise triggers revision rather than rhetorical defense. [investor-stated · strong] Philosopher-Entrepreneur (“How Philosophy Works”; Aristotle section).
• Historical and vehicle context — The earliest artifact is a consumer-software plan with staged network and financing tests; later applications moved into Greylock, boards, SPACs, autonomy, and AI, where feedback latency and irreversible harm differ. [researcher inference · strong] LinkedIn pitch (financing slides); Aurora/Joby worksheets.
• When to use — Use for initial selection, follow-ons, pivots, board decisions, and scenario planning. Do not use as a substitute for security terms, base rates, or scientific/regulatory evidence. [researcher inference · strong] Game On (“Why Games Are Serious Strategy”).
• How to apply it — Write: objective; agents/incentives; scarce resources; causal mechanism; prior; three disconfirmers; update interval; kill/reopen threshold; and next financing action. Ask, “What observation would make us change the game?” [researcher inference · strong]
• Examples —
  • 2004 LinkedIn — documented founder-planning artifact: the fundraising deck staged professional-network proof and financing; it is not a Hoffman investment decision. [contemporaneous record · strong] original deck (network/financing slides).
  • 2018 Aurora — researcher-applied illustration: team/safety/scale theory can be tested against filed milestones and later distinct vehicles; the original IC alternative set is missing. [researcher inference · strong] Aurora case.
  • 2021–2023 Neeva — researcher-applied adverse case: the aligned-payer/differentiation theory met default distribution and willingness-to-pay failure; acquisition salvaged assets, not the consumer thesis. [researcher inference · strong] Neeva postmortem.
• Evidence — The 2004 company deck, 2021 game essay, and Stanford founder lecture are distinct placements, not independent Hoffman evidence. Independent records fix bounded facts: Aurora’s S-1 records financing, risks and later company history; Snowflake’s 10-K records the Neeva acquisition. Neither validates theory quality or causal efficacy. Aurora S-1 (business/risk factors); Snowflake 10-K (business-combinations note).
• Limits, failure modes, and exception ledger — A neat game can launder a bad prior; probabilities may be performative; slow science lacks fast feedback; coalitions can change rules. No public full IC memo/pass corpus tests calibration. The record attempted direct writings, 37 appearances, filings, and 71 deep-core rows; private IC minutes and full funnels remain unavailable. [researcher inference · strong]
• Evolution and contradictions — The 2004 plan is company-specific; 2021 philosophy/game essays generalize it; 2024 AI screens ask founders for a theory but add amplification and conflict. Coherence may be later synthesis rather than continuous prior doctrine. [researcher inference · moderate] Possible Riffs (Hoffman turns beginning “Well, there’s two parts…”).
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] AI can draft plausible narratives cheaply, so ICs should require dated priors, disconfirmers, data lineage, and an update log. It applies when feedback arrives within runway; likely misuse is calling a story a theory. Grounded case: Joby’s certification stages provide real updates while passenger targets remain forecasts. Joby Q2 2026 (certification table).

Select learning velocity and complementary teams

• Rule — Prefer founders who learn faster than the uncertainty changes and whose team covers the critical capabilities; verify through behavior and references, not charisma or credentials. Model type: people-selection model. Falsifier: learning/reference signals do not predict decision quality in a full funnel. [investor-stated · strong]
• Evidence status — [investor-stated · strong] for adaptability, complementary-team, and reference tests; [researcher inference · insufficient public record] for predictive power because rejected-company base rates and independent process evidence are absent.
• Provenance and origin — The 2014 founder lecture is the earliest located compact public selection rule; Hoffman’s short learning formulation is “infinite learning curve” (3 words, 34:21–34:27). Later SocialNet failure accounts are candidate biographical antecedents, not proof of causal origin. Candid transcript (SocialNet passage); How to Be a Great Founder (10:02–15:06; 34:16–34:27).
• Mechanism and reasoning — Startups face unknowns; changed-mind behavior, fast customer contact, truth under bad news, recruiting, and complementary roles increase error-correction capacity. References reveal repeated behavior more reliably than a pitch. [researcher inference · strong]
• Historical and vehicle context — The rule fits seed/Series A where mature metrics are sparse and partnership/board capacity matters. It is weaker when scientific, regulatory, or manufacturing proof dominates personality evidence. [researcher inference · strong] From Seed (18:03–21:14; 25:23–34:00).
• When to use — Use before seed/Series A, CEO succession, and follow-ons after surprises. Do not use to excuse missing domain capability, integrity failures, or an unfundable market. [researcher inference · strong] How to Be a Great Founder (10:02–15:06; 34:16–34:27).
• How to apply it — Ask for one consequential changed mind, one failure update, the last three references, role gaps, recruiting evidence, current learning loop, and the next result that would force team redesign. Kill on concealed bad news, no customer feedback, or an unowned critical function. [researcher inference · strong]
• Examples —
  • 1997–2000 SocialNet — retrospective failure case: delayed feedback and team isolation became later evidence for learning/support tests; contemporaneous teammate records are absent. [retrospective witness · moderate] Entrepreneurship and Loneliness (“Facing Loneliness at SocialNet”).
  • 2008–2009 Jeff Weiner — retrospective documented process: references, observation, Slavet’s introduction, and an interim role reduced succession uncertainty. [retrospective witness · strong] Transformational Moments (02:01–15:43).
  • 2010 Airbnb — researcher-applied illustration: sparse metrics did not end diligence because founder learning/product causality mattered; Slavet/Sze/partnership credit stays separate. [researcher inference · moderate] From Seed (18:03–21:14); case.
• Evidence — The founder lecture, SocialNet account, Weiner succession account, and Series A discussion are distinct placements; Aurora/Nauto theses add technical-team boundaries. They remain Hoffman/Greylock self-reports rather than independent proof that the screen predicts outcomes. Model-specific attempted routes: creator captions for the 2014 lecture (“reference,” “learn,” “adapt”); the Candid and Weiner transcripts; LinkedIn and Aurora S-1 management/risk sections; Nauto/Aurora issuer and firm theses; and name/date queries preserved in the 164-family terminal screen. Results established statements, roles and outcomes but no public reference logs, rejected-founder cohort, or matched independent outcome study. The 2-source efficacy gate therefore remains an explicit exception. Aurora S-1 (management/risk factors); Nauto (founder/wedge paragraphs).
• Limits, failure modes, and exception ledger — Retrospective accounts select coherent lessons; adaptability can mean undisciplined pivoting; references reproduce network homogeneity; complementary teams can mask unresolved authority. Original reference logs, pass cohorts, demographic funnels, and team-performance comparisons were not public after writings/talks/filing review. [researcher inference · strong]
• Evolution and contradictions — The 2014 rule de-emphasizes credentials; the 2021 MBA essay calls an MBA an adverse entrepreneurship signal but Chris Yeh supplies a countercase. Behavior and network composition, not credential, is the defensible core. [researcher inference · strong] Philosopher Versus MBA (“Top Drawbacks…”; “The Case…”).
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Require learning artifacts—customer transcripts, model-error logs, experiment decisions—and reference diversity. In AI, fast demo iteration is not customer or safety learning. Likely misuse: selecting articulate “learners” without domain proof. Manas AI is a current boundary: scientific/clinical milestones must outrank founder narrative. Manas launch (opening).

Map the network product and distribution loop

• Rule — Underwrite who joins, what recurring job creates value, how relevant participation raises value, how the loop distributes, how trust/safety degrade it, and how value is captured. Model type: network-product/distribution model. Falsifier: density does not improve retention/value or economics after channel effects. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for LinkedIn/network-first and distribution screens; [researcher inference · moderate] for portable efficacy.
• Provenance and origin — LinkedIn’s underlying 2004 founder fundraising deck states, “Where the network is the key, the first mover is critical to revenue” (13 words, slide “Establishing the Network: Our First Priority”); the 2006 social-computing discussion and later Rule of Three extend the record. The 2013 annotated publication is retrospective commentary. original deck (“Establishing the Network: Our First Priority” slide); annotations; CHM Social Computing (44:21–45:39).
• Mechanism and reasoning — Repeat value builds trusted participation; relevant density improves discovery/context; the resulting loop may reduce acquisition cost or create data/capability advantage. Fraud, multi-homing, platform dependence, concentration, and poor capture can reverse it. [researcher inference · strong]
• Historical and vehicle context — The 2004 LinkedIn plan centers staged network proof and distribution; Nauto, Coda, Blockstream and AI later extend the logic to data, collaboration, protocols, and capability networks, where safety and capture differ. [researcher inference · strong] original deck (network/strategy slides); written synthesis.
• When to use — Use for networks, marketplaces, collaboration, data flywheels, protocols, and ecosystems. Do not use when participation is incidental, supply is subsidized, or user growth does not improve retained value. [researcher inference · strong] Rule of Three (distribution/distinctiveness screen).
• How to apply it — Map nodes/sides, job, cold start, contribution, retained cohort, density unit, channel, multi-homing, trust/safety, capture, and degradation. Kill if the “network” is only audience size or paid acquisition. [researcher inference · strong]
• Examples —
  • 2004 LinkedIn — documented founder-planning application: professional identity and connections were staged before mature monetization; this is a company plan, not a Hoffman investment decision. [contemporaneous record · strong] original deck (product/network slides).
  • 2017 Nauto — investor-stated application: an immediate fleet-safety wedge generates data for a larger autonomy system; miles are exposure, not proved safety. [investor-stated · strong] Nauto (¶¶4–6).
  • 2021–2023 Neeva — adverse boundary: differentiated paid search lacked a sufficient distribution/willingness-to-pay loop. [researcher inference · strong] Taking on Tech Titans (Neeva model/investor turns); postmortem.
• Evidence — The 2004 company deck, 2006 Hoffman talk, 2009 Hoffman screen, and Greylock Nauto thesis are distinct placements, not independent proof. LinkedIn’s S-1 independently fixes company growth/risk history; Snowflake’s 10-K independently fixes Neeva’s acquisition. Neither supplies cross-company density, CAC, or retention evidence that validates the mechanism. LinkedIn S-1 (business/risk factors); Snowflake 10-K (business-combinations note).
• Limits, failure modes, and exception ledger — “Network effects” is often post-hoc; total users can hide low relevant density; data may not be lawful/defensible; subsidies can imitate compounding. No complete cohort, CAC, multi-homing, or network-attributed return series exists across the 71 rows. [researcher inference · strong]
• Evolution and contradictions — The rule moves from professional identity to marketplaces, protocols, data and AI; Hoffman explicitly says network effects are a bias, not a hard gate. [investor-stated · strong] 20VC (00:46–03:05).
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Separate model/data scale from user network effects and test platform/API dependence, privacy, provenance, safety, and unit economics. Applicable signal: retention/value rises with a defined density measure. Likely misuse: calling AI usage or proprietary data a moat. Niteshift’s Greylock-led seed and separate Hoffman backing are attribution facts, not proof of a network. Niteshift (financing paragraph).

Exploit constrained incumbents, not merely weak ones

• Rule — Invest against an incumbent only when a structural incentive, architecture, channel, margin, regulation, or customer conflict prevents an effective response. Model type: competitive-strategy model. Falsifier: incumbents can copy/distribute economically before the startup reaches a defensible wedge. [investor-stated · strong]
• Evidence status — [investor-stated · strong] for the constrained-incumbent rule; [retrospective witness · moderate] for SpaceX; [researcher inference · strong] for Neeva as falsifying control.
• Provenance and origin — The 2020-11-10 Choosing Bad Competition essay gives the explicit rule—“The great strategy is not to try to find a competition-free market.” (Hoffman, 12 words, opening)—and uses his initial SpaceX pass/later invest as the changed-mind example. Choosing Bad Competition (opening and SpaceX passages).
• Mechanism and reasoning — A startup wins not because the incumbent is stupid but because serving the new job would cannibalize economics, violate architecture, disrupt channel/status, or require a different risk tolerance. [researcher inference · strong]
• Historical and vehicle context — The 2020 essay addresses incumbent response; regulated physical markets and AI add partner, acquisition, bundling, financing, and regulatory response paths. [researcher inference · moderate] Choosing Bad Competition (competition passages); Inflection decision (paras.29–34).
• When to use — Use in incumbent-challenge theses and follow-on reviews. Do not use when the only claim is better UX, founder speed, or moral superiority. [researcher inference · strong] Choosing Bad Competition (competition screen).
• How to apply it — Name incumbent; new job; exact constraint; likely response; time-to-copy/distribute; startup wedge; capital needed before response; and evidence that the constraint persists. Kill if incumbent response is profitable and compatible. [researcher inference · strong]
• Examples —
  • SpaceX — decision date not publicly reconstructed: Hoffman says he initially overweighted incumbent strength, then recognized incumbent misalignment; the public account omits the original pass date, vehicle and memo. [retrospective witness · moderate] source (SpaceX passage).
  • 2021 Neeva — documented thesis/adverse outcome: ad-funded search conflict was real, but subscription/default distribution did not produce a sustainable consumer company. [researcher inference · strong] Neeva lab.
  • 2014 Blockstream — researcher-applied illustration: open trust-minimized infrastructure was positioned against centralized control; the source does not connect itself to the later constrained-incumbent rule, and capture/governance/economics remain unresolved. [researcher inference · moderate] Blockstream (thesis paragraphs).
  • 2024 Inflection/Microsoft — documented incumbent-response case: licensing and personnel transfer show an incumbent can respond without conventional product copying or acquisition. [researcher inference · strong] CMA decision (paras.29–34).
• Evidence — The competition essay, Neeva transcript, Blockstream thesis, and SpaceX account are distinct Hoffman/firm placements, not independent records. Snowflake independently fixes Neeva’s acquisition; the CMA independently documents Microsoft’s nontraditional Inflection response. Neither proves the constrained-incumbent rule predicts returns. Snowflake 10-K (Neeva acquisition note); CMA decision (paras.29–34).
• Limits, failure modes, and exception ledger — Constraints can disappear through reorganization/acquisition; regulation can protect incumbents; a startup may face a different bottleneck. Survivor examples do not show base rates. No original SpaceX pass memo or portfolio-wide incumbent-response comparison was public. [researcher inference · strong]
• Evolution and contradictions — The rule corrects a simple “big company is slow” heuristic. AI creates tension because Microsoft and other incumbents can partner and hire rather than build conventionally, as the Inflection transaction shows. [researcher inference · strong] Microsoft proxy (Inflection transaction).
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Require a written constraint and response tree; stress-test bundling, compute, data, distribution, regulatory and M&A options. Likely misuse: celebrating disruption without estimating incumbent response. Current Inflection is a grounded example of a nontraditional response—license plus personnel—rather than clean displacement. CMA decision (paras.29–34).

