Bob Nelsen

1. Snapshot and testable edge thesis

Nelsen is a cofounder/managing director of ARCH, formed from the University of Chicago/Argonne commercialization system. The proposed edge is early scientific access × platform judgment × company-formation capability × recruiting/boards × long-duration pooled capital × follow-on capacity. [researcher inference · strong] institutional origin current shared IC

Scope: strongest in US lab-derived genomics/platform biology/manufacturing from the 1990s–2020s. Predicted unseen evidence: named science before consensus, direct data access, several independent experiments, explicit governance owners, and financing matched to proof. Falsifiers: no matched ex-outlier outperformance, indistinguishable platform failures, follow-ons driven by sunk costs, or partner/founder contributions explaining the apparent personal record. Only 10/55 current person-linked names are sufficiently reconstructed for bounded cases; exact returns remain unknown. [researcher inference · strong] denominator

2. Core philosophy

Nelsen says the work is to spend time with scientists and usually create companies with others; the goal is cure, prevention, or system disruption, with money following impact. He calls ARCH “gut investors” and frames multiplicative uncertainty through real options rather than conventional NPV. [investor-stated · strong] BioVenture, 00:01:13, 00:18:03–00:19:03, 01:07:48–01:10:16

Synthesis: this is construction, not ordinary selection. Impact supplies ambition; experiments supply falsification; pooled capital buys options. “Money follows” fails when payer/adoption/commodity/security economics break, as Arivale, Sapphire, and deCODE show. [researcher inference · strong] failure labs

3. Sourcing

ARCH's original advantage was a mandate and network spanning university/lab scientists, IP, students, trustees, and business formation. Later sourcing adds repeat founders, scientific advisers, proactive themes, and institutional reputation. [contemporaneous record · strong] UChicago 2009 testimony, pp. 26–34

Repeatability requires recording the lab, scientist, introducer, internal sponsor, and other partner. Access is not selection skill. Theranos supplies a negative sourcing rule: if technical data cannot be inspected, pass. [investor-stated · moderate] partial Xconomy republication

4. Picking

Prefer enabling platforms with multiple assets, very large patient/industry consequences, direct experimental access, and exceptional people. Karuna is an exception because prior human evidence, Steve Paul, and unmet need made a single product unusually legible. [investor-stated · strong] BioVenture, 00:23:02–00:25:09

“Gut” should be treated as expert evidence integration, not a transferable score. Write causal claim, independent critic, falsifier, market/payer mechanism, and why now. Sapphire and mRNA show bold judgment can be wrong; Rubius/Codiak/Unity show “platform” is not validation. [researcher inference · strong] prediction audit

5. Deal and portfolio mechanics

ARCH forms early, seeks meaningful fund ownership, uses boards, syndicates broadly, reserves heavily, and often holds through IPO. Public examples: Illumina Fund III at 14.2% pre-IPO, Karuna 12.9% in 2020, GRAIL 9.35% as-converted, EQRx 7.4%. These are dated shared-vehicle holdings, not universal targets. [documented behavior · strong] field ledger

Price/terms discipline is not public. A reproducible memo should specify security, ownership under dilution, milestone tranches, reserve path, conflicts across vehicles, board/safety rights, and stop rules. Acquisition counts only after legal close and cash distribution; private value is a mark. [researcher inference · strong] denominator/economics boundary

6. Vehicle and incentive fit

ARCH is a sequence of LP-backed closed-end funds and overage vehicles, not family/evergreen capital. Scale rose from Fund VIII >$400M to XIII >$3B; Fund X's overage separated broad formation from fewer large follow-ons. Current shared control includes five IC members. [contemporaneous record · strong] Fund X Form 4

Enablement: long private duration, multiple assets, manufacturing/data build, ownership retention. Constraints: fundraising, fee/carry and distributions, LP confidence, allocation conflicts, liquidity, geographic/regulatory authority, key-person succession. Fund XIII Form D gives offering and fee timing, not rate/carry/term. [researcher inference · moderate] vehicle analysis

7. Team attribution and work with founders

Nelsen favors hungry/impatient leaders and uses executive chairs/boards to provide experience. That is not “young founder at any cost”; it is a paired design. [investor-stated · strong] BioVenture, 00:09:51–00:11:42

Credit is distributed: founders/scientists invented; partners sourced/sponsored; IC approved; funds owned; executives operated; boards governed; acquirers exited. Nelsen credits Burow/Berns for Metsera; filings link Burow to Receptos and Gillis to Codiak. [documented behavior · strong] people map Unknown roles stay unknown.

8. Views on the craft

Venture outcomes are power-law-like real options under multiplicative uncertainty. Failure tolerance matters, and an IPO finances the next stage. Fundraising is not success, despite Nelsen's own fundraising-share boast. [investor-stated · strong] BioVenture, 00:43:56–00:45:02 and 01:07:48–01:10:16

A great investor in this system identifies important capability early, recruits truth-seeking people, funds decision-changing evidence, preserves ownership when evidence improves, and stops assets when it does not. Access and capital scale are advantages to control, not proof. [researcher inference · strong] repeated rules

9. Signature bets and decision process

Use the frozen-record exercises rather than summaries:

• Sapphire failure exercise: pilot versus continuous commodity economics.
• Illumina: enabling platform without founder conflation.
• Juno: response, lethal toxicity, program/company distinction.
• Karuna: asset exception and provenance.
• GRAIL: clinical utility, vertical incentives, regulator route.
• deCODE: science/company/security separation.

Each asks: Would you invest then? What kills it? What ownership/terms fit? What unlocks the next check? The actual outcome follows separately.

