Brook Byers

1. Snapshot and testable edge thesis

2. Core philosophy

• Begin with the consequential problem — [investor-stated · moderate]. Byers’s oral history describes working backward from unmet medical needs rather than beginning with an available technology. Bancroft, printed pp. 27–36
• Make specialization cumulative — [investor-stated · moderate]. Domain knowledge, scientific and clinical relationships, and reputation form an ecosystem for finding and evaluating new companies. Repetition shows consistency, not efficacy. Stanford ecosystem clip
• Create and build, not merely select — [documented behavior · strong]. Hybritech and the 1986 cohort show temporary operating roles, recruiting, chair work, and company formation. Bancroft, printed pp. 18–36 The Genomic Health filing independently verifies only the four selected founding-president/chair titles it names. Genomic Health S-1, director biography
• Buy evidence in stages — [researcher inference · moderate]. Hybritech’s technical proof, IDEC’s program re-ranking, and diagnostics’ clinical/reimbursement gates suggest that each check should change a decision. decision labs
• Treat the investor–founder relationship as an unresolved written theme — [researcher inference · insufficient public record]. Bibliographic records establish that the Pratt chapter existed and recurred, but no controlled body or pinpointed quotation survives; REVIEW therefore assigns it no substantive relationship or governance proposition. Pratt source analysis

Operating-system synthesis — [researcher inference · moderate]. Need → specialist network → missing-company thesis → complementary team → staged proof → active governance → evidence-conditioned follow-on or stop. The chain is falsifiable at each arrow; a famous outcome does not retroactively validate the process.

3. Sourcing

Sourcing diligence should preserve name/date variants, founders, scientific institutions, legal vehicles, and rejected opportunities. No public rejection ledger survives, so false-negative rates and funnel conversion are insufficient public record.

4. Picking

• Clinical decision first — [investor-stated · moderate]. Define the patient/physician decision, present workaround, and measurable benefit before admiring the science. Bancroft, pp. 27–36
• Complementary team — [investor-stated · strong]. Look for scientists, an entrepreneurial leader, and operating functions whose strengths are nonredundant; the Stanford lecture makes team complementarity explicit. Stanford “Ideal Team” transcript
• Several independent paths — [investor-stated · moderate]. In biotechnology, multiple programs can diversify biological risk when their mechanisms are genuinely distinct. Bancroft, pp. 39–46
• Milestone and dependency map — [researcher inference · strong]. Test science, clinical action, regulation, reimbursement, manufacturing, financing, and partner dependence as one causal chain. CardioDx shows that diagnostic validity without sufficient utility/coverage can still fail. CMS evidence review
• Red flags — [researcher inference · moderate]. Kill or redesign when science is not reproducible, no operator can be recruited, a program lacks an independent proof path, the clinical decision will not change, one payer controls viability without durable coverage, or partner time/capital cannot reach the next proof.

The record does not expose a universal founder score, price ceiling, formal reference protocol, or complete rejected-deal sample. Picking rules are reconstructed from direct process statements and cases; they are not a validated predictive model.

5. Deal and portfolio mechanics

Portfolio practice should be analyzed as partner-time allocation plus ownership and reserves, not company count alone. The 40-family discovered ledger is exhaustive only for named public evidence; it is not the unitemized 60-plus universe. company ledger

6. Vehicle and incentive fit

Vehicle-fit conclusion — [researcher inference · moderate]. Long scientific development, active boards, multi-program companies, and follow-ons require durable pooled capital, reserves, and partner capacity. Finite fund lives, fundraising, fees/carry, distributions, allocation conflicts, regulation, liquidity, succession, and key-person continuity may constrain that ideal; the public record is too incomplete to show how KPCB resolved each dimension. A modern fund must inspect its governing documents rather than assume historical flexibility.

7. Team attribution and work with founders

Founder-working rule — [researcher inference · moderate]. Before investment, write who owns scientific truth, operating execution, investor sponsorship, board work, recruiting, follow-on advocacy, and the kill case. The partial Pratt chapter and Perkins introduction favor active relationships, but no public corpus reveals how Byers handled founder conflict or management replacement across the portfolio. writings synthesis

8. Views on the craft

• Power law — [researcher inference · insufficient public record]. The case list appears outcome-concentrated, but no complete cash flows permit a literal split or ex-top-one result. Pareto ranking
• Fund size and partner time — [investor-stated · strong]. Firm scale should follow the number of partners who can do real company work; ownership should justify that scarce time. Stanford 2005
• Access versus selection — [researcher inference · moderate]. Byers’s scientific and founder networks generated opportunities, but missing passes and ordinary outcomes prevent isolating selection skill. Bancroft, printed pp. 18–36 · searched denominator
• Governance — [investor-stated · moderate]. Close working relationships, active boards, and complementary teams are central; intervention without capacity, consent, or attribution can become theater. Stanford “Developing Close Working Relationships”
• Exits — [researcher inference · strong]. A public listing is financing/liquidity, an acquisition is an issuer event, and neither establishes investor cash without ownership and disposition records. return discipline
• Great investor test — [researcher inference · moderate]. An analyst-created standard should require repeated pre-outcome problem framing, complementary teams, informative milestones, action on adverse evidence, honest capacity allocation, and risk-adjusted economics across a denominator. The public record partly supports the process but not the performance test. talks frameworks · denominator

9. Signature bets and decision process

Use the frozen-record exercises before reading outcomes:

1. Hybritech — decide whether a $300,000 formation check and six-month technical proof justify institution and team risk.
2. IDEC — decide whether a different program has enough independent evidence to reinvest after failure.
3. Genomic Health — connect assay validity to a changed treatment decision and reimbursement.
4. Pacific Biosciences — distinguish financing a capital-intensive platform from realizing an exit.
5. CardioDx — failure lab: require clinical utility and coverage, not only diagnostic accuracy.
6. Genentech — distinguish assistance and learning from founder, lead-investor, and economic credit.
7. Enjoy — failure lab: test unit economics and liquidity without an original Waycross memo.

For each: invest/pass, named kill criteria, ownership/security/rights, next-check milestone, partner capacity, decision owner, and actual outcome. These are analyst-created reconstructions, not recovered Byers IC memos. [researcher inference · strong]

10. Mental models and maxims

These nine are supportable recurring rules. “Magic Mondays,” public/private ecosystems, neglected problems, and category-versus-project fit are useful routines or context but should not be split into extra models without distinct portfolio application evidence. [researcher inference · moderate]

11. Evolution and contradictions

Durable public themes are specialization, complementary teams, active governance, and evidence staging. The unresolved contradictions are intensity versus partner capacity, persistence versus sunk cost, ecosystem benefit versus private moat, mission versus project economics, and collective firm identity versus individual credit. [researcher inference · strong] talks frameworks · deal-credit controls · failure ledger

12. Failures and limits

• CardioDx outcome — [contemporaneous record · strong]. The company shut in 2019 after its withdrawn IPO; the exact Byers/KPCB loss is not public. San Francisco Chronicle, 8 January 2019 Analyst lesson — [researcher inference · moderate]: treat changed clinical action and durable coverage as early gates, while preserving that no public record shows Byers adopted this process correction. CMS evidence review
• Nanogen outcome — [contemporaneous record · strong]. Bankruptcy asset sale was reported at $25.685 million and management expected common holders to receive nothing; KPCB cost/distributions are missing. issuer release The Byers association is separately [retrospective witness · moderate].
• InSite Vision — [documented behavior · moderate]. A long development arc ended in a modest reported transaction; time, dilution, and proceeds remain unknown. transaction release
• Enjoy outcome and holder boundary — [contemporaneous record · strong]. Byers managed Waycross’s disclosed stake; Chapter 11 followed the public combination, but original cost, thesis, follow-ons, and loss are insufficient public record. S-1 · docket
• Selection and survivorship — [researcher inference · strong]. The current firm biography is selective, while the 40 named families do not reconcile its unitemized 60-plus claim, passes, shutdowns, or full economics. Kleiner Perkins biography · denominator
• Era dependence — [researcher inference · strong]. The oral record describes sparse early-biotech information, small funds, local networks, and strategic-pharma financing; transfer to any later case requires those conditions to be retested rather than assumed. historical context

[researcher inference · strong] No public evidence reviewed proves a formal sell discipline, reserve policy, standardized postmortem, price ceiling, or false-negative rate. These are limits of the evidence and possibly of the articulated philosophy; they should not be filled by inference. source-map coverage assessment · uncertainty ledger

13. Historical operating environment

In the 1970s, the oral record describes roughly $1 million rounds assembled by four or five venture firms, limited specialist information, and personal or small partnership capital. Bancroft, pp. 10–24 Commercial biotechnology had limited templates, so a narrowly falsifiable technical proof and recruited operator could be more valuable than a detailed long-range forecast. [researcher inference · moderate]

[retrospective witness · moderate] Byers describes the mid-1980s formation pattern as university and clinical networks, multiple programs, staged financing, and pharmaceutical strategic relationships; together they made specialized company formation feasible but also created financing and partner-dependency risk. Bancroft, printed pp. 31–46 The later IDEC/Biogen release supports the Genentech relationship and merger outcome, not a period-wide university or public-market claim. IDEC/Biogen merger release No reliable numerical formation or success base rate was found; qualitative context is preferable to a manufactured benchmark.

The diagnostics era added archived tissue, genomics, sequencing, payer evidence, clinical guidelines, laboratory operations, and public-policy infrastructure. Genomic Health and CardioDx show why scientific validity, changed clinical action, reimbursement, and unit economics must clear together. Genomic Health S-1 · CardioDx S-1/A

Later company values and current datasets must not leak into ex-ante judgment. Compare cases only after matching stage, modality, vintage, capital intensity, regulatory/payer path, ownership, and vehicle; otherwise a historical lesson becomes hindsight. [researcher inference · strong]

14. Modern VC translation (as of 2026-08-23)

These are dimension-matched procedural comparisons, not evidence of current performance. Contemporary facts come from official FDA, NVCA, NIH, and ILPA materials; every proposed transfer, failure condition, signal, and application remains an analyst-created hypothesis [researcher inference · moderate] unless separately labelled. No matched 2026 portfolio or fund-return cohort was found.

Public/private ecosystem dependencies remain context across specialization and staging, not a tenth model. CardioDx remains the grounded historical control for clinical utility and coverage; it is not presented as a contemporary example. CMS evidence review

15. Comparison with completed peers

Pitch Johnson is the direct apprenticeship and early-vehicle comparator; Arthur Rock is the same-era active-board and founder-network comparator; Bob Nelsen is the closest later life-sciences company-factory comparator. All three are marked complete in the repository index. Comparison is mechanism- and evidence-matched, not a performance ranking. [researcher inference · strong]

16. The strongest case against greatness

Skeptical thesis — [researcher inference · strong]. Byers may be an excellent institution builder whose personal-investor reputation is inflated by KPCB access, favorable early-biotech conditions, selective famous-company narratives, team achievements, company-level transaction values, and missing failure/economic denominators.

1. Access and era. Few venture firms, elite Stanford/Bay Area science, KPCB brand, and maturing biotech/public markets created opportunities unavailable to a typical investor. historical environment
2. Team misattribution. Scientists, founders, executives, partners, boards, co-investors, strategics, universities, and public institutions produced the companies; Genentech and Align are explicit exclusion controls. team map
3. Missing denominator. Forty named families do not reconcile the unitemized 60-plus claim, and no pass ledger exists. ledger
4. Weak economics. Zero cases reveal a complete cash-flow chain. Reported issuer values and pre-IPO fund stakes cannot establish Byers’s personal or fund return. return discipline
5. Outlier and survivorship risk. Hybritech/IDEC/Genomic Health may dominate memory while CardioDx, Nanogen, InSite, Enjoy, quiet failures, and ordinary cases remain economically opaque.
6. Retrospective doctrine. The strongest narrative source was edited decades after early decisions; the written record exposes little on price, reserves, exits, conflict, or failure. writings boundary
7. Nonrepeatability. An elite partner’s network, KPCB platform, and decades of domain reputation are not a procedure a new analyst can simply execute.
8. Contradiction. Persistence can become sunk cost; ecosystem advocacy can mask public/private dependency; hands-on governance can exceed capacity; current portfolio metadata contains a demonstrably impossible Crescendo date. data-integrity controls

Evidence favors the favorable thesis if original memos, partner votes, role-specific interventions, rejected deals, matched ordinary cases, and full fund cash flows show the predicted process and better ex-outlier outcomes. It favors the skeptic if opportunities were already preselected, Byers-specific actions were limited, follow-ons lacked new evidence, failures were undercounted, or vintage/team/KPCB effects explain results. Until then, greatness is supportable as craft and institution building, not audited alpha. [researcher inference · strong]

17. Analyst study guide, glossary, and go deeper

Recommended reading path

1. Read the biography; write the edge, mechanism, prediction, alternative explanation, and falsifier.
2. Read the Bancroft oral-history analysis, then Stanford 2005; separate retrospective doctrine from contemporaneous fragments.
3. Read the written-corpus boundary so sparse authorship is not mistaken for missing research.
4. Reconcile the 40-family denominator, vehicles, and people map before ranking companies.
5. Complete the Hybritech, IDEC, Genomic Health, CardioDx failure, and Enjoy failure labs without reading the reveal first.
6. Apply every mental model to one unseen current case, then reread the skeptical thesis.

Investor-specific glossary

Reusable checklists

Sourcing

• Name the unmet decision/problem, scientific source, introducer, rights, institution, internal sponsor, and competing formation route.
• Separate privileged access from evidence that selection is better.
• Record passes and rejected formations, not only funded survivors.

Diligence and IC

• Map science, clinical action, regulator, payer, manufacturing, data/IP, capital, and strategic dependencies.
• Define independent evidence, alternatives, dissent, security/ownership, dilution, reserves, partner capacity, next milestone, and kill criteria.
• Assign scientific, operating, investment, governance, recruiting, follow-on, and exit credit before the outcome.

Governance and follow-on

• Put one accountable owner, time budget, decision calendar, and evidence threshold beside every promised intervention.
• Re-underwrite company, program, security, vehicle, and opportunity cost separately after failure.
• Distinguish an IPO from liquidity and a company transaction from fund cash.

Go deeper

1. Bancroft oral history, printed pp. 1–86 — richest primary formation narrative; paraphrase because rights are reserved and cross-check retrospective claims.
2. Stanford 2005 lecture family — clearest direct rules on partner capacity, syndication, teams, customers, and decisions.
3. Fogarty 2019 — mature team and persistence account; best used against the IDEC no-hindsight lab.
4. SEC filings in the investments ledger — strongest ownership, board, vehicle, and attribution controls.
5. Los Angeles Times Hybritech reconstruction — best independent formation narrative, still incomplete for cash returns.

The highest-value next records are the full Pratt chapter; original Hybritech/IDEC/Genomic Health/CardioDx decision and board packets; KPCB fund agreements and cash flows; the complete 60-plus ledger and pass list; recoverable PMWC media and a fully speaker-labeled Charlie Rose transcript; and role-specific partner records. Their absence limits performance claims rather than preventing an auditable account of the craft.

Pareto 80/20 — what drove the record

The 80/20 frame is a prioritization lens, not a claim that a measurable percentage of Brook Byers’s or any KPCB fund’s return came from these companies. No public case supplies the complete cash-flow chain needed for deal MOIC or fund IRR. [researcher inference · insufficient public record] The firm’s live biography claims involvement with more than 60 companies without itemizing them, so even a company-outcome concentration statistic would have an incomplete denominator. Kleiner Perkins biography, biography paragraphs and selected portfolio

Economic contribution ranking

Dominant-outlier test. Removing Hybritech from a directional company-outcome ranking leaves several large reported issuer outcomes, but it does not create an ex-Hybritech fund-return series. The same missing cash-flow fields prevent both the headline and ex-outlier performance tests. [researcher inference · strong]

Historical-importance ranking

Decision-learning ranking

Denominator, attribution, and return discipline

Scope and reconciliation

The searched universe covers Brook Byers/Brook H. Byers name variants; Asset Management and Kleiner & Perkins/KPCB vehicle references; the live partner page and partnership archive; Bancroft and other oral histories; firm histories; SEC issuer filings; acquisition and bankruptcy controls; and candidate companies named by those records. A “company family” deduplicates renamed entities and follow-ons—for example Arris/Axys and XDx/CareDx. A row is included when a credible source supports capital, governance, operating, formation, or explicit control relevance; a role-only association remains separate, and an explicit misattribution is retained as an exclusion control. [researcher inference · strong]

Coverage fractions. Named-source coverage is 40/40 of the discovered named universe. Lab-depth coverage is 7/40 of that same named universe. Coverage of the firm’s more-than-60 claim is not computable because the denominator is unitemized and may contain aliases/follow-ons. Economically complete coverage is 0/40: no case supplies all initial/follow-on capital, ownership through time, dilution, cash distributions, residual value, fees/carry, and dates. [researcher inference · strong]

Therefore no personal Byers MOIC, IRR, DPI, TVPI, hit rate, wealth-created estimate, or KPCB fund return is computed; no issuer acquisition value is called investor proceeds; no IPO is treated as realization; no manager-shared SEC position is divided among managers; and no undisclosed company is imputed as a win or loss. All outcome values remain reported.

Where a filing lists Byers beside a partnership stake, the vehicle owns the securities. Shared voting or dispositive power is governance evidence, not personal ownership. Direct holdings are separated only when the filing does so—for example 61,435 Veracyte shares and the disclosed CardioDx securities. Veracyte proxy, security-ownership footnote · CardioDx S-1/A, principal stockholders and footnotes [contemporaneous record · strong]

Historical and institutional context

Byers’s oral history says Asset Management used Pitch Johnson’s personal capital and that roughly $1 million early rounds were often syndicated among four or five firms; it dates Byers’s first-employee role to 1972. Bancroft, printed pp. 10–14 [retrospective witness · moderate] It says he joined Kleiner & Perkins in 1977 and reports, rather than audits, an $8 million first Kleiner & Perkins fund and $15 million KPCB I in 1978, with Hybritech and Home Healthcare consuming a large portion. Bancroft, printed pp. 18–24 and 46–54 [retrospective witness · provisional] The partnership’s ability to combine formation labor with pooled capital is supported as described behavior, but the causal effect on returns is unproved. [researcher inference · moderate]

For Hybritech, the oral history and contemporary reporting support a narrow, time-bounded technical proof, interim investor operation, and later recruitment rather than a complete company at inception. Bancroft, printed pp. 18–27 · Los Angeles Times, 19 April 1988 [documented behavior · moderate] The same oral history describes later formation work, multi-program biological-risk thinking, financing-cycle exposure, pharmaceutical bargaining constraints, and Monday partnership review; those are retrospective descriptions, not verified portfolio-wide compliance or proof of efficacy. Bancroft, printed pp. 31–54 and 70–79 [retrospective witness · moderate]

Vehicle economics, incentives, and control map

The public record identifies changing capital forms and selected security positions but does not disclose fund agreements or a Byers compensation record. Vehicle effects below are therefore dimension-by-dimension rather than a synthetic claim that one structure caused performance.