Buy speed only when critical scale pays

• Rule — Accept temporary efficiency debt only when a large market, product-market fit, relative competition, durable scale advantage, survival capital, operational scalability, and bounded harm make delay more costly. Model type: capital-tempo/scaling model. Falsifier: speed does not improve defensibility or increases irreversible harm/financing risk. [investor-stated · strong]
• Evidence status — [investor-stated · strong] for conditional doctrine; [researcher inference · moderate] for efficacy because systematic comparison and cash returns are absent.
• Provenance and origin — The 2016 HBR interview gives the earliest mature formulation: Hoffman calls it “the science and art of rapidly building out a company” (10 words, p.4, “What is blitzscaling?”). 2018 talks clarify relative speed and inefficiency; Yeh’s 2018 intellectual history documents revision. HBR (pp.4–8); How We Wrote (¶¶10–20).
• Mechanism and reasoning — Speed can capture density, learning, channel, standards or supply before rivals; capital buys time but also technical, organizational, governance and reputational debt. [researcher inference · strong]
• Historical and vehicle context — The 2016 formulation addressed software companies pursuing critical scale; the 2020 shock and Hoffman’s 2022 bear-market essay made runway and efficiency constraints explicit. Deep tech adds physical/safety milestones; SPACs add redemption and public-governance incentives. [researcher inference · strong] HBR (pp.4–8); Bear Market (capital sections).
• When to use — Use after causal product evidence when scale advantage is perishable. Do not use before product learning, without runway, or where harm/technical debt is irreversible. [researcher inference · strong] HBR (pp.4–8).
• How to apply it — Name critical scale, competitor clock, scale advantage, evidence threshold, inefficiencies purchased, runway, debt repayment date, 3S harms, and stop rule. Kill if scale is not durable or financing fails before proof. [researcher inference · strong]
• Examples —
  • Airbnb 2009–2010 — researcher-applied sequence: handcrafting preceded acceleration; manual learning is the boundary, not contradiction. [researcher inference · strong] Chesky transcript (pp.2–4); Airbnb case.
  • 2020 Covid — contemporaneous stress test: Hoffman made speed relative to changed demand/capital, not a universal accelerator. [investor-stated · strong] Covid audio (25:50–31:46).
  • 2019 WeWork / 2021 Hippo — adverse boundaries: WeWork’s moat/margins/operations and Hippo’s early underwriting/repricing show headline growth/public capital cannot prove scale quality or sponsor return. [researcher inference · strong] WeWork note; Hippo case.
• Evidence — HBR, CHM 2018, Covid audio, and the Chesky transcript are distinct doctrine/application placements, not independent proof. O’Reilly and Lee/Kim provide independent criticism/analysis; issuer filings independently fix Joby/Hippo company states. Those sources bound rather than validate causal efficacy. O’Reilly; Lee/Kim (abstract, §§4.5–4.6); Joby Q2.
• Limits, failure modes, and exception ledger — Critical scale is often asserted after success; growth can be subsidy; capital can postpone falsification; harm may arrive after market capture. O’Reilly and Lee/Kim supply criticism but not a matched Hoffman-portfolio test. [contemporaneous record · moderate] O’Reilly; Lee/Kim (abstract, §§4.5–4.6).
• Evolution and contradictions — Doctrine moved from speed to responsibility, relative competition and scarce-capital efficiency. The conditionality is genuine, yet many controls were articulated after public criticism. [researcher inference · moderate] Responsible Blitzscaling.
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Charge every inefficiency to density, learning, distribution or standards; track runway and debt repayment. AI compute can amplify both speed and harm. Likely misuse: treating model usage or financing as critical scale. Joby’s certification path shows capital cannot buy regulatory completion instantly. Joby Q2.

Plan A/B/Z and finance across states

• Rule — Maintain a current thesis (A), adjacent alternative (B), and survival/restart state (Z), each with triggers, actions, milestones, and financing. Model type: scenario/optionality model. Falsifier: scenarios do not change decisions or consume runway without retiring uncertainty. [investor-stated · strong]
• Evidence status — [investor-stated · strong] for staged plans, permanent-beta risk, and 2020 volatility planning; [researcher inference · moderate] for portfolio performance.
• Provenance and origin — LinkedIn’s underlying 2004 founder fundraising deck uses staged financing scenarios; the 2012 Permanent Beta talk generalizes intelligent risk, closing with Hoffman’s “being in permanent beta” (4 words, closing Q&A). Planning for Volatility makes states operational. Hoffman’s 2013 annotated pitch is retrospective commentary. original deck (financing slides); annotations; Permanent Beta (closing Q&A); Planning (00:59–06:10).
• Mechanism and reasoning — A state plan preserves survival and upside while preventing one median forecast from controlling hiring/capital. Triggers convert uncertainty into pre-agreed action. [researcher inference · strong]
• Historical and vehicle context — LinkedIn's staged software plan differs from Aurora/Joby capital intensity and from SPAC deadlines; scenario value depends on capital duration, security, reserves and feedback latency. [researcher inference · strong] LinkedIn pitch (financing slides); Joby worksheet.
• When to use — Use at investment, annual planning, shocks, follow-ons and financing. Do not use when options are not operationally or legally feasible. [researcher inference · strong] Planning (00:59–06:10).
• How to apply it — For A/B/Z specify probability range, cash date, milestone, hiring, product/channel action, follow-on amount, covenant/board approval, trigger and owner. Kill plans without decision-linked signals. [researcher inference · strong]
• Examples —
  • 2004 LinkedIn — documented founder-planning application: the fundraising deck staged alternate plans and capital needs; it is not an investment decision by Hoffman. [contemporaneous record · strong] original deck (financing slides).
  • 2020 pandemic portfolios — contemporaneous application: planning and Virtual VC adapted runway, diligence and portfolio allocation under changed assumptions. [investor-stated · strong] Planning (00:59–06:10); Virtual VC (23:09–25:30).
  • 2021 Joby — researcher-applied boundary: certification stages, cash burn and future capital must replace a single launch forecast. [researcher inference · strong] Joby worksheet.
• Evidence — The company pitch, Stanford transcript, and two Hoffman pandemic recordings are distinct placements, not independent evidence. Joby and Hippo filings independently fix later milestone, liquidity and operating states; they do not show that A/B/Z planning caused results. No matched forecast-accuracy or reserve-allocation series was found. Joby Q2; Hippo Q2.
• Limits, failure modes, and exception ledger — Scenarios can be unfalsifiable or too broad; option value costs dilution/carry/time; Plan Z may harm stakeholders. No public forecast-accuracy or reserve-allocation series exists. The corpus searched Covid audio, writings, filings and 71 rows; private planning/LP records remain absent. [researcher inference · strong]
• Evolution and contradictions — The early career metaphor becomes a board/portfolio financing discipline in 2020; the 2022 bear-market essay then stresses runway and efficiency. Permanent beta is merged here and in the pivot model, not counted separately. [researcher inference · strong] Bear Market (capital sections).
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Tie downside/base/upside to cash, compute, regulation, hiring and follow-ons. Observable signal: a trigger changes action within a board cycle. Likely misuse: decorative scenario decks. Current Hippo improvement after early repricing shows why states and time horizons must stay separate from sponsor return. Hippo Q2.

Trade decision speed against reversibility, harm, and inaction

• Rule — Decide faster when delay is costly and action is reversible; slow down or escalate as severity, breadth affected, systemic cascade, irreversibility, and conflict rise. Model type: decision-window/risk-governance model. Falsifier: faster decisions do not improve outcomes after reversibility and harm controls. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for decision windows, risk of inaction and 3S; [researcher inference · moderate] for a unified rule and efficacy.
• Provenance and origin — Calculating Risk (2020-07-30) is the earliest located compact timing rule; Talking Politics (2020-08-13) adds severity, scope and systemic effects, and How to Decide (2021-03-23) operationalizes speed/accountability. Hoffman writes: “better” doesn’t just mean “more accurate.” It also means “faster.” (10 words, “What is Good Decision-Making?”). Later AI talks compare action and inaction. Calculating Risk (risk/runway sections); How to Decide (“What is Good Decision-Making?”); 3S (“EVALUATING IMPACT…”).
• Mechanism and reasoning — Reversible experiments generate information; irreversible harms require evidence and independent control. Delay also has costs, but naming them prevents “uncertainty” from becoming paralysis. [researcher inference · strong]
• Historical and vehicle context — Software deployment encouraged fast iteration; autonomy, biotech, elections and AI raise third-party/systemic stakes. Director and sponsor conflicts make recusal part of risk design. [researcher inference · strong]
• When to use — Use for deployments, financing, public positions, boards and crisis actions. Do not use “risk of inaction” when beneficiaries, red lines or stop authority are undefined. [researcher inference · strong] How to Decide (decision-window/accountability).
• How to apply it — Score reversibility; severity; affected population; cascade; delay cost; information gain; owner; conflict; red line; rollback. Set a deadline proportional to reversibility and harm. [researcher inference · strong]
• Examples —
  • 2020 pandemic — contemporaneous application: Hoffman advocated scenario-linked action amid rapidly changing assumptions. [investor-stated · strong] Planning (00:59–06:10).
  • 2021 Aurora de-SPAC — documented governance control: Hoffman recused from both sides’ valuation/terms discussions, separating conflict from speed. [contemporaneous record · strong] S-4 (“Background…Hoffman’s role”).
  • 2023–2026 AI — boundary case: deployment-first advocacy acknowledges amplification but does not identify a portfolio investment rejected for safety. [investor-stated · moderate] CHM AI (24:39–25:17); NBIM (11:04–13:42).
• Evidence — Decision/risk writings, the 3S essay, Quartz rebuttal, and AI appearances are distinct Hoffman/coauthored placements, not independent proof. Aurora’s S-4 and Microsoft’s proxy independently record two recusals; they establish conflict controls, not the unified risk rule’s efficacy. Aurora S-4 (“Background…Hoffman’s role”); Microsoft proxy (Inflection transaction/recusal).
• Limits, failure modes, and exception ledger — Scores can hide value judgments; reversibility for company may be irreversible for users; inaction risk can rationalize acceleration; conflicted owners may set weak red lines. No adjudicated portfolio-wide risk-decision log exists. [researcher inference · strong]
• Evolution and contradictions — Early intelligent-risk language becomes explicit decision windows and 3S; AI increases urgency while leaving stop rules less specific. The tension is unresolved, not averaged away. [researcher inference · moderate]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Require independent escalation for AI, biotech, autonomy, security and civic systems; pilot only if rollback is real. Likely misuse: equating competitive urgency with social permission. Grounded example: Microsoft’s Inflection recusal record shows conflict controls can coexist with transaction speed. Microsoft proxy.

Pivot the whole system when the fatal premise breaks

• Rule — When evidence kills a foundational premise, revisit product, market, distribution, team, capital, governance, and timing together; preserve what still works and stop defending the original label. Model type: system-pivot model. Falsifier: coordinated pivots do not improve learning or survival relative to narrower changes. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for testing and changing the thesis; [retrospective witness · strong] for SocialNet/LinkedIn; [researcher inference · strong] for the whole-system synthesis; [researcher inference · insufficient public record] for efficacy.
• Provenance and origin — The earliest located public articulation is the 2014 founder lecture’s theory-and-flexibility rule; later failure and pivot essays make the change mechanism explicit. Hoffman’s pivot formulation is “Failure isn’t the goal; learning is the goal.” (8 words, “What Leads to Pivots?”). SocialNet is a retrospective candidate antecedent, not proven causal origin. Founder lecture (10:02–15:06); Startup Pivot (“What Leads to Pivots?”); Myths About Failure (SocialNet/LinkedIn passages).
• Mechanism and reasoning — A fatal premise often links several design choices; changing only product can leave the same broken channel, authority, or financing clock. A coordinated reset converts sunk-cost defense into a new falsifiable system. [researcher inference · strong]
• Historical and vehicle context — Software pivots may change within one runway; regulated or capital-intensive companies face slower evidence, security-holder consent, and higher shutdown costs. Acquisition can preserve assets while falsifying the original standalone thesis. [researcher inference · strong] Neeva outcome.
• When to use — Use after a falsified demand, distribution, technical, regulatory, leadership, or financing premise. Do not use when evidence only calls for normal iteration. [researcher inference · strong] Startup Pivot (pivot framework).
• How to apply it — Name the killed premise; freeze hindsight; map dependencies; retain validated assets; rewrite product/market/channel/team/governance/capital; set a new proof date and kill rule. [researcher inference · strong]
• Examples —
  • 1998–2000 PayPal — retrospective application: the pivot essay attributes multiple product/market changes to the PayPal team; it is a later account, not a contemporaneous decision log. [retrospective witness · strong] Startup Pivot (PayPal pivot passage).
  • 1997–2000 SocialNet — adverse boundary: later accounts say feedback/distribution problems persisted; no contemporaneous memo proves a coordinated pivot process. [retrospective witness · moderate] Myths About Failure (SocialNet passage).
  • 2021–2023 Neeva — adverse boundary: the consumer search thesis ended; Snowflake acquired technology and people, but the original product/distribution model did not survive. [researcher inference · strong] Neeva lab.
• Evidence — The founder lecture and pivot/failure essays supply doctrine and the PayPal/SocialNet retrospective accounts; Snowflake’s filing independently fixes the Neeva outcome without proving the pivot mechanism. Model-specific attempted routes included the direct writings, 37 retained appearances, PayPal/LinkedIn/Aurora/Joby/Smartmatic filings, and 71-row deep-core outcome ledger. They produced no second independent decision-process record or matched pivot study, so the 2-source application gate remains an explicit exception. Snowflake 10-K (business-combinations note).
• Limits, failure modes, and exception ledger — “Pivot” can rename drift, preserve weak teams, or evade a stop; acquisition may hide failure economics. Private board/IC records and matched pivot-versus-stop outcomes were unavailable after direct writings, 37 appearances, and 71 deep-core reviews. [researcher inference · strong]
• Evolution and contradictions — Permanent beta is an individual/organizational posture; this model is the capital-decision rule activated by a killed premise. Later doctrine is more coherent than the sparse contemporaneous SocialNet record, so continuity remains unproved. [researcher inference · moderate]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Require a dependency map and fresh financing plan, not an AI feature relabel. Applicable signal: one failed premise forces coordinated changes with a new proof date. Likely misuse: preserving valuation while the mechanism stays broken. Neeva is the grounded warning. Neeva lab.

Build boards as help, control, and succession systems

• Rule — Design the board around stage-specific capability, independent control, conflict handling, CEO support and succession; distinguish advice, observation, approval, and fiduciary authority. Model type: governance-capability model. Falsifier: capability/independence design does not improve intervention or succession quality. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for help/control and board-composition doctrine; [retrospective witness · strong] for Weiner/Airbnb; [contemporaneous record · strong] for Aurora recusal; [researcher inference · insufficient public record] for board/succession efficacy.
• Provenance and origin — The earliest located compact articulation is the 2017 Better Board essay: “A board of directors is a team, not a collection of ‘all-stars’” (Hoffman, 12 words, ¶7). Written Part I appeared in 2020; the acquired public Part II appearance is dated 2021-01-26 and adds stage, composition, and control. Better Board (¶7); Board Part I (composition/control sections); Board Part II (00:34–04:43).
• Mechanism and reasoning — Complementary directors widen information and recruiting; independent authority constrains conflicts; staged succession avoids binary founder-versus-board shocks. Confusing help with control obscures who can decide. [researcher inference · strong]
• Historical and vehicle context — Personal angels and observers may help without votes; Greylock directors act for fund/company governance; SPAC sponsors and affiliate investors create recusal needs. [researcher inference · strong] vehicle evidence.
• When to use — Use at investment, new stage, CEO transition, conflict, follow-on, and exit. Do not use celebrity names as substitutes for missing capability or independence. [researcher inference · strong] Board Part I (composition/control sections).
• How to apply it — Map seats, votes, observer rights, fiduciary duty, capability, information, conflicts, recusal, CEO feedback, succession triggers, and one owner per governance gap. [researcher inference · strong]
• Examples —
  • 2008–2009 LinkedIn — retrospective succession case: Slavet’s introduction, interim operating work, references, and observation preceded Weiner’s CEO appointment. [retrospective witness · strong] Transformational Moments (02:01–15:43).
  • 2010 Airbnb — role boundary: [retrospective witness · strong] Slavet supported the deal, Sze was skeptical, and the partnership approved it. Airbnb Reflections (partner-process paragraphs). [retrospective witness · moderate] Greylock's current profile separately says, “Reid was a board observer for Airbnb”; founder/team results do not become observer results. Greylock profile (role-summary paragraph beginning “Reid was a board observer for Airbnb”; accessed 2026-08-27).
  • 2021 Aurora/Reinvent — contemporaneous conflict control: the S-4 records Hoffman’s recusal from both sides’ valuation and terms discussions. [contemporaneous record · strong] Aurora S-4 (“Background of the Business Combination — Hoffman’s role”).
• Evidence — Board essays/audio and the Weiner/Airbnb accounts are distinct Hoffman/firm placements, not independent proof. Aurora’s S-4 and Microsoft’s proxy independently record conflict recusals; they do not validate board capability or succession efficacy. Searches across LinkedIn/Aurora filings, board transcripts, 37 appearances and the 71-row ledger found no public board minutes or matched transition study, so the 2-source mechanism gate remains an explicit exception. Aurora S-4; Microsoft proxy.
• Limits, failure modes, and exception ledger — Public board lists rarely reveal votes, dissent, intervention quality, or succession alternatives; retrospective accounts center visible actors. No complete board-minute or CEO-transition comparison was public. [researcher inference · strong]
• Evolution and contradictions — The doctrine moves from director quality to explicit help/control and conflict design. Airbnb shows influence without a vote; Aurora shows a famous participant stepping out of the decision. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Add AI/safety/regulatory capability only where authority, information, and independence are real. Applicable signal: every critical gap and conflict has a named owner and escalation path. Likely misuse: advisory theater. Current application is a hypothetical IC rule; Aurora’s filed recusal remains the grounded historical minimum for conflict clarity. Aurora note.