10. Mental models and maxims

11. Evolution and contradictions

Observed process updates are limited: kill bad drugs, business-model attention after Arivale, system manufacturing after pandemic, regulatory/AI caveats. No public standardized postmortem or improved hit rate is proven. [researcher inference · moderate] Xconomy note Arivale response WMIF note Cambridge note

12. Failures and limits

Sapphire falsifies commercial extrapolation; mRNA falsifies modality confidence; Xcyte/Codiak/Rubius/Unity test platform proof; Arivale tests payer/unit economics; original deCODE tests security; EQRx tests capital/operating success; Sana/Lyell test burn/dilution; GRAIL tests exit regulation. [researcher inference · strong] adverse controls

Limits: 55-name curated denominator; no reject set; incomplete holdings/distributions; survivorship; partner/founder/team misattribution; private marks; era dependence; complaints versus findings. “Failure is acceptable” does not make a failed price rational ex ante. [researcher inference · strong] denominator and uncertainty adversarial boundary

13. Historical operating environment

The method emerged when universities/labs lacked mature transfer markets and biotechnology capital was scarcer. Illumina/deCODE preceded modern genomics infrastructure; Juno formed during CAR-T emergence and a receptive IPO cycle; Funds X–XIII operated amid massive private-platform capital, pandemic, AI excitement, China tension, IRA incentives, and complex regulator/payer paths. [researcher inference · moderate] historical context

No stable quantitative cross-era base rate was found. Use modality, stage, vintage, total capital, ownership/dilution, and exit-market matched comparisons; do not compare headline sale values. [researcher inference · strong] uncertainty/retrieval boundary

14. Modern VC translation — as of 2026-08-22

Grounded current example: Prime/Sana/Lyell are explicitly live could-win/fail programs, so reserve decisions should depend on asset evidence rather than sunk platform identity. [investor-stated · moderate] BioVenture, 01:02:22 onward

Hypothetical: an AI-biology seed requests $300M. Require proprietary experimental data, three independent asset tests, model-commoditization falsifier, milestone tranches, named critic, and reserve/dilution map. Do not infer quality from capital raised. [researcher inference · provisional] Cambridge AI evidence limits Fierce capital doctrine

15. Comparison with completed peers

Valid comparison is mechanism/outlier control, not era-adjusted performance ranking. None is a modern matched biotech-company factory. [researcher inference · moderate]

16. The strongest case against greatness

The skeptical thesis: Nelsen's reputation may be a selected narrative built from Illumina/Juno/Karuna and transaction values; ARCH's university access, partner expertise, scientist/founder execution, Alaska/LP capital, biotech exit cycles, and luck may explain it. The current 55-name page omits a full loss denominator, sometimes assigns team/other-origin cases (Receptos, Rubius), and includes private marks. “Gut” is not independently replicable, Fund/IC holdings are not personal, and scale can amplify failure. [researcher inference · strong] denominator/attribution

Evidence that decides: complete audited cash flows and matched modality/stage/vintage/ownership cohort; ex-outlier performance; contemporaneous memos across wins/losses/passes; partner-role/vote ledger; milestone efficiency of megarounds; documented postmortem changes. Until then, greatness is plausible institution-building skill, not proved personal alpha. [researcher inference · strong] denominator and retrieval boundary adversarial falsifiers

17. Analyst study guide, glossary, and go deeper

Reading path

1. Profile for formation/edge thesis.
2. Best writings and five talks for direct voice.
3. Denominator, then Sapphire, Juno, GRAIL, deCODE.
4. Models and review.

Glossary

• Company formation: create legal/team/capital structure around science before a conventional startup exists.
• Platform: shared capability producing multiple assets/experiments; not generic optionality.
• Real option: staged right, not obligation, to invest after evidence.
• Overage fund: parallel pool for fewer/larger follow-ons.
• Shared control: several IC members may vote/dispose fund shares; not personal ownership.
• IPO book at seed: early syndicate with public-scale follow-on capacity.
• Private mark: financing valuation, not cash return.

Reusable checklists

• Sourcing: lab/scientist/IP/introducer/sponsor; negative searches; conflict/identity check.
• Diligence: raw data, independent critic, mechanism, payer, manufacturing, regulator, falsifier.
• IC: vehicle/security/ownership/dilution/reserves; opportunity cost; dissent; next evidence; stop rule.
• Governance: founder gaps, executive chair, safety escalation, asset/company kills, conflicts, succession.
• Outcome: track security layer, gross versus net, closing/legal/regulatory, distributions, attribution.

Best sources

Primary: Illumina S-1, Juno S-1/13G, GRAIL S-4/FTC, 2009 testimony, 2026 video/transcript. Secondary: Forbes 2016 for deep narrative, always filing-checked.

Pareto 80/20 — what drove the record

A literal 80/20 return split is not computable: no public deal-level cost/distribution denominator exists. The three defensible Pareto rankings use different measures and must not be merged. [researcher inference · strong] bounded denominator

Economic headline ranking — transaction values, not ARCH/Nelsen returns

Illumina may be the largest historical compounding association, but without a disposition or transaction denominator it is not forced into a sale-value table. GRAIL's nominal acquisition unwound. Project Prometheus's reported $41B post-money is an unranked private financing mark, not economic contribution. [researcher inference · provisional] Illumina case GRAIL case Prometheus financing

Historical-importance ranking

Learning-value ranking

Bounded denominator and retrieval ledger

Evidence-case fraction: 10/55 = 18.2%. Economic coverage: not computable. “Searched” means reconciled to both current page lists and firm PDF, not reconstructed cap tables. The ten cases are Illumina, Juno, Karuna, GRAIL, deCODE, Denali, EQRx, Rubius, Homology, and bluebird; sufficiency concerns role/outcome boundaries, not returns. profile firm PDF

Historical market, regulatory, and base-rate context

ARCH's 1986 university affiliate preceded today's specialized biotech-company factories. Illumina formed during early genomics/public-market development; deCODE depended on a national dataset and repeated financing; Juno rode early CAR-T human signals and a 2013–14 biotech window; Fund X–XIII grew through abundant private capital, pandemic urgency, and platform/AI enthusiasm. Cross-era transaction values and survival cannot be compared without vintage, modality, capital, dilution, and exit-market controls. [researcher inference · strong] institutional chronology

Biotech base rates are harsh because biology, trial design, safety, manufacturing, regulation, reimbursement, adoption, and capital structure are multiplicative. A correct scientific direction can produce a bad security (deCODE), an approved/important product can produce distress (bluebird), and a valuable formation can have a destructive exit path (GRAIL). Nelsen's “real options” framing fits these base rates, but public evidence does not show calibration or matched-cohort superiority. [researcher inference · strong] BioVenture, 01:07:48–01:10:16

Regulatory context also changes. 2009 climate bets depended on patents, grants, commodity/project finance; modern therapeutics face FDA/IRB/trial, manufacturing, IRA exclusivity, payer, antitrust, and China/geopolitical branches. These are part of the ex-ante tree, not explanations added after outcome. [researcher inference · moderate] historical source map

Complete 55-name person-linked field ledger

Role tiers are secondary shorthand, not evidence grades: Tier A = Bob-specific board/formation evidence; Tier B = shared vehicle or ownership evidence; Tier C = partner/team evidence; Tier D = ARCH-profile self-attribution only. Every final cell begins with the canonical attribution/support label. “Unknown” is an answer, not an invitation to infer.