People, capital, and decision-rights map

“Unknown” is a finding. The table records only the seven lab cases because they are the sufficiently reconstructed subset; broad biographies do not establish who negotiated, approved, followed on, or exited each company.

Comprehensive discovered-company ledger

The ledger is exhaustive only for the 40 named families recovered in the scoped search. “Candidate” means credible capital, governance, operating, or formation evidence; it does not mean a complete decision record or return. Every row carries a claim-local source and two-axis evidence label. Unknown fields remain unknown rather than being inferred from proximity to KPCB.

Field-completeness supplement for all 40 named families

This row-keyed supplement is part of the ledger, not a second denominator. It makes every §2.4.6 field explicit for every named family. “Unknown” and “insufficient public record” mean that the same-numbered sourced ledger row above did not support the field; they are not filled from analogy. Each row repeats its controlling locator and two-axis label so the supplement remains claim-local and auditable.

Ex-ante decision lab 1 — Hybritech

Exercise status — [researcher inference · strong]. The prompts below are analyst-created from records available before or at the early decision; they are not claimed KPCB questions or recovered IC minutes. Read through “Follow-on/governance/exit-stop” before the outcome.

• Known then: In 1978 a scientific team and monoclonal-antibody diagnostic concept existed, while a complete operating company and established biotechnology template did not; the proposed first plan centered on a narrow six-month technical proof. Bancroft, printed pp. 14–24 · Los Angeles Times, 19 April 1988
• Unknown then: reproducibility, regulatory path, customer adoption, manufacturing, permanent leadership, follow-on capital, ownership/dilution, and eventual liquidity.
• Disputed/retrospective: the early plan and role sequence are substantially retrospective; reported transaction values vary by definition, and no source converts them into KPCB proceeds. [retrospective witness · moderate]
• Alternatives/dissent: license the science to an incumbent, fund only academic work, syndicate a small company, or pass. No contemporaneous dissent memo or alternatives ranking survives.
• Ownership/terms: the startup check was reported at $300,000, but security, cap table, reserves, rights, and target ownership are insufficient public record. Los Angeles Times, 19 April 1988
• Would you invest then? Conditional invest only if the assay test is technically discriminating, the team can reproduce it, and permanent-operator recruiting starts immediately.
• Kill criteria: fail the defined proof; irreproducible assay; blocked rights; no credible permanent leader; or runway unable to reach the result.
• Evidence next check: replicated assay performance, customer/problem validation, regulatory plan, permanent executive, and an ownership/reserve plan.
• Follow-on/governance/exit-stop: use a temporary operator with an explicit handoff; assign scientific, operating, board, and financing owners; stop after failed proof; do not prewrite an acquisition as the exit.
• Actual decision and sourced outcome: KPCB funded; Byers served temporarily as president and then chair after permanent leadership arrived; the proof was reported in roughly three months. Lilly’s transaction was reported near $485 million in contemporaneous coverage, but follow-ons, dilution, proceeds, and time remain unknown. Los Angeles Times, 19 April 1988, formation/role sequence [documented behavior · insufficient public record]

Ex-ante decision lab 2 — IDEC

Exercise status — [researcher inference · strong]. This is a no-hindsight reconstruction from a later board story plus transaction evidence, not a recovered original memo. Stop before the outcome.

• Known then: the company faced monoclonal-antibody biology, clinical development, manufacturing, financing, and strategic-partner risk; a first approach disappointed and a financing window narrowed. Fogarty recording, 25:24–28:52
• Unknown then: whether anti-CD20 would work, Genentech would sustain development, manufacturing and trials would succeed, or a later market/transaction would exist.
• Disputed/retrospective: the reinvestment narrative was told after Rituxan and the Biogen combination; original evidence, exact failure, vote, terms, and dissent are not public. [retrospective witness · moderate]
• Alternatives/dissent: stop the company, sell/license assets, narrow to another program, or bridge to a defined readout. No contemporaneous alternatives memo or dissenter is known.
• Ownership/terms: the later account refers to roughly $6 million of financing, but KPCB’s security, ownership, dilution, reserves, and individual allocations are insufficient public record. Fogarty recording, 25:24–28:52
• Would you invest then? Conditional follow-on only if a biologically distinct program retains independent evidence and the financing reaches a decisive readout.
• Kill criteria: same failed mechanism, no strategic commitment, runway short of proof, manufacturing infeasibility, or opportunity cost superior elsewhere.
• Evidence next check: target validation, preclinical/clinical signal, funded development plan, manufacturing feasibility, and an explicit revised thesis.
• Follow-on/governance/exit-stop: version the thesis; assign board/program owners; separate rescue from new investment; stop if the next readout fails; do not make a future strategic deal the underwriting premise.
• Actual decision and sourced outcome: the board reinvested; IDEC advanced anti-CD20 with Genentech support, and IDEC holders later received 50.5% of the 2003 Biogen combination. That issuer result is not KPCB proceeds. IDEC/Biogen exhibit, transaction terms and Rituxan history [contemporaneous record · moderate]

Ex-ante decision lab 3 — Genomic Health

Exercise status — [researcher inference · strong]. The exercise reconstructs a 2001-era formation decision from the S-1 and later evidence; no original IC packet is claimed.

• Known then: gene-expression measurement could potentially inform breast-cancer treatment decisions; Byers joined the board in 2001, while analytical validity, clinical validity, utility, physician action, reimbursement, and scalable lab operations still required proof. Genomic Health S-1, business and management
• Unknown then: prospective adoption, coverage, unit economics, guideline incorporation, dilution, IPO timing, and exit.
• Disputed/retrospective: “work backward from the clinical decision” is supported as Byers doctrine, but no recovered memo proves every original Genomic Health criterion. [researcher inference · moderate]
• Alternatives/dissent: invest in a platform, partner with an incumbent lab, build a narrower test, wait for validation, or pass. No dissent/alternatives record survives.
• Ownership/terms: KPCB X vehicles held 12.6% immediately before the IPO under shared control; initial check, security sequence, target ownership, and reserves are unavailable. Genomic Health S-1, principal-stockholder table/footnote 9
• Would you invest then? Conditional invest if the assay can change a named treatment decision and a staged plan tests validity, utility, adoption, coverage, and lab execution.
• Kill criteria: no decision impact, nonreproducible signal, clinically weak incremental value, structurally unavailable reimbursement, or unfinanceable evidence path.
• Evidence next check: validated recurrence prediction, demonstrated actionability, payer evidence, lab quality, physician adoption, and a financed next study.
• Follow-on/governance/exit-stop: board milestones should separate assay, clinical, payer, and commercial risk; compare XDx and CardioDx contemporaneously; stop if clinical utility cannot be established.
• Actual decision and sourced outcome: KPCB X vehicles held 12.6% before the IPO; Exact later announced a reported roughly $2.8 billion transaction. The stake at closing and KPCB proceeds are not established. Genomic Health S-1 · transaction filing [contemporaneous record · insufficient public record]

Ex-ante decision lab 4 — Pacific Biosciences

Exercise status — [researcher inference · strong]. This is a platform-underwriting exercise based on the prospectus, not a recovered KPCB memo.

• Known then: single-molecule sequencing promised a differentiated capability but required instrument, chemistry, accuracy, read-length, throughput, workflow, manufacturing, and customer proof. PacBio prospectus, business and risk factors
• Unknown then: technical convergence, cost curve, application mix, adoption speed, capital needs, dilution, and durable competitive position.
• Disputed/retrospective: the public filing records risks and holdings, not Byers’s original rationale, thresholds, alternatives, or dissent. [researcher inference · moderate]
• Alternatives/dissent: fund the platform, wait for integrated performance, invest in applications, syndicate more heavily, or pass. Dissent is insufficient public record.
• Ownership/terms: KPCB Holdings had 3,317,378 shares/8.67% pre-IPO; Byers separately purchased 14,658 Series E shares for $102,606 and 6,724 Series F shares for $51,304. A distinct 47,410-share principal-stockholder row/footnote difference is not assigned to him. PacBio prospectus, recent sales and principal stockholders
• Would you invest then? Conditional invest only with sequenced technical/commercial milestones and capital sufficient to reach each independent falsifier.
• Kill criteria: performance plateau, unreliable manufacturing, cost curve outside target applications, no repeat customer use, or financing unable to reach product proof.
• Evidence next check: repeated accuracy/read-length/throughput results, validated use cases, instrument reliability, customer retention, and manufacturing yield.
• Follow-on/governance/exit-stop: require milestone-specific releases, preserve reserves for technical delay, separate direct from vehicle exposure, and treat IPO as financing rather than liquidity.
• Actual decision and sourced outcome: KPCB held the disclosed pre-IPO position and Byers made the disclosed Series E and Series F purchases; the company became public, while disposition and return remain unknown. PacBio prospectus, recent sales and principal stockholders [contemporaneous record · insufficient public record]

Ex-ante decision lab 5 — CardioDx

Exercise status — [researcher inference · strong]. This is the required failure lab. It freezes an early diagnostic decision using the filing and later evidence controls; no original memo, payer file, or board packet is claimed.

• Known then: a blood-based coronary-disease test addressed a consequential diagnostic decision; Byers served as CEO from July 2003 to August 2004 and as director from July 2003. Clinical validity, changed physician action, patient outcomes, guidelines, payer coverage, and test economics still required proof. CardioDx S-1/A, business and management
• Unknown then: prospective utility, reimbursement durability, guideline uptake, commercial scale, financing, dilution, and stop timing.
• Disputed/retrospective: the attractive unmet-need narrative does not prove the original thesis or each milestone; the surviving record does not expose board dissent or Byers’s later process response. [researcher inference · strong]
• Alternatives/dissent: wait for stronger utility evidence, partner with an incumbent, narrow the intended-use population, redesign the study, or pass. No contemporaneous dissent record survives.
• Ownership/terms: the S-1/A separates direct Byers securities from KPCB X positions; cost basis, later dilution, reserve decisions, recovery, and realized loss are unavailable. CardioDx S-1/A, principal stockholders and footnotes
• Would you invest then? Conditional invest only if a staged plan makes clinical utility and payer acceptance early go/no-go evidence, not post-launch assumptions.
• Kill criteria: no incremental decision value, irreproducible validation, adverse evidence review, coverage structurally unavailable, or runway unable to reach a decisive study.
• Evidence next check: blinded validation, changed physician action, outcome/utility evidence, payer criteria, test economics, and sufficient runway.
• Follow-on/governance/exit-stop: give the board an evidence calendar; zero-base each bridge; stop or redesign after a failed utility/coverage threshold; do not use IPO readiness as proof of clinical value.
• Actual decision and sourced outcome: CardioDx was funded, filed for an IPO, withdrew it, and shut in 2019 after evidence and reimbursement problems. The filing-backed direct/vehicle positions do not reveal the loss. CMS coverage analysis, evidence review · San Francisco Chronicle, 8 January 2019 [contemporaneous record · insufficient public record]

Ex-ante decision lab 6 — Genentech as an attribution control

Exercise status — [researcher inference · strong]. This lab tests credit assignment, not whether Genentech was a good company. It uses the firm’s role history and excludes hindsight performance.

• Known then: Bob Swanson and Herbert Boyer founded the company; early needs included scientific translation, product planning, market analysis, capital, licensing, recruiting, and governance. Kleiner Perkins Genentech history, founders and formation
• Unknown then: product feasibility, regulation, manufacturing, licensing, market adoption, future financing, and value.
• Disputed/retrospective: later firm pages place Byers near the company, but that proximity cannot rewrite founder, board, investment-authority, or ownership roles. [documented behavior · strong]
• Alternatives/dissent: K&P could finance, syndicate, delay, or pass; Byers could assist without owning the investment decision. Original dissent and terms are not public.
• Ownership/terms: Byers-attributable ownership, check authority, and economics are insufficient public record; no position is assigned to him.
• Would you invest then? The analyst may invest in the company, but must exclude the case from a Byers-originated ranking absent attributable decision and vehicle evidence.
• Kill criteria: for attribution, stop personal-credit assignment when founder, sponsor, board, vehicle, or decision authority belongs elsewhere or remains unknown.
• Evidence next check: contemporaneous cap table, investment memo, partner approval, board minutes, and a role-specific work record.
• Follow-on/governance/exit-stop: maintain a credit ledger separating scientific invention, company formation, investment decision, board work, operations, follow-ons, and exit influence.
• Actual decision and sourced outcome: the firm history names Swanson/Boyer as founders, Perkins/Kleiner as original board principals, and Byers as assisting product planning/market analysis. Genentech remains a learning case and an excluded Byers-economic case. Kleiner Perkins Genentech history [documented behavior · strong]

Ex-ante decision lab 7 — Enjoy and follow-on discipline

Exercise status — [researcher inference · strong]. Because the original Waycross entry thesis is missing, this lab freezes only the public-combination/follow-on decision visible in the S-1. It does not retrofit Byers’s historical doctrine to the company.

• Known then: the filing disclosed a service-heavy model, operating losses, negative operating cash flow, partner concentration, capital need, a Waycross holder managed by Byers, and a separately controlled KPCB XVI position. Enjoy S-1, business, risk factors, liquidity, and selling-securityholder footnotes
• Unknown then: market-level contribution margin, durable repeat behavior, future partner terms, labor productivity, financing availability, Waycross cost, and loss severity.
• Disputed/retrospective: the public record does not prove Byers sponsored the original investment or approved a follow-on; company failure alone cannot establish personal loss or process failure. [contemporaneous record · insufficient public record]
• Alternatives/dissent: sell, hold without adding, add only after unit-economic proof, restructure, or pass on the public transaction. Original dissent is unavailable.
• Ownership/terms: the filing identifies Waycross and KPCB XVI separately, but does not reconstruct Waycross entry security, cost, later dispositions, or Byers economics. Enjoy S-1, selling-securityholder footnotes
• Would you invest then? Pass on new capital unless cohort/market contribution margins, labor productivity, partner durability, and liquidity create a credible path to self-funding.
• Kill criteria: persistently negative contribution after allocated labor, concentrated partner power without protection, cash runway below a credible proof period, or financing that only delays insolvency.
• Evidence next check: contribution margin by market, repeat economics, partner sensitivity, labor utilization, liquidity runway, and a downside restructuring plan.
• Follow-on/governance/exit-stop: separate existing-holder incentives from fresh underwriting; assign a kill advocate; define a liquidity floor; choose sale/restructure/stop before emergency financing.
• Actual decision and sourced outcome: the public transaction implied reported enterprise value near $1.18 billion; Enjoy entered Chapter 11 about eight months after listing. That establishes company failure, not Waycross cost or loss. Enjoy S-1, transaction summary and risk factors · official restructuring docket [contemporaneous record · insufficient public record]

Portfolio forensics: observed structure and missing denominator

This is a forensic description of the 40 discovered names, not an estimate of the unitemized 60-plus firm claim. The searchable record over-selects companies mentioned in Byers’s oral history, current firm marketing, surviving filings, and known outcomes; failed passes, quiet write-offs, unpublicized follow-ons, and companies absent from those routes are likely missing. Consequently, observed counts are evidence of source coverage, not portfolio weights or a return distribution. [researcher inference · strong] Denominator reconciliation

Failure and adverse-evidence postmortems

CardioDx — clinical utility and reimbursement failure

• Original thesis and evidence: the filing describes a blood-based test for coronary-disease assessment and documents Byers’s early CEO/director role, but no original IC memo or Byers-authored thesis survives. CardioDx S-1/A, business and management [contemporaneous record · provisional]
• Warning signs: clinical-utility and coverage scrutiny became decision-critical; the CMS evidence review and later shutdown reporting control the adverse record. CMS coverage analysis, evidence review · San Francisco Chronicle, 8 January 2019
• Capital response: the filed direct and KPCB securities show exposure, but follow-on checks, reserve decisions, bridge terms, and any recovery are insufficient public record.
• Governance response: Byers’s board role is public; votes, study redesign, payer strategy, CEO interventions, and dissent are not.
• Stop/exit decision and outcome: the IPO was withdrawn and the company shut in 2019; cost and realized loss are unavailable. CardioDx S-1/A, offering status and ownership · shutdown report
• Claimed lesson and later process change: no located Byers statement documents a CardioDx-specific lesson or changed process. The package’s rule—underwrite clinical utility and payment as product milestones—is [researcher inference · strong], not investor testimony.
• Attribution and counterfactual: Byers’s operating/board role makes this a strong accountability case, but management, clinical evidence, payers, KPCB partners, and the board share causation; it is unknown whether earlier stopping, another study, or different coverage policy would have changed the result.

Nanogen — bankruptcy after public-market access

• Original thesis and evidence: Byers’s oral history supports a portfolio/board association, but no original thesis, entry security, target ownership, or milestone plan survives. Bancroft, portfolio discussion [retrospective witness · provisional]
• Warning signs: the decisive public record is bankruptcy, not a recovered earlier warning chronology.
• Capital response: KPCB checks, follow-ons, refusal to fund, dilution, and recovery are insufficient public record.
• Governance response: board role is reported retrospectively; votes, management changes, restructuring alternatives, and dissent are unknown.
• Stop/exit decision and outcome: Nanogen reported a $25.685 million Section 363 sale of substantially all assets and said management expected no value for common stockholders. Nanogen issuer release, paragraphs 2 and 4 [contemporaneous record · insufficient public record]
• Claimed lesson and later process change: no Nanogen-specific Byers lesson or documented portfolio-rule change was located. “IPO/public visibility is not realization or durable value” is [researcher inference · strong].
• Attribution and counterfactual: the company, creditors, customers, management, board, and capital structure shaped the loss; without KPCB cash flows and board records, neither personal loss nor a better stop date is supportable.