Assemble ecosystems for missing capabilities

• Rule — Underwrite the company and the external system needed to make it work: partners, regulators, suppliers, data, standards, customers, and public institutions; allocate interfaces, incentives, dependencies, and failure ownership. Model type: ecosystem-capability model. Falsifier: partner assembly adds dependency without accelerating proof or defensibility. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for co-design and capability-network claims; [contemporaneous record · strong] for named Nauto/Aurora partnerships; [researcher inference · insufficient public record] for general efficacy.
• Provenance and origin — Nauto’s 2017 fleet/data thesis and Aurora’s 2018 partner thesis are early located investment applications; the 2020 adaptability conversation makes co-design explicit in Hoffman’s phrase “co-design your go-to-market” (3 words, 09:33–09:47). Nauto (fleet/data paragraphs); Aurora (partnership paragraphs); Adaptability (09:33–09:47).
• Mechanism and reasoning — Frontier products often lack an internal capability, legitimacy, channel, or data source. Complementary partners can close the gap, but concentration and conflicting incentives can relocate rather than remove risk. [researcher inference · strong]
• Historical and vehicle context — Network software emphasized users and distribution; autonomy, health, energy, and AI require physical, regulatory, scientific, and public partners. A fund or board can convene but does not own partner execution. [researcher inference · strong]
• When to use — Use where one company cannot independently supply certification, manufacturing, distribution, data, standards, or trust. Do not call ordinary vendors an ecosystem. [researcher inference · strong] Adaptability (08:27–11:40).
• How to apply it — Map missing capability, partner, contract/right, interface, incentive, dependency, substitute, data ownership, regulator, failure owner, and financing milestone. [researcher inference · strong]
• Examples —
  • 2017 Nauto — investor-stated application: fleet safety was the immediate wedge; fleet data and automaker relationships were necessary to the wider autonomy thesis. [investor-stated · strong] Nauto thesis (¶¶4–6).
  • 2018 Aurora — investor-stated application: experienced founders plus automaker/transport partners were presented as a safe path to scale; later securities and transaction economics remain separate. [investor-stated · strong] Aurora thesis (team/partner paragraphs).
  • 2017 Entrepreneur First — investor-stated selection application: Greylock’s thesis treated talent, cohort, mentorship and capital as an entrepreneur-formation system before a company idea; later outcome economics are not public. [investor-stated · strong] EF thesis (talent/platform paragraphs).
• Evidence — Two firm theses and a timestamped Hoffman interview are distinct placements; an independent host does not make Hoffman’s claim independent. Aurora’s S-1 independently records partnerships/risks and Joby’s Q2 filing independently records the company/FAA dependency. Neither establishes that ecosystem assembly caused performance. Searches across issuer risk factors, firm theses, the Adaptability transcript and 71 rows found no matched partnership-outcome study, so the 2-source efficacy gate remains an explicit exception. Aurora S-1 (business/risk factors); Joby Q2 (certification table).
• Limits, failure modes, and exception ledger — Partner logos can mask unsigned rights, channel dependence, data capture, or veto risk. The public record lacks contract terms and matched partnership outcomes across the 71 rows. [researcher inference · strong]
• Evolution and contradictions — The network idea expands from participant value to institutional capability. That is not the same as a user network effect and is kept separate; dependency may weaken rather than strengthen a moat. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] For AI/biotech/autonomy/energy, diligence the dependency graph with contracts and substitutes. Applicable signal: each external capability has an enforceable interface and failure owner. Likely misuse: counting partnerships as traction. Joby’s Q2 2026 certification table is a grounded current example of company/FAA dependency with separately measured progress. Joby Q2 2026 (certification table).

Fence actor, vehicle, objective, and conflict

• Rule — Attribute every capital decision to actor × legal vehicle × transaction × beneficiary × decision right; separate investment, founder equity, board service, donation, advocacy, sponsorship, and counterparty roles. Model type: attribution/incentive-control model. Falsifier: role separation does not change decision or economic interpretation. [researcher inference · strong]
• Evidence status — [contemporaneous record · strong] for filed securities, sponsor and recusal facts; [investor-stated · strong] for some personal-versus-firm disclosures; this is mainly a researcher audit rule reconstructed from repeated role conflicts.
• Provenance and origin — No investor-origin quote applies: this is a researcher-reconstructed audit rule, not a Hoffman maxim. Its components recur in the 2014 Blockstream/Xapo distinction, Reinvent filings, OpenAI/Inflection records, and nonprofit controls. Blockstream (personal/firm vehicle paragraph); vehicle note.
• Mechanism and reasoning — Incentives, ownership, approval, conflict and proceeds attach to a legal actor, not a famous name. The fence prevents firm/company value, donations, or board influence from becoming personal investment return. [researcher inference · strong]
• Historical and vehicle context — Hoffman’s record spans founder equity, personal angels, Greylock funds, Discovery, Reinvent sponsors/PIPE, Reprogrammed, Aphorism, Microsoft, Kiva, Callisto, and Endeavor. Terms and beneficiaries differ. [contemporaneous record · strong] vehicle comparison.
• When to use — Use before selection credit, governance analysis, valuation, ranking, conflict judgment, and return calculation. Do not infer vehicle from “Hoffman-backed.” [researcher inference · strong] people map.
• How to apply it — Record legal owner; capital source; security; check; allocation; approval; board/observer right; recusal; beneficiary; follow-on; proceeds; current holding; unknowns. [researcher inference · strong]
• Examples —
  • 2014 Blockstream/Xapo — investor-stated role split: Hoffman described personal Blockstream participation and a distinct Greylock Xapo investment. [investor-stated · strong] Blockstream thesis (vehicle paragraph).
  • 2018/2021 Aurora — contemporaneous vehicle split: Greylock Series A, Reinvent sponsor/PIPE, and affiliates are distinct; the S-4 records recusal. [contemporaneous record · strong] Aurora case.
  • 2015–2025 OpenAI/Inflection — mixed-role boundary: donation, reported Aphorism investment, OpenAI board, Microsoft board/counterparty, cofounding, and recusal cannot be pooled. [contemporaneous record · strong] AI capital note.
• Evidence — The direct thesis and analyst notes are distinct placements, not independent. SEC, issuer/company and nonprofit records independently establish vehicles, ownership language, roles and recusals; they support the attribution fence but cannot supply missing private economics. Reinvent case; Microsoft proxy (Inflection transaction/recusal).
• Limits, failure modes, and exception ledger — Private SPVs, carried-interest allocation, side letters, beneficial ownership, and donation instruments are often missing. The fence can bound attribution but cannot manufacture cash flows. [researcher inference · strong]
• Evolution and contradictions — Role multiplicity grows after Greylock through sponsors, affiliates, nonprofits and AI counterparties. Public storytelling often compresses these roles; filings reopen them. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Put the actor/vehicle/conflict block at the top of every memo and AI-generated portfolio report. Applicable signal: cap table, filing, board record, and narrative reconcile. Likely misuse: “strategic investor” without owner or beneficiary. The Inflection proxy is the grounded recusal example. Microsoft proxy.

Audit outcomes with denominators and cash flows

• Rule — Never infer investment performance from company outcome, round size, valuation, IPO price, current holding, or aggregate fund value; require a declared denominator and actor/vehicle security, cost, dilution, proceeds, remaining position, waterfall, and time. Model type: evidence/performance-audit model. Falsifier: audited actor-level cash flows match the shortcut conclusions. [researcher inference · strong]
• Evidence status — This is a researcher-reconstructed audit rule, [researcher inference · strong], supported by repeated filing boundaries; Hoffman does not publish it as a standalone doctrine.
• Provenance and origin — No investor-origin quote applies: the rule emerged from researcher reconciliation as of 2026-08-27 of the 164-name screen, 81 deep-core dispositions, 71 included relationships, and only four strict reconstructions; it is not back-attributed to Hoffman. coverage; economics gaps.
• Mechanism and reasoning — Security-holder economics require linked cash flows. A headline company value can coexist with dilution, preferences, sponsor warrants, partial sales, retained shares, fund waterfalls, or no actor ownership. Denominator choice also changes apparent hit rates. [researcher inference · strong]
• Historical and vehicle context — Founder equity, personal angels, LP funds, SPAC sponsors/PIPEs, affiliates, donations and fund-backing controls have incompatible economics; the 71-row set is selection-biased and not career coverage. [researcher inference · strong] vehicle table.
• When to use — Use for rankings, “wins,” losses, ownership, exits, attribution, and peer comparisons. Do not report return when any required cash-flow link is absent. [researcher inference · strong] ranking boundary.
• How to apply it — Freeze population; dedupe company but separate transactions; identify actor/vehicle/security; collect cost/follow-ons/dilution/proceeds/remaining value/waterfall/time; label unknowns; sensitivity-test outlier removal. [researcher inference · strong]
• Examples —
  • 2016 LinkedIn — measured anchor: 14,489,899 scheduled shares at $196 imply gross cash merger consideration for those shares; basis, taxes, prior sales, and full proceeds remain unknown. [contemporaneous record · strong] final proxy (“Security Ownership,” consideration table).
  • 2021 Hippo/Joby — adverse public-value boundary: issuer results and security prices do not establish sponsor or personal return without ownership and cash flows. [researcher inference · strong] Reinvent note.
  • 2023 Neeva — acquisition boundary: Snowflake recorded $185.4m consideration, but no retained source links that amount to Hoffman proceeds or return. [researcher inference · strong] Snowflake 10-K (business-combinations note).
• Evidence — SEC proxy/10-K/S-4 records, vehicle documents, and the explicit count reconciliation provide independent placements. SEC SPAC rule (p.402, Figure 2); denominator note.
• Limits, failure modes, and exception ledger — Cash flows, carry, taxes, cost basis, reserves and rejected-company denominators are mostly private; conservatism can understate real contribution. The rule bounds claims rather than calculating missing alpha. [researcher inference · strong]
• Evolution and contradictions — Public profiles and retrospectives compress company success into named-investor narratives; later filings disaggregate ownership and transactions. The model is an analyst control, not a later Hoffman conversion. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Make generated portfolio summaries fail closed when actor/security/denominator is missing. Applicable signal: a reviewer can trace every value to a security and owner. Likely misuse: ranking investors by headline exits. Wealthfront’s terminated UBS sale followed by a 2025 IPO shows why event identity must also be checked. Wealthfront filings (row 65).

Use alliances and network intelligence as reciprocal information systems

• Rule — Build time-bounded, reciprocal missions among founders, employees, partners, and investors; use diverse relationships to acquire information, test judgment, recruit, and transition without pretending employment is family. Model type: alliance/information model. Falsifier: explicit reciprocal missions and diverse reference networks do not improve information or transitions. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for alliance and network-intelligence doctrine; [retrospective witness · strong] for LinkedIn/Weiner; [researcher inference · insufficient public record] for investment-selection efficacy.
• Provenance and origin — The 2012 Permanent Beta talk states network intelligence; the 2014 Alliance excerpt formalizes tours of duty, mutual benefit and network intelligence as “a mutually beneficial deal, with explicit terms, between independent players” (Hoffman, Ben Casnocha, and Chris Yeh, 10 words, p.8). Later talent writing extends it. Permanent Beta (network/permanent-beta passages); Alliance excerpt (p.8); Networked Age (alliance/network sections).
• Mechanism and reasoning — Explicit missions reduce false permanence; reciprocal networks create references, introductions, external signals and future collaboration. Diversity and consent are necessary because dense networks can become echo chambers or extraction systems. [researcher inference · strong]
• Historical and vehicle context — The rule began as career/employment doctrine and appears in sourcing, diligence, recruiting, partnership and succession. It is not itself a network effect or proof of investment alpha. [researcher inference · strong]
• When to use — Use for references, recruiting, founder/investor expectations, partnerships, board transitions, and portfolio communities. Do not use to demand one-way loyalty. [researcher inference · strong] Alliance excerpt (pp.8–18).
• How to apply it — Write mission, horizon, mutual contributions, information rights, network access, checkpoints, conflicts, transition, alumni relationship, and a diversity test for references. [researcher inference · strong]
• Examples —
  • 2008–2009 LinkedIn/Weiner — retrospective application: Slavet’s introduction, interim operating role, observation, and references created a staged CEO transition. [retrospective witness · strong] Transformational Moments (02:01–15:43).
  • 2010 Airbnb — researcher-applied sourcing case: repeated trusted referrals and references preceded the meeting; this is not evidence of a long prior founder relationship. [researcher inference · strong] Airbnb Reflections (referral/reference-check paragraphs).
  • 2018 Aurora — researcher-applied partnership illustration: founders and transport/automaker relationships were part of the published thesis, but the source does not call this an application of the Alliance framework; contract economics remain unknown. [researcher inference · moderate] Aurora thesis (team/partnership paragraphs).
• Evidence — Stanford transcript, book excerpt, talent essay, Weiner account and firm/company case records provide more than four placements. They are repeated Hoffman/coauthored/Greylock accounts, not independent proof of causal efficacy. Independent filings can confirm dates and roles but not that the Alliance mechanism caused outcomes; searches of company filings and the completed corpus found no matched alliance-performance study, so the 2-source application gate remains an explicit model-specific shortfall. LinkedIn S-1 (management/ownership); Aurora S-1 (business/partnership risks).
• Limits, failure modes, and exception ledger — Networks reproduce privilege, references can collude, “mutuality” can mask employer/investor power, and tours can normalize disposability. No complete sourcing funnel or matched alliance-outcome dataset exists. [researcher inference · strong]
• Evolution and contradictions — Permanent beta’s individual adaptability becomes a negotiated employer/employee alliance and then a diligence/governance tool. This model merges generic “network intelligence” slogans into a reciprocal decision rule. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Record consent, reciprocity and network diversity; distinguish human references from AI-generated reputation signals. Applicable signal: both parties can name value, horizon, checkpoint and exit. Likely misuse: unpaid portfolio labor or “family” rhetoric. This is a hypothetical current IC application; the staged 2008–2009 Weiner process is the grounded historical example. Transformational Moments (02:01–15:43).