Vehicle economics and incentives

Every “Plausible effect” cell below is a bounded causal interpretation rather than a disclosed ARCH rule. [researcher inference · provisional]

One LP's report reports, as of 2025-12-31, Fund VII 37.77% IRR/4.89x TVPI/4.60x DPI; VIII 20.77%/3.19x/2.44x; IX 19.14%/2.16x/1.53x; and IX Overage 22.69%/2.56x/1.89x. [contemporaneous record · provisional] Provider/date/basis caveat: these are North Carolina Investment Authority table entries for one LP's exposures as of that date; the table does not establish gross versus net basis, a firm-wide audited result, or Nelsen's personal performance. NC report

People, capital, and influence map

Decision labs — no-hindsight reconstruction

Lead failure exercise — Sapphire Energy

Frozen information set: July 29, 2009

Sapphire displayed algae oil; Nelsen argued sunlight, CO₂, non-arable land, IP, venture, and deployment finance could produce domestic fuel within one-to-ten years. Known: pilot possibility, huge commodity market, policy support. Unknown: continuous yield, contamination, harvesting, capex, water/nutrients, oil-price resilience, project finance. [researcher inference · strong] hearing, pp. 26–34

Reader exercise

• Would you invest then? At what pilot evidence, check size, and vehicle?
• What would kill the deal? Set yield, contamination, energy-balance, capex, and unsubsidized-cost thresholds.
• What ownership/terms fit the risk? Would milestone tranches/project-asset protection/non-dilutive conditions fit?
• What evidence unlocks the next check? Specify independently verified run duration, scale, barrels/day, and economics.

Sourced actual decision and outcome

ARCH backed Sapphire; DOE later projected 5,000 barrels/day by 2018. A DOE-commissioned retrospective says Sapphire ceased by 2017 and the facility sold for pennies; ORNL records shift to nutrition/feed algae. [documented behavior · strong] DOE 2014 retrospective p. 9-22 ORNL Exact ARCH loss/process change unknown.

Case 1 — Illumina: platform formation without founder conflation

Frozen information set: 1998–2000

Optical arrays/sequencing promised cost/throughput advantage. Fund III invested $5.05359M and held 3,615,299 shares; Nelsen boarded June 1998. Risks: performance, yield, standards, adoption, capital, incumbents. Founders/scientists supplied origin. [contemporaneous record · strong] S-1

Reader exercise

• Would you invest then? What separates a platform from a clever instrument?
• What would kill the deal? Set yield, cost/data point, accuracy, adoption, and incumbent thresholds.
• What ownership/terms fit the risk? Match technical/manufacturing/financing risk.
• What evidence unlocks the next check? Define reproducible benchmark and customer workflow.

Sourced actual decision and outcome

Fund III invested and Nelsen boarded; Illumina became a category leader. “Seeded/pioneered” is ARCH marketing, not scientific-founder evidence. Dispositions/cash flows are missing. [contemporaneous record · strong] S-1, history/ownership/board biography

Case 2 — Juno: formation, urgency, clinical failure, and sale

Frozen information set: 2013–2014

Fred Hutch/MSK/Seattle Children's supplied CAR-T work with striking early responses. ARCH/Alaska financed; Nelsen boarded. Known risks: small cohorts, lethal toxicity, manufacturing/site variability, scaling, competition. [contemporaneous record · strong] S-1/A

Reader exercise

• Would you invest then? Weigh early responses against tiny cohorts/toxicity.
• What would kill the deal? Define death, neurotoxicity, manufacturing, and efficacy gates.
• What ownership/terms fit the risk? Specify board/safety rights and stages.
• What evidence unlocks the next check? Require multi-site efficacy, manufacturing consistency, and safety.

Sourced actual decision and outcome

Juno disclosed two deaths/hold in July 2016 and five treatment-related deaths by November; ROCKET stopped. Celgene/BMS paid $9.101B outside-holder cash inside $10.434B accounting consideration. Fund VII tender gross was $918,057,930 before cost/tax, not personal/net. [contemporaneous record · strong] July Nov accounting Form 4

Case 3 — Karuna: a product exception

Frozen information set: 2018–2019

PureTech founded Karuna; Steve Paul championed xanomeline+trospium. Prior human evidence and unmet need competed with tolerability, adherence, interaction, and replication risk. Fund IX/overage invested $35M in disclosed A/B rounds; Nelsen boarded. [contemporaneous record · strong] S-1

Reader exercise

• Would you invest then? Does prior pharmacology justify an asset exception?
• What would kill the deal? Set efficacy, side-effect, adherence, replication gates.
• What ownership/terms fit the risk? Reflect asset concentration in terms/tranches.
• What evidence unlocks the next check? Define Phase 2 effect/safety threshold.

Sourced actual decision and outcome

ARCH invested; Nelsen boarded through 2021. 2020 proxy: ARCH 3,362,872/12.9%, Nelsen 16,250 direct/options. BMS announced $330/share, $14B equity/$12.7B net and later completed the acquisition; ARCH closing holding/proceeds remain unresolved. [contemporaneous record · strong] proxy announcement completion

Case 4 — GRAIL: platform value and exit governance

Frozen information set: 2016

Illumina formed GRAIL with $100M+ from Illumina, ARCH, Bezos, Gates, Sutter Hill. Promise: falling sequencing cost/early cancer detection. Risks: false positives, prevalence, tissue-of-origin, utility, reimbursement, study scale, vertical supply conflicts. [contemporaneous record · strong] formation

Reader exercise

• Would you invest then? What analytical/utility evidence justifies population capital?
• What would kill the deal? Set specificity, stage, tissue-origin, payer, study gates.
• What ownership/terms fit the risk? Address strategic supply/control conflict.
• What evidence unlocks the next check? Require prospective screening/utility/reimbursement proof.