InSite Vision — long duration and a modest endpoint

• Original thesis and evidence: a public filing supports Byers’s founding-president/chair role, while the original ophthalmic-development thesis, terms, and milestones remain unavailable. Genomic Health S-1, Byers director biography [contemporaneous record · insufficient public record]
• Warning signs: the long development history and modest endpoint are visible retrospectively; a dated warning-signal sequence is not public.
• Capital response: KPCB initial/follow-on capital, reserve decisions, security changes, dilution, and disposition are unknown.
• Governance response: formation title does not reveal later board votes, recruiting, program stops, financing alternatives, or dissent.
• Stop/exit decision and outcome: Sun Pharma’s 2015 transaction was reported at about $48 million equity value and $58.9 million transaction value. InSite issuer release, transaction terms [contemporaneous record · insufficient public record]
• Claimed lesson and later process change: no InSite-specific Byers postmortem was found. The time-to-liquidity and opportunity-cost lesson is [researcher inference · moderate] because invested capital and duration-adjusted cash flows are incomplete.
• Attribution and counterfactual: Byers’s formation role supports accountability but not sole credit; whether earlier program pruning, financing, or sale would have improved the investor outcome is unknown.

IDEC — failed program, re-underwritten continuation

• Original thesis and evidence: the public role record supports formation leadership; the original antibody-program memo and exact criteria do not survive. Genomic Health S-1, Byers director biography [contemporaneous record · insufficient public record]
• Warning signs: Byers later described an initial clinical failure and financing constraint. Fogarty recording, 25:24–28:52 [retrospective witness · moderate]
• Capital response: the later story describes roughly $6 million and renewed backing, but security, KPCB share, reserves, and alternative use of capital remain unknown.
• Governance response: the board reportedly voted to reinvest; members’ arguments, dissent, program-level evidence, and conditions are not public.
• Stop/exit decision and outcome: the failed path was not treated as the end of the company; IDEC advanced anti-CD20 with Genentech and IDEC holders later received 50.5% of the Biogen combination. IDEC/Biogen exhibit, transaction terms and Rituxan history
• Claimed lesson and later process change: Byers’s late lesson emphasizes tenacity; the narrower “fresh thesis plus new evidence” rule is [researcher inference · strong]. No portfolio-wide continuation ledger demonstrates a formal process change.
• Attribution and counterfactual: scientists, executives, Genentech, the board, and KPCB supplied critical work. IDEC’s favorable outcome creates survivorship bias; whether stopping after the first failure would have been rational ex ante cannot be decided without the missing packet.

Enjoy — public-combination failure with incomplete investment attribution

• Original thesis and evidence: the S-1 identifies Byers as Waycross’s managing member, but it does not recover the original entry thesis, sponsor, security, cost, or approval. Enjoy S-1, selling-securityholder footnotes [contemporaneous record · insufficient public record]
• Warning signs: the filing disclosed operating losses, negative operating cash flow, partner concentration, labor intensity, and continuing capital needs. Enjoy S-1, risk factors and liquidity
• Capital response: the public combination supplied company capital, but Waycross follow-ons, sales, reserves, and loss are not established.
• Governance response: Byers’s Waycross authority does not prove a company board vote or KPCB XVI authority; KPCB XVI’s disclosed control group was separate. Enjoy S-1, selling-securityholder footnotes
• Stop/exit decision and outcome: Enjoy entered Chapter 11 about eight months after listing; the docket establishes the case, not the amount of any Waycross loss. official restructuring docket, case landing page [contemporaneous record · insufficient public record]
• Claimed lesson and later process change: no located Byers postmortem or documented process change exists. The unit-economics/follow-on rule is [researcher inference · moderate].
• Attribution and counterfactual: severe company failure is clear; original Byers sponsorship and personal economics are not. A different public transaction, earlier sale, or no follow-on cannot be scored without holder cash flows and decision records.

Attribution controls that prevent false failures or wins

Genentech remains excluded because official history allocates founding and original investment leadership elsewhere while bounding Byers’s assistance. Kleiner Perkins Genentech history Align remains excluded because firm history allocates investment leadership and board service to Joe Lacob. Kleiner Perkins Align history These controls are [documented behavior · strong] and prevent proximity to KPCB from becoming personal credit.

Adverse-source and data-integrity controls

The record includes allegations or interested narratives that cannot be converted into findings. A PacBio derivative complaint named Byers, but the court approved a voluntary dismissal without prejudice; the order does not adjudicate the allegations. N.D. Cal. order, Case 4:11-cv-06703-CW, pp. 1–3 [contemporaneous record · strong] WIRED’s Pao-trial report says Trae Vassallo sent her complaint to Byers and other partners and that the firm investigated; it does not establish personal misconduct by Byers. WIRED, 25 February 2015, trial report paragraphs 12–13 [contemporaneous record · moderate] Proposition 71 financing criticism and later CIRM governance criticism establish public/private-benefit and governance scrutiny; neither cited source shows that Byers personally received a grant. Los Angeles Times, 18 October 2004 · Nature commentary, 2020, pp. 1–2 [contemporaneous record · moderate]

The current KP bio also labels Crescendo “partnered since 2018,” although Myriad announced and completed the acquisition in 2014. Myriad later recorded reported final consideration of $259 million. This impossible date is a direct warning not to infer entry years or vehicles from portfolio cards. Kleiner Perkins biography, Crescendo card · Myriad announcement, 4 February 2014 · Myriad 10-K, acquisition note [contemporaneous record · strong]

Uncertainty ledger and retrieval agenda

The record supports a hands-on company-formation practice with several major reported company outcomes. It does not support a personal return series, a fund return series, or a numerically defensible hit rate.

Coverage and evidence base

The public written record is genuinely sparse while the oral record is rich. Four underlying works are personally credited or signed: the Pratt chapter, the Perkins introduction, the Fountain Wind comment, and the Dovey foreword. Two are fully accessible and two are partial. Two further decision-relevant documents are collective: the NCATS working-group report and the Nature Biotechnology correspondence. They establish participation or assent, not individual drafting. [researcher inference · strong]

All six have source notes. The underlying-work count deduplicates Pratt editions and campaign reposts. CIRM testimony, Georgia Tech commencement, Stanford, Fogarty, PMWC, and Byers's own oral history are speeches or interview transcripts and belong in talks, not this denominator. The Genentech, Electronic Arts, and Frank Caufield pages are unbylined firm histories; SEC signatures, board consents, book blurbs, and press quotations are adjacent evidence rather than authored investment writing. [researcher inference · strong]

Additional signature records—Y2K liability, a teacher-policy campaign, DACA, and drug-pricing letters—were inventoried but excluded from deep analysis because the public record establishes at most cosignature and in some cases only secondary/platform evidence. A 1978 Hybritech business plan is also excluded: later recollections conflict over whether Byers, Ivor Royston, Howard Birndorf, or Howard Greene wrote or refined it, and the archived document was not inspected. [researcher inference · strong] contemporary Hybritech account

Source-by-source analysis

1. Relationship Between Venture Capitalist and Entrepreneur — by 1983, editions through 2006

The strongest bibliographic title omits an initial “The.” A 1983 Georgia Tech profile already described Byers as a guide contributor; a World Bank bibliography explicitly locates the chapter in the 1986 edition; later citations recur; and a 2006 thirtieth-edition table of contents places it at pp. 105–108. These records establish one repeatedly published underlying work, not multiple essays, and do not prove that every edition is textually identical. [contemporaneous record · moderate] 1986 citation route · 2006 contents route

The full chapter was not legitimately accessible through Google Books, Open Library, Internet Archive, or catalog/citation routes. An earlier discovery note referred to a secondary quotation but retained no canonical URL and pinpoint locator; REVIEW therefore excludes that fragment and assigns no content proposition to it. The surviving evidence establishes bibliography, title, edition recurrence, and the 2006 contents placement—not argument, reasoning, examples, ownership policy, or governance doctrine. [researcher inference · insufficient public record] source note

The operational implication is provisional: a venture relationship may be underwritten as a continuing service and governance commitment rather than a financing event. Whether Byers applied that rule consistently must be tested against company and board records rather than inferred from a partial text. [researcher inference · provisional]

2. Introduction to Thomas J. Perkins oral history — 2001

In a signed one-page introduction, Byers praises a path from operating experience to active venture work. Sales, product, marketing, and general-management exposure supposedly made Perkins useful on technology and market problems; intellectual curiosity enabled cross-domain engagement; mentoring later partners extended the practice institutionally. Perkins oral history, printed p. iv [investor-stated · strong]

The implied chain is: operating depth improves diagnosis; technical curiosity broadens intelligible opportunities; active board participation converts capital into company-building help; mentorship makes judgment less dependent on one person. That is evidence of what Byers admired, not proof that he reproduced it or that it caused returns. [researcher inference · moderate] source note

The piece is commemorative and favorable. Its round firm-wide totals for companies, revenue, market capitalization, and employment lack a calculation, ownership denominator, or performance methodology and are not used as return evidence. It gives no failure case, founder conflict, price rule, or falsifier. [researcher inference · strong]

3. NCATS working-group summary of findings — 2011

The NIH Advisory Committee to the Director names Byers among fifteen working-group members. The report addresses translational bottlenecks, precompetitive infrastructure, public/private partnerships, coordination, and avoiding duplication. NCATS report, cover, p. 1, findings and Appendix A [contemporaneous record · strong]

The group logic treats translational science as infrastructure: shared methods and coordination can address failure points that no single lab or company has sufficient incentive or scope to solve. That is decision-relevant context for Byers's life-sciences network, but no recommendation, sentence, edit, or dissent can be assigned personally without minutes or a separate signed statement. [documented behavior · moderate] source note

Operationally, the report invites a public/private dependency map: identify which precompetitive capability should be shared, which company-specific risk private capital should bear, and where a new institution would duplicate existing work. Treating the whole consensus as Byers's proprietary investment philosophy would be false precision. [researcher inference · moderate]

4. Fountain Wind scoping comment — 2019

Byers's signed California filing supports renewable energy in general but opposes this project's siting near Moose Camp because of visual, shadow, and community effects. California Energy Commission filing, Letter P24 [investor-stated · strong]

The letter is full and individually attributable, but it is outside investing. It shows project-level opposition coexisting with mission-level support; that can be read as attention to local externalities or as self-interested NIMBYism. The document cannot adjudicate between those interpretations and supplies no portfolio application. [researcher inference · moderate] source note

Its limited decision lesson is to separate agreement with a category from underwriting a particular project. Applying that boundary to venture decisions is an analyst-created analogy, not a documented Byers rule. [researcher inference · provisional]

5. Supporting US public health experts — 2020

The Nature Biotechnology correspondence supports Anthony Fauci, NIAID, and public-health institutions during COVID-19 and presents biotechnology as a strategic national innovation asset. Byers appears in a large cosignatory list, not among the nine named lead authors. Open Springer Nature text and signatories [contemporaneous record · insufficient public record]

The signature is consistent with his institutional biomedical activity, but it cannot establish personal drafting, a distinct investment rule, or how he allocated capital during the pandemic. Bibliographic systems that enumerate all signers as “authors” do not override the document's cosignatory structure. [researcher inference · strong] source note

6. Foreword to The Idea Is the Easy Part — 2023

Simon & Schuster metadata and the book's table of contents credit Byers with the foreword to Brian Dovey's startup book. Publisher record The foreword body was not accessible through the legitimate public routes checked. [contemporaneous record · insufficient public record]

The official endorsement page indicates approval of realistic, myth-resistant startup advice, but an endorsement cannot stand in for the foreword's reasoning or examples. The only safe biographical inference is that Byers remained involved in transmitting founder/investor lessons in 2023. [researcher inference · moderate] source note

Cross-source synthesis

The controlled personal writing supports a narrow durable emphasis: the Perkins introduction praises operating competence, active partnership, and mentorship. The Pratt chapter’s bibliographic existence is established, but its inaccessible body contributes no substantive proposition. [researcher inference · moderate] Pratt note · Perkins introduction

The institutional record adds a different scale. NCATS addresses shared translational infrastructure, while the Nature letter endorses public-health expertise. These are group positions with limited personal attribution, but they place Byers inside public/private biomedical coordination as well as company boards. [documented behavior · moderate] NCATS · Nature

The corpus says almost nothing reliable about price, ownership targets, reserves, concentration, follow-on rules, exits, or fund economics. It cannot independently supply a complete investing system; those elements must come from the spoken record and deal filings. [researcher inference · strong]

Intellectual chronology, evolution, and contradictions

Continuity from relationship → operating mentorship → translational infrastructure is a plausible researcher synthesis, not a proven intellectual progression. The sharpest tension is between repeated celebration of active engagement and the absence of public writing on board conflict, management replacement, failed governance, or accountability. [researcher inference · moderate]

Decision playbook and modern translation

Current applications must re-test fund size, partner capacity, clinical/regulatory timelines, data rights, AI-model validation, reimbursement, and public-institution dependencies. The historical sources do not justify copying KPCB-era board intensity or Bay Area networks into a modern fund without those conditions. [researcher inference · strong]

Best available writings and unresolved record

1. Relationship Between Venture Capitalist and Entrepreneur is most relevant to investment craft, but remains partial and therefore cannot anchor a complete doctrine. Catalog
2. Introduction to Thomas J. Perkins oral history is the best fully accessible personal writing and the clearest statement of admired operating and institutional traits. Official PDF, printed p. iv
3. NCATS working-group findings is the best decision-relevant collective text, with the strict boundary that none of its recommendations is individually attributable. NIH report

Only three are ranked; expanding to five would pad the list with a non-investing local comment, an inaccessible foreword, or a collective signature. The highest-value missing records are a controlled Pratt chapter with edition comparison, the Dovey foreword, public NCATS minutes/drafts, the 1978 Hybritech plan and letter, and any legitimate partner memos, LP letters, board letters, or annual communications. No personal blog, book, shareholder letter, or broader signed investment-essay archive was found after the documented archive, catalog, publisher, firm, regulatory, and web searches. [researcher inference · strong]

Coverage denominator and access

The defensible spoken corpus contains eight core appearance families, not the 30–80 items expected for a podcast-era investor. All eight have notes and were inspected at the strongest legitimate public route found. Two are fully text-controlled—the Bancroft oral history and 2005 CIRM transcript. Stanford's 2005 lecture has a full audio shell and official transcript fragments; Stanford 2007 and Fogarty 2019 have official media plus research-generated or automatic captions with speaker/timestamp maps. Charlie Rose now has a site-provided English VTT and two explicitly cued Byers turns; the two PMWC panels remain participation-only because no legitimate caption/audio acquisition succeeded. [researcher inference · strong]

Stanford's “Ideal Team,” “Magic Mondays,” “Multiple Investors,” “Decision-Making,” “Customers,” “Getting Funding,” and related pages are clips from one April 2005 lecture and are deduplicated. The two Berkeley catalog URLs likewise point to the same 2006 oral history. Metadata-only Georgia Tech, UCSF, SIEPR, Personalized Medicine Coalition, Duke, and cancer-panel events are useful archive leads but not treated as completed substantive appearances. Searches of major venture/podcast venues found no verified Byers appearance on Invest Like the Best, 20VC, Acquired, Tim Ferriss, Knowledge Project, Masters of Scale, a16z, C-SPAN, CNBC, or Bloomberg. [researcher inference · strong]

Source-by-source guide

1. Bancroft oral history — the formation and doctrine spine

The oral history covers Byers's path from Georgia Tech and Stanford to Pitch Johnson's Asset Management Company, then KPCB, Hybritech, the life-sciences specialization, later diagnostics, public policy, and forecasts. Bancroft, printed pp. 1–86 It is the only source where his formation, company stories, risk taxonomy, financing-cycle view, partnership model, and personalized-medicine thesis appear in one sustained record. [retrospective witness · moderate]

The core process is not “pick great science.” It is to work backward from unmet medical needs; build competence and networks in the relevant science and clinic; form companies where an important need lacks an existing vehicle; recruit scientific and operating leaders; stage capital against technical and market risk; and govern actively. Bancroft, printed pp. 27–46 The interview also says large pharmaceutical companies tend to wait for de-risked programs and that multi-program biotechnology portfolios can diversify biological risk. [investor-stated · moderate]

Because the account was recorded decades after early investments and reviewed by Byers, filings and independent founders control role, ownership, and outcome disputes. No quotation is reproduced: the volume's rights notice requires permission. source note

2. Stanford 2005 — the clearest operating lecture

The Stanford lecture supplies the most decision-ready direct rules. Byers connects firm scale to partner time, suggests roughly two new company commitments per investor per year, describes two or three co-investors as workable, and says ownership must compensate for the intensity of engagement. He also emphasizes an internal champion, actual customer resonance, complementary teams, and close investor–founder work. Official lecture family and transcript [investor-stated · strong]

These are dated 2005 claims, not universal constants. The archive provides a full audio/podcast shell but only fragmented official clip transcripts. The fragments imply testable capacity and portfolio questions: how many boards can a partner serve well; what ownership and reserves justify that time; when does syndication improve judgment versus blur responsibility; and who inside the partnership owns the decision? The lecture itself constrains hero attribution by identifying Align as Joe Lacob's deal. [researcher inference · strong] source note

3. Charlie Rose 2004 — a contemporaneous policy checkpoint

The official archive identifies Byers, Susan Desmond-Hellmann, and Robert Klein on a medical-technology panel shortly before California's stem-cell implementation debates. Charlie Rose archive A 2026-08-23 caption probe recovered the site-provided English VTT. After an explicit host cue, Byers discusses systems biology, network/pathway complexity, and the long database-building horizon at 10:34.934–11:36.434; after a second explicit cue, he discusses software, biological complexity, information access, health-care delivery, records, and privacy at 18:04.400–20:24.534. [investor-stated · moderate] caption retrieval and speaker map Uncued panel dialogue remains unassigned.