Preserve exit optionality; do not confuse listing with return

• Rule — Compare private continuation, financing, M&A, IPO, shutdown and asset sale as capability-and-stakeholder choices; an exit event is not a return until security-holder cash flows are linked. Model type: strategic-option/exit model. Falsifier: explicit option comparison does not improve mission, financing or realized-value decisions. [researcher inference · strong]
• Evidence status — [investor-stated · strong] for considering M&A versus IPO; [contemporaneous record · strong] for filed transaction mechanics; [researcher inference · strong] for the unified optionality/return fence.
• Provenance and origin — The 2020-07-23 M&A-or-IPO essay states that “the question of M&A or IPO is a question of means, not ends.” (Hoffman, 13 words, ¶47); LinkedIn, Wealthfront, Neeva and SPAC records expose different outcomes. M&A or IPO (¶47); outcome ledger.
• Mechanism and reasoning — Financing and governance preserve or close future paths. A buyer can add distribution/capability but impose mission/integration costs; a listing raises capital and liquidity but creates public obligations; shutdown or asset sale may be rational after a fatal premise. [researcher inference · strong]
• Historical and vehicle context — Founder equity, fund securities, sponsor warrants and personal angels receive different consideration and timing. Public listing, merger announcement, close, and shareholder liquidity are distinct events. [contemporaneous record · strong] LinkedIn final proxy (“Security Ownership,” merger-consideration table).
• When to use — Use in financing, strategic reviews, buyer offers, IPO readiness, shutdown and postmortem. Do not let headline valuation substitute for stakeholder allocation or mission cost. [researcher inference · strong] M&A or IPO (¶¶33–48).
• How to apply it — Compare standalone capital/runway, buyer capability, integration, control, employee/customer impact, security waterfall, taxes, timing, regulatory risk, mission, and reopening conditions; then separately calculate actor-level realization. [researcher inference · strong]
• Examples —
  • 2016 LinkedIn/Microsoft — documented strategic and economic event: merger consideration is filed and actor-linked scheduled shares are visible, but basis/tax/full realized return remain unknown. [contemporaneous record · strong] LinkedIn final proxy (“Security Ownership,” merger-consideration table).
  • 2022–2025 Wealthfront — documented path change: the UBS acquisition was terminated; Wealthfront later completed a Nasdaq IPO, so an announcement was not the exit and an IPO was not actor-level return. [contemporaneous record · strong] UBS termination; 424B4.
  • 2023 Neeva/Snowflake — adverse strategic salvage: the consumer thesis ended while technology/team were acquired; company consideration is not Hoffman proceeds. [researcher inference · strong] Neeva lab; Snowflake 10-K.
• Evidence — Direct exit doctrine, two SEC transaction records, an issuer termination, and the Neeva acquisition note exceed four placements and two independent sources. M&A or IPO (¶¶12–18, 33–48).
• Limits, failure modes, and exception ledger — Boards may lack a real alternative; retention/mission claims are hard to verify; preferences, taxes and fund waterfalls remain private. Public outcomes cannot establish counterfactual value. [researcher inference · strong]
• Evolution and contradictions — The strategic essay emphasizes capability and mission; later records show partial, terminated, recapped and asset outcomes. Exit optionality complements A/B/Z but is kept distinct because it concerns control and security-holder paths. [researcher inference · strong]
• Modern VC translation (as of 2026-08-27) — [researcher inference · strong] Model acqui-hire, licensing, recapitalization and secondary liquidity separately from M&A/IPO. Applicable signal: the board compares paths and stakeholder waterfalls before exclusivity. Likely misuse: “IPO-ready” as thesis validation. OpenAI’s 2025 recap and Wealthfront’s terminated-sale/IPO sequence are grounded current boundaries. OpenAI structure; Wealthfront 10-K.

Evidence gaps and candidate models

The inventory reconciles 14 recurring decision rules. Candidates were merged or excluded rather than split to inflate the count:

• Permanent beta is provenance/posture for A/B/Z and whole-system pivot, not a separate capital rule. [researcher inference · strong] Permanent Beta (risk/permanent-beta passages); Startup Pivot (pivot framework).
• Rule of Three supplies mass-audience/distribution evidence for model 3, categorical distinctiveness as a precursor or boundary—not the full constrained-incumbent mechanism—in model 4, and capital-efficiency/later-financing evidence for models 5–6. It contains no founder-learning or team-complementarity test and lacks a distinct mechanism after its actual components are separated. [researcher inference · strong] Rule of Three (numbered screen).
• Blitzcapital, relative speed and responsible blitzscaling are stages/boundaries of conditional speed and state financing. [researcher inference · strong] HBR (pp.4–8); Responsible Blitzscaling.
• Superagency, professional-VC defense and general public-good rhetoric have public arguments but not three distinct documented capital applications plus independent outcome evidence. [researcher inference · moderate] AI access boundary; Professional VC (claimed advantages); Sunny Day Fund (proposal).
• MBA adverse signal, “make everyone a hero,” philosophy credentials, RuneQuest/Catan metaphors and isolated slogans inform selection or provenance but do not independently clear the 3-case/4-placement/2-source gate. [researcher inference · strong] MBA essay (“Top Drawbacks…”; “The Case…”); Game On (named games).
• Impact capital is not treated as one mechanism: Kiva, Callisto, Endeavor and OpenAI have different instruments, beneficiaries and decision rights and belong under the actor/vehicle fence. [researcher inference · strong] impact/vehicle note; AI capital note.

[researcher inference · strong] No model has portfolio-level causal validation: private memos, full pass funnels, decision minutes, ownership/cash flows and matched base rates remain unavailable. “Independent” here means a substantively independent issuer, filing, witness, or analysis—not a mirror, host, or repeated Hoffman interview.

What the evidence supports

The package supports a falsifiable process thesis, not a finding of investment alpha: Hoffman’s public method combines relationship-derived information with explicit theories of founder learning, product/distribution compounding, incumbent constraint, complementary capability, conditional speed, scenario planning, governance, and revision. The method is visible across dated writings and talks and is consistent with selected decisions, but no complete opportunity set, matched comparison group, or actor-level return series proves that it outperformed. [researcher inference · moderate] Philosophy thesis; 14-model field guide; economic boundary.

Three conclusions survive adversarial review:

1. The doctrine is real but evolved. The 2004 LinkedIn deck is a founder-planning artifact; later writings articulate theory-of-game, permanent beta, conditional speed, boards, alliances, and risk. The record does not prove those mature rules existed at Stanford, SocialNet, or PayPal. [researcher inference · strong] Profile chronology; 2004 deck (strategy/network/financing slides); chronology.
2. Role and vehicle separation materially changes the record. David Sze/Greylock, not Hoffman, receive LinkedIn and Facebook institutional-round credit; Airbnb's interested retrospective records Slavet support and Sze skepticism while Greylock's current profile separately lists Hoffman as observer; Aurora’s Greylock financing and Reinvent transaction are separate; Microsoft recorded an Inflection-related recusal. [researcher inference · strong] LinkedIn/Facebook audit; Airbnb/Aurora audit; Greylock profile (role-summary paragraph beginning “Reid was a board observer for Airbnb”; accessed 2026-08-27); Microsoft proxy (Related Person Transactions, Inflection).
3. Visible economic significance is concentrated but personal performance is unmeasured. LinkedIn’s final proxy schedules 14,489,899 Hoffman shares at $196, or $2,840,020,204 gross cash consideration for shares. Basis, taxes, transfers, liabilities, earlier sales, and complete cash flows are missing; outside LinkedIn no large actor-linked cash-flow anchor permits a defensible return ranking. [contemporaneous record · strong] Final merger proxy (“Security Ownership” and merger-consideration table); Pareto boundary.

The strongest adverse cases are not footnotes. Neeva shows that product/business-model differentiation can fail against default distribution and acquisition economics; Hippo separates later company improvement from sponsor return; Stripe and SpaceX show domain priors can cause misses; SocialNet is an operating failure recalled later. [researcher inference · strong] Failure ledger; Neeva postmortem.

What could falsify the thesis

The process thesis would be strengthened by dated memos across winners, ordinary cases, passes, and failures that repeatedly specify actors, founder-learning evidence, product/distribution causality, incumbent constraint, critical scale, alternative states, vehicle/beneficiary, governance gaps, and predeclared stop or pivot triggers. It would be weakened or overturned by any of the following. [researcher inference · strong]

• Access explains selection: a complete sourced/opportunity funnel shows that PayPal, LinkedIn, and Greylock access—not the reconstructed decision rules—explains entry and outcomes. [researcher inference · strong] Sourcing limits.
• Rules arrived after success: contemporaneous memos omit the later mechanisms or contradict retrospective winner accounts. [researcher inference · strong] Profile causal boundary.
• No economic edge: audited, stage/sector-matched actor-and-vehicle cash flows show ex-LinkedIn performance is ordinary or worse after fees, dilution, write-offs, and opportunity cost. [researcher inference · strong] Retrieval ledger.
• Partner/team attribution dominates: IC minutes and founder records show partners sourced, diligenced, approved, governed, or exited the material cases while later narratives overassign Hoffman. [researcher inference · strong] People map.
• Mechanisms do not predict: prospective use of the 14 rules does not improve decision calibration, milestone quality, or stop discipline relative to a preregistered baseline. Models 2, 8, 9, 10, and 13 already disclose insufficient independent efficacy evidence. [researcher inference · insufficient public record] Mental models.
• Harm overwhelms scale benefits: matched evidence shows speed systematically degrades governance, safety, or stakeholder welfare even when critical-scale premises appear present. [researcher inference · moderate] O’Reilly critique; Lee/Kim (abstract; §§4.5–4.6).

Attribution and denominator audit

The arithmetic reconciles, but none of the populations is a career-performance denominator:

The 14-case Pareto set is 19.7% of 71 and is explicitly explanatory, not 80% of realized gains. Company financing, enterprise value, acquisition consideration, affiliate holdings, and public share prices remain company/entity measures unless linked to the actor’s security, basis, proceeds, and waterfall. [researcher inference · strong] Pareto opening.

The review corrected one repeated attribution defect in the source guides: the Airbnb record supports David Sze’s skepticism and partnership debate, not formal dissent. The corrected wording now appears in WRITINGS and TALKS, consistent with the direct retrospective Airbnb account (partner-discussion paragraphs). [retrospective witness · moderate]

Transfer limits for modern VCs

The modern translation is dated 2026-08-27 and should be used as hypotheses/checklists, not as Hoffman doctrine or guaranteed playbooks. [researcher inference · strong] Modern translation.

• Network effects: require cohort retention, interaction density, trust/safety costs, and a causal distribution loop. A famous network or user count is not the mechanism. [researcher inference · strong] Network model.
• Blitzscaling: buy speed only after demonstrating critical-scale payoff, relative urgency, financing capacity, and bounded operational/harm debt. The doctrine’s historical cases do not supply a 2026 market base rate. [researcher inference · moderate] Conditional-speed model; Greylock 18 (opening/model paragraphs).
• AI/deep tech: separate scientific milestones, regulation, counterparties, and public-benefit claims from financing or valuation. Joby certification percentages and OpenAI Foundation control are current company facts, not Hoffman returns. [researcher inference · strong] Joby Q2 exhibit (certification table); OpenAI structure (structure/ownership).
• Governance: board help, director control, observer access, sponsor incentives, and recusal are different. Record the conflict, beneficiary, approval owner, and abstention before assigning credit. [researcher inference · strong] Board model.
• Vehicles: personal angel, LP fund, Discovery pool, sponsor/PIPE, affiliate, foundation, donation, impact support, and fund backing cannot be pooled. The public record lacks most fees, carry, reserves, and distributions. [researcher inference · strong] Vehicle comparison.
• Failures: predeclare the fatal premise and next evidence. Neeva suggests that a structurally differentiated product still needs paid-retention and default-distribution tests; public evidence does not show Greylock formally adopted this prescription. [researcher inference · moderate] Neeva exercise.
• Peer comparison: the Hoffman/Sze/Conway/Feld comparison is dimension-specific researcher synthesis, not a rank or alpha table. Different capital forms, partnership processes, and documentation levels constrain transfer. [researcher inference · moderate] Peer table.

Unresolved evidence

The highest-value unresolved records are:

1. Cash flows: personal/foundation/affiliate ledgers; Greylock/Discovery/Reinvent capital accounts; security cost, dilution, sales, distributions, fees, carry, taxes, write-offs, and remaining positions. Without these, DPI, TVPI, MOIC, IRR, PME, hit rate, and true concentration are not computable. [researcher inference · strong] Economic retrieval map.
2. Decision records: original memos, IC votes, alternatives, partner-role logs, board consents, intervention logs, reserve decisions, and stop/reopen dates. Airbnb’s retrospective process is unusually detailed but still lacks original terms and lifecycle economics. [researcher inference · strong] Airbnb reconstruction.
3. Opportunity denominator: complete sourced, diligenced, passed, lost, and failed populations. The 164-name bounded screen cannot be promoted to career coverage. [researcher inference · strong] Denominator limits.
4. Model efficacy: independent prospective or matched studies for founder learning, pivot quality, boards/succession, ecosystems, and alliances. Repeated Hoffman interviews are placements, not independent validation. [researcher inference · insufficient public record] Evidence-gap ledger.
5. Contested/current records: underlying calendar or institutional records for the 2026 Epstein-contact correction, adjudicated OpenAI conflict/governance records, and live role/title updates. The package preserves procedural and evidence boundaries rather than inferring knowledge or guilt. [researcher inference · strong] Profile contestation.
6. Outer 83: per-name source routes were not retained or revalidated. None of that screening-only pool is used in evidence, rankings, reconstruction, or performance analysis. [researcher inference · strong] Denominator note.

Central-claim audit trail

Every high-risk thesis, quotation family, economic figure, role/vehicle attribution, denominator, failure, peer comparison, and modern translation was rechecked against the stated source class and locator.

Reproducible ordinary-claim audit trail

Sampling rule. For each of the eight core documents, divide nonblank claim-bearing body lines into quartiles and inspect the first externally checkable ordinary claim at or after the 25% and 75% marks, advancing past a claim already audited as central. For this review itself, start in its first and fifth substantive H2s and advance forward across later H2s when the next claim is already in the central audit; this yields the source-count claim in H2 3 and the outer-83 claim in H2 5. This deterministic spread produced 16 ordinary claims—two per core document—rather than hand-picking only famous cases. [researcher inference · strong]

Independent material deal and team cross-checks

Six rows above have substantively independent issuer, filing, regulator, participant, witness, or analysis routes; Airbnb is retained as an additional bounded witness/vehicle check and is not misrepresented as independently proven partnership dialogue. [researcher inference · strong]

§11.1 acceptance-evidence matrix

Review QA evidence and verdict

• Eight core documents inspected in intended order: profile, sources, writings, talks, investments/cases, philosophy, mental models, and this review.
• High-risk central audit: 24 claim families covering economic figures, denominators, roles, vehicles, failures, peers, modern translation, and current-role/contestation boundaries; the separate quotation manifest audits 110 atomic-note quote bullets plus reader-core duplicate/origin placements row by row.
• Ordinary spread: 16 deterministic claims, two per core document.
• Material attribution checks: 7 rows; at least 6 use substantively independent records, with Airbnb explicitly bounded.
• Acceptance evidence: 20/20 rows pass after the Airbnb wording correction.
• Reconciled corpus arithmetic: WRITINGS 45/42 = 33 full + 5 partial + 4 inaccessible with 3 indexes; TALKS 73 = 37 + 36; DEALS 164 = 81 + 83, core 81 = 71 + 10, included 71 = 67 investment claims + 4 controls, strict 4 + unresolved 67; sources 195 processed / 0 unchecked.