Sourced actual decision and outcome

ARCH reached 63,710,240 as-converted shares/9.35% in the 2020 transaction record, under shared fund control; that dated stake does not yield ARCH cash proceeds or Bob-personal economics. Illumina announced the reacquisition at $8 billion of cash-and-stock consideration in 2020. Closing purchase accounting was different: Illumina's 2021 10-K recorded a $9.745 billion total purchase price—$2.862 billion cash, $4.975 billion stock, $757 million contingent consideration, $1.149 billion fair value of its previously held investment, and $2 million settlement of pre-existing relationships. These are acquirer accounting categories, not an ARCH return. [contemporaneous record · strong] S-4, ownership table 2021 10-K, note 4

The US and EU paths must be separated. The FTC found likely vertical harm and ordered divestiture. In December 2023, the Fifth Circuit held that substantial evidence supported the FTC's market, prima-facie harm, and efficiencies conclusions, but vacated and remanded because the Commission applied the wrong legal standard to Illumina's Open Offer rebuttal. Illumina nevertheless proceeded to separate GRAIL. It completed the spin on June 24, 2024, distributing about 85.5% of GRAIL to Illumina holders and retaining about 14.5%; the FTC dismissed its administrative complaint on August 15 after the separation. [contemporaneous record · strong] Fifth Circuit, pp. 34–35 FTC final opinion FTC dismissal

The spin also required a $974 million disposal-funding contribution for 2.5 years of operations. Illumina's Q3 2024 filing recorded $1.466 billion of GRAIL goodwill impairment and $420 million of IPR&D impairment in Q2 2024, while noting $4.626 billion of accumulated GRAIL-reporting-unit goodwill impairment at 2023 year-end; its retained 14.5% interest was initially recorded at $397 million and then carried a $332 million unrealized loss through 2024-09-29. Those are Illumina accounting losses/funding, not ARCH fund losses. [contemporaneous record · strong] Illumina Q3 2024 10-Q, notes 2, 4, 6

Separately, the CJEU held that the European Commission lacked jurisdiction. On September 6, 2024 the Commission withdrew its prior investigation, prohibition, interim-measures, divestiture, and €432 million fine decisions; Illumina reversed the fine accrual. The fine therefore cannot be presented as a sustained payable cost, while the independent US antitrust path and completed spin remain adverse facts. [contemporaneous record · strong] CJEU release EC withdrawal, ¶¶10–14 Illumina Q3 2024 10-Q, legal proceedings

Case 5 — deCODE: right science, fragile security

Frozen information set: 1996 and 2009 rescue

Icelandic population/genetic data offered discovery advantage; privacy, trust, monetization, leverage, financing, and product timing were risks. By 2009: $69.9M assets/$313.9M debt; rescue assets were a different security decision. [contemporaneous record · strong] 10-K

Reader exercise

• Would you invest then? Answer separately for original equity and rescue assets.
• What would kill the deal? Set privacy, partner revenue, burn, leverage, and data gates.
• What ownership/terms fit the risk? Specify asset priority/debt/governance/liquidation terms.
• What evidence unlocks the next check? Require validated discovery cash flow/refinancing.

Sourced actual decision and outcome

Original company entered bankruptcy; Saga/Polaris/ARCH acquired assets. Amgen announced $415M subject to adjustments and recorded $401M cash at close. Never blend original equity and rescue outcomes. [contemporaneous record · strong] announcement 10-K

Failure and adverse-control labs

Xcyte — platform/regulatory/financing interaction

Original thesis and evidence. Xcyte sought to turn ex-vivo T-cell activation/expansion into a repeatable therapeutic platform for cancer and immune disease. The early promise was biological and platform-level, but public hindsight does not recover the original ARCH memo, clinical probability, vehicle cost, or Nelsen-specific sponsorship. [retrospective witness · provisional] Forbes profile company proxy

Warning signals and capital escalation/refusal. Regulatory objection to the proposed pivotal program, weak financing capacity, and the gap between a broad platform story and an approvable protocol were decision-changing warnings. A retrospective reports about nine years and roughly $150 million financed, but the public record does not allocate that sum to ARCH or show which follow-on checks ARCH approved or refused. Treat it as reported company financing, not an audited fund loss. [retrospective witness · moderate] Forbes

Governance, stop, and outcome. The company cut staff, terminated its lead program, reviewed strategic alternatives, and sold substantially all T-cell technology assets for $5 million; the remaining public shell pursued another transaction. These are documented board/company responses, but board minutes do not show who argued for each step. [contemporaneous record · strong] proxy, “Background” and asset-sale sections

Claimed lesson, later change, and uncertainty. Nelsen later used Xcyte as part of a persistence/learning narrative that conceptually preceded Juno. That is retrospective narration, not proof that the same asset survived or that a specific new rule caused Juno's outcome. No public document shows a standardized later ARCH protocol, follow-on threshold, or measured process improvement. The counterfactual is unresolved: earlier refusal might have saved capital but also eliminated learning that the team later says mattered; continued funding might still have failed because regulatory and clinical uncertainty were fundamental. [researcher inference · provisional] retrospective account company record

Exercise: Would you invest then? What would kill the deal? What ownership/terms fit the risk? What evidence unlocks the next check? [researcher inference · provisional] Actual outcome revealed only after the exercise: regulatory/program/financing failure and a $5 million asset sale; no ARCH-specific loss or documented process change is recoverable. [contemporaneous record · strong] company proxy

Codiak — platform breadth without proof

Original thesis and evidence. ARCH and MD Anderson formed Codiak around engineered exosomes as a delivery/therapeutic platform. The prospectus identifies ARCH incorporation/licensing/recruiting and Steve Gillis, not a Bob-specific originating decision; Flagship joined through VL27 and Lonza supplied in-kind support. The thesis offered multiple payload/assets, but IPO-stage evidence remained precommercial and clinical proof sparse. [contemporaneous record · strong] prospectus, “Our history”/related parties/principal stockholders