4. CIRM testimony 2005 — ecosystem model and attribution warning

Speaking for San Francisco, Byers describes five complementary inputs for a biomedical cluster: basic science, clinical translation, companies, capital, and favorable policy; he offers institutions, scientists, NIH funding, companies, and trained labor as practical measures. CIRM transcript, pp. 192–193 [investor-stated · strong]

The testimony is contemporaneous and explicit, but it is also an interested site-selection pitch. Its loose statement that he was involved in starting Genentech demonstrates why public rhetoric needs role controls: Swanson and Boyer founded the company, while Tom Perkins led the early Kleiner & Perkins relationship. Kleiner Perkins Genentech history [documented behavior · strong]

5. Stanford 2007 — broader mission and neglected-problem rule

The 1:21:28 official panel places Byers with John Hennessy, Kavita Ramdas, Paul Yock, KR Sridhar, and others on entrepreneurship and world problems. Stanford page A research diarization of the official audio maps his principal turns: interdependent global problems and public-private collaboration [06:54–09:08], early-detection health technologies [09:09–15:24], public policy and science as leverage [67:10–69:09], and the rule to attack neglected problems instead of duplicating an existing organization [72:09–72:55]. [investor-stated · moderate]

This adds a mission-selection rule and a contemporaneous personalized-medicine forecast, but it is research ASR rather than an official transcript. The panel's public-interest setting also cannot establish that these topics became underwriting criteria. source note

6. PMWC 2011 — forecast checkpoint

The 2011 panel places Byers with George Church, Lee Hood, Kathy Hudson, and Ralph Snyderman discussing the next decade of personalized medicine. PMWC/Vimeo It sits after Genomic Health, XDx, and CardioDx and should eventually be tested against 2019 claims and realized diagnostics outcomes. Caption and speaker timestamps were not surfaced, so no exact forecast is personalized here. [contemporaneous record · insufficient public record]

7. PMWC 2019 — later forecast endpoint

The 2019 panel with J. Craig Venter and Ralph Snyderman is a natural endpoint for the forecast sequence that begins in the oral history and the 2011 PMWC session. PMWC/Vimeo It establishes continuing public association with personalized medicine; it does not, without speaker locators, establish which expectations Byers retained or abandoned. [contemporaneous record · insufficient public record]

8. Fogarty 2019 — mature retrospective and persistence case

The Fogarty appearance emphasizes teams [12:14], psychological safety [23:00], and persistence. Its most decision-relevant story begins around 25:24: the board's decision to reinvest in IDEC after an initial clinical failure, followed by a persistence lesson around 28:52. Later Oculeve, OptiMedica, CareDx, and Verana stories widen the case set. KP official recording · Fogarty recap [retrospective witness · moderate]

The lesson is not “always persist.” It is to identify what evidence remained after the failed trial, what mechanism or team belief changed, what kill criteria applied, and why another check had positive expected value. Those details require filings and the deal reconstruction; the late favorable outcome can otherwise turn persistence into a tautology. [researcher inference · strong] source note

Recurring stories and frameworks

Deal-credit controls

Evolution and biographical turning points

• 1970–1977: the oral history links a Stanford VC panel to apprenticeship under Pitch Johnson and then KPCB entry. The sequence is documented; motivation and causality remain retrospective. [retrospective witness · moderate]
• 1978–1986: Hybritech plus a study period precede Gen-Probe and a five-company formation cohort. The public record supports a shift toward exclusive life-sciences specialization, while the exact partner roles differ by case. [documented behavior · strong]
• 2002–2007: the oral history, Charlie Rose, Stanford, and CIRM show the company-building model widening into personalized medicine and public biomedical infrastructure. CIRM's broad Genentech language also shows narrative-credit drift. [researcher inference · moderate]
• 2011–2019: PMWC and Fogarty shift the public emphasis toward forecast review, teams, psychological safety, and persistence. Without complete PMWC speaker transcripts, intellectual change cannot be cleanly separated from topic and venue. [researcher inference · provisional]

Recollection versus contemporaneous records

1. Byers began venture capital at Asset Management in 1972 and joined KPCB in 1977; biographies that place him at KPCB from 1972 conflate career and firm dates. Pitch Johnson oral history
2. Genentech is a learning and early-assistance case for Byers, not his founding investment. Swanson, Boyer, Perkins, and Kleiner retain origin credit. Firm history
3. The Stanford archive identifies Align as Joe Lacob's deal; do not convert a KPCB example into Byers sourcing. Stanford 2005
4. Fogarty's IDEC persistence story is outcome-aware. The original board evidence, operating-team role, alternatives, security, and follow-on terms must control the no-hindsight case. [researcher inference · strong]
5. The eight core appearances are a selected public denominator. Repetition within them cannot reveal omitted failures, passes, private disagreement, or ordinary board work. [researcher inference · strong]

Read, watch, or listen to these five first

1. Bancroft oral history — the complete formation and doctrine spine; read with the rights and retrospective-bias warnings. PDF
2. Stanford 2005 lecture — the clearest operating rules for capacity, syndication, customers, teams, and partnership decisions. Lecture family
3. Fogarty 2019 — best late-career team and persistence synthesis; use the IDEC story as a prompt for forensic verification. Video
4. CIRM 2005 testimony — strongest concise ecosystem framework and a live demonstration of attribution risk. Transcript, pp. 192–193
5. Stanford 2007 panel — best contemporaneous bridge from personalized medicine to neglected-problem selection and public-policy leverage; read with the ASR caveat. Audio

Unresolved spoken record

Searches of Invest Like the Best, 20VC, Acquired, Tim Ferriss, Knowledge Project, Masters of Scale, a16z, C-SPAN, CNBC, Bloomberg, official firm archives, Stanford, Vimeo/PMWC, and title/year variants found no additional verified Byers appearance with a recoverable transcript. Duplicate Stanford clips, mere event listings, interviewer roles, documentary supporter credits, and third-party mentions were excluded from the eight-family denominator. [researcher inference · strong]

Work backward from unmet medical need

• Rule: Start with a consequential clinical decision or patient need; specify what must change in care, then work backward to the science, evidence, product, regulation, reimbursement, and company required to make that change.
• Evidence status: [investor-stated · moderate]. Byers describes the process directly, while the personalized-diagnostics sequence supplies researcher-applied comparisons rather than recovered decision memos; the public portfolio has neither a complete denominator nor attributable returns.
• Provenance and origin: The earliest controlled articulation is Byers's retrospective account of his 1983–1985 study period and company formation, especially the unmet-need discussion at Bancroft, printed pp. 27–36. The archive's rights notice requires paraphrase, so no quotation is reproduced.
• Mechanism and reasoning: A clinical-decision anchor reverses technology-push diligence. It makes product utility, adoption, payment, and implementation falsifiable before an analyst mistakes scientific novelty for demand. [researcher inference · moderate] The causal claim remains unproven because selection, team, vehicle, and era effects cannot be isolated.
• Historical and vehicle context: Byers's retrospective account describes sparse late-1970s biotechnology precedent, small syndicated rounds, and KPCB capital paired with formation labor (Bancroft, printed pp. 18–24 and 39–46). [researcher inference · moderate] Those conditions favored a narrow proof plan over an all-at-once platform promise; later diagnostics also depended on clinical validation and payer coverage, as CardioDx demonstrates (shutdown account).
• When to use: Use at pre-seed through early clinical diligence when a technology claims medical value. Do not use it as a slogan when the buyer, care decision, regulatory path, or measurable patient benefit cannot be named; and do not infer willingness to pay merely from disease burden.
• How to apply it:
  1. Write the present care decision, responsible user, harmed population, and counterfactual without the product.
  2. Name the minimum change in diagnosis, treatment, outcome, or cost that would matter.
  3. Map analytical validity, clinical validity, clinical utility, workflow, regulation, and payment; mark the first fatal dependency.
  4. Ask the IC: “What result would change care?”, “Who must believe it?”, and “What result kills the need thesis?”
  5. Fund only to the next decision-changing proof; re-underwrite the company after that result.
• Examples:
  • Genomic Health — researcher-applied illustration: the public product record is consistent with a decision-first thesis—whether gene-expression information could alter breast-cancer treatment—but the filing does not connect Byers's rule to the investment. KPCB X vehicles held 12.6% before the IPO; Exact Sciences later announced a reported roughly $2.8 billion transaction, but KPCB proceeds remain unknown (Genomic Health S-1; transaction filing).
  • XDx/CareDx — researcher-applied illustration: the transplant-care product is consistent with decision-first selection, but the filing documents Byers as chair and KPCB X nominee ownership—not the original rule or a personal stake (XDx S-1).
  • CardioDx — researcher-applied counterexample: a blood test addressed a consequential coronary-disease decision, yet evidence and reimbursement pressure contributed to the 2019 shutdown. The outcome tests the rule's limit; neither filing nor reporting recovers the entry rationale (CardioDx S-1; shutdown account).
  • Hybritech — researcher-applied illustration: the initial narrow diagnostic proof is consistent with the rule, but the surviving contemporary account does not establish the full unmet-need memo or every contributor's rationale (Los Angeles Times, 1988).
• Evidence: The rule is explicit in Bancroft, printed pp. 27–36. The Genomic Health, XDx, and CardioDx filings independently document products, roles, and vehicle positions used for researcher-applied comparison—not recovered uses of the rule; the adverse outcome is independently reported by the San Francisco Chronicle.
• Limits, failure modes, and exception ledger: Important needs can lack a reimbursable product, validated biomarker, scalable workflow, or reachable buyer. CardioDx is the central exception; Genomic Health is the favorable selected case. The 0/40 economically measurable denominator prevents a performance claim. The archive sweep found no attributable rejected-need ledger, scoring rubric, or contemporaneous IC memo; Bancroft, SEC filings, company outcomes, and Stanford/PMWC routes were checked, while PMWC speaker-level captions remain unavailable.
• Evolution and contradictions: The early formation narrative widened into personalized medicine and early detection by 2007 (Stanford panel, 09:09–15:24). The later comparison is bounded: Genomic Health's filing documents the company/vehicle and its later transaction was reported at roughly $2.8 billion (Genomic Health S-1; transaction filing); XDx's filing establishes a separate transplant-diagnostics company and Byers/KPCB roles without a complete return (XDx S-1); and CardioDx shut after evidence and reimbursement problems (CardioDx shutdown). The contradiction resolves only if “need” begins underwriting rather than proving a diagnostic will work.
• Modern VC translation (as of 2026-08-23): [researcher inference · moderate] Transfers: preserve the decision-backward map. Does not transfer: 1980s local-network access, check sizes, and assay-first proof cannot be presumed adequate for an AI-enabled diagnostic; add prospective utility, data provenance, change control, implementation cost, and payer/procurement evidence. FDA's 2025 draft treats AI-enabled devices across the product lifecycle and is explicitly non-binding, establishing a changed regulatory-documentation premise without dictating an investment result (FDA guidance, status and lifecycle scope). Signals: a named decision owner, measurable counterfactual, and near-term proof. Likely misuse: wrapping a biomarker or AI model in a large-disease narrative. Current evidence boundary: UCSF documents Epiodyne formation help but not an investment memo or milestones, so no grounded modern application is asserted (UCSF, 2024).

Specialize until learning and network compound

• Rule: Concentrate deeply enough in one technical and clinical domain that accumulated pattern recognition, trusted specialists, founder references, and operating talent improve the next decision; expand only when a comparable learning network exists.
• Evidence status: [investor-stated · moderate]. Byers repeatedly explains life-sciences specialization as a learning/network system, but no matched comparison proves that specialization caused superior returns.
• Provenance and origin: The supportable origin is the study period after Hybritech and the move into a life-sciences formation cohort, reconstructed in Bancroft, printed pp. 27–39. Stanford's 2005 ecosystem clip is a later contemporaneous articulation (VC Industry: An Ecosystem, transcript p. 1).
• Mechanism and reasoning: Repeated domain work lowers search and reference costs, lets an investor compare scientific and clinical claims, and creates a reusable pool of scientists, executives, co-investors, and strategics. [researcher inference · moderate] The signal is not number of contacts; it is whether the network produces independent disconfirmation, talent, and faster risk retirement.
• Historical and vehicle context: Contemporary reporting describes the San Diego biotechnology cluster and scientist-commercialization network (Los Angeles Times, 1991). Byers's CIRM testimony named science, clinical translation, companies, capital, and policy as complementary Bay Area inputs, although it was an interested site pitch (CIRM, printed pp. 191–193). The inference that KPCB labor and long boards enabled relationship compounding is [researcher inference · moderate], supported by the active-board account rather than a measured network-return study (Bancroft, printed pp. 39–68).
• When to use: Use when expertise and relationships are cumulative, technical claims need specialist challenge, and follow-on talent can move between companies without conflicts. Do not use it to justify geographic insularity, consensus capture, portfolio conflicts, or investing outside competence merely because a known person endorses the deal.
• How to apply it:
  1. Define the domain narrowly enough to name its decision-makers, scientific failure modes, and regulatory path.
  2. Map independent experts, founders, operators, customers, payers, and strategics; record conflicts and missing viewpoints.
  3. Ask what the last three investments taught that changes this decision.
  4. Require one network-delivered disconfirming test and one reusable talent or capability advantage.
  5. Kill the specialization claim if it produces only warm introductions, repeated consensus, or no better-calibrated milestone plan.
• Examples:
  • Hybritech to Gen-Probe — documented behavior: Byers's oral history links the diagnostic-company experience, people network, and later Gen-Probe formation; the ledger preserves Hybritech conflict-clearance and board attribution rather than collapsing it into a solo chain (Bancroft, printed pp. 18–39).
  • 1986 life-sciences cohort — researcher-applied grounded illustration: the oral record connects the cohort to concentrated study and networks; the cited filing verifies only selected formation roles, so it cannot prove the rule drove every company (Bancroft, printed pp. 27–39; Genomic Health S-1, director biography).
  • Genomic Health / XDx / CardioDx — researcher-applied within-domain test: the three-company diagnostics set shows useful learning breadth precisely because one company failed; it cannot establish that every company used the same rationale (Genomic Health S-1; CardioDx shutdown).
  • Electronic Arts — boundary case: Byers also served on EA's board for thirteen years. It shows range, but not that life-sciences specialization was universal or that its mechanisms transfer unchanged to software (Kleiner Perkins EA history).
• Evidence: The origin narrative appears in Bancroft, printed pp. 27–39; the direct 2005 rule appears in Stanford's ecosystem clip; CIRM, printed pp. 191–193 supplies a contemporaneous cluster framework; and Ivor Royston's UCSD oral history provides an independent scientific-founder countervoice to the Byers-centered formation story.
• Limits, failure modes, and exception ledger: Networks may entrench consensus, exclusion, high-cost geography, and shared blind spots. The unitemized “more than 60 companies” claim and 0/40 complete economic chains prevent a specialization-return estimate. The sweep checked Bancroft, Stanford, CIRM, firm histories, filings, Pitch Johnson and Royston oral histories, and critical policy/trial routes; it found no network map, rejected-deal sample, or non-Bay-Area matched comparison.
• Evolution and contradictions: The oral record moves from broad early technology exposure to deliberate life-sciences study, while the EA history preserves a cross-sector board case (Bancroft, printed pp. 10–39; KP Electronic Arts history). CIRM in 2005 and the collective NCATS report in 2011 widen the public unit from company networks to translational infrastructure (CIRM, pp. 191–193; NCATS report). [researcher inference · moderate] The tension resolves only if specialization means a competence system, not exclusivity.
• Modern VC translation (as of 2026-08-23): [researcher inference · moderate] Transfers: build a conflict-aware knowledge graph and test whether it improves questions, independent challenge, recruiting, and milestones. Does not transfer: elite Bay Area access, physical proximity, and accumulated Byers/KPCB reputation cannot be copied or treated as selection alpha; the current selected portfolio discloses neither sourcing nor efficacy (Kleiner Perkins biography). Changed assumption/evidence boundary: no current matched network or pass ledger is public, so application efficacy is insufficient public record. Signals: repeated expert access plus demonstrated belief updating. Likely misuse: confusing a closed network with calibrated judgment; no current case is labeled a grounded application.

Create the missing company

• Rule: When an important validated problem has no credible company able to pursue it, assemble the scientific founders, interim operating plan, governance, and staged capital to create one; do not incubate merely because a technology is available.
• Evidence status: [documented behavior · moderate]. Filings verify four founding-president/chair roles, while motives and repeatability rely substantially on Byers's later account.
• Provenance and origin: Hybritech in 1978 is the earliest supportable use. Contemporary reporting and the oral history show Byers helping write/refine the plan, serving temporarily as president, and recruiting permanent leadership, with scientific and partner credit preserved (Los Angeles Times, 1988; Bancroft, printed pp. 18–24).
• Mechanism and reasoning: Formation can capture opportunities that selection-only investors cannot fund because the essential vehicle, team, and initial experiment do not yet coexist. The method works only if the investor can define a narrow proof, recruit complementary operators, and relinquish interim operating control. [researcher inference · moderate]
• Historical and vehicle context: The oral history describes small early funds and Hybritech/Home Healthcare consuming a large portion of capacity, while contemporary reporting records a reported $300,000 Hybritech startup check (Bancroft, printed pp. 18–24; Los Angeles Times, 1988). [researcher inference · moderate] Incubation therefore depended on partner time, syndication, and permanent-executive recruitment. Analyst-created hypothetical: a hypothetical high-volume studio model would require a separate, matched test of staffing, ownership, reserves, and outcomes; no current factual comparison is asserted.
• When to use: Use when the need and enabling science are credible but the organizational vehicle is genuinely missing. Do not use when a capable company already exists, when the investor lacks operating bandwidth, when conflicts block the best team, or when the only thesis is ownership capture.
• How to apply it:
  1. Write the missing-vehicle proof: why no existing company can or will pursue the problem.
  2. Secure scientific-founder ownership and conflict clearance before writing a company mythology.
  3. Define the first six-to-twelve-month falsifier, budget, and financing contingency.
  4. Recruit a permanent operator and pre-commit the investor's transition from interim execution to governance.
  5. Kill or redirect the formation if the technical proof, team assembly, or conflict test fails.
• Examples:
  • Hybritech — documented application: Byers served as temporary founding president and later chair; Royston, Birndorf, David, Greene, Miller, Perkins, and the syndicate retain scientific and operating credit. Lilly's roughly $485 million reported transaction value is not KPCB proceeds (Los Angeles Times, 1988).
  • IDEC — researcher-applied grounded illustration: the filing identifies Byers as founding president/chair and the merger record assigns operating/product credit across a wider team; neither source exposes the missing-company selection memo (Genomic Health S-1; IDEC/Biogen release).
  • InSite Vision and Ligand — researcher-applied role illustrations: the director biography verifies formation roles, not why each company was created; outcome and economics differ and remain incomplete (Genomic Health S-1; InSite transaction release).
  • Genentech — documented non-application: Swanson and Boyer had already created the company; Byers assisted product planning and market analysis, while Perkins and Kleiner held the early board relationship. It is a learning case, not his incubation (Kleiner Perkins history).
• Evidence: The formation sequence is narrated in Bancroft, printed pp. 18–39; contemporary Hybritech reporting is in the Los Angeles Times; the exact four founding-president/chair companies are independently fixed by the Genomic Health S-1; and the Kleiner Perkins Genentech history supplies the exclusion control.
• Limits, failure modes, and exception ledger: Formation magnifies attribution risk, founder conflict, capital concentration, and key-person dependency. InSite's modest reported transaction endpoint and unresolved BioSurface/Progenx outcomes warn against selecting only later successes. Public routes checked include Bancroft, SEC biographies, early-company filings, firm histories, LA Times, and Royston/Pitch Johnson oral histories; missing are the 1978 plan, cap tables, formation memos, failed-incubation ledger, and role-by-role decision packets.
• Evolution and contradictions: Hybritech supplies the early temporary-operating/handoff case, and Byers later described a 1986 formation cohort plus Healtheon and Drugstore.com (Los Angeles Times, 1988; Bancroft, printed pp. 27–39 and 57–68). The filing evidence is strongest for exactly four founding-president/chair roles, so later “incubated” language cannot establish identical mechanics (Genomic Health S-1, director biography).
• Modern VC translation (as of 2026-08-23): [researcher inference · moderate] Transfers: use a formation charter naming need, IP, conflicts, milestone, recruiting owner, rights, reserves, and handoff. Does not transfer: Hybritech's reported check size, six-month plan, local network, or interim-investor role is not a universal studio template. Changed assumption/evidence boundary: UCSF says Byers helped establish Epiodyne in 2024 but discloses no terms, decision process, milestones, or outcome, so current application evidence is insufficient (UCSF). Signals: validated problem plus orphaned capability and an identified permanent operator. Likely misuse: manufacturing a company to maximize ownership; no analyst hypothetical is substituted for a grounded case.