The acceptance verdict is PASS with material uncertainty preserved. The package supports a coherent, testable investing-process reconstruction and a rigorous teaching architecture. It does not establish career completeness, comparative alpha, or ex-LinkedIn actor-level performance. [researcher inference · strong]

Direct writings, books, and named investment theses

• ★★★ | Reid Hoffman writing index | https://www.reidhoffman.org/writing/ | 279 dated entries, 2012-10-01 through 2025-06-15; accessed 2026-08-24 | first-party archive | Inventory route, not content evidence. WRITINGS selected explicit investing, governance, risk/failure, named-deal, and blitzscaling-evolution works; non-investment topical/political/promotional posts, duplicates, and audio/video were excluded reproducibly.
• ★★★ | Reid Hoffman books index | https://www.reidhoffman.org/books/ | current; accessed 2026-08-24 | first-party bibliography | Canonical six-book map with co-authors; distinguishes authored doctrine from speeches and editorial summaries.
• ★★★ | Impromptu: Amplifying Our Humanity Through AI | https://www.impromptubook.com/wp-content/uploads/2023/03/impromptu-rh.pdf | 2023 | free first-party book PDF | Full 247-page text with printed-PDF-page locators; a dated GPT-4 travelogue and the only complete book body reached through the priority bibliography in this phase.
• ★★★ | Superagency: What Could Possibly Go Right with Our AI Future | https://www.simonandschuster.com/books/Superagency/Reid-Hoffman/9798893310108 | 2025-01-28 | official publisher page; Reid Hoffman with Greg Beato | Full 288-page body not acquired; title, authors, date, length, and publisher synopsis accessible, so no book-body claim is permitted.
• ★★★ | Masters of Scale: Surprising Truths from the World’s Most Successful Entrepreneurs | https://www.penguinrandomhouse.com/books/669987/masters-of-scale-by-reid-hoffman-with-june-cohen-and-deron-triff/ | 2021-09-07 | official publisher page; Reid Hoffman with June Cohen and Deron Triff | Full 304-page body not acquired; description and official reader’s guide accessible and analyzed only as partial publisher material.
• ★★★ | Blitzscaling: The Lightning-Fast Path to Building Massively Valuable Companies | https://www.penguinrandomhouse.com/books/557261/blitzscaling-by-reid-hoffman-and-chris-yeh/ | 2018-10-09 | official publisher page; Reid Hoffman and Chris Yeh | Full 336-page body not acquired; description is accessible, while direct doctrine is tested through the separately mapped 2016 HBR text and 2022 Hoffman update.
• ★★★ | The Alliance: Managing Talent in the Networked Age | https://www.theallianceframework.com/wp-content/uploads/2021/10/the-alliance_chapter.pdf | 2014-07-08 | official 20-page Chapter 1 excerpt; Reid Hoffman, Ben Casnocha, and Chris Yeh | Partial body: Chapter 1 accessible with physical/printed page locators; Chapters 2–8 and appendices were not acquired.
• ★★★ | The Startup of You (Revised and Updated) | https://www.penguinrandomhouse.com/books/212453/the-startup-of-you-revised-and-updated-by-reid-hoffman-cofounder-and-chairman-of-linkedin-and-ben-casnocha/ | original 2012-02-14; revised edition released 2022 | official publisher page; Reid Hoffman and Ben Casnocha | Revised body not acquired; publisher synopsis and co-author’s 2022 edition note accessible, so body-level evolution claims remain out of scope.
• ★★★ | LinkedIn’s Series B Pitch to Greylock | https://www.reidhoffman.org/linkedin-pitch-to-greylock/ | current; accessed 2026-08-24 | annotated first-party deck | Detailed retrospective annotations on product, network effects, pitch strategy, and partnership; pair with the original deck and S-1.
• ★★★ | My Rule of Three for Investing | https://techcrunch.com/2009/04/19/reid-hoffman-my-rule-of-three-for-investing/ | 2009-04-19 | first-person essay/interview | Early, pre-Greylock investor criteria; the 60-plus web-venture figure is a single-source TechCrunch host characterization, not a Hoffman self-report or auditable denominator.
• ★★★ | Stimulus 2.0: It’s the Startups, Stupid | https://techcrunch.com/2009/03/04/stimulus-20-it%E2%80%99s-the-startups-stupid/ | 2009-03-04 | first-person policy essay | Decision-era capital-allocation argument during the financial crisis; government-match proposal and portfolio-breadth premise require policy and adverse-selection scrutiny.
• ★★★ | If, Why, and How Founders Should Hire a “Professional” CEO | https://www.linkedin.com/pulse/20130123161202-1213-if-why-and-how-founders-should-hire-a-professional-ceo | 2013-01-23 | first-person operating essay | Full text on founder self-assessment, staged CEO succession, product ownership, and the founder/new-CEO relationship; retrospective LinkedIn evidence, not universal causal proof.
• ★★★ | The Future of the Bitcoin Ecosystem and “Trustless Trust” — Why I Invested in Blockstream | https://www.linkedin.com/pulse/20141117154558-1213-the-future-of-the-bitcoin-ecosystem-and-trustless-trust-why-i-invested-in-blockstream/ | 2014-11-17; company reprint https://blog.blockstream.com/en-reid-hoffman-on-the-future-of-the-bitcoin-ecosystem/ | full first-person named-investment thesis | Original body accessible in guest SSR; explicit vehicle fence: personal Blockstream first-round investment after Greylock discussion versus separately identified Greylock investment in Xapo; no amount, ownership, or return disclosed.
• ★★★ | Managing Talent in a Networked Age, Part I | https://techcrunch.com/2014/11/09/managing-talent-in-a-networked-age/ | 2014-11-09 | first-person operating essay | Full accessible derivative application of The Alliance to Greylock portfolio advice; not independent corroboration of the co-authored book.
• ★★★ | In Praise of Professional Venture Capital | https://www.linkedin.com/pulse/praise-professional-venture-capital-reid-hoffman | 2019-12-17 | full first-person investing essay | Defends professional, incentive-aligned capital allocation and board assistance against government/family-capital alternatives; advocacy by an interested VC, not comparative performance evidence.
• ★★★ | Is Blitzscaling Over? | https://www.linkedin.com/pulse/blitzscaling-over-reid-hoffman/ | 2020-05-07 | full Hoffman/Chris Yeh essay; original guest-accessible body | Pandemic-era boundary conditions: relative speed, winner-take-most dynamics, and WeWork as a misapplication; Airbnb prediction is dated ex-ante evidence, not proof of the general rule.
• ★★ | Blitzscaling During Covid | https://www.reidhoffman.org/books/blitzscaling/ | 2020-06-18 | official archive record and landing-page metadata; Hoffman/Chris Yeh audio | WRITINGS remains partial (title/subtitle only); TALKS completed the official Greymatter RSS audio with OpenAI API transcription and timestamped claims in t-blitzscaling-during-covid.md.
• ★★★ | VC Investing During Covid | https://www.linkedin.com/pulse/vc-investing-during-covid-reid-hoffman | 2020-07-15 | full first-person Greylock-practice note | Written preview of Hoffman, Saam Motamedi, and Chris Yeh discussion: remote diligence, geographic reach, changed runway targets, and incubation; claims are firm self-report without cohort outcomes.
• ★★★ | Calculating Risk | https://greylock.com/greymatter/calculating-risk/ | 2020-07-30 | full Hoffman/Chris Yeh Greylock essay | Decision windows, pre-emptive financing, people risk, and severity/scope/systematicity framework; LinkedIn and Aurora examples are retrospective.
• ★★★ | Myths About Failure | https://greylock.com/greymatter/reid-hoffman-myths-about-failure/ | 2020-08-20 | full first-person Greylock essay | Replaces “fail fast” with thesis-led learning; SocialNet failure and PayPal pivot preserve team/board attribution but remain retrospective cases.
• ★★★ | The Startup Pivot | https://greylock.com/reid-hoffman/reid-hoffman-the-startup-pivot/ | 2020-09-08 | full first-person Greylock essay | Explicit investment-thesis updating, A/B/Z planning, capability refactoring, and five-step PayPal retrospective; no PayPal co-founder credit imputed.
• ★★★ | Building a Great Startup Board, Part 1 | https://greylock.com/reid-hoffman-building-a-great-startup-board-pt-i/ | 2021-01-19 | full first-person Greylock essay | Board-as-supportive-catalyst doctrine, stage model, attribution rule, and SocialNet/PayPal learning; not minutes or outcome evidence.
• ★★ | Building a Great Startup Board, Part 2 | https://greylock.com/building-a-great-startup-board-pt-2/ | 2021-01-26 | official page introduction plus embedded audio | WRITINGS remains partial (three introductory paragraphs); TALKS completed the public SoundCloud/YouTube mirror with audio-checked automatic-caption timestamps in t-building-board-part2.md.
• ★★★ | Bear Market Blitzscaling | https://greylock.com/reid-hoffman/reid-hoffman-bear-market-blitzscaling/ | 2022 | Greylock essay/transcript | Full direct-voice update to the growth doctrine under constrained capital; headings are stable locators.
• ★★★ | Airbnb Reflections | https://greylock.com/portfolio-news/airbnb-reflections/ | 2020 | Greylock transcript | Best first-party route for the first Greylock deal, earlier non-investment, Sze skepticism, and Slavet support; the separate current Greylock Hoffman profile supplies the observer-not-director status.
• ★★★ | Our Investment in Coda | https://greylock.com/portfolio-news/our-investment-in-coda/ | 2017-10-19 | named firm investment thesis | Hoffman-authored Series A rationale and board role; direct thesis, not outcome or return proof.
• ★★★ | Our Investment in Nauto | https://greylock.com/portfolio-news/our-investment-in-nauto/ | 2017-07-19 | named firm investment thesis | Decision-era autonomous-systems rationale; useful for Web 2.0-to-deep-tech evolution.
• ★★★ | Our Investment in Aurora: A Safe Path to Scale Autonomous Vehicles | https://medium.com/greylock-perspectives/our-investment-in-aurora-a-safe-path-to-scale-autonomous-vehicles-bd0680fa4fe0 | 2018-02-28 | named thesis/repost | Hoffman-authored thesis mirror; pair with Aurora filing for director, financing, and vehicle facts.
• ★★★ | Congrats, Aurora | https://greylock.com/portfolio-news/congrats-aurora/ | 2021-11-04 | Greylock retrospective/transcript | Public-debut retrospective and financing sequence; susceptible to success-selection bias.
• ★★ | A New Paradigm in Human-Machine Interaction | https://greylock.com/blog/a-new-paradigm-in-human-machine-interaction/ | 2022-03-08 | company-formation announcement | Suleyman-authored Inflection launch naming Hoffman and Simonyan; compare with CMA and Microsoft wording.
• ★★★ | My New Investment in Change.org, The Global Hub for Positive Social Impact | https://www.linkedin.com/pulse/my-new-investment-changeorg-global-hub-positive-social-reid-hoffman/ | 2017-05-26 | full first-person named personal impact-investment thesis | Explicitly personal rather than Greylock; Hoffman says any realized equity appreciation will go to nonprofits, but supplies no investment amount, instrument, ownership, enforcement, or realized outcome.
• ★★ | Our Investment in Entrepreneur First | https://news.greylock.com/our-investment-in-entrepreneur-first-5a0903f25bd4/ | 2017-09-11 | Hoffman-authored Greylock named-investment thesis; partial primary access | Official archive exposes the opening excerpt; same-day TechCrunch, EF, and Mosaic records preserve Greylock lead, Hoffman board role, operating-company/fund boundary, and direct interview language, but do not restore the missing body.
• ★★ | Our Investment in Callisto | https://news.greylock.com/our-investment-in-callisto-950d74bfd43f | 2018-03-20 | Hoffman-authored Greylock named-support post; partial primary access | Official archive exposes the opening excerpt; same-day VentureBeat preserves a Hoffman quotation and identifies Greylock's support as a donation to a nonprofit, not personal or conventional equity capital.
• ★★★ | Theranos: Blitzscaling, Blitzfailing, or Blitzfrauding? | https://www.linkedin.com/pulse/theranos-blitzscaling-blitzfailing-blitzfrauding-reid-hoffman/ | 2018-08-28 | full first-person anti-case essay | Distinguishes iterative product embarrassment from deception and scales verification to harm severity/scope; retrospective public critique, not evidence of a Hoffman investment or pre-exposure detection.
• ★★★ | Responsible Blitzscaling | https://www.linkedin.com/pulse/responsible-blitzscaling-reid-hoffman/ | 2018-11-28 | full Hoffman/Chris Yeh excerpt adapted from Blitzscaling | Joins velocity to moral, cultural, and regulatory responsibility; full short excerpt but partial book access, not independent corroboration of the inaccessible book.
• ★★★ | WeWork: Blitzscaling or Blitzflailing? | https://www.linkedin.com/pulse/wework-blitzscaling-blitzflailing-reid-hoffman/ | 2019-10-21 | full first-person anti-case essay | Explicit market/distribution/margin/moat screen plus product-market-fit and operational limiters; ex-post analysis after the failed offering, not an ex-ante investment decision.
• ★★★ | M&A or IPO? | https://greylock.com/reid-hoffman/reid-hoffman-ma-or-ipo/ | 2020-07-23 | full first-person exit-strategy essay | Strategic value creates IPO/acquisition optionality; PayPal price sequence and LinkedIn integration claims are retrospective, with vehicle economics and a complete Greylock outcome denominator undisclosed.
• ★★★ | Re-inventing the SPAC | https://reid.medium.com/re-inventing-the-spac-3fcbbd941aec | 2020-10-05 | full first-person Reinvent vehicle thesis | Hoffman identifies Reinvent Technology Partners with Mark Pincus and Michael Thompson as a SPAC offering “venture capital at scale”; separate from Greylock and personal angels, with sponsor terms and later outcomes requiring filings.
• ★★★ | Choosing Bad Competition | https://greylock.com/greymatter/choosing-bad-competition/ | 2020-11-10 | full first-person market-selection essay | Three-part incumbent-failure taxonomy with LinkedIn and a candid initial SpaceX rejection/later-investment anti-case; later SpaceX vehicle and economics are not disclosed.
• ★★★ | How to Decide | https://greylock.com/reid-hoffman/how-to-decide/ | 2021-03-23 | full first-person decision-process essay | Defines quality as speed plus accuracy, makes learning visible in repeat-founder selection, and uses SocialNet/PayPal as retrospective evidence; embedded audio not needed for the complete written body.
• ★★★ | Sunny Day Fund: Why We Need to Invest In Our Shared Energy Future Now | https://www.linkedin.com/pulse/sunny-day-fund-why-we-need-invest-our-shared-energy-future-hoffman/ | 2015-12-08 | full first-person allocation essay | Public-private climate-energy capital doctrine: patient private investment at the laboratory-to-commercialization gap; coalition participation is not a disclosed personal check, vehicle interest, or return.
• ★★★ | Blitzcapital — sometimes crazy, sometimes essential to great technology businesses | https://www.linkedin.com/pulse/blitzcapital-sometimes-crazy-essential-great-reid-hoffman/ | 2017-05-15 | full first-person allocation essay | Capitalization continuum and critical-scale test: finance speed only when winner-take-most competition makes time decisive; bootstrap when economics and competition permit, with no financing/outcome denominator disclosed.
• ★★★ | Building a better board of directors | https://www.linkedin.com/pulse/building-better-board-directors-reid-hoffman | 2017-09-12 | full first-person governance essay | Distinct director-selection doctrine: capability matrix, stage fit, time commitment, executive relationships, succession, and CEO-undermining boundary; examples remain retrospective.
• ★★ | How We Wrote Blitzscaling | https://www.linkedin.com/pulse/how-we-wrote-blitzscaling-reid-hoffman/ | 2018-10-03 | full guest intellectual-history essay by Chris Yeh on Hoffman's profile | Participant account of framework formation, Stanford teaching test, discarded manuscript, responsibility expansion, and network feedback; doctrine history, not efficacy evidence.
• ★★ | Greylock legacy blog index | https://greylock.com/blog-old/ | current; accessed 2026-08-24 | archive index | Discovery route, not content evidence. WRITINGS followed canonical bodies for the selected governance, risk, failure, pivot, and blitzscaling works and excluded duplicate, non-Hoffman, and audio-only records.