Warning signals and capital escalation/refusal. Platform breadth, manufacturing needs, early clinical uncertainty, continuing losses, and dependence on external finance were visible. Public filings reconstruct capital providers and risk factors better than internal decisions: exact ARCH follow-ons, reserve refusals, IC votes, and cost basis remain unknown. The reported $30–$40 million ARCH loss is a later Nelsen estimate, not audited, not allocated across vehicles, and not personal. [retrospective witness · provisional] reported estimate/boundary

Governance, stop, and outcome. Codiak filed Chapter 11 in March 2023, initiated a sale process, and faced Nasdaq delisting; the court later ordered dissolution and cancellation of common shares. The issuer record establishes company outcome, not which ARCH partner drove governance or when the fund stopped. [contemporaneous record · strong] bankruptcy 8-K dissolution order, docket 289

Claimed lesson, later change, and uncertainty. The supported lesson is that a multi-asset “platform” and strategic manufacturing help do not substitute for clinical proof, runway, or asset-specific stop rules. No public source shows a documented ARCH process change, improved follow-on rule, or Bob-specific postmortem. Counterfactually, smaller/staged financing might have reduced dollar loss but could also have prevented adequate platform/clinical testing; the public record cannot distinguish bad thesis, bad price, execution failure, or an unlucky biological draw. [researcher inference · provisional] prospectus bankruptcy record

Exercise: Would you invest then? What would kill the deal? What ownership/terms fit the risk? What evidence unlocks the next check? [researcher inference · provisional] Actual outcome revealed only after the exercise: Chapter 11, dissolution, and cancellation; Bob-specific sourcing remains unsupported. [contemporaneous record · strong] bankruptcy 8-K dissolution order, docket 289

Arivale — prevention without unit economics

Original thesis and evidence. Lee Hood's systems-biology vision and Clayton Lewis's company sought to combine genomics, repeated biomarker assays, and coaching to prevent disease rather than wait for treatment. ARCH led financing; the service ultimately reached roughly 5,000 customers. That establishes real usage and measurement, not medical efficacy or sustainable willingness to pay. [retrospective witness · moderate] postmortem

Warning signals and capital escalation/refusal. The company raised more than $50 million, employed about 120 people, and faced costly assays/coaching, hard customer acquisition, pricing resistance, and a prevention market that had not matured. Public sources do not provide cohort economics, burn by round, ARCH ownership, follow-on approvals/refusals, or the precise point at which payer/consumer evidence failed the underwriting. [retrospective witness · moderate] GeekWire

Governance, stop, and outcome. Unable to secure additional capital or an economic model matching customer willingness to pay, Arivale shut in April 2019. Management described closure rather than a sale; exact creditor/shareholder recoveries and ARCH loss remain unknown. The record does not expose board votes or who decided to stop. [retrospective witness · moderate] GeekWire

Claimed lesson, later change, and uncertainty. Nelsen said ARCH would continue science-driven consumer wellness but pay more attention to the business model. That is a documented verbal update, not evidence of a formal new rule or improved later results. [investor-stated · moderate] BioCentury response Counterfactually, lower-cost assays, employer/payer distribution, or later consumer timing might have changed economics; they do not prove the original price or scaling decision was rational. [researcher inference · provisional] reported operating constraints

Exercise: Would you invest then? What would kill the deal? What ownership/terms fit the risk? What evidence unlocks the next check? [researcher inference · provisional] Actual outcome revealed only after the exercise: consumer/payer/CAC/cost failure and shutdown; ARCH loss is not public. [retrospective witness · moderate] postmortem

Current-union and board adverse controls

The response/outcome cells below are issuer- or docket-supported [contemporaneous record · strong]; the attribution/lesson boundaries are [researcher inference · moderate]. Each row carries its controlling source.

Portfolio-level forensics

Observed scope is 55 current self-attributions, not complete history/cash flows; it overweights survivors/current marketing. Formation/seed dominates but public follow-ons matter. Sector evolved from software/tools/energy to therapeutics, genomics, manufacturing, AI/care. Visible stakes can be meaningful (Illumina 14.2% pre, Karuna 12.9% in 2020, GRAIL 9.35%, EQRx 7.4%) and shared. Public checks range $50K–$250M; launch syndicates reach $1B. Hit rate is not computable. [researcher inference · strong] searched denominator 55-name field ledger vehicle evidence

The reported one-LP Fund VII–IX snapshots (North Carolina Investment Authority, figures as of 2025-12-31, table basis not independently shown as gross/net or firm-wide) are consistent with power-law success, not proof of Nelsen selection. Removing Illumina/Juno/Karuna/Array/Receptos leaves live/adverse cases but no cash-flow residual-alpha test. [researcher inference · provisional] provider/date/basis control

Completed-peer comparison

Controversy and allegation controls

Juno issuer deaths are facts; merger complaint is allegation. GRAIL US/EU paths are distinct and current. Sana action is pending; cited issuer text does not identify Nelsen as defendant. No searched source located an SEC enforcement or adjudicated personal misconduct finding against Nelsen—a bounded negative, not proof none exists. [researcher inference · strong] Juno issuer and complaint controls GRAIL regulatory records Sana issuer boundary search saturation

Uncertainty and retrieval ledger

• Private: LP letters, IC memos/votes/dissent, board minutes, source/sponsor/negotiator/follow-on/exit attribution, rejects/passes.
• Missing economics: costs, dilution, dispositions, recycling, distributions, fees/carry/PME, pecuniary shares, taxes/proceeds.
• Missing media/private-company records: full TWST/Xconomy/Long Run/READOUT texts; Prometheus/Stability/Altos/Xaira/Resilience cap tables.
• Decision-changing future evidence: original ARCH documents; audited fund cash flows; matched Flagship/Third Rock/5AM cohort; adverse postmortems; private governance.
• Public evidence supports a consequential company-formation institution and selected board/vehicle participation, not an audited personal track record. [researcher inference · strong] searched denominator adversarial boundary

Coverage and evidence base

The bounded priority writing corpus is five items: 2009 prepared congressional testimony; a 2014 email reproduced by a journalist; 2020 public posts preserved within a STAT interview; and Nelsen-attributed statements in the 2020 Fund X and 2024 Fund XIII releases. Nelsen has no located public book, regular blog/newsletter, shareholder/LP letters, or durable personal archive. Interviews belong in talks.md, not silently in the authored count. [researcher inference · strong] corpus reconciliation

Authorship/access are unequal. Congress is full primary text; the email/posts survive in reputable journalist custody; firm releases are primary for stated intent and promotional for outcomes. Each item has a dedicated note. Private memos and deleted social material remain unknown.