Stage capital against decision-changing evidence

• Rule: Size and sequence each check so that it buys an independently assessable technical, clinical, market, regulatory, reimbursement, or manufacturing result that can justify stopping, redirecting, or financing the next stage.
• Evidence status: [investor-stated · moderate]. Byers describes technical and market risk directly, and Hybritech documents an early proof stage; private tranche contracts, reserve schedules, and votes remain unavailable.
• Provenance and origin: Hybritech in 1978 is the earliest supported use: a reported $300,000 startup check funded a narrow six-month technical test, with a result in roughly three months (Los Angeles Times, 1988). Byers later generalized technical and market risk in Bancroft, printed pp. 39–46, paraphrased under the archive's rights notice.
• Mechanism and reasoning: A discriminating milestone turns uncertainty into a reversible capital decision. [researcher inference · strong] It works only when the result changes the decision; a ceremonial or convenient milestone delays recognition of failure.
• Historical and vehicle context: The oral history describes small early funds and syndicated rounds that made capital and partner time scarce (Bancroft, printed pp. 18–24). Later platforms and diagnostics added instrument, clinical-utility, regulatory, and payer dependencies, visible in the PacBio prospectus and CardioDx S-1/A, pp. 4–9 and 109–110.
• When to use: Use when a major uncertainty can be isolated, tested within runway, and interpreted before the next irreversible commitment. Do not use staging when dependencies are inseparable, the test lacks power, delayed financing destroys value, or tranching substitutes control for adequate capitalization.
• How to apply it:
  1. Draw the dependency chain from science through clinical/customer action, regulation, payment, manufacturing, and capital.
  2. Rank uncertainties by decision impact; define the observation that would update the IC.
  3. Budget runway through the readout plus contingency; state ownership, dilution, reserves, and syndication effects.
  4. Ask what belief changes on success, what dies on failure, and whether the result is independently reproducible.
  5. Record stop, redirect, and follow-on actions before the result; re-underwrite at the milestone.
• Examples:
  • Hybritech — documented application: the seed bought a six-month technical proof, which arrived early; operating recruitment and further capital followed. The record supports staging, not a complete cap table or KPCB return (Los Angeles Times, 1988).
  • Pacific Biosciences — researcher-applied illustration: accuracy, throughput, cost, workflow, and manufacturing create separable readouts; the filing documents KPCB XI and direct Byers purchases, not the thresholds authorizing them (prospectus, principal-stockholder and related-party sections).
  • Genomic Health — researcher-applied illustration: analytical validity was insufficient without clinical validity, treatment-changing utility, adoption, and coverage. The S-1 documents the company and 12.6% KPCB X position, not a milestone contract (S-1, business and principal-stockholder sections).
  • CardioDx — documented boundary/failure, not an application: technical progress did not overcome clinical-utility and reimbursement failure. The outcome makes payer evidence a proposed pre-scaling gate; it does not prove KPCB used that gate at entry (CMS review; shutdown reporting).
• Evidence: The rule appears in Bancroft, printed pp. 39–46; early behavior is independently reconstructed by the Los Angeles Times; later boundaries appear in PacBio, Genomic Health, and CardioDx filings. This establishes citation breadth, not a universal KPCB tranche policy.
• Limits, failure modes, and exception ledger: Milestones can be gamed, correlated, underpowered, late, or economically irrelevant; bridges can preserve optionality while hiding sunk cost. Bancroft, Stanford 2005 fragments, Hybritech reporting, issuer filings, and outcomes were checked; no standardized check sizes, reserve ratios, milestone covenants, board votes, or portfolio stop-rate ledger surfaced.
• Evolution and contradictions: Hybritech documents a narrow technical test; the later oral record generalizes technical and market risk (Los Angeles Times, 1988; Bancroft, printed pp. 39–46). Fogarty's IDEC account raises post-failure staging, while CardioDx shows technical progress without sufficient utility/coverage (Fogarty, 25:24–28:52; CMS review). No public source demonstrates a formal portfolio-wide process change.
• Modern VC translation (as of 2026-08-23): [researcher inference · strong] Transfers: make each check buy the next causal proof. Does not transfer: Hybritech's reported check size, six-month horizon, and assay-centered gate cannot be copied into capital-intensive platforms or AI-enabled diagnostics. Changed conditions: add data provenance, replication, change control, security, compute/manufacturing scale, prospective utility, and payment; FDA's non-binding 2025 draft frames AI-device risk across the lifecycle (FDA guidance). Signals: a near-term falsifier and pre-agreed action. Likely misuse: vanity readouts. Analyst-created hypothetical: require prospective workflow evidence and a named buyer action before scaling an AI diagnostic; this is not a Byers case.

Build several independent shots on goal

• Rule: Within a biotechnology company, fund multiple programs only when their mechanisms and failure paths are genuinely independent; let evidence re-rank or stop programs without treating company survival as proof every shot deserves continuation.
• Evidence status: [investor-stated · provisional]. Byers states the concept and later connects it to IDEC, but IDEC is the only clearly documented company application found. The public record misses the normal three-case bar, so this model has a retrieval ledger rather than padding.
• Provenance and origin: The earliest located articulation is the retrospective three-program biotechnology model in Bancroft, printed pp. 39–46, especially p. 43. It is paraphrased because Bancroft reserves quotation permission. No contemporaneous IC memorandum was recovered.
• Mechanism and reasoning: Distinct programs can reduce single-target dependence and preserve a company option after one fails. [researcher inference · moderate] The benefit disappears when programs share assay bias, biology, manufacturing, payer, regulator, or capital; nominal count is not diversification.
• Historical and vehicle context: Early biotech faced high biological uncertainty, episodic financing, and strategic-pharma dependence. IDEC's anti-CD20 path and Genentech relationship show how another mechanism plus external capital could preserve an option (IDEC/Biogen release, founders and Rituxan discussion). Fund reserves and partner time still constrained program count.
• When to use: Use when a team and vehicle can finance mechanistically distinct programs to informative readouts and one failure need not invalidate the rest. Do not use it for correlated variants, pipeline sprawl, shared fatal dependencies, or insufficient talent and capital.
• How to apply it:
  1. Map each program's target, modality, assay, manufacturing, clinic, regulator, payer, partner, and capital dependencies.
  2. Mark shared nodes; count independence only when one plausible failure does not decide the others.
  3. Give each program its own evidence threshold, budget, stop rule, and readout.
  4. Ask what survives failure, which shared dependency makes this secretly one bet, and whether reserves reach the next readout.
  5. Re-rank after results; do not retain failure to preserve an appearance of diversification.
• Examples:
  • IDEC — sole clearly documented application: Byers's late account describes an initial clinical disappointment and board reinvestment; the company advanced a distinct anti-CD20 program with Genentech. Missing board materials prevent a clean reconstruction of independence and thresholds (Fogarty, 25:24–28:52; IDEC/Biogen release).
  • Hybritech — explicit non-example: the located account describes one narrow technical proof, not several internal programs (Los Angeles Times, 1988).
  • Genomic Health, XDx, and CardioDx — portfolio boundary, not company applications: separate companies with divergent outcomes cannot be relabeled as internal shots. Their filings do not connect this rule to entry or follow-on (Genomic Health S-1; CardioDx S-1).
• Evidence: Bancroft, printed pp. 39–46 establishes the rule; Fogarty, 25:24–28:52 supplies the selected IDEC application; the IDEC/Biogen release controls the program/outcome; Genomic Health and CardioDx define non-application boundaries. Citation depth clears four placements and two independent sources; application breadth remains one.
• Limits, failure modes, and exception ledger: Correlated programs can increase burn without reducing risk; strategic partners may support only what they control. Retrieval exception ledger: Bancroft technical/market-risk pages; Fogarty video, captions, and recap; IDEC/Biogen exhibit; Genomic Health/XDx/CardioDx/PacBio filings; current portfolio; processed writings; and name/date/“three program” searches recorded in sources.md were inspected. They yielded only IDEC. Missing are board minutes, pipeline reviews, IC memos, budgets, reserves, rejected designs, and failed multi-program comparators.
• Evolution and contradictions: The rule appears in the oral history as a biotechnology risk heuristic and in Fogarty as an outcome-aware IDEC story (Bancroft, printed pp. 39–46; Fogarty, 25:24–28:52). CardioDx's single-product vulnerability is a boundary, not evidence of application (CardioDx S-1). [researcher inference · moderate] More shots preserve options only by spending focus, reserves, and partner capacity.
• Modern VC translation (as of 2026-08-23): [researcher inference · provisional] Transfers: replace program count with a dependency-correlation matrix and reserve-weighted tree. Does not transfer: “three” is not a universal optimum, and IDEC's strategic-financing/outcome path is not a generic base rate. Changed assumption/evidence boundary: no current Byers-linked program-level application or comparable public denominator was found, so efficacy under current capital, modality, and development conditions is insufficient public record. Signals: distinct biology and separately financeable readouts. Likely misuse: calling minor variants or one shared model diversification. No hypothetical is offered as application evidence.

Pair scientists with complementary operators

• Rule: Give scientific truth, entrepreneurial direction, operating execution, commercial or clinical translation, and governance named owners with nonredundant strengths; do not let reputation or advisers substitute for accountable operators.
• Evidence status: [investor-stated · moderate]. Stanford states complementarity, Fogarty repeats a team-first doctrine, and records show differentiated roles; no matched sample isolates team construction from science, capital, market, or era.
• Provenance and origin: Hybritech in 1978 is the earliest supported practice: scientific founders and an interim investor-president were followed by permanent operating leadership (Los Angeles Times, 1988). The direct public formulation appears in Stanford's “Ideal Team,” official transcript p. 1.
• Mechanism and reasoning: Specialized science and commercialization require different judgment; explicit ownership reduces single-person bottlenecks and lets governance test handoffs. [researcher inference · strong] It works only if roles carry time, authority, conflicts, and replacement triggers.
• Historical and vehicle context: The oral history describes investor formation labor and executive recruiting, while contemporary reporting documents scientist-commercialization tension (Bancroft, printed pp. 18–39; Los Angeles Times, 1991). [researcher inference · moderate] Given long boards and finite partner capacity, permanent operators were necessary to avoid investor bottlenecks.
• When to use: Use when invention, regulated development, manufacturing, or go-to-market require different expertise. Do not use complementarity to rationalize unclear founder authority, adviser-heavy charts, duplicate executives, or investor operation without a handoff.
• How to apply it:
  1. List the next twelve-to-eighteen-month decisions and required expertise.
  2. Assign one owner for science, product, operations, adoption, financing, and governance.
  3. Reference off-list; test dissent, conflict, time, and cross-functional translation.
  4. Ask which critical decision lacks an owner, which roles overlap, and who replaces the interim operator.
  5. Put hiring milestones, rights, and replacement triggers into financing and board plans.
• Examples:
  • Hybritech — researcher-applied grounded illustration: Royston, Birndorf, and David supplied science; Byers served temporarily as president/chair; Miller and Greene took operating roles. The sources document role differentiation but do not connect it to the later explicit complementarity rule (Los Angeles Times; Royston oral history).
  • IDEC — researcher-applied grounded illustration: the formation network did not substitute for Hanna, Rastetter, Levy, scientists, executives, Genentech, and the board. The record documents differentiated contributions, not a recovered team-selection memo (IDEC/Biogen release).
  • Electronic Arts — researcher-applied cross-sector illustration: Byers and Doerr contacted Hawkins, Sequoia and Sevin Rosen co-financed, and Hawkins's team built the company; this supports shared roles, not a life-sciences formula or sole credit (KP history).
  • Genentech — documented attribution boundary, not an application: Swanson and Boyer founded it, Perkins and Kleiner held the early board role, and Byers assisted product and market work (KP history).
• Evidence: The rule is in Stanford “Ideal Team”; team/psychological-safety emphasis is in Fogarty, 12:14 and 23:00; Hybritech roles are independently supported by the Los Angeles Times and Royston; IDEC and Genentech add attribution controls.
• Limits, failure modes, and exception ledger: Complementarity can hide founder conflict, unclear authority, investor dependency, or decorative advisers. Stanford, Fogarty, Bancroft, Hybritech reporting, Royston, company histories, and SEC biographies yielded no portfolio hiring scorecard, management-replacement ledger, or causal-return evidence.
• Evolution and contradictions: Hybritech shows interim operation and handoff; Stanford 2005 states complementarity; Fogarty 2019 adds psychological safety (Los Angeles Times, 1988; Stanford “Ideal Team”; Fogarty, 12:14 and 23:00). [researcher inference · moderate] Active investor help can conflict with operator authority, and no public founder-dispute record resolves that tension.
• Modern VC translation (as of 2026-08-23): [researcher inference · strong] Transfers: assign nonredundant owners with time, authority, and replacement triggers. Does not transfer: routine investor operation, a local executive network, or adviser prestige cannot replace accountable full-time leadership. Changed conditions: AI-enabled medical devices add data, model, security, quality, regulatory, and reimbursement ownership across a lifecycle; FDA's cited draft is non-binding but supports that documentation premise (FDA guidance). Signals: independent judgment plus constructive conflict. Likely misuse: an adviser slide. Analyst-created hypothetical: separately assign model-validation and clinical-action owners; it is not a Byers case.

Budget partner time, ownership, and board load together

• Rule: Accept a new lead investment only when a named partner has the time and competence to perform the promised company work, and when ownership, reserves, governance rights, and syndication justify that scarce commitment.
• Evidence status: [investor-stated · provisional]. Stanford directly links firm scale, roughly two new companies per investor per year, ownership, and engagement; no complete board-load, time, ownership, or reserve ledger tests KPCB's practice.
• Provenance and origin: The earliest direct formulation found is Byers's 20 April 2005 Stanford lecture, especially “Size of VC Firms,” official transcript p. 1. The older Pratt chapter is bibliographically established, but its inaccessible body contributes no substantive proposition to the capacity rule (Pratt catalog).
• Mechanism and reasoning: Partner attention is a finite production input. Excess board load turns recruiting, governance, financing, and problem-solving promises into unowned work; insufficient ownership or reserves may not compensate the vehicle for that labor and risk. [researcher inference · strong] Ownership is a fund-economics variable, not a moral entitlement.
• Historical and vehicle context: The oral and Stanford records describe small early funds, intensive board work, and two or three co-investors as a practical range (Bancroft, printed pp. 18–24; Stanford “Multiple Investors,” transcript p. 1). Later KPCB vehicles pooled securities under shared manager control, so a partner's name beside a position is not personal ownership (Veracyte proxy, security-ownership table and footnotes).
• When to use: Use before leading, taking a board seat, incubating, or promising operating help. Do not apply the 2005 ratios mechanically across stages, geographies, team structures, or fund sizes; do not demand ownership for brand access or equate a board seat with useful work.
• How to apply it:
  1. Inventory each partner's boards, formations, recruiting searches, financings, crises, travel, and committee work.
  2. Estimate routine and crisis load; name the specific partner and backup owner.
  3. Model initial and diluted ownership, reserves, fund contribution, syndicate roles, and rights under downside cases.
  4. Ask what existing obligation loses time, who owns the next bottleneck, and whether runway reaches it.
  5. Pass, syndicate differently, or reduce the promise when time, competence, ownership, and reserves do not fit.
• Examples:
  • Hybritech — researcher-applied illustration: temporary presidency, recruiting, chair work, and a large early fund commitment exemplify intense formation, but no time sheet or ownership bridge proves compliance with the later rule (Bancroft, printed pp. 18–24).
  • Genomic Health — researcher-applied illustration: a board role from 2001 and a 12.6% pre-IPO KPCB X position show durable governance plus vehicle ownership; shared control and missing time records prevent capacity or personal-economic claims (S-1, management and principal stockholders).
  • Pacific Biosciences — researcher-applied illustration: the filing documents Byers's board role, an 8.67% pre-IPO KPCB XI position, and direct purchases. It shows exposure and governance, not well-budgeted workload (prospectus).
  • CardioDx — documented boundary/failure, not an application: Byers originated the concept, served as CEO/director, and held direct securities, yet the company shut. Intensity and ownership did not overcome clinical-utility and payer risk (S-1; shutdown).
• Evidence: The direct rule and heuristics are in Stanford 2005; relationship intensity is partially supported by the Pratt record; independent records include Genomic Health, PacBio, and CardioDx. Citation breadth is sufficient; compliance remains unmeasured.
• Limits, failure modes, and exception ledger: The ratios are dated and firm-specific; platforms may distribute work or obscure accountability. Large ownership can increase attention or escalation bias. Stanford, Pratt, Bancroft, SEC holdings, portfolio cards, and board biographies yielded no partner calendars, complete annual board counts, reserve schedules, time allocation, opportunity costs, or workload-matched outcomes.
• Evolution and contradictions: Hands-on apprenticeship and Hybritech operation precede the explicit 2005 capacity/ownership rule (Bancroft, printed pp. 10–24; Stanford “Size of VC Firms”). Later filings show shared manager groups rather than a single holder (Veracyte proxy). [researcher inference · strong] Formation intensity and opportunity volume create an unresolved capacity contradiction.
• Modern VC translation (as of 2026-08-23): [researcher inference · moderate] Transfers: budget workload by stage/bottleneck and name who has authority. Does not transfer: the 2005 approximate companies-per-partner ratio, firm-size heuristic, or ownership logic is not a current universal. Changed assumption/evidence boundary: platforms may distribute tasks, but no public Byers/KPCB time-and-outcome ledger shows that headcount substitutes for partner attention; current efficacy is insufficient public record. Signals: a named partner with calendar capacity and a measurable intervention. Likely misuse: presenting platform headcount as unlimited attention. Treat ownership only as one input after stage, dilution, reserves, and fund size; no grounded current case is claimed.