Talks, interviews, podcasts, and oral histories

• ★★★ | Investing in Long-Term Ideas | https://stvp.stanford.edu/node/12521/printable/print | 2011-06-30 | Stanford printable transcript; mapped 1:48 clip | Live transcript replaces a dead PDF; analysis is fenced to the long-term-investing clip.
• ★★★ | Live Life in Permanent Beta | https://stvp.stanford.edu/node/12431/printable/print | 2012-02-22 | Stanford printable transcript | Full direct-voice route on intelligent risk, networks, adaptation, and career strategy.
• ★★★ | Candid Fireside Chat on the Entrepreneurial Journey | https://stvp.stanford.edu/node/10701/printable/print | 2024 | Stanford printable fireside transcript | Current retrospective on SocialNet failure, PayPal, LinkedIn, risk, resilience, and AI.
• ★★★ | Economic Club of New York conversation | https://www.econclubny.org/documents/10184/109144/2018HoffmanTranscript.pdf | 2018 | event transcript PDF | Forty-page direct interview on PayPal, networks, and blitzscaling.
• ★★★ | How to Be a Great Founder | https://www.youtube.com/watch?v=dQ7ZvO5DpIw | 2014-11-04 | Stanford/YC video; creator captions | Founder-selection lecture with creator-provided English subtitles and timestamps.
• ★★★ | Blitzscaling | https://assets-global.website-files.com/6446dd70aaa07871c795c474/6537e2547084885cc62e3375_HBR%20-%20Blitzscaling.PDF | 2016-04 | HBR edited interview PDF | Eight-page direct interview on market, capital, and organizational conditions.
• ★★★ | Social Computing: From Message Boards to Social Networks | https://www.youtube.com/watch?v=PZJvgR3NM18 | 2006 | Computer History Museum video; automatic captions | Contemporaneous LinkedIn trust, identity, and network evidence.
• ★★ | Blitzscaling: How Silicon Valley Learned to Grow | https://www.youtube.com/watch?v=iHpt877wmHE | 2018 | Computer History Museum video; automatic captions | Mature public doctrine, with Chris Yeh’s co-authorship preserved.
• ★★ | AI + Humanity | https://www.youtube.com/watch?v=9sCUFUtjmoE | 2023 | Computer History Museum video; automatic captions | Current AI amplification, risk, and governance arguments.
• ★★★ | Reid Hoffman oral history | https://achievement.org/achiever/reid-hoffman/ | interview 2014-09-13; page later updated | dated Q&A inside living biography | Q&A processed; contaminated host biography excluded.
• ★★★ | Possible Riffs 007 — AI Investments | https://www.possible.fm/podcasts/riffs007/ | 2024-07-31 | official AI-generated, lightly edited transcript | Current AI-founder tests with the publisher’s machine-text caveat.
• ★★ | Strategy Session: Early Adopters | https://mastersofscale.com/strategy-session-early-adopters/ | 2019-09-26 | official automatic transcript | Transferable early-adopter attributes and competition-dependent product-market-fit/distribution sequencing, audio spot-checked.
• ★★ | Reid Hoffman Confronts the AI Critics | https://mastersofscale.com/reid-hoffman-confronts-the-ai-critics/ | 2023-11-07 | official automatic transcript acquired before current Cloudflare block; RSS duration 23:36 | Hoffman’s responses on comparative risk and the pause letter; missing-opposition and affiliation caveats preserved.
• ★★★ | The Reid Hoffman Story (Part 2): Make everyone a hero | https://rss.art19.com/masters-of-scale | 2019-02-26 | official ART19 public audio; OpenAI API transcript | Exact GUID gid://art19-episode-locator/V0/asxj18qKrADugx8qjg2CB5mG46vms_ixtffmXdysEs4; PayPal relative speed, Friendster angel thesis, LinkedIn conviction/acquisition loops, CEO succession, and Greylock entrepreneur-first culture. The CMS route misroutes to Part 1, so official RSS audio controls.
• ★★★ | Brian Chesky — Handcraft the Core Experience | https://mastersofscale.com/wp-content/uploads/2018/09/mos-episode-transcript_-brian-chesky.pdf | 2017–2018 | speaker-attributed transcript PDF | Chesky owns operating evidence; Hoffman owns narrative frame.
• ★★ | Ten Commandments for Startup Success | https://tim.blog/2017/06/26/10-commandments-startup-success/ | 2017-06-26 | publisher transcript | Founder and risk heuristics; publisher error warning retained.
• ★★★ | The Tim Ferriss Show #101 — Reid Hoffman and Michael McCullough | https://tim.blog/wp-content/uploads/2018/08/101-reid-hoffman-and-michael-mccullough.pdf | 2015 | transcript PDF | Network, founder, succession, and operating evidence with co-speaker credit.
• ★★ | The Knowledge Project #147 — Better Decisions, Fewer Mistakes | https://fs.blog/knowledge-project-podcast/reid-hoffman/ | 2022-09-20 | public audio; OpenAI API transcript | Members-only written transcript replaced by lawfully public audio and timestamps.
• ★★★ | Washington Post Live — The Path Forward: Artificial Intelligence | https://www.washingtonpost.com/washington-post-live/2023/05/10/transcript-path-forward-artificial-intelligence-with-reid-hoffman/ | 2023-05-10 | publisher-edited event transcript | Full AI deployment, risk, regulation, and incentive discussion.
• ★★ | In Good Company — Shaping the AI Era | https://www.nbim.no/en/news-and-insights/podcast/2026/reid-hoffman-shaping-the-ai-era-investing-in-transformation-and-calling-on-europe/ | 2026-02-25 | NBIM/Acast audio; OpenAI API transcript | Current AI investing and incumbent-adoption evidence with role fences.
• ★★★ | Video: Reid Hoffman pitches LinkedIn to himself | https://www.youtube.com/watch?v=z0-wSApbhq4 | 2024-10-03 | Hoffman video; automatic captions | Retrospective/AI reconstruction kept separate from the 2004 deck.
• ★★★ | Taking on Tech Titans | https://greylock.com/greymatter/taking-on-tech-titans/ | 2021-08-17 | official full transcript | Ex-ante Neeva thesis and structural-competition test; later failure retained as anti-case.
• ★★ | Philosopher Versus MBA | https://greylock.com/reid-hoffman/philosopher-versus-mba/ | 2021-07-27 | official companion essay | Distinct episode’s written body; no oral locators claimed.
• ★★ | The Philosopher-Entrepreneur | https://greylock.com/reid-hoffman/the-philosopher-entrepreneur/ | 2021-07-20 | official companion essay | Philosophy-to-practice and Coda intellectual-history evidence.
• ★★★ | People-First Capitalism | https://greylock.com/greymatter/airbnb-brian-chesky-people-first-capitalism/ | 2021-06-08 | official full transcript | Distinct Airbnb crisis, IPO, stakeholder, and board-cadence evidence.
• ★★ | Game On | https://greylock.com/greymatter/reid-hoffman-game-on/ | 2021-05-18 | official companion essay | Strategy-game origin of uncertainty, feedback, and coalition models.
• ★★★ | Reid Hoffman #AMA with 20VC | https://soundcloud.com/greylock-partners/ama-with-20-min-vc-reid-hoffman | 2021-01-13 | official RSS audio; OpenAI API transcript | Airbnb partnership skepticism and network-effects investment thesis.
• ★★ | Entrepreneurship and Loneliness | https://greylock.com/reid-hoffman/entrepreneurship-and-loneliness/ | 2020-12-15 | official podcast-derived essay | SocialNet failure versus LinkedIn support-network biography.
• ★★★ | Building Ecosystems of Adaptability | https://soundcloud.com/greylock-partners/reid-hoffman-interview-greymatter-audio | 2020-08-27 | official RSS audio; OpenAI API transcript | Product/GTM co-design and crisis adaptation.
• ★★ | Talking Politics | https://greylock.com/greymatter/talking-politics/ | 2020-08-13 | Hoffman prose associated with episode | 3S governance rule; written route exception, not oral evidence.
• ★★★ | From Seed to Series A | https://soundcloud.com/greylock-partners/the-secrets-of-raising-a-series-a | 2020-08-06 | official RSS audio; OpenAI API transcript | Hoffman/Guo early-stage selection, diligence, and board-partner evidence.
• ★★★ | The Virtual VC | https://soundcloud.com/greylock-partners/the-virtual-vc | 2020-07-16 | official RSS audio; OpenAI API transcript | Hoffman/Motamedi remote-diligence and firm-process evidence.
• ★★★ | Transformational Moments | https://soundcloud.com/greylock-partners/transformational-moments-and-learnings-from-jeff-weiner | 2020-06-11 | official RSS audio; OpenAI API transcript | Weiner succession, references, reversible trial, and team credit.
• ★★★ | Planning for Volatility | https://soundcloud.com/greylock-partners/reid-hoffman-on-planning-for-volatility | 2020-06-09 | official RSS audio; OpenAI API transcript | Contemporaneous portfolio, planning, and capital doctrine.
• ★★ | Quartz fast-culture rebuttal | https://qz.com/1572866/reid-hoffman-and-chris-yeh-defend-their-book-blitzscaling | 2019-03-19 | Hoffman/Yeh written rebuttal | Archive-route written exception on blitzscaling responsibility and ethics.

Official profiles, company records, filings, and transaction evidence

• ★★★ | PayPal 2002 Form 10-K | https://www.sec.gov/Archives/edgar/data/1103415/000091205702009834/a2073071z10-k405.htm | filed 2002-03-13 | SEC filing | Primary role boundary: Confinity/PayPal director and PayPal EVP for external relationships; does not label Hoffman a PayPal co-founder.
• ★★★ | Greylock partner profile — Reid Hoffman | https://greylock.com/team/reid-hoffman/ | current; accessed 2026-08-24 | firm biography/portfolio | Official current/prior portfolio union and education; dynamic and time-sensitive, not a career denominator.
• ★★★ | Greylock master portfolio | https://greylock.com/portfolio/ | current; accessed 2026-08-24 | firm portfolio directory | Named-partner attribution route; essential control showing David Sze for LinkedIn/Facebook and Hoffman for selected later companies.
• ★★ | Greylock gives Hoffman a $20 million Discovery Fund | https://techcrunch.com/2010/09/27/greylock-gives-super-angel-turned-vc-reid-hoffman-a-20-million-seed-fund/ | 2010-09-27 | contemporaneous transaction report | Routes the special seed allocation inside Greylock Fund XIII and names reported pre-Greylock angels; secondary report requiring deal-by-deal verification.
• ★★★ | LinkedIn Form S-1 | https://www.sec.gov/Archives/edgar/data/1271024/000119312511016022/ds1.htm | filed 2011-01-27 | SEC filing | Primary chronology, management, risks, launch metrics, ownership, and Greylock/Hoffman control fence; use printed sections and HTML lines.
• ★★★ | LinkedIn–Microsoft merger proxy | https://www.sec.gov/Archives/edgar/data/1271024/000104746916014430/a2229104zdefm14a.htm | filed 2016-07-22 | SEC transaction filing | Sale process, Hoffman/Nadella/Gates contacts, board deliberation and ownership; transaction consideration is not a personal-proceeds ledger.
• ★★★ | LinkedIn secures $10 million Series B led by Greylock | https://news.linkedin.com/2004/10/linkedin-secures-10-million-in-series-b-funding-led-by-greylock | 2004-10 | issuer release | Contemporaneous round and David Sze board attribution.
• ★★★ | Original LinkedIn Series B pitch deck | https://archive.computerhistory.org/resources/access/text/2023/08/102803543-05-01-acc.pdf | 2004 deck; archived 2023 | museum archive PDF | Decision-era artifact; compare slide-by-slide with Hoffman's later annotations and actual S-1 outcomes.
• ★★★ | Microsoft to acquire LinkedIn | https://news.microsoft.com/source/2016/06/13/microsoft-to-acquire-linkedin/ | 2016-06-13 | issuer transaction release | $26.2 billion announced consideration, governance support, and role labels; transaction value is not personal proceeds.
• ★★★ | Microsoft 2017 proxy | https://www.sec.gov/Archives/edgar/data/789019/000119312517310951/d461626ddef14a1.pdf | 2017 | SEC proxy PDF | Joined Microsoft board in March 2017; use pp. ii and 14.
• ★★★ | Microsoft 2024 proxy | https://www.sec.gov/Archives/edgar/data/789019/000119312524242883/d858775ddef14a.htm | 2024 | SEC proxy | Lines 1827–1829: Inflection license/hiring deal, sub-10% Reprogrammed and Greylock stakes, deemed beneficial ownership, and Hoffman recusal.
• ★★★ | Microsoft 2026 Form 8-K | https://www.sec.gov/Archives/edgar/data/789019/000119312526258667/d26760d8k.htm | filed 2026-06-05; event 2026-06-02 | SEC filing | Item 5.02: will not stand for re-election, remains through annual meeting, no disagreement; controls stale live bios.
• ★★★ | Aurora Form S-1 | https://www.sec.gov/Archives/edgar/data/1828108/000119312521321663/d230050ds1.htm | 2021 | SEC filing | Exact February 2018 Greylock financing, director grant/transfer, board biography, and risk route.
• ★★★ | Aurora 2023 proxy | https://www.sec.gov/Archives/edgar/data/1828108/000182810823000066/aurorainnovationinc2023pro.htm | 2023 | SEC proxy | Hoffman director biography and issuer claim of 100-plus technology investments; company claim is not a denominator.
• ★★★ | Joby 2023 annual report/proxy | https://www.sec.gov/Archives/edgar/data/1819848/000181984823000154/joby-2023xars.pdf | 2023 | SEC filing PDF | Board role since 2021 and SPAC/public-company governance route; large PDF requires printed-page locators.
• ★★ | Airbnb Form S-1/A | https://www.sec.gov/Archives/edgar/data/1559720/000119312520306257/d81668ds1a.htm | 2020 | SEC filing | Canonical issuer route for Greylock holdings and company history; full HTML exceeded research-tool retrieval limits.
• ★★★ | Facebook secures $25 million investment | https://about.fb.com/news/2006/04/facebook-secures-25m-investment-2/ | 2006-04 | issuer release | Greylock-led Series B attributed to David Sze; prevents improper Hoffman deal credit.
• ★★ | Announcement: fbFund | https://about.fb.com/news/2007/09/announcement-fbfund/ | 2007-09 | issuer release | Contemporaneous adviser role and social-platform ecosystem context.
• ★★★ | Introducing OpenAI | https://openai.com/index/introducing-openai/ | 2015-12-11 | organization announcement | Names Hoffman among donors within a collective $1 billion commitment; does not disclose his paid amount.
• ★★★ | OpenAI and Elon Musk | https://openai.com/index/openai-elon-musk/ | 2024 | interested-party documentary response | OpenAI says Hoffman bridged a funding gap; use as party evidence and pair with pleadings/court reporting.
• ★★★ | Microsoft invests in and partners with OpenAI | https://openai.com/index/microsoft-invests-in-and-partners-with-openai/ | 2019-07-22 | organization announcement | $1 billion Microsoft partnership and cloud relationship; relevant to later board-conflict analysis.
• ★★★ | Microsoft / Inflection AI inquiry | https://www.gov.uk/cma-cases/microsoft-slash-inflection-ai-inquiry | 2024 | UK regulatory case page/decision route | Formation and transaction chronology plus September clearance; its narrower transaction chronology defines the deal record but does not override an explicit public co-founder label.
• ★★★ | CMA Microsoft / Inflection full decision | https://assets.publishing.service.gov.uk/media/6719ff5f549f63039436b3c8/__Full_text_decision__.pdf | decision 2024-09-04; PDF published 2024-10-24 | UK regulator decision PDF | Fifty-six-page primary record of personnel, license, founder/board chronology, jurisdiction and clearance; competition disposition is not an investment-quality finding.
• ★★★ | Introducing Manas AI | https://manasai.co/news/introducing-manas-disrupting-drug-discovery | 2025-01-26 | co-founder announcement | Hoffman and Siddhartha Mukherjee describe the formation and AI-biopharma thesis; claims require later scientific and commercial outcome tests.
• ★★ | Manas AI company and team | https://manasai.co/company | current; accessed 2026-08-24 | company record | Date-sensitive page adds Ujjwal Singh as co-founder and names current backers; not proof of role chronology before September 2025 or investment economics.
• ★★★ | DOJ Epstein Library | https://www.justice.gov/epstein | current; accessed 2026-08-24 | official archive/search portal | Primary document-discovery route; DOJ warns that searchability is imperfect, so missing search hits are not evidence of absence and document mentions are not misconduct findings.