Climate for Innovation testimony — July 29, 2009

Historical context, argument, and reasoning

Before ARCH's later biotech concentration, Nelsen argued that federal research, university/lab commercialization, patents, venture capital, and separate deployment finance formed one innovation system. Established firms would not accept the earliest technical risk, so investors had to create companies. [investor-stated · strong] House hearing, printed pp. 26–34

Evidence, examples, assumptions, and actual decisions

He said he had helped found 30 companies over 23 years and used solar and Sapphire algae fuel as live examples. The count establishes his dated self-description, not success. The reasoning assumes IP can secure investment and pilot technologies can traverse deployment economics. ARCH's Sapphire commitment makes the statement decision-relevant rather than abstract policy. [documented behavior · moderate] note

Intellectual development, implications, and contradiction

The enduring rule is institutional sourcing plus explicit financing layers. The crucial contradiction is forecast calibration: Nelsen said the algae-fuel time was “now,” while a DOE retrospective says Sapphire ceased by 2017 and the facility shifted away from fuel. [documented behavior · strong] DOE retrospective, printed p. 9-22 A demonstration bottle validated possibility, not continuous commercial economics.

Email on Janet Yellen's biotech comments — July 17, 2014

Historical context, argument, and reasoning

During a debate over stretched biotech valuations, Nelsen told Matthew Herper that sector picking should be left to domain experts and pointed to real cures. The public body is the reproduced email, not an investor-owned archive. [investor-stated · strong] Forbes, “Nelsen wrote” paragraph

Evidence, assumptions, and relationship to decisions

The email has no price, probability, dilution, expected-value, or trade evidence. Herper names Illumina, Alnylam, and Agios for context, but visible winners cannot validate a whole sector. The decision implication is modest: company science should not be replaced by a macro label. [researcher inference · moderate] reproduced email

Intellectual development and contradiction

This is an early public version of domain-first “gut.” The assumption that insiders are better judges may be true and conflicted. The later real-options account explains why precision is hard but does not remove price or opportunity cost. [researcher inference · moderate] 2014 email 2026 transcript, 01:07:48–01:10:16

Early COVID posts and March 2020 interview

Historical context, argument, and reasoning

Posts from January 20/31 warned that reported cases understated transmission and highlighted asymptomatic spread; the March interview advocated masks, testing, distancing, and therapeutic development. Nelsen labeled his estimates “back-of-the-envelope,” separating urgent action from numerical certainty. [documented behavior · strong] STAT chronology

Evidence, actual decisions, and assumptions

Infectious-disease networks and Vir proximity plausibly supplied early sensing, but the article does not document a fund trade. The low-regret action logic was sound: cheap prevention against asymmetric delay cost. It also creates a portfolio-bias risk. [researcher inference · moderate] STAT chronology

Intellectual development and contradiction

The same interview records skepticism that mRNA vaccines would work well and a preference for antibodies. mRNA became decisive. The balanced learning is strong early pandemic-risk recognition plus a material technology-modality miss—not a one-sided “called it right” story. [documented behavior · strong] STAT chronology note

Fund X statement — April 2, 2020

Historical context, argument, and reasoning

ARCH announced $1.46B across Fund X and an overage fund. Nelsen said biology was becoming an engineering science through data/compute. The structure made the claim operational: a core vehicle could form companies broadly; an overage pool could make fewer, larger follow-ons. [contemporaneous record · strong] ARCH release

Evidence, assumptions, and relationship to decisions

The release is primary for amount/structure and promotional for “transformative” outcomes. It exposes a real capital decision, not company allocation. Its assumption is that reserves preserve nonlinear ownership and financing continuity rather than subsidize weak programs. [researcher inference · moderate] Fund X release

Intellectual development and contradiction

The overage design anticipates Nelsen's 2024 desire for an “IPO book” at seed. Larger reserves reduce financing risk and increase sunk-cost/governance risk. No DPI, TVPI, fee, carry, allocation, or stop rule is published. [researcher inference · strong] Fierce note

Fund XIII statement — September 26, 2024

Historical context, argument, and reasoning

Fund XIII above $3B repeats “great science and great teams,” early co-formation, and long-duration capital, while adding AI, prevention, cures, and access. The Form D corroborates a $3B offering, not performance or Bob-only control. [contemporaneous record · strong] release Form D

Evidence, assumptions, and relationship to decisions

The capital jump changes feasible formations and loss capacity. The release assumes large platforms need institutional coalitions at inception. “Equitable” and “transformative” remain unmeasured aspirations. [researcher inference · moderate] Fund XIII release Form D

Intellectual development and contradiction

By 2024, company formation has become financing architecture: team, data, manufacturing, and many clinical shots can be capitalized before conventional proof. Scale is not success; Project Prometheus's reported $41B private mark remains unrealized. [researcher inference · strong] Axios

Cross-source evolution

Cross-source decision synthesis

1. [researcher inference · strong] [sourcing] Find a technical capability at its institutional origin. 2009 testimony
2. [researcher inference · strong] [picking] Require inspectable data and write the expert's causal claim/falsifier. Theranos pass
3. [researcher inference · strong] [risk] Act before perfect data only when delay cost is asymmetric; score every forecast later. STAT
4. [researcher inference · strong] [portfolio] Estimate full capitalization and vehicle before formation. Fund X
5. [researcher inference · strong] [risk] Define program, payer, manufacturing, regulatory, and capital-structure kill gates. failure labs
6. [researcher inference · strong] [people] Attribute science, sourcing, vote, board, operations, and capital separately. shared IC

Author-like decision playbook

The nine prompts below are an analyst reconstruction, [researcher inference · strong], grounded in the cross-source decision synthesis, shared-control filing, and failure labs, not a verbatim ARCH checklist.

• Problem: Is the impact measurable as patient/industry change, not fundraising?
• Origin: Which lab, scientist, introducer, sponsor, vehicle, and partner owns each decision?
• Evidence: What direct data is inspectable, and what cheapest experiment kills the thesis?
• Platform: What independent shots exist? Why does an asset-only exception qualify?
• People: What urgency is valuable, and what board/operator scaffolding is missing?
• Capital: What do proof, clinic, manufacturing, and adoption cost under delay/failure?
• Market: Who pays now? What CAC, price, reimbursement, and switching behavior must hold?
• Stop: What result ends an asset, platform, or company despite ample cash?
• Economics: Financing value is a mark; return requires dated cost and cash distribution.