Require a partnership champion and explicit shared decision

• Rule: Give each opportunity a named internal champion accountable for the thesis and work, then expose it to an explicit partnership decision that records evidence, dissent, attribution, conditions, and the owner of the next check.
• Evidence status: [investor-stated · insufficient public record]. Stanford describes a sector champion, subgroup review, broader second meeting, off-list references, and collective approval; private vote records are unavailable.
• Provenance and origin: The controlled articulation is the 20 April 2005 Stanford clip “Decision-Making at Kleiner Perkins,” official transcript p. 1. “Magic Mondays” documents a subgroup-to-all-partner routine, one component rather than a separate model.
• Mechanism and reasoning: A champion prevents responsibility from dissolving into committee process; collective review challenges domain blind spots and allocates shared capital. Recording dissent and next ownership makes learning falsifiable. [researcher inference · strong] The mechanism fails when advocacy becomes immunity or consensus hides who did the work.
• Historical and vehicle context: Stanford describes a sector champion, collective review, and syndication, while later filings document pooled vehicle authority (Stanford “Decision-Making”; Stanford “Multiple Investors”; Foundation Medicine S-1/A). [researcher inference · moderate] Specialization can improve one partner's information while shared review and legal ownership constrain lone-narrator credit.
• When to use: Use for entries, follow-ons, extensions, pivots, and policy exceptions. Do not use it to rubber-stamp a powerful partner, force unanimity, assign sole credit for collective work, or treat subgroup familiarity as independent diligence.
• How to apply it:
  1. Name the champion, thesis, source, conflicts, time budget, and evidence owner before full review.
  2. Circulate the disconfirming case, off-list references, alternatives, security/ownership, reserves, and kill criteria.
  3. Record approve/pass/conditional approval, dissent, abstentions, unresolved facts, and decision rights.
  4. Ask who changes their mind on what evidence, who owns failure, and which claim the group independently tested.
  5. At the milestone, compare the original memo with reality before the same champion argues for more capital.
• Examples:
  • Hybritech — researcher-applied illustration: Byers was a visible sponsor and interim operator, but Perkins, scientists, executives, co-investors, and the board had separate roles. The absent partner vote prevents calling this a documented use of the later process (Los Angeles Times, 1988).
  • IDEC — researcher-applied illustration: the late story identifies a board decision to reinvest after failure, while operating and strategic work was collective. Missing minutes, dissent, and terms prevent assigning the decision entirely to Byers (Fogarty, 25:24–28:52; IDEC/Biogen release).
  • Electronic Arts — documented shared-sourcing boundary, not a process application: the firm says Byers and Doerr contacted Hawkins and Sequoia/Sevin Rosen co-financed. Sourcing, advocacy, syndication, board work, and economics remain distinct (KP history).
  • Align Technology — documented exclusion control, not an application: Byers used it as a teaching example but identified it as Joe Lacob's deal; the firm assigns the lead and board seat to Lacob (KP history).
• Evidence: The process appears in Stanford “Decision-Making” and “Magic Mondays”; EA and Align constrain champion/credit claims; Foundation Medicine controls shared vehicle authority. These sources do not prove every vote followed the teaching rule.
• Limits, failure modes, and exception ledger: Champion systems create confirmation bias, bargaining, suppressed dissent, and key-person dependency; consensus can diffuse accountability. Stanford fragments/audio, SEC manager groups, firm histories, Fogarty, and cases yielded no IC minutes, vote tallies, dissent ledger, champion-by-company map, or outcome-linked postmortem.
• Evolution and contradictions: Apprenticeship precedes Stanford's 2005 subgroup/all-partner routine, while later filings make shared manager authority visible (Bancroft, printed pp. 10–14; Stanford “Magic Mondays”; Foundation Medicine S-1/A). [researcher inference · strong] Repetition is not learning without a feedback record, and no public vote/postmortem ledger proves errors changed later decisions.
• Modern VC translation (as of 2026-08-23): [researcher inference · strong] Transfers: keep a named sponsor, independent skeptic, conflict register, versioned memo, and outcome review. Does not transfer: Monday cadence, subgroup architecture, or implicit consensus is not a universal IC design. Changed assumption/evidence boundary: organization and vehicle rights must be mapped anew; no current KPCB vote/dissent ledger is public, so modern compliance is insufficient public record. Signals: evidence-driven dissent plus a clear owner. Likely misuse: charisma-led advocacy followed by collective amnesia. Distinguish recommendation, approval, board work, and risk; no grounded current case is asserted.

Re-underwrite persistence after failure

• Rule: After a failed program or adverse result, treat more capital as a new investment: continue only if a distinct thesis retains independent evidence, defined kill criteria, adequate runway, and better expected value than stopping or reallocating.
• Evidence status: [retrospective witness · provisional]. IDEC is the sole clearly documented application found. The favorable story is outcome-aware, and no follow-on ledger shows consistent use in ordinary or failed companies.
• Provenance and origin: The earliest public articulation located is Byers's 2019 Fogarty account of a board vote to reinvest after IDEC's first clinical failure at 25:24–28:52. The underlying decision was decades earlier; no contemporaneous board packet, financing memo, or caption-safe quotation surfaced.
• Mechanism and reasoning: Failure can eliminate one path while leaving another mechanism, team capability, asset, or strategic option. A fresh memo separates that option from sunk cost and status defense. [researcher inference · strong] Persistence adds value only when new evidence changes expected value; endurance is not evidence.
• Historical and vehicle context: IDEC faced biological uncertainty, constrained financing, and strategic-pharma dependence; Genentech helped fund anti-CD20 development, and IDEC holders ultimately received 50.5% of the Biogen combination, while the original KPCB security, alternatives, and return remain unknown (IDEC/Biogen release). Finite reserves and insider signaling make a follow-on different from abstract patience. [researcher inference · moderate]
• When to use: Use after a failed trial, missed milestone, financing shortfall, lost partner, regulatory setback, or broken market premise when a separable residual thesis remains testable. Do not use it when the same mechanism failed, runway cannot reach proof, governance cannot change, stakeholder harm is unacceptable, or opportunity cost dominates.
• How to apply it:
  1. Freeze the original thesis, prediction, security, ownership, reserves, and kill criteria; state what failure invalidated.
  2. Build a new thesis using only current evidence; name what is independent of the failed path.
  3. Compare stop, sale, wind-down, recap, pivot, and follow-on on equal capital and time assumptions.
  4. Require an independent kill advocate; ask whether the IC would invest today without the existing stake, what new fact changes value, and what result ends support.
  5. Version the decision and review it at the next milestone without hindsight rewriting.
• Examples:
  • IDEC — sole clearly documented application: Byers retrospectively describes board reinvestment after an initial clinical failure, followed by an anti-CD20 path and Genentech development. Original evidence, dissent, terms, and opportunity cost are unavailable (Fogarty, 25:24–28:52; IDEC/Biogen release).
  • CardioDx — boundary, not documented application: filings and evidence review show continued activity before withdrawal and shutdown, but no board record proves zero-base re-underwriting or when support stopped (S-1/A; CMS review).
  • Nanogen — adverse control, not documented application: bankruptcy materials report a $25.685 million asset sale and expected no common-holder value; KPCB follow-ons and stops are absent (issuer release).
  • Enjoy — adverse control, not documented application: Byers-managed Waycross holdings and Chapter 11 are public; entry, later capital decisions, and loss are not, so the case cannot be retrofitted (S-1, selling-securityholder footnote; docket).
• Evidence: Fogarty is direct but late; its institutional recap is a navigation aid, not independent transaction proof; the IDEC/Biogen exhibit independently controls the program/outcome; CardioDx, Nanogen, and Enjoy define unproven boundaries. Citation depth clears four placements and two sources; application breadth remains one.
• Limits, failure modes, and exception ledger: Persistence invites sunk cost, reputation defense, bridge signaling, inside-round distortion, and survivorship bias. Retrieval exception ledger: Fogarty video/captions/recap; Bancroft company, financing-cycle, and board sections; IDEC/Biogen exhibit; CardioDx S-1/S-1A, CMS review, and shutdown report; Nanogen release; Enjoy S-1/docket; current portfolio; processed writings; and failure/lawsuit/name-date searches in sources.md were inspected. They yielded one explicit application. Missing are board minutes, IC memos, terms, reserve histories, dissent, wind-down analyses, and a funded-versus-stopped denominator.
• Evolution and contradictions: Bancroft states general technical/market risk, and Fogarty supplies the 2019 IDEC persistence lesson (Bancroft, printed pp. 39–46; Fogarty, 25:24–28:52). CardioDx, Nanogen, InSite, and Enjoy disclose adverse outcomes but no corresponding Byers process correction (CardioDx shutdown; Nanogen release; InSite release; Enjoy docket). [researcher inference · strong] Tenacity and capital discipline remain unresolved without fresh decision records.
• Modern VC translation (as of 2026-08-23): [researcher inference · provisional] Transfers: require a versioned zero-base follow-on memo, independent kill advocate, security-aware ownership analysis, and opportunity-cost comparison. Does not transfer: IDEC's favorable outcome cannot justify generic persistence, its historical financing path is not a current base rate, and endurance is not an evidence threshold. Changed assumption/evidence boundary: present security, reserves, runway, alternatives, and stakeholder effects must be rebuilt; no second grounded Byers application or current follow-on ledger exists. Signals: genuinely new evidence for a separable path with runway to a falsifier. Likely misuse: labeling every bridge “conviction.” No current example is invented.

Evidence gaps and candidate models

• Magic Mondays: supported as a 2005 KPCB review routine, but not a tenth model because it is part of champion/shared decision and no company-level application or measured learning loop is public (Stanford).
• Public/private ecosystem: CIRM supplies a science–clinic–company–capital–policy framework, and NCATS supplies collective infrastructure context, but neither ties it to distinct company selections, returns, or a repeatable Byers underwriting rule. It remains context, not a portfolio model (CIRM, printed pp. 191–193; NCATS).
• Attack neglected problems / avoid the fifth copy: Stanford 2007 contains one speaker-mapped articulation at 72:09–72:55, but no memo, pass, or funded case connects it to Byers's portfolio. It is excluded pending application evidence (Stanford 2007).
• Separate category conviction from project underwriting: the Fountain Wind letter supports a non-investing category/project distinction, but no portfolio application was found; it remains an analyst analogy (CEC Letter P24).
• Customer resonance and two-or-three-investor syndication: both are direct Stanford practices, but their distinct content is already carried inside unmet-need diligence, staging, capacity, and shared-decision procedures. No separate application denominator justifies model proliferation (Stanford 2005).
• Active founder relationship and psychological safety: these signals remain inside complementary teams, capacity, and shared decision. The full Pratt chapter and portfolio conflict/management-replacement record are missing, so a separate model would split synonyms and overstate evidence (Pratt; Fogarty, 12:14–23:20).

Across the nine included models, the decisive missing records are original company and IC memoranda, partner votes and dissent, board packets, partner-time and reserve ledgers, rejected deals, multi-program reviews, follow-on/stop decisions, and complete vehicle cash flows. Until those appear, the field guide supports a documented craft—formation, evidence staging, teams, governance, and bounded persistence—not audited investment alpha.

What the evidence supports

The strongest defensible conclusion is about craft, not measured performance. Brook Byers’s surviving record supports an unusually hands-on US life-sciences formation practice: begin with an unmet clinical need, acquire enough specialist competence to frame the scientific and market risks, create a company when no suitable vehicle exists, recruit complementary scientific and operating leaders, stage capital against evidence, and make board work part of the investment product. The mechanism is clearest in the Hybritech formation record, the sustained Bancroft oral history, and the 2005 Stanford lecture. Hybritech reconstruction · Bancroft source analysis · Stanford operating lecture [researcher inference · strong]

The chronology also supports accumulated specialization rather than instantaneous insight. Byers first learned in Pitch Johnson’s personal-capital Asset Management structure in 1972, joined Kleiner & Perkins in 1977, helped form Hybritech, and later concentrated on life sciences. chronology · Pitch Johnson oral history, PDF pp. 11–20 That sequence is evidence of apprenticeship and domain compounding; it is not evidence that every later KPCB life-sciences result was personally sourced or caused by Byers. [documented behavior · strong]

The mixed diagnostics record prevents a victory-only reading. Genomic Health reached a large announced transaction, XDx/CareDx became public, while CardioDx withdrew its offering and shut after clinical-utility and reimbursement pressure. Genomic Health S-1 and transaction control · CardioDx failure postmortem The evidence supports a repeatable question—whether a clinical need can become a validated, reimbursable workflow—not a claim that need-first selection guarantees success. [researcher inference · strong]

The people evidence further supports a team-dependent model. Scientists, founders, permanent executives, partners, syndicates, strategic acquirers, payers, regulators, and boards repeatedly supply distinct inputs. Genentech, Collagen, and Align are explicit controls against converting a firm relationship or teaching story into Byers-originated economics. people/capital/decision-rights map · attribution controls [documented behavior · strong]

What could falsify the thesis

The edge thesis predicts contemporaneous evidence of a consistent sequence: problem selection before technology enthusiasm; specialist-network advantage; deliberate company formation; milestone-based risk reduction; capacity-aware ownership and board allocation; and zero-based follow-on decisions. It also predicts that ordinary and failed cases—not only Hybritech, IDEC, and Genomic Health—should show the same process. edge thesis and predictions [researcher inference · strong]

Five evidence packages could overturn or materially weaken it:

1. Original investment memoranda showing generic opportunity chasing, a technology-first rationale, or a rationale written only after a favorable outcome.
2. Partner-role and board records showing that others sourced, underwrote, recruited, governed, or chose follow-ons while later narratives shifted credit toward Byers.
3. A complete pass/failure ledger showing no repeated process in ordinary or adverse cases, or showing that follow-ons followed sunk cost rather than new evidence.
4. Complete deal/fund cash flows showing no life-sciences or ex-outlier advantage after vintage, stage, ownership, risk, partner, and team contributions are controlled.
5. Matched completed-peer or market evidence showing the same outcomes without the claimed formation/specialization mechanism.

The strongest skeptical account remains plausible: Byers had privileged access to a rare scientific cluster and an elite partnership; power-law outcomes, partner/team work, survivorship, and issuer-scale storytelling then made the formation practice look more individually causal and economically measurable than the public record permits. The record excluding any dominant return outlier cannot be computed because no case has a complete cash-flow chain. skeptical case · Pareto denominator warning [researcher inference · strong]

Attribution and denominator audit

The firm’s live biography claims involvement with more than 60 companies but does not itemize them. The scoped search recovered 40 named families; all 40 were searched and researched, seven were sufficiently reconstructed for no-hindsight labs, 33 remain below that lab-depth threshold, and three of the 40 are explicit exclusions/credit controls. Named-family coverage is 40/40, but coverage of the unitemized 60-plus claim is not computable. Economically complete cases are 0/40, so hit rate, deal MOIC, personal return, fund IRR, and economic concentration are not computable. denominator reconciliation [researcher inference · strong]

The ledger now has two complementary layers: 40 claim-local sourced summary rows and a 40-row field-completeness supplement. The supplement exposes, for every named family, chronology, stage, vehicle, contemporaneous thesis, check/security/ownership, team, board/governance, follow-on/dilution, exit/failure, reported economics, retrospective status, and uncertainty. Mechanical review counted each of the twelve field labels exactly 40 times. main ledger · field-completeness supplement [documented behavior · strong]

Ownership was re-audited at the vehicle level. Shared voting or dispositive power is governance evidence, not personal ownership. PacBio is the sharpest numeric check: the prospectus supports KPCB Holdings at 3,317,378 shares/8.67% pre-IPO and separately supports Byers purchasing 14,658 Series E shares for $102,606 and 6,724 Series F shares for $51,304; a distinct 47,410-share row/footnote difference remains unassigned. PacBio prospectus, recent sales and principal-stockholder table [contemporaneous record · strong]

The source-map parity audit extracted external URLs from the six substantive pre-review roots (_profile.md, writings.md, talks.md, investments.md, investmentphilosophy.md, and mental-models.md), normalized trailing punctuation and URL fragments, and compared them with sources.md. Every Brook Byers-specific root URL is mapped to a processed row or normalized canonical route. Twelve unmatched normalized URLs belong exclusively to the completed Pitch Johnson, Arthur Rock, and Bob Nelsen comparison and are controlled by those peer dossiers; they are excluded from the T021-specific source-map denominator and disclosed rather than silently counted as Brook Byers sources. T021-specific result: 0 missing. source map · completed-peer comparison [researcher inference · strong]

Transfer limits for modern VCs

The transferable unit is a decision system: define the unmet need and reachable user, map technical/clinical/regulatory/payment risk, name team and decision owners, stage capital to falsifiable evidence, reserve partner capacity, separate vehicle economics from personal credit, and re-underwrite follow-ons. The historical ratios and context do not transfer mechanically. A 2005 suggestion of roughly two new commitments per investor per year is a dated capacity heuristic, not a universal board-load rule; Bay Area/San Diego cluster access is not a current geographic mandate; and a favorable IDEC persistence story is not permission to bridge every struggling company. investment philosophy · follow-on model

Modern regulated and AI-enabled health companies add lifecycle documentation, security, data and model-change control, clinical utility, and reimbursement work that the historical cases did not fully expose. Current FDA guidance is product-specific and partly non-binding; NIH licensing routes are case-dependent; NVCA forms are voluntary models; ILPA principles are best-practice guidance rather than proof of a fund’s actual terms. modern controls and limits Every transfer in the synthesis is therefore labeled researcher inference and must be tested against the actual product, jurisdiction, vehicle, ownership, reserves, and team. [researcher inference · strong]

The closest completed peers sharpen those limits without creating a performance ranking. Pitch Johnson is the direct personal-capital apprenticeship comparator; Arthur Rock is the active-board/founder-network comparator; Bob Nelsen is the later institutional life-sciences formation-platform comparator. The comparison supports differences in capital form, formation scale, reserves, governance, and failure visibility, but none of the four dossiers supplies matched personal-alpha measurement. peer comparison [researcher inference · strong]

Unresolved evidence

The highest-value unresolved records are a controlled full Pratt chapter; original Hybritech, IDEC, Genomic Health, and CardioDx decision/board packets; KPCB partnership agreements, carry and capital-account records; company-by-company initial/follow-on/disposition cash flows; the complete 60-plus company list; an attributable pass and failure ledger; and matched sector/vintage comparisons. Their absence blocks personal-performance, return, concentration, reserve, and causal-alpha claims. retrieval agenda

Direct-voice limits are specific. Charlie Rose supplied a site-provided English VTT on 2026-08-23; two host-cued Byers turns are mapped, while uncued dialogue remains unassigned. caption retrieval note The 2011 and 2019 PMWC Vimeo pages preserve event custody, but yt-dlp --simulate --list-subs failed before format enumeration with Vimeo OAuth HTTP 401 on both routes. No legitimate audio was acquired for the configured OpenAI transcription path, so no PMWC proposition is personalized. 2011 attempt ledger · 2019 attempt ledger Georgia Tech media returned HTTP 403; other metadata-only panels name their exact route and barrier in the spoken-record ledger.