Biography, press, books, and historical context

• ★★★ | LinkedIn’s Reid Hoffman: The network philosopher | https://money.cnn.com/2009/08/24/technology/linkedin_reid_hoffman.fortune/ | 2009-08-24 | contemporaneous profile/interview | Early career, SocialNet, PayPal, LinkedIn, angel investing, and missed-deal context before later fame.
• ★★ | LinkedIn founder Reid Hoffman: the venture capitalist with a Rolodex brain | https://www.latimes.com/archives/la-xpm-2008-jul-08-fi-hoffman8-story.html | 2008-07-08 | contemporaneous profile | Early biography, network reputation, and reported wins/misses; archive access is partial.
• ★★★ | Reid Hoffman, LinkedIn co-founder, sees AI extending human agency | https://apnews.com/article/reid-hoffman-linkedin-artificial-intelligence-trump-aa9600b3c82f886cbe336f519672f9d7 | 2025 | independent profile | Current AI/political context and biography from a high-reliability wire service.
• ★★★ | Venture capitalist Reid Hoffman pares Greylock role | https://www.wsj.com/public/resources/documents/JJ1zqihQoLz0w7wMXtJk-WSJNewsPaper-8-31-2023.pdf | 2023-08-31 | newspaper PDF, pp. B1/B4 | Near-contemporaneous Fund XVII role report: not a GP in the new fund, remained strategic partner/colleague, possible occasional investments; reconcile with live firm title.
• ★★ | Reid Hoffman on his billion-dollar impact-investing bet | https://www.forbes.com/sites/aislinnmurphy/2018/10/03/linkedin-cofounder-reid-hoffman-on-his-billion-dollar-impact-investing-bet/ | 2018-10-03 | reported interview | Self-reported $1.5 billion commitment and criteria; grants, investments, vehicles, and results are not enumerated.
• ★★★ | The Founders | https://www.simonandschuster.com/books/The-Founders/Jimmy-Soni/9781501197246 | 2022/2024 edition | official publisher/book route | Reported PayPal history based on interviews and internal material; acquire and cite exact chapters/pages later.
• ★★ | The Power Law | https://www.penguinrandomhouse.com/books/580120/the-power-law-by-sebastian-mallaby/ | 2022-02-01 | official publisher/book route | Institutional venture-history and network context; full book, exact Hoffman passages, and source notes needed later.

Criticism, adverse outcomes, litigation, and contested routes

• ★★★ | Hoffman apologizes for funding group behind Alabama disinformation | https://www.washingtonpost.com/technology/2018/12/26/internet-billionaire-reid-hoffman-apologizes-funding-group-behind-disinformation-alabama-race/ | 2018-12-26 | independent reporting | Direct-apology and funding-chain route; access is partial and operational knowledge remains disputed.
• ★★ | House report route to Project Birmingham | https://www.govinfo.gov/content/pkg/CRPT-116hrpt657/pdf/CRPT-116hrpt657.pdf | 2020-12-03 | congressional report PDF, p. 14 | Primary government publication recording the spoofing-campaign allegation and cited reporting; the report itself does not establish Hoffman directed the operation.
• ★★★ | Hoffman apologizes for role in Epstein-linked MIT donations | https://www.axios.com/2019/09/12/reid-hoffman-jeffrey-epstein-mit-donations | 2019-09-12 | independent reporting | Apology and MIT fundraising-introduction route; do not extend beyond supported conduct.
• ★★★ | MIT fact-finding report on Jeffrey Epstein relationships | https://web.mit.edu/fact2020/files/MIT-report.pdf | 2020-01-10 | institution-commissioned investigation PDF | Documentary chronology of Media Lab fundraising contacts, with stated scope and institutional-interest limits; not an exoneration or evidence of unrelated misconduct.
• ★★★ | Hoffman corrects his Epstein-contact chronology | https://x.com/reidhoffman/status/2018107902486020319 | 2026-02-01 UTC | direct public statement; platform access may be partial | Hoffman corrects his 2019 last-contact account after reviewing calendar records; self-report and disclosed records require independent document-level corroboration.
• ★★ | Why Hippo’s stock fell after its SPAC | https://www.forbes.com/sites/jeffkauflin/2021/09/30/why-hippos-stock-has-fallen-more-than-50-since-going-public-through-a-spac/ | 2021-09-30 | adverse-outcome analysis | Early public-market deterioration and underwriting criticism; does not calculate sponsor or investor realized returns.
• ★★ | Joby Aviation to go public via Reinvent SPAC | https://www.axios.com/2021/02/24/flying-taxis-linkedin-cofounder-spac | 2021-02-24 | contemporaneous transaction report | Valuation, sponsor, capital needs, and execution risk; use with filings.
• ★★★ | Crowder v. LinkedIn preliminary-settlement order | https://law.justia.com/cases/federal/district-courts/california/candce/4%3A2022cv00237/390494/135/ | 2025-12-17 | federal order mirror, ECF 135 | Denies preliminary settlement approval; LinkedIn is the defendant, pleaded API/Azure assertions remain allegations, and the order is not a merits or liability finding.
• ★★★ | Musk v. Altman second amended complaint | https://www.courthousenews.com/wp-content/uploads/2026/04/SAC-Musk-Altman.pdf | 2026 | court pleading PDF | Ninety-three-page allegation route naming Hoffman; allegations are not findings and require docket/disposition checks.
• ★★★ | Musk/OpenAI claims dismissed as time-barred | https://apnews.com/article/musk-openai-trial-verdict-0b9b0bfaffe96f2c930341f52dfe4f8c | 2026-05-18 | independent court report | Advisory jury and court dismissed remaining claims on limitations grounds; procedural disposition is not a merits finding or vindication.
• ★★★ | Crowder v. LinkedIn discovery order | https://docs.justia.com/cases/federal/district-courts/california/candce/4%3A2022cv00237/390494/233 | 2026-08-04 | federal discovery order mirror, ECF 233 | Allows a limited three-hour Hoffman deposition for possible unique knowledge; LinkedIn is defendant and the order is not liability evidence.
• ★★ | Irish DPC decision on LinkedIn Ireland | https://www.dataprotection.ie/en/news-media/press-releases/irish-data-protection-commission-fines-linkedin-ireland-eu310-million | 2024-10-24 | regulator release | Corporate GDPR findings and €310 million fine, shown as pending appeal; do not personalize post-Microsoft company conduct without temporal evidence.
• ★★ | Reported 2026 DOJ probe of American Future Republic | https://www.cbsnews.com/news/justice-dept-reid-hoffman-e-jean-carroll-trump-lawsuits/ | 2026-05-28 | independent reporting | Active nonprofit-funding investigation; at publication no charge was reported and Hoffman reportedly had not been contacted.
• ★★★ | The fundamental problem with Silicon Valley’s favorite growth strategy | https://www.oreilly.com/radar/the-fundamental-problem-with-silicon-valleys-favorite-growth-strategy/ | 2019-03-21 | independent practitioner critique | Tests blitzscaling’s platform externalities, winner-take-all assumptions and boundary conditions while acknowledging the authors’ stated conditions.
• ★★★ | When do startups scale? Large-scale evidence from job postings | https://doi.org/10.1002/smj.3596 | published 2024-03-07; Strategic Management Journal 45(9):1633–1669 | peer-reviewed open-access study by Saerom Lee and J. Daniel Kim | Large-sample early-scaling counterevidence with explicit measurement, A/B-testing moderation, platform result, and limitations; association is not a causal verdict on every blitzscaling case.