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

What survives

Early access, technical transparency, company creation, named decision owners, staged evidence, syndicate endurance, and manufacturing/regulatory design remain useful. A modern AI-biotech memo should show a proprietary data-generation loop and an experiment that improves or falsifies the model. [researcher inference · strong] Cambridge transcript, pp. 3–5, 13–14

What fails or requires adaptation

Patent/IP confidence does not solve commodity deployment; capital abundance does not prove platform quality; insider conviction does not replace price; direct-to-consumer prevention needs willingness-to-pay; and “gut” without ARCH's networks/boards/long money is cargo cult. Sapphire, Arivale, Rubius/Unity, and mRNA provide concrete controls. [researcher inference · strong] failure controls

Changed assumptions and live signals

Fund sizes, AI compute/data access, global capital, biotech public-market liquidity, China risk, IRA incentives, accelerated/conditional approval debate, clinical decentralization, manufacturing concentration, and payer pressure differ materially from 2009. Track experiment-cycle time, capital per decisive result, syndicate reserve coverage, split-adjusted dilution, payer evidence, and regulator theory—not private valuation alone. [researcher inference · moderate] vehicle chronology policy record

Likely misuse and examples

Grounded: Karuna qualifies as a product exception because prior human pharmacology, a side-effect mitigation, Steve Paul, and unmet need jointly narrowed uncertainty; ARCH did not originate PureTech's company/molecule. [contemporaneous record · strong] Karuna S-1

Hypothetical: a seed AI-biotech claims a foundation model and asks for $300M. Translation requires proprietary experimental data, independent asset tests, milestone tranches, named technical critics, and a commoditization falsifier. Misuse is “Bob funds platforms big, therefore raise big.” [researcher inference · provisional] Cambridge AI evidence limits Fierce capital doctrine

Best writings and unresolved record

No public book, essay collection, annual/shareholder/LP letter, personal blog, or complete social archive was found. Private IC material could materially change the synthesis.

Authored-corpus reconciliation

Coverage and transcript denominator

The priority denominator is 11 substantive ★★★/★★ appearances with notes and deep sections: TWST 2003; Xconomy/BaseLaunch 2018; STAT 2020; C&EN, Long Run, and WMIF 2021; Founder Stories 2023; BioCentury, Fierce, and Cambridge 2024; BioVenture VoiCes 2026. Axios Re:Cap and READOUT LOUD remain ★ mapped-only routes: searches found no public content-grade transcript, so neither supplies a processed Nelsen argument. Mirrored HTML/PDF routes do not create independent appearances. source inventory

Access is deliberately uneven: TWST/Xconomy are partial; Long Run lacks a transcript/quote timestamps; Founder Stories is ASR-assisted; WMIF uses official recap plus mirror; BioCentury lacks a reliable full transcript; Cambridge has an edited full PDF; BioVenture has primary video and ASR averaging 0.77, with cited spans audio-checked. No source is upgraded beyond its access.

Listen to these five first

2003 — TWST: formation before biotech concentration

Accessible opening material describes university/lab company creation, >$900M under management, formation/initial funding of four companies, and early participation in roughly 20. It establishes an early mixed-technology institutional method, not a complete interview or personal ledger. [contemporaneous record · moderate] TWST note

Decision implication: underwrite licensing, recruiting, syndication, and governance needed to turn research into a company. Contradiction: access and selection skill cannot be separated, and paywall limits detail. [researcher inference · moderate] note

2018 — Xconomy/BaseLaunch: inspect, terminate, and hold through IPO

Nelsen says IPO is financing rather than exit, weak drug candidates should be killed, and Theranos secrecy blocked diligence. Appearing after Juno's public rise and before sale, it supplies relatively clean evidence of long-hold intent. [investor-stated · moderate] partial republication note

Decision implication: technical access is a kill gate; company/platform conviction can survive an asset termination. Contradiction: the original full transcript/memo is absent, and long holding without explicit sell rules can become escalation. [researcher inference · moderate] note

2020 — STAT: early action and a modality miss

Nelsen's spread/asymptomatic warnings and low-regret public-health measures were directionally strong. In the same record he doubted mRNA vaccines would work well and preferred antibodies. [documented behavior · strong] STAT chronology note

Decision implication: act on asymmetric delay while recording confidence separately; score technology predictions independently from macro sensing. No disclosed investment follows mechanically from the forecast. [researcher inference · strong] note

March 2021 — C&EN: aging as mission and mechanism

Nelsen frames aging as investigable biology and admits personal fear of death. This explains persistence and attention, not portfolio efficacy; self-experimentation is N-of-one motivation evidence. [investor-stated · strong] C&EN speaker transcript note

Decision implication: personal curiosity can sustain long validation; require stronger controls from companies than from self-experiment. Mission intensity is also a confirmation-bias risk. [researcher inference · moderate] note

March 2021 — The Long Run: manufacturing and mental-health search space

Show notes/audio place manufacturing, mental health, pandemic lessons, and Fund XI in one episode. Without transcript or quote-grade timestamps, this package uses it only to date themes. [contemporaneous record · provisional] episode note

Decision implication: themes organize sourcing, not selection. Company science, payer, execution, and economics still determine a decision. Exact Nelsen-versus-partnership allocation remains unknown. [researcher inference · provisional] note

May 2021 — World Medical Innovation Forum: manufacturing before success

At the May 21 event, Nelsen describes physical/regulatory scalability as a cell/gene-therapy constraint and Resilience as proactive infrastructure. The official recap controls event/date; a Dec. 14 mirror supplies partial timestamped wording. [investor-stated · moderate] official recap mirror note

Decision implication: finance a shared bottleneck only with utilization/customer/obsolescence falsifiers. Large fixed assets can become a burden after pandemic demand normalizes. [researcher inference · moderate] note

2023 — Founder Stories: urgency with scaffolding

Nelsen returns to university science, important clinical problems, instincts, recruiting, and early risk. Video/auto captions support paraphrase; no unchecked ASR quote is used. [retrospective witness · moderate] video note