The package also preserves negative evidence: the Pratt chapter’s existence and pages are bibliographically established, but its inaccessible body supplies no relationship, governance, or capacity proposition; GRAIL’s filing supports former board service and compensation/options, not a Byers-origin investment or KPCB vehicle; and Forbes’s 1999 profile is a contemporaneous promotional artifact with chronology errors, not a chronology control. Pratt boundary · GRAIL ledger row · Forbes source-map limit

Twenty-point acceptance-evidence matrix

Central and high-risk claim audit

Method: inventory every claim class named by §2.6—edge/falsifiers, causal biography, quoted or numeric economics, Pareto, signature rationales, team/vehicle, failures, peers, and modern premises—then inspect the adjacent source and locator. Results below record the disposition; “corrected” means the root document was changed before this acceptance record.

Ordinary-claim audit

Reproducible sampling rule: for each of the eight core documents in reading order, select the first and last externally checkable non-central factual/access claim after the document’s first substantive H2, excluding headings, definitions, and claims already in the central audit. If the last candidate repeats a central claim, move upward to the nearest noncentral claim. This yields 16 claims—two per core document.

Mechanical and reachability checks

• H1/H2 structure: each of the eight core documents has exactly one H1 and a substantive H2 as the next nonblank line; investments.md opens with the required Pareto H2.
• Evidence vocabulary: root-document two-axis labels were parsed against the five allowed attribution labels and four support grades; no custom label remains.
• Local links: relative Markdown targets and heading anchors, including the three completed peer dossiers, were checked after heading changes. Repairs included the failure-postmortem, vehicle-map, people/capital-map, and portfolio-forensics anchors.
• Hallucination tells: scans for placeholders/status stubs, unsupported Pratt fragment content, uncited literal return percentages, suspicious round-return claims, and raw quotation marks found no accepted unsupported claim. Issuer values are tagged reported and separated from investor proceeds.
• Source reachability: priority public routes were reopened where feasible. All ★★★/★★ source rows are processed; access-only rows state the attempted route and barrier and make no inaccessible-body claim.
• Project-native validation: three sequential attempts were recorded. The first site/npm test exited 127 at astro: not found because the isolated worktree had no node_modules; a PATH-only retry exited 1 because ESM could not resolve astro/config. A temporary site/node_modules symlink to the main workspace’s already-installed locked dependency tree was then created; the exact npm test exited 0, Astro built 37 pages (including /investors/021-brook-byers/), and Node TAP reported 15 tests, 15 passed, 0 failed, 0 skipped, duration 1092.968 ms. The temporary symlink was removed after the run; no dependency installation occurred and no disk-heavy task overlapped the test. After the source-wording and run-ID integration edits, the same exact symlink-scoped npm test was rerun sequentially: exit 0, 37 pages built, 15 tests passed, 0 failed or skipped, TAP duration 1772.876905 ms; the symlink was again removed.
• Final mechanical totals: 219 relative links/anchors checked with 0 failures; 382 two-axis evidence labels checked with 0 vocabulary failures; 0 unchecked ★★★/★★ rows; the ten required lab content field labels each appear seven times, plus seven exercise-status lines; and all twelve ledger field labels appear exactly 40 times.
• Integration reconciliation: an initial fetch at 5c0df6285cdf40b8cc32f51b2e9a261dec9bc59c exposed live maximum 0197; a first mandatory refresh at eee6fbeff9cd8f9223ec3d9e1b730de3fe5c9680 found T048 occupying 0198–0203; and a second pre-commit refresh at f417b746dc97547dbe1a15e6e7373146119bd811 found T024 occupying 0204–0209. Only T021 run bindings were remapped, finally to the next live-free contiguous range 0210–0215, in TODO.md and logs/run-log.md. Its twelve claim/done events are one chronological block; a task-binding duplicate audit confirms each final ID belongs exclusively to the matching T021.1–T021.6 task and no old T021 binding remains. Twenty-four unrelated pre-existing collision IDs remain unchanged. The independent adversary passed all content and mechanical gates before REVIEW close; the index and theme inbox were then finalized.

Direct writings, letters, and public testimony

• ★★★ | Relationship Between Venture Capitalist and Entrepreneur | https://books.google.com/books/about/Pratt_s_Guide_to_Venture_Capital_Sources.html?id=-9kRAQAAMAAJ | by 1983; explicitly located 1986; reprinted through 2006 | attributed essay; metadata/partial | Core investment writing; 2006 edition pp. 105–108, but full text and first appearance remain unresolved.
• ★★★ | Introduction to Tom Perkins oral history | https://digicoll.lib.berkeley.edu/record/218111/files/klienerperkins00perkrich.pdf | 2001-12-30 | signed introduction in oral-history volume | Direct written assessment of Perkins and the partnership; Bancroft rights notice requires paraphrase.
• ★★ | Foreword to The Idea Is the Easy Part | https://www.simonandschuster.com/books/The-Idea-Is-the-Easy-Part/Brian-Dovey/9781637744048 | 2023 | credited foreword; body inaccessible | Genuine Byers credit; content cannot be reconstructed from endorsement copy.
• ★★ | Fountain Wind scoping comment, Letter P24 | https://efiling.energy.ca.gov/GetDocument.aspx?DocumentContentId=82694&tn=248313 | 2019-02-10 | signed public comment | Full non-investing position on renewable energy and local externalities.
• ★★ | NCATS working-group summary of findings | https://acd.od.nih.gov/reports/ACD-NCATSWGR_09212011.pdf | 2011-09-21 | collective NIH advisory report; verified canonical route | Locator: cover/member list, p. 1, findings, and Appendix A; Byers is one of 15 members, so group findings cannot be personalized.
• ★★★ | CIRM/ICOC public-meeting testimony | https://www.cirm.ca.gov/wp-content/uploads/archive/files/agenda/transcripts/05-06-05.pdf | 2005-05-06 | official government transcript | Contemporaneous public-policy argument for locating the stem-cell institute in San Francisco; PDF p. 192 onward.
• ★★ | Supporting US public health experts | https://communities.springernature.com/posts/supporting-us-public-health-experts | 2020 | co-signed letter | Full collective text; signature does not establish individual authorship.
• ★★ | Georgia Tech commencement address | https://repository.gatech.edu/entities/archivalmaterial/54b0e188-908b-48f3-9402-f574c492fa3c | 2010-05 | official archival metadata/access-boundary processing | Official record and linked media/body route were attempted on 2026-08-23; metadata resolved, while the linked body/video returned HTTP 403. No speech-content claim is made.

Oral histories, lectures, and interviews

• ★★★ | Brook Byers: Biotechnology Venture Capitalist, 1970–2006 | https://digicoll.lib.berkeley.edu/record/218191/files/byers_brook.pdf | interviews 2002–2005; published 2006 | 95-page oral-history transcript | Core direct-voice chronology and investment doctrine; quotation requires Bancroft permission, so paraphrase only.
• ★★★ | Bancroft catalog record 218191 | https://digicoll.lib.berkeley.edu/record/218191 | 2006 | archive metadata | Controls custody, interview dates, and file route; catalog/transcript interviewer metadata differ.
• ★★★ | Decoding How Venture Capitalists Really Work Today | https://stvp.stanford.edu/videos/size-of-vc-firms | 2005-04-20 | Stanford lecture/audio family | Full 46-minute audio shell with fragmented official clip transcripts.
• ★★★ | Size of VC Firms transcript | https://stvp.stanford.edu/wp-content/uploads/sites/3/2024/09/size-of-vc-firms-transcript.pdf | 2005-04-20 | official clip transcript PDF | One transcript fragment; clip pages are not separate appearances.
• ★★★ | Medical Technology Panel | https://charlierose.com/videos/536 | 2004-11-18 | archived video plus site-provided English VTT | Caption enumeration/download succeeded on 2026-08-23. Only the two host-cued Byers turns at 10:34.934–11:36.434 and 18:04.400–20:24.534 are personalized; uncued dialogue remains unassigned and caption wording is not a certified transcript. retrieval note
• ★★★ | 21st Annual Fogarty Lecture | https://www.kleinerperkins.com/perspectives/21st-annual-fogarty/ | 2019-10-11 | official video/automatic captions | Late-career philosophy and IDEC/Oculeve/OptiMedica/CareDx/Verana stories with navigation timestamps.
• ★★★ | Teamwork, Tenacity and the Value of a Bet | https://www.fogartyinnovation.org/teamwork-tenacity-and-the-value-of-a-bet/ | 2019-11-18 | institutional recap | Detailed independent event recap and retrieval aid for the Fogarty recording.
• ★★★ | Personalized Medicine: The Next Decade? | https://vimeo.com/19846731 | 2011-01 | conference video | Direct forecast record with George Church, Lee Hood, Kathy Hudson, and Ralph Snyderman.
• ★★★ | Evolving Perspectives From the Beginning of Personified Medicine | https://vimeo.com/313795186 | 2019 | conference video | Later forecast audit with J. Craig Venter and Ralph Snyderman.
• ★★★ | The Role of Entrepreneurship in Solving World Problems | https://stvp.stanford.edu/av/role-entrepreneurship-solving-world-problems | 2007-02-28 | 81-minute Stanford panel audio | Research diarization and speaker mapping completed; ASR timestamps remain approximate.
• ★★ | VC Industry: An Ecosystem | https://stvp.stanford.edu/av/vc-industry-ecosystem | 2005-04-20 | Stanford clip/transcript | Direct rule on networks and specialization; one clip within the 2005 lecture denominator.
• ★★ | Decision-Making at Kleiner Perkins | https://stvp.stanford.edu/av/decision-making-kleiner-perkins | 2005-04-20 | Stanford clip/transcript | Partnership process; one clip within the 2005 lecture denominator.
• ★★ | The Ideal Team | https://stvp.stanford.edu/av/ideal-team | 2005-04-20 | Stanford clip/transcript | Team construction; one clip within the 2005 lecture denominator.
• ★★ | Multiple Investors | https://stvp.stanford.edu/av/multiple-investors | 2005-04-20 | Stanford clip/transcript | Syndication rule; one clip within the 2005 lecture denominator.
• ★★ | Developing Close Working Relationships | https://stvp.stanford.edu/av/developing-close-working-relationships | 2005-04-20 | Stanford clip/transcript | Board/founder relationship; one clip within the 2005 lecture denominator.
• ★★ | Magic Mondays | https://stvp.stanford.edu/videos/magic-mondays | 2005-04-20 | Stanford clip/transcript | Internal review and pattern-recognition routine; one appearance, not a separate denominator item.
• ★★ | Importance of Stem Cell | https://stvp.stanford.edu/videos/importance-of-stem-cell | 2005-04-20 | Stanford clip/transcript | Policy thesis; one appearance, not a separate denominator item.
• ★★ | Industry-University Collaboration | https://www.ucsf.edu/news/2010/07/101191/industry-university-collaboration-key-translating-discovery-care | 2010-06 | official recap/access-boundary processing | Recap and title/person recording/transcript routes inspected on 2026-08-23; no full recording or speaker-level transcript located. Panel context is not personalized to Byers.
• ★★ | Future of Personalized Medicine | https://siepr.stanford.edu/events/summit/siepr-economic-summit-2010 | 2010-03-12 | official program/access-boundary processing | Program and title/person media routes inspected on 2026-08-23; no stable recording/transcript located. Participant metadata is not evidence of investment, endorsement, or proposition.
• ★★ | Perspectives on Personalized Medicine: A Conversation | https://www.personalizedmedicinecoalition.org/Userfiles/PMC-Corporate/file/2011-conference-program.pdf | 2011-11-09 | official program PDF/access-boundary processing | Program session and title/person media routes inspected on 2026-08-23; no recording or speaker-level transcript located. Session metadata only.
• ★★ | Invention 101 | https://today.duke.edu/2006/02/invention.html | 2006-02-17 | official recap/access-boundary processing | Published recap and full-transcript/media routes inspected on 2026-08-23; no complete talk record located. This package assigns the fragment no core doctrine.
• ★ | Ventured: General Keith Alexander | https://techcrunch.com/2016/03/01/an-interview-with-general-keith-alexander-on-cybersecurity-snowden-and-ironnet/ | 2016-03-01 | interview/partial audio | Byers is interviewer, not subject; useful only for questioning style.

Firm, company, and institutional records

• ★★★ | Brook Byers official biography | https://www.kleinerperkins.com/people/brook-byers/ | live | official biography/portfolio | Current roles and the 60-plus-company, 1984-practice, four-company/$8bn promotional claims; Crescendo metadata fails a filing cross-check.
• ★★★ | Kleiner Perkins partnerships archive | https://www.kleinerperkins.com/partnerships/ | live | interactive firm portfolio | Candidate firm-wide universe; partner tags do not prove personal sourcing or ownership.
• ★★★ | Genentech: Pioneering Medical Breakthroughs | https://www.kleinerperkins.com/perspectives/genentech/ | retrospective | firm case history | Controls Swanson/Boyer/Perkins/Kleiner attribution and narrows Byers's early assistance.
• ★★★ | Electronic Arts: Bringing Storytelling to Video Games | https://www.kleinerperkins.com/perspectives/electronic-arts/ | 2020-02-02 | firm case history | Says Byers and John Doerr called Trip Hawkins; partner metadata conflicts with the narrative.
• ★★★ | Frank J. Caufield memorial | https://www.kleinerperkins.com/perspectives/frank-caufield/ | 2019-11-27 | firm history | Cross-checks Byers/Caufield joining Kleiner and Perkins in 1977.
• ★★★ | Stanford STVP biography and media index | https://stvp.stanford.edu/people/brook-byers | archival snapshot; inspected 2026-08-23 | institutional biography/media index | Locator: biography lines and Related Media cards. Historical claims are promotional/dated and cross-checked; authoritative routes identify the 2005/2007 appearance families.
• ★★★ | Stanford Biodesign history | https://biox.stanford.edu/highlight/stanford-biodesign-marks-15-years-new-name-focus | 2016; inspected 2026-08-23 | institutional history | Direct fetch timed out, but Stanford-indexed body was recovered. Locator: renaming/Byers mentoring paragraph and program-history sections; supports mentoring/support only, not investment performance or sole authorship of the process.
• ★★★ | UCSF Medal announcement | https://www.ucsf.edu/news/2007/04/102596/chancellor-award-ucsf-medals-four-leaderson-may-3 | 2007-04-27; inspected 2026-08-23 | institutional award record | Locator: archived notice paragraphs naming the medal, UCSF Foundation/campaign roles, and dated portfolio biography. Supports institutional roles; promotional venture metrics require independent controls.
• ★★ | Cracking Codes in Cellular Communication | https://www.ucsf.edu/news/2024/08/428211/cracking-codes-cellular-communication | 2024-08 | institutional company history | Witness account of Byers helping establish Epiodyne.
• ★★ | Silicon Valley Archives | https://library.stanford.edu/libraries/silicon-valley-archives | live; inspected 2026-08-23 | archive gateway/access-boundary processing | Gateway, medical-innovation description, catalog links, and exact-name page search inspected; no Brook Byers text match or dedicated personal-papers collection identified. This is a future archival route, not content evidence.