Deal, filing, vehicle, and outcome sources processed in DEALS

• ★★★ | Greylock legacy Reid Hoffman profile | https://greylock.com/wp-content/uploads/2020/04/Reid-Hoffman.pdf | 2020 archive | official firm PDF | Physical/PDF p. 2 reconciles prior portfolio names and the live TrialPlay/legacy TrialPay inconsistency; inventory evidence, not economics.
• ★★★ | The network philosopher | https://www.wired.com/story/reid-hoffman-network-philosopher/ | 2012-03 | independent profile/interview | Paragraph beginning “Hoffman’s fortunes as an investor” supplies named early rounds, Facebook five-figure/secondary boundary, and no complete cash flows.
• ★★ | Why Microsoft bought LinkedIn | https://www.wired.com/2017/03/now-we-know-why-microsoft-bought-linkedin/ | 2017-03 | independent profile | Discovery Fund paragraph reports Dropbox/Pandora/Tumblr/Groupon/Shopkick/Airbnb; secondary vehicle claim, not audited attribution.
• ★★ | Joi Ito life update | https://joi.ito.com/weblog/2010/05/13/life-update-and.html | 2010-05-13 | participant account | Paragraph beginning “With Reid I invested” supplies co-investment names; retrospective witness, not terms/outcomes.
• ★★★ | Airbnb Form 3 | https://www.sec.gov/Archives/edgar/data/1474746/000089924321019955/0000899243-21-019955-index.htm | 2021-06-10 | SEC filing index | Establishes Greylock XIII reporting entities; not Hoffman ownership or return.
• ★★★ | Aurora S-4 | https://www.sec.gov/Archives/edgar/data/1828108/000119312521216134/d184562ds4.htm | 2021 | SEC registration statement | “Background of the Business Combination — Hoffman’s role” records recusal from both sides and no valuation/terms consultation.
• ★★★ | Aurora Sponsor Agreement | https://www.sec.gov/Archives/edgar/data/1828108/000119312521215991/d326572dex103.htm | 2021 | SEC exhibit | §§1.3–1.6 expose sponsor holdings, lockups and performance vesting; entity economics, not personal proceeds.
• ★★★ | Joby transaction presentation | https://www.sec.gov/Archives/edgar/data/1819848/000119312521205648/d24019d425.htm | 2021 | SEC Rule 425 | Transaction overview, sponsor alignment, projections and risk; forecast evidence, not outcome.
• ★★★ | Joby Q2 2026 shareholder exhibit | https://www.sec.gov/Archives/edgar/data/1819848/000181984826000432/joby-8kx20260805xexx99_2.htm | 2026-08-05 | SEC exhibit | Certification-stage, liquidity and net-loss current state; passenger targets forward-looking.
• ★★★ | Hippo S-4/A | https://www.sec.gov/Archives/edgar/data/1828105/000119312521210870/d137441ds4a.htm | 2021 | SEC registration statement | Transaction background, pro forma ownership, PIPE, sponsor vesting/redemption mechanics.
• ★★★ | Hippo Q2 2026 earnings | https://www.sec.gov/Archives/edgar/data/1828105/000182810526000041/q226earningsreleasefinal.htm | 2026-08 | SEC exhibit | Current company income/loss-ratio/combined-ratio evidence; not sponsor return or original-valuation validation.
• ★★★ | Snowflake FY2024 10-K | https://www.sec.gov/Archives/edgar/data/1640147/000164014724000101/snow-20240131.htm | 2024 | SEC filing | Business-combinations note records $185.4m Neeva cash consideration; purchase price is not investor proceeds.
• ★★★ | SEC SPAC disclosure guidance | https://www.sec.gov/rules-regulations/staff-guidance/disclosure-guidance/disclosure-special-purpose-acquisition | 2020-12-22 | regulator guidance | Sponsor compensation, PIPE, conflict and dilution dimensions.
• ★★★ | SEC SPAC liability/risk statement | https://www.sec.gov/newsroom/speeches-statements/spacs-ipos-liability-risk-under-securities-laws | 2021-04-08 | regulator statement | Contemporary SPAC surge and investor-protection context; not a finding on Reinvent.
• ★★★ | OpenAI LP | https://openai.com/index/openai-lp/ | 2019-03-11 | organization announcement | Names Hoffman’s charitable foundation as investor in then-capped-profit structure; no amount/instrument.
• ★★★ | OpenAI current structure | https://openai.com/our-structure/ | current after 2025-10-28 recapitalization | organization record | OpenAI Group PBC controlled by OpenAI Foundation; foundation stake is not Aphorism/Hoffman value.
• ★★ | Aphorism Foundation filings | https://projects.propublica.org/nonprofits/organizations/464347021 | through current available filings | Form 990 index | Foundation/controlled-entity route; schedules needed for exact OpenAI instrument/disposition.
• ★★★ | Helion investors | https://www.helionenergy.com/about | current; accessed 2026-08-27 | issuer page | “Our Investors” names Hoffman; exact round/vehicle/economics absent.
• ★★★ | Astra acquires Apollo Fusion | https://wp.astra.com/news/astra-acquires-apollo-fusion-to-reach-new-orbits/ | 2021-06-07 | acquirer announcement | Names Hoffman lead investor and $50m base + up to $95m earnouts; company consideration not investor proceeds.
• ★★★ | Crown Agents/Helios acquires Segovia assets | https://www.heliosinvestment.com/uploads/files/Crown-Agents-Bank-acquires-Segovia-04-04-2019.pdf | 2019-04-04 | acquirer release PDF | p. 2: Hoffman decision attribution, reported $15m Series A, Taptap spinout; acquisition price unknown.
• ★★ | Taptap Send Series A | https://techcrunch.com/2021/06/24/taptap-send-gets-13-4m-for-a-no-fee-money-transfer-service-aimed-at-price-conscious-emerging-market-users/ | 2021-06-24 | independent round report | $13.4m round co-led by Canaan/Hoffman; allocation/vehicle unknown.
• ★★ | Taptap Send Series B | https://techcrunch.com/2021/12/20/taptap-send-raises-65m-to-build-cross-border-remittances-focused-on-the-most-underserved-markets/ | 2021-12-20 | independent round report | Participation/direct thesis; aggregate round not check.
• ★★★ | Introducing Niteshift | https://niteshift.dev/blog/introducing-niteshift | 2026-06 | issuer announcement | Closing financing paragraph separates Greylock-led $7m seed from Reid Hoffman backing.
• ★★ | Sanmai financing/board announcement | https://www.linkedin.com/posts/jay-sanguinetti-b49a0a17_billionaire-hoffman-backs-brain-startup-using-activity-7344798693729976320-x8uO | 2025 | founder/issuer post | Says Hoffman led $12m and joined board; vehicle/terms/clinical outcomes absent.
• ★★★ | Vor Biopharma 2026 proxy | https://www.sec.gov/Archives/edgar/data/1817229/000119312526181883/vor-20260423.htm | 2026 | SEC proxy | Beneficial-ownership footnote 7 and related-party/director-designee route; no personal-return inference.
• ★★★ | Wealthfront IPO prospectus | https://www.sec.gov/Archives/edgar/data/1524566/000162828025056780/wealthfront-424b4.htm | 2025 | SEC filing | IPO/current outcome route; holdings or IPO value do not establish Hoffman proceeds.
• ★★★ | Hoffman invests in Care.com | https://om.co/gigaom/care-com/ | 2007-07-30 | contemporaneous report | Opening/final financing paragraph supports 2007 Series A participation; later aggregate-size reports conflict.
• ★★★ | Letters to a Young Investor: Airbnb and Stripe | https://www.generalist.com/p/letters-reid-hoffman-1 | Hoffman response 2023-10-11; published 2023-10-31 | first-person letter | “Reid’s response” records 2012 Stripe pass, decision rationale and retrospective lesson.
• ★★ | Nanosolar founder history | https://blog.rmartinr.com/nanosolar/ | historical founder account | Names Hoffman among angel friends; terms/failure economics absent.
• ★★★ | Brain Co. Series A | https://www.prnewswire.com/news-releases/brain-co-launches-from-stealth-with-30-million-series-a-302554423.html | 2025-09 | issuer release | $30m participant list names Hoffman; no allocation/lead inference.
• ★★ | CardSpring launches | https://medium.com/@greylockvc/cardspring-launches-payment-network-platform-260cece49da3 | 2012 | Greylock post | Discovery Fund seed and Slavet/Hoffman/Rosenthal authorship; no sole attribution/economics.
• ★★ | Every Media financing | https://www.linkedin.com/posts/everyinc_its-a-big-day-at-every-weve-raised-2-activity-7330999384572112898-0FzE | 2025 | company post | $2m co-lead claim; vehicle/terms unknown.
• ★★★ | LTSE regulatory comment | https://www.sec.gov/comments/sr-iex-2018-06/iex201806-3492821-162270.pdf | 2018 | SEC comment PDF | Signature labels Hoffman “Investor, LTSE”; no investment terms.
• ★★★ | PureTech H1 2017 report | https://news.puretechhealth.com/static-files/b26d90fe-4d75-4ae1-96d3-578d95c04c17 | 2017 | issuer report PDF | Physical p. 8 identifies Sync convertibles/outsider group; $5m aggregate not allocation.
• ★★ | Hoffman funds Smartmatic litigation | https://www.washingtonpost.com/style/media/2024/07/08/reid-hoffman-funds-smartmatic-fox-lawsuit/ | 2024-07-08 | independent report | Multimillion-dollar litigation-capital purpose; instrument/return unknown.
• ★★ | Greylock Discovery invests in Artfinder | https://www.fastcompany.com/1795694/linkedin-cofounder-reid-hoffman-uk-silicon-valley-can-be-exported | 2011 | direct interview | ArtFinder Q&A supplies vehicle attribution; no terms/outcome.
• ★★ | Shopkick rounds and Hoffman board | https://techcrunch.com/2010/07/14/shopkick-greylock-reid-hoffman/ | 2010-07-14 | independent report | Personal/Kleiner then Greylock round sequencing; allocation/economics incomplete.
• ★★ | Swipely Series A | https://techcrunch.com/2010/05/11/swipely/ | 2010-05-11 | independent report | Greylock participation and observer route; personal-versus-firm boundary remains source-sensitive.
• ★★ | Wrapp Series A and board | https://www.fastcompany.com/1809188/wrapp-mogul-why-linkedins-reid-hoffman-investing-social-gifting-site | 2012 | independent/direct report | Greylock participation and Hoffman board; no personal check inference.
• ★★★ | Greylock 18 | https://greylock.com/firm-news/introducing-greylock-18/ | 2026-07-14 | official fund announcement | $1.5b early-stage LP-backed fund; opening says partners make 1–2 new investments/year; no fee/carry/reserve terms.
• ★★★ | Greylock 18 Form D | https://www.sec.gov/Archives/edgar/data/2136563/000213656326000001/0002136563-26-000001-index.htm | filed 2026-06-09 | SEC filing index | Exempt-offering/entity route for current LP-backed fund comparison.
• ★★★ | Adobe acquires Aviary | https://express.adobe.com/page/lJINB/ | 2014 | acquirer announcement | Company outcome; transaction allocation and Hoffman proceeds absent.
• ★★★ | Twitter welcomes CardSpring | https://blog.x.com/en_us/a/2014/welcoming-cardspring-to-the-twitter-team | 2014 | acquirer announcement | Company acquisition outcome; consideration and investor allocation undisclosed.
• ★★★ | Flexport trucking-business update | https://www.flexport.com/blog/update-on-flexports-trucking-business/ | 2025 | acquirer/company update | Distinguishes Convoy’s shutdown and acquired technology stack from company/liability acquisition, then records platform sale to DAT.
• ★★★ | Quotient merger announcement | https://www.sec.gov/Archives/edgar/data/1115128/000119312523170226/d473281dex991.htm | 2023 | SEC exhibit | Neptune acquisition and approximately $430m company equity value; not investor proceeds.
• ★★★ | Dropbox Q1 2018 results | https://investors.dropbox.com/news-releases/news-release-details/dropbox-announces-fiscal-2018-first-quarter-results | 2018 | issuer release | “Initial Public Offering and Private Placement” fixes March 23 trading and March 27 IPO closing.
• ★★★ | NetDragon acquires Edmodo | https://www.netdragon.com/press/2018.shtml | 2018 | acquirer release | 100% acquisition and reported $137.5m company consideration; not Hoffman proceeds.
• ★★★ | NetDragon Edmodo shutdown filing | https://file.download.99.com/down/ir_en_20220818.pdf | 2022 | issuer filing PDF | Records planned September 2022 closure of Edmodo’s free B2C platform.
• ★★ | Funny Or Die ownership history | https://funnyordie.com/about/ | current | company history | Records Henry Muñoz acquisition in 2021; terms undisclosed.
• ★★★ | Cisco completes IronPort acquisition | https://www.sec.gov/Archives/edgar/data/858877/000119312507202552/dex131.htm | 2007 | SEC exhibit | Company acquisition outcome; investor allocation absent.
• ★★★ | IAC completes Care.com acquisition | https://www.sec.gov/Archives/edgar/data/891103/000110465920015988/tm1926761d17_ex99-1.htm | 2020 | SEC exhibit | Closing and approximately $500m company value; not Hoffman proceeds.
• ★★★ | GameStop acquires Kongregate | https://www.kongregate.com/en/announcements/big-news-kongregate-has-been-acquired-by-gamestop | 2010 | company announcement | Company acquisition outcome; terms and investor allocation absent.
• ★★★ | CBS acquires Last.fm | https://ir.paramount.com/node/58586/html | 2007 | acquirer release | Approximately $280m company consideration; not Hoffman proceeds.
• ★★★ | Nanosolar facility-closure auction | https://www.hgpauction.com/wp-content/uploads/2014/03/NanoSolar_8_13_13_4web.pdf | 2013 | auction catalog PDF | “Complete Facility Closure” and asset auction; no bankruptcy, write-off or investor recovery inference.
• ★★★ | Bed Bath & Beyond One Kings Lane disposition | https://www.sec.gov/Archives/edgar/data/886158/000088615820000008/bbby201910k.htm | 2020 | SEC 10-K | Records sale and says proceeds were not material to issuer; no Hoffman return.
• ★★★ | SiriusXM/Pandora transaction | https://www.sec.gov/Archives/edgar/data/908937/000090893720000011/siri-20191231x10k.htm | 2019 | SEC 10-K | Completed acquisition and $2.879b company consideration; not investor proceeds.
• ★★ | Seesmic acquires Ping.fm | https://techcrunch.com/2010/01/04/seesmic-acquires-ping-fm/ | 2010 | contemporaneous report | Company acquisition outcome; terms undisclosed.
• ★★ | EA/Rupture 10-Q analysis | https://www.cbsnews.com/news/10-q-watch-eas-acquisition-of-rupture-15m-not-30m/ | 2008 | filing-based independent report | $15m cash including transaction costs, correcting later $30m claim; not Hoffman proceeds.
• ★★★ | SK Planet acquires Shopkick | https://www.sec.gov/Archives/edgar/data/1015650/000119312515158586/d912764d20f.htm | 2014 | SEC filing | KRW230.9b acquisition plus KRW18.7b assumed current liabilities; company economics only.
• ★★★ | VideoEgg acquires Six Apart | https://www.newswire.ca/news-releases/videoegg-acquires-six-apart-to-create-say-media-545527492.html | 2010 | company release | SAY Media combination; terms undisclosed.
• ★★★ | Lightspeed acquires Upserve | https://www.lightspeedhq.com/news/lightspeed-acquisition-of-upserve/ | 2020 | acquirer release | Estimated $430m company consideration; not Hoffman proceeds.
• ★★★ | PureTech Sync Project asset sale | https://www.sec.gov/Archives/edgar/data/1782999/000162828021006938/ptv-20201231.htm | 2020 | SEC 10-K | Bose asset purchase and $4.5m aggregate consideration; holder allocation absent.
• ★★★ | Synacor/Technorati asset transaction | https://www.annualreports.com/HostedData/AnnualReportArchive/s/NASDAQ_SYNC_2015.pdf | 2015 | issuer annual report PDF | $3m cash asset acquisition; not investor proceeds.
• ★★★ | Automattic acquires Tumblr | https://help.tumblr.com/es/25-de-septiembre-de-2019-automattic-adquiere-tumblr/ | 2019 | company announcement | Later company ownership outcome; terms and earlier investor allocation absent.
• ★★★ | Alibaba acquires Vendio | https://www.vendio.com/ecommerce/press_releases/2010/pr06232010 | 2010 | company announcement | Company acquisition outcome; consideration undisclosed.
• ★★★ | Meniga acquires Wrapp | https://www.meniga.is/en-gb/about-us/ | 2019 | company history | Company acquisition outcome; terms and investor allocation absent.
• ★★★ | Take-Two completes Zynga acquisition | https://www.sec.gov/Archives/edgar/data/946581/000119312522160004/d306239dex999.htm | 2022 | SEC exhibit | Closing terms of $3.50 cash plus 0.0406 Take-Two share per Zynga share; Hoffman holdings/proceeds absent.

Coverage assessment

• 195 ranked source rows: 45 direct writings/books/theses + 35 talk/interview/oral-history rows + 26 official profiles/filings/transactions + 7 biography/press/history + 15 criticism/adverse/litigation + 67 de-duplicated DEALS filing/vehicle/outcome rows = 195. The DEALS additions are sources actually processed, not an assertion that the public deal universe is complete.
• TALKS universe: 73 raw routes = 20 seeded rows + 2 direct-section audio routes + all 51 writing-index TALK-routed rows. The TALKS ledger assigns these once to 37 retained priority records and 36 exclusions. The retained set contains 34 actual appearances/episodes and three written route exceptions; version pages and companion essays never become independent corroboration.
• Classification and scope: modern, internet-native investor with unusually dense direct voice; multi-vehicle and high-contestation. The evidence set spans formative biography, operator roles, personal angels, Greylock, founder equity, Reinvent SPACs, Reprogrammed/family-affiliated holdings, philanthropy/impact, boards, AI company formation, and political/reputational routes.
• Names and date variants searched: Reid Hoffman, Reid G. Hoffman, SocialNet, Socialnet, Relationships.com, LinkedIn 2002, LinkedIn March/May 2003, Greylock November 2009, personal angel, Greylock investor, Reinvent Capital, Reprogrammed Interchange, foundation, OpenAI board, Inflection founder/cofounder, and current Microsoft board 2026 variants.
• Direct-voice and archive routes tried: ReidHoffman.org writing/books, Greylock author/portfolio/legacy indexes, Masters of Scale, Possible, Stanford eCorner, Stanford/YC, TED, Harvard IOP, Computer History Museum/OpenCHM, Academy of Achievement, Tim Ferriss, Knowledge Project, Washington Post Live, and Norges Bank Investment Management.
• Pinpoint locator quality: TALKS uses human transcript pages/headings, creator captions, automatic-caption timestamps checked against audio, or timestamps from ten OpenAI API transcriptions of official public audio. Companion essays and the three written route exceptions use heading/paragraph locators and are never assigned spoken timestamps.
• Access barriers: Knowledge Project’s written transcript remains members-only, but the public audio was fully transcribed; old Stanford transcript PDFs and several legacy Greylock pages return 404/HTML, but live Stanford printables, official RSS enclosures, creator captions, or public video mirrors supplied the retained bodies. The Information, Le Monde, and Los Angeles Times bodies remain partial/paywalled; some book bodies require acquisition. These are retrieval constraints, not negative evidence.
• Dynamic/stale-source warning: Greylock’s live partner bio still lists Microsoft board service, while the June 2026 8-K says Hoffman will remain only through the 2026 annual meeting and will not seek re-election. Cite the filing for current status and timestamp the profile.
• DEALS denominator: the bounded named screen contains 164 company/capital-relationship families: 81 pre-query deep-core candidates plus 83 query-screened outer candidates. The core is fully dispositioned as 71 included and 10 excluded/control; the included set contains 67 public investment claims plus LinkedIn founder equity, Callisto donation, Kiva microlending support and Endeavor Catalyst backing. Four cases are strictly reconstructed and 67 remain unresolved at strict lifecycle depth. The outer 83 is an exact screening-only inventory; per-name routes were not preserved/revalidated and none enters evidence or rankings. This is not a complete public or career universe; the 60+ host and 100+ issuer characterizations are non-enumerated and incompatible. Denominator ledger
• Attribution controls: David Sze/Greylock led LinkedIn and Facebook Series B and Sze controlled Greylock’s LinkedIn shares in the 2011 filing; Hoffman had no voting/dispositive power over those Greylock shares. Airbnb is the clearest Hoffman-proposed Greylock case, with Slavet support and Sze skepticism in the retrospective account; Greylock's current Hoffman profile separately says, “Reid was a board observer for Airbnb” (role-summary paragraph; accessed 2026-08-27). Greylock and Microsoft call Hoffman an Inflection co-founder; the CMA’s narrower transaction chronology foregrounds Suleyman and Simonyan and identifies Hoffman as a board member, so both descriptions remain fenced.
• Historical context and completed peers: the LinkedIn S-1/deck, Meta releases, CHM social-computing archive, The Founders, and The Power Law provide era context. Completed canon comparisons should prioritize David Sze (same partnership/attribution control), Ron Conway (personal angel breadth), Brad Feld and Fred Wilson (public doctrine), John Doerr and Vinod Khosla (operator-to-investor/deep tech), and Masayoshi Son (scale and governance); comparisons require vehicle, era, stage, and opportunity-set normalization.
• Highest-value retrieval leads after DEALS: original Greylock/Discovery/Reinvent IC and board materials; audited LP/fund and personal cash-flow schedules; Aphorism/OpenAI instruments and conversion records; cap tables, preferences, reserve/follow-on schedules and distribution notices; original pass memos; exact sponsor pecuniary allocations; current failure/write-off ledgers; and primary records resolving the 83 screened-outer candidates beyond screening depth.
• Hard gaps after DEALS: no audited actor-level cash-flow ledger, complete opportunity/pass/failure set, IC votes, systematic decision memos, ownership/dilution/reserve series, LP letters, fees/carry, board minutes, philanthropic repayment/beneficiary ledger, or complete adjudicated OpenAI record. The 71-row core is a research denominator, not a performance denominator; company consideration, public share prices and affiliate holdings never substitute for Hoffman proceeds or returns.

Source notes