Decision implication: identify both the young founder's advantage and the experienced board/operator plan. Contradiction: urgency can degrade clinical controls; the talk does not define when the board slows a team. [researcher inference · moderate] note

February 2024 — BioCentury: AI, China, IRA, and approval

Nelsen says AI requires biological data/experiments; prevention lacks aligned payment; IRA molecule windows affect incentives; China combines science with strategic risk; conditional approval/payer evidence may accelerate learning. [investor-stated · moderate] article/audio route note

Decision implication: attach payer, regulation, and geopolitics to the experimental thesis. Access limit: no reliable full public transcript, so exact policy quotation is excluded. [researcher inference · moderate] note

March 2024 — Cambridge: AI biology is not ordinary software

Nelsen distinguishes biology's sparse/fragmented data and absent unified model from easier AI care-delivery applications; he sees excessive regulation as a nearer risk than existential AI. [investor-stated · strong] PDF, printed pp. 3–5, 13–14 note

Decision implication: write what generates proprietary experimental data, how the loop improves, and what commoditizes it. The multi-speaker panel shows the expert network but not its disagreements. [researcher inference · strong] note

May 2024 — Fierce: the seed-stage IPO book

Discussing roughly $1.7B across Xaira, Mirador, and Metsera, Nelsen says seed should include investors capable of financing billions and admits detailed early budgets can be false precision. [investor-stated · strong] Fierce, seed-syndicate paragraphs note

Decision implication: model the full journey, vehicle, dilution, reserves, governance, and stop tranches at formation. Contradiction: upfront abundance can hide weak milestones; private rounds are not returns. [researcher inference · strong] note

January 2026 — BioVenture VoiCes: the explicit operating system

Nelsen describes forty years as spending time with scientists and usually creating companies with others (00:01:13), advocates hungry/impatient leaders with executive-chair/board support (00:09:51–00:11:42), and says impact should precede money (00:18:03–00:19:03). [investor-stated · strong] primary video

He calls partners “gut investors,” says multiplicative uncertainty defeats conventional NPV/IRR, and frames a real-options portfolio (01:07:48–01:10:16). Platform preference, Karuna exception, Burow/Berns Metsera credit, live Prime/Sana/Lyell uncertainty, and a fundraising-share boast immediately undercut by “fundraising is a bad metric” make it the richest contradiction source. [investor-stated · strong] timestamped ASR note

Decision implication: gut must be made auditable through a written causal claim, scientific critic, falsifier, milestone, ownership/vehicle, and attribution record. ASR confidence averages 0.77; cited spans were audio-checked, not the entire video. [researcher inference · strong] note

Cross-talk evolution

Repeated rules and documented contradictions

• [investor-stated · strong] [sourcing] Spend time with scientists before company formation. 2003 2026, 00:01:13
• [investor-stated · strong] [people/boards] Combine hunger with governance scaffolding. 2026, 00:09:51–00:11:42
• [investor-stated · moderate] [risk/exits] Kill bad candidates, retain company/platform options, treat IPO as financing. 2018 partial
• [researcher inference · strong] [portfolio] Match syndicate/reserves to full path, then preserve stop discipline. Fierce
• [documented behavior · strong] [risk] Sapphire and mRNA show boldness can become narrative overconfidence. DOE STAT

Appearance reconciliation

Axios Re:Cap and READOUT LOUD remain metadata-only and are not counted as processed content. No ★★★/★★ appearance lacks a note or deep section.

Create the missing company

• Rule — When important, inspectable science lacks a fit company, form the company; do not wait for a packaged pitch.
• Evidence status — [investor-stated · strong] across 2003, 2009, and 2026; applications usually partnership/team-produced.
• Provenance and origin — UChicago's 1986 commercialization affiliate is the earliest institutional use; Nelsen managed new-company formation from 1987. UChicago Illumina S-1 biography
• Mechanism and reasoning — Combine lab capability, IP, recruited executives, governance, and capital before market consensus; capture early ownership and design choices.
• Historical and vehicle context — Emerged when university transfer was immature; LP funds could hold illiquid equity. Fund XIII scale changes feasible company size but not the formation sequence. Fund XIII
• When to use — Use for protectable science with a credible experiment and no suitable home; do not use when licensing to an incumbent is superior or formation costs exceed information value.
• How to apply it — Name lab/scientist/IP/introducer/sponsor; define first product/experiment; recruit CEO/chair; map vehicle/security; set a 12–24 month kill milestone; ask why a new entity beats license/partnership.
• Examples —
  • Illumina — documented formation participation: Fund III/Nelsen board around an optical/genomics platform; scientific founders retain credit. S-1
  • Juno — documented multi-institution formation: Fred Hutch/MSK/Seattle Children's science plus ARCH/Alaska capital/board. S-1/A
  • Sapphire — counterexample/documented: formation and capital did not solve commercial deployment economics. DOE retrospective, p. 9-22
• Evidence — 2009 testimony, pp. 26–34; BioVenture, 00:01:13; TWST partial; issuer cases above.
• Limits, failure modes, and exception ledger — Access may be mistaken for judgment; de novo overhead is large; invention credit belongs to scientists; Rubius profile placement conflicts with Flagship origin.
• Evolution and contradictions — Mixed small formations evolved into $800M–$1B systems; scale may undermine experimental thrift.
• Modern VC translation (as of 2026-08-22) — [researcher inference · moderate] Transfers: form around a proprietary experiment/data loop when no incumbent is the better home. Changed assumption/failure boundary: abundant AI capital and reusable models make an “AI” label less defensible as standalone IP; Cambridge's 2024 discussion stresses fragmented biological data and the absence of a unified model. Signals: exclusive data-generation access, named IP, a recruited operator, and a 12–24 month falsifier. Likely misuse: confusing a financing announcement with proof that a new entity was necessary. Grounded current example: Xaira's reported $1B launch makes the organizational experiment observable but remains too young/private to validate the rule. Cambridge, printed pp. 3–5 Fierce, Xaira paragraphs

Treat a platform as a portfolio of real options

• Rule — Prefer a shared capability that creates several genuinely independent assets, and fund each option only after decision-changing evidence.
• Evidence