Filings, ownership, and outcome controls

• ★★★ | Genomic Health Form S-1 | https://www.sec.gov/Archives/edgar/data/1131324/000089161805000492/f09100orsv1.htm | 2005-07-15 | SEC issuer filing | Director biography names the exact four founding-president/chair companies; KPCB X ownership and team boundaries.
• ★★★ | Veracyte Form S-1 | https://www.sec.gov/Archives/edgar/data/1384101/000104746913009208/a2216719zs-1.htm | 2013-09-20 | SEC issuer filing | Board chronology, vehicle ownership, and shared KPCB control.
• ★★★ | Veracyte definitive proxy | https://www.sec.gov/Archives/edgar/data/1384101/000104746914003731/a2219362zdef14a.htm | 2014-04 | SEC issuer filing | Separates 61,435 direct shares from KPCB XII vehicle holdings and control.
• ★★★ | Pacific Biosciences prospectus | https://www.sec.gov/Archives/edgar/data/1299130/000119312510237221/d424b1.htm | 2010 | SEC issuer filing | Board role, KPCB XI ownership, Byers direct purchases, and row/footnote mismatch.
• ★★★ | Foundation Medicine S-1/A | https://www.sec.gov/Archives/edgar/data/1488613/000119312513373140/d553644ds1a.htm | 2013-09-20 | SEC issuer filing | Director role, KPCB XIV entities, and nominee/beneficial-ownership limits.
• ★★★ | Foundation Medicine/Roche proxy | https://www.sec.gov/Archives/edgar/data/1488613/000162612915000119/brp65951-def14a_061815.htm | 2015 | SEC issuer filing | Transaction and Byers resignation; company value is not KPCB or personal proceeds.
• ★★★ | Five Prime Therapeutics S-1 | https://www.sec.gov/Archives/edgar/data/1175505/000119312513304936/d536956ds1.htm | 2013 | SEC issuer filing | Correct filing: KPCB IX/X vehicle holdings and shared-manager limits; the prior 2007 route was XDx.
• ★★★ | XDx Form S-1 | https://www.sec.gov/Archives/edgar/data/1217234/000104746907007848/a2178902zs-1.htm | 2007 | SEC issuer filing | Chair tenure, KPCB X nominee ownership, and beneficial-ownership disclaimer.
• ★★★ | IDEC/Biogen merger release | https://www.sec.gov/Archives/edgar/data/875045/000104746903021993/a2113458zex-99_1.htm | 2003-06-23 | SEC company exhibit | Founder/operating credit and Rituxan history.
• ★★★ | Novavax financing announcement | https://www.sec.gov/Archives/edgar/data/1000694/000129993306001380/exhibit1.htm | 2006 | SEC company exhibit | Contemporaneous Byers vaccine thesis and Prospect co-investment.
• ★★ | GRAIL draft S-1 | https://www.sec.gov/Archives/edgar/data/1699031/000162827920000160/filename1.htm | confidentially submitted 2020-07-31; inspected 2026-08-23 | SEC issuer filing | HTTP 200 official body. Locators: Management/Director Compensation pp. 136–143 and footnote 5; Byers resigned 2019 and had disclosed board compensation/options. Principal Stockholders begins p. 162. Filing does not establish a Byers-origin investment thesis or KPCB vehicle.
• ★★ | Enjoy Technology S-1 | https://www.sec.gov/Archives/edgar/data/1830180/000119312521311525/d187911ds1.htm | 2021-10-28 | SEC issuer filing | Separates Byers-managed Waycross Ventures from KPCB XVI's control group.
• ★★ | Enjoy Technology restructuring docket | https://cases.stretto.com/enjoytechnology/ | 2022 onward | official claims-agent docket | Bankruptcy and outcome control; exact Waycross loss remains unknown.

Contemporary reporting, witness accounts, and skeptical evidence

• ★★★ | The Faces Behind Biotech | https://www.latimes.com/archives/la-xpm-1991-05-26-me-3678-story.html | 1991-05-26 | contemporary newspaper feature | San Diego cluster, development cycles, and scientist-commercialization tension.
• ★★★ | Brook Byers in The Innovators | https://www.forbes.com/asap/1999/0531/090a_print.html | 1999-05-31; inspected 2026-08-23 | contemporary magazine profile | Locator: Byers entry paragraphs. It incorrectly places him at KPCB in 1972 and dates IDEC formation to 1998; used only as promotional/narrative evidence, never as chronology or economics control.
• ★★★ | Pitch Johnson oral history | https://archive.computerhistory.org/resources/access/text/2020/05/102795131-05-01-acc.pdf | 2019 | CHM transcript | Independent witness for Asset Management, Tandem, and Genentech boundaries.
• ★★★ | Ivor Royston oral history | https://library.ucsd.edu/assets/sdta/transcripts/royston-ivor_20081014.pdf | 2008 | institutional oral history | Scientific-founder countervoice for Hybritech and San Diego biotech.
• ★★★ | Align Technology firm history | https://www.kleinerperkins.com/perspectives/align-technology/ | retrospective | firm case history | Assigns belief, investment leadership, and board service to Joe Lacob; excludes Align from Byers credit.
• ★★★ | Crescendo / Myriad transaction control | https://www.sec.gov/Archives/edgar/data/899923/000119312515287589/d907829d10k.htm | 2014–2015 | SEC acquirer filing | Records $259 million final consideration and disproves the current profile's 2018 date.
• ★★ | Herbert Boyer on Genentech's founding | https://dnalc.cshl.edu/view/15029-The-meeting-that-started-the-first-biotech-company-Herbert-Boyer.html | archival interview; inspected 2026-08-23 | institutional transcript/video | Locator: transcript and metadata lines under “meeting with Bob Swanson.” Independent control for Boyer/Swanson founding; says nothing about a Byers-led original investment.
• ★★ | How Sand Hill Road Became Venture Capital's Main Street | https://www.wired.com/story/how-sand-hill-road-became-the-main-street-of-venture-capital/ | 2017-10-24; inspected 2026-08-23 | historical feature | Locator: Byers/Heidrich exchange on cluster proximity and syndication, then multi-speaker criticism. Byers recollection is retrospective; Wadhwa/Gouw industry criticism is not personalized to him.
• ★★ | Ellen Pao trial day two | https://www.wired.com/2015/02/ellen-pao-trial-day-two/ | 2015-02 | trial reporting | Byers received Trae Vassallo's complaint; does not establish misconduct by him.
• ★★ | PacBio derivative dismissal order | https://cases.justia.com/federal/district-courts/california/candce/4%3A2011cv06703/251826/40/0.pdf | 2013-04-12 | court-order mirror | Dismissal without prejudice; allegations are not findings.
• ★★ | Proposition 71 financing criticism | https://www.latimes.com/archives/la-xpm-2004-oct-18-me-prop71money18-story.html | 2004-10-18 | contemporary reporting | Campaign-finance and private-benefit context.
• ★★ | CIRM governance retrospective | https://media.nature.com/original/magazine-assets/d41586-020-03141-3/d41586-020-03141-3.pdf | 2020 | Nature commentary | Later governance/conflict critique; no evidence found that Byers personally received a grant.
• ★ | Proposition 211 opposition | https://www.sfgate.com/business/article/Lines-Drawn-On-Securities-Initiative-2977191.php | 1996 | contemporary reporting | Byers's board-liability activism and rhetoric.
• ★ | 2006 pandemic fund retrospective | https://www.axios.com/2020/04/21/the-legacy-of-kleiner-perkins-2006-pandemic-fund | 2020-04-21 | press retrospective | Firm-level strategy and later healthcare retreat; attribution requires caution.

Supplemental bibliographic, deal, and outcome controls

These rows close root-document/source-map URL parity for Brook Byers-specific claims. They were inspected as bibliographic, filing, outcome, payer, or transaction controls; they do not create a complete portfolio or return denominator.

• ★★ | Pratt chapter 1986 citation route | https://documents1.worldbank.org/curated/en/833631468764983560/pdf/multi-page.pdf | bibliography publication route | World Bank document | Locator: bibliography entry only; establishes citation/title/edition route, not chapter body or wording.
• ★★ | Pratt 2006 thirtieth-edition contents route | https://www.kriso.lt/db/9780914470090.html | 2006 edition metadata | bookseller catalog/table of contents | Locator: contents entry pp. 105–108; metadata only, not controlled chapter text.
• ★★★ | Hybritech formation and financing reconstruction | https://www.latimes.com/archives/la-xpm-1988-04-19-fi-1332-story.html | 1988-04-19 | contemporary newspaper | Formation roles, reported startup check/proof timing, and team-credit control; no cap table or KPCB proceeds.
• ★★ | Hybritech transaction-value contemporaneous control | https://www.latimes.com/archives/la-xpm-1987-07-14-fi-3739-story.html | 1987-07-14 | contemporary newspaper | Reported issuer-transaction calculation; not investor proceeds and value definitions vary.
• ★★★ | Gen-Probe acquisition control, Hologic Form 10-K | https://www.sec.gov/Archives/edgar/data/859737/000119312512482840/d393079d10k.htm | 2012 | SEC acquirer filing | Locator: Gen-Probe acquisition note; purchase price is issuer consideration, not KPCB return.
• ★★★ | Genomic Health / Exact transaction filing | https://www.sec.gov/Archives/edgar/data/1131324/000110465919043803/a19-16265_17425.htm | 2019-07-29 | SEC issuer exhibit | Locator: merger announcement/transaction value; closing stake and KPCB proceeds remain unknown.
• ★★★ | CardioDx Form S-1 | https://www.sec.gov/Archives/edgar/data/1304909/000104746913009619/a2216950zs-1.htm | 2013 | SEC issuer filing | Business, role, and preliminary ownership route used in model examples; later S-1/A controls updated details.
• ★★★ | CardioDx Form S-1/A | https://www.sec.gov/Archives/edgar/data/1304909/000104746914004275/a2218080zs-1a.htm | 2014 | SEC issuer filing | Locator: business pp. 4–9, management pp. 109–110, principal stockholders/footnotes; cost and loss remain unknown.
• ★★★ | CardioDx CMS evidence review | https://www.cms.gov/medicare-coverage-database/view/article.aspx?articleId=56142&ver=2 | coverage-analysis version 2 | official payer evidence review | Locator: evidence/analysis sections; controls clinical-utility and coverage concerns, not investment attribution.
• ★★★ | CardioDx shutdown reporting | https://www.sfchronicle.com/business/article/CardioDx-maker-of-heart-disease-test-shutting-13518778.php | 2019-01-08 | contemporary newspaper | Shutdown and reimbursement/evidence context; paywall/access may be partial and no investor loss is disclosed.
• ★★★ | Nanogen bankruptcy asset-sale release | https://www.sec.gov/Archives/edgar/data/1030339/000119312509145960/dex991.htm | 2009-07-06 | SEC issuer exhibit | Locator: paragraphs 2 and 4; $25.685 million Section 363 sale and management expectation of no common value.
• ★★★ | InSite Vision / Sun Pharma transaction release | https://www.sec.gov/Archives/edgar/data/802724/000119312515320586/d32215dex991.htm | 2015-09-16 | SEC issuer exhibit | Locator: transaction terms; issuer values are reported, while KPCB ownership/proceeds remain unknown.
• ★★ | Oculeve consideration, Allergan Form 10-Q | https://www.sec.gov/Archives/edgar/data/1578845/000156459015009844/agn-10q_20150930.htm | 2015 | SEC acquirer filing | Locator: acquisitions note; consideration includes contingent fair value and is not investor proceeds.
• ★★ | OptiMedica consideration, Abbott Form 10-Q | https://www.sec.gov/Archives/edgar/data/1800/000110465913082270/a13-19626_110q.htm | 2013 | SEC acquirer filing | Locator: acquisitions note; reported upfront/milestones are issuer consideration, not KPCB proceeds.
• ★★ | Crescendo acquisition announcement | https://www.sec.gov/Archives/edgar/data/899923/000119312514034939/d669060dex991.htm | 2014-02-04 | SEC acquirer exhibit | Controls 2014 announcement date against the firm page’s impossible 2018 partner date; final consideration is controlled by Myriad 10-K.

Current modern-translation controls (accessed 2026-08-23)

These official materials control only the changed-condition premises in the 2026 translations. They are not evidence of Byers’s historical process, the efficacy of a proposed modern application, or venture returns.

• ★★★ | Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations — Draft Guidance for Industry and Food and Drug Administration Staff (January 2025) | https://www.fda.gov/regulatory-information/search-fda-guidance-documents/artificial-intelligence-enabled-device-software-functions-lifecycle-management-and-marketing | issued 2025-01; accessed 2026-08-23 | FDA draft-guidance landing page | Locator: title/subtitle and status banner identify lifecycle-management scope, draft status, and non-binding recommendations; the “Issued by” description addresses marketing-submission documentation, safety/effectiveness evaluation, and risk across the total product life cycle. Controls the AI-device lifecycle, documentation, and non-binding-status premises in mental models 1, 4, and 6. Limits: not for implementation; does not determine a product-specific regulatory path, clinical utility, reimbursement, security staffing, or an investment result.
• ★★★ | Guidances with Digital Health Content | https://www.fda.gov/medical-devices/digital-health-center-excellence/guidances-digital-health-content | live index; accessed 2026-08-23 | FDA guidance index | Locator: “List of FDA Guidance Documents with Digital Health Content,” specifically the dated rows for AI-enabled device lifecycle management (2025-01-07, draft), medical-device cybersecurity (2025-06-27, final), predetermined change-control plans (2025-08-18, final), and clinical-decision-support software (2026-01-29, final). Controls the philosophy’s premise that current digital-health guidance separates lifecycle, cybersecurity, change-control, and clinical-decision-support obligations. Limits: an index does not make every guidance applicable to every product and does not establish clinical utility, payer acceptance, or investment merit.
• ★★★ | Model Legal Documents | https://nvca.org/model-legal-documents/ | page modified 2026-08-14; accessed 2026-08-23 | industry-association model-document index | Locator: opening “What’s new” paragraph identifies Outbound Investment Security Program, bulk-data-rule, and tranched-financing updates; “FREE DOWNLOADS” → “NVCA Model Legal Documents” → “Financing Documents” lists the Voting Agreement; “Model University Biotech Licensing Term Sheet and Agreement” lists the early-stage life-science term sheet and sample license. Controls the philosophy’s current university-biotech model availability, tranche mechanics, named legal-update, and existence-of-voting/financing-document premises. Limits: voluntary models are neither law nor evidence of universal adoption or governance authority; they require transaction-specific counsel and do not show that tranching, rights, or governance improve outcomes.
• ★★★ | Forms and Model Agreements | https://www.techtransfer.nih.gov/partnerships/forms-model-agreements | live inventory; accessed 2026-08-23 | NIH Technology Transfer forms/model-agreement index | Locator: in-page sections “License Application,” “Model License Agreements,” and “Start-Up Model License Agreements,” including NIH License Application, Commercial Evaluation License Agreement, NCI Start-Up 3.0 License Application, and NCI Start-Up 3.0 Model. Controls the premise that current NIH materials expose distinct application, evaluation, commercialization, and start-up-license routes. Limits: the inventory does not promise that a route is available, confer rights, or establish that a start-up is preferable to another commercialization vehicle.
• ★★★ | NIH Start-Up Exclusive License Agreements | https://www.techtransfer.nih.gov/partnerships/nih-start-exclusive-license-agreements | live page; accessed 2026-08-23 | NIH Technology Transfer program explainer | Locator: opening current-program warning and the paragraphs/bullets on applicant companies, a product-development business plan, the Start-Up Evaluation License Agreement, and the Start-Up Exclusive Commercial License Agreement. Controls only that the described NIH start-up program used organizational applicants and development plans and offered evaluation or commercialization routes. Limits: the page says current administration is decentralized and availability is decided by each institute or center case by case; many detailed features describe the former centralized program in the past tense. It does not support a universal claim that NIH’s current licensing process requires an applicant organization and development plan.
• ★★★ | ILPA Principles 3.0: Fostering Transparency, Governance and Alignment of Interests for General and Limited Partners | https://ilpa.org/wp-content/uploads/2019/06/ILPA-Principles-3.0_2019.pdf | 2019; accessed 2026-08-23 | institutional best-practice PDF | Locator: printed p. 5 “Overview”; p. 6 “Intended Scope”; p. 19 “Key Person” and “Time and Attention”; pp. 21, 23, 27, and 31 for fund governance, conflicts, approval/removal/suspension rights, and LPAC authority; p. 40 “LP Disclosures.” Controls the philosophy’s premise that ILPA guidance makes alignment, transparency, key-person provisions, time allocation, conflicts, governance, and LP/LPAC rights explicit fund dimensions. Limits: 2019 non-binding best-practice guidance, not proof of 2026 market practice, any particular fund agreement, or company-level authority; it cannot establish Byers’s fund terms or investment efficacy.

Coverage assessment

The sweep searched Brook H. Byers/Brook Byers with year, interview, speech, oral history, archive, manuscript, foreword, essay, letter, memo, annual report, shareholder, LP, portfolio, board, SEC, S-1, 13D, acquisition, failure, lawsuit, complaint, enforcement, Proposition 71, CIRM, PacBio, Hybritech, IDEC, Genentech, Genomic Health, and KPCB/Kleiner Perkins name variants. Identity noise from Thomas H. Byers, Brook Thomas Byers, Ken Byers, Brook Porter, and an unattributed Byers Capital was excluded.

Direct routes attempted include Bancroft, Stanford STVP/ETL, Georgia Tech, UCSF, CIRM, Charlie Rose, Vimeo/PMWC, Fogarty, Stanford Libraries, SEC EDGAR, NIH, California Energy Commission, Computer History Museum, UCSD, and official firm/company/publisher archives. The defensible spoken denominator is eight core appearance families plus several metadata-only events—not dozens of duplicate Stanford clips. The strict personally attributed written denominator is four underlying works: Pratt chapter, Perkins introduction, Dovey foreword, and Fountain Wind comment. NCATS is collective institutional work; Nature and other campaigns are signatures. The Pratt and Dovey bodies, Georgia Tech media, and several conference recordings remain restricted, missing, or metadata-only. Charlie Rose supplied a site-provided English VTT on 2026-08-23; only explicitly cued Byers turns are personalized.

Candidate deal denominators are: the oral history's self-enumerated set through 2005; SEC filings under Brook H. Byers and KPCB Holdings; the Genomic Health director biography; the current selected Kleiner portfolio; and the firm's unitemized “more than 60 companies” claim. The audit mapped 40 named company families: 35 candidate capital/governance relationships, two role-only associations, and three explicit attribution controls. Named-family coverage is 40/40, but coverage of the firm's unitemized 60-plus claim is not computable. No case supplies a complete cost, ownership, dilution, cash-flow, pass, or loss ledger, so the economically measurable denominator is 0/40 and neither hit rate nor portfolio-return multiple is computable.

Historical controls include contemporary Los Angeles Times and Forbes profiles, Pitch Johnson/Ivor Royston/Herbert Boyer witness records, company filings, CIRM records, and Pao/PacBio/Proposition 71 adverse routes. Three completed peers were verified in the live repository index and used in the synthesis: Pitch Johnson (direct apprenticeship/personal-capital comparator), Arthur Rock (same-era active-board/founder-network comparator), and Bob Nelsen (later life-sciences formation-platform comparator). Their T021 comparison cites both the completed peer dossiers and underlying evidence. Highest-value retrieval leads are a controlled Pratt edition, recoverable PMWC audio/captions, early-company IPO/merger filings, KPCB vintage records, and an attributable failure/pass ledger.

Source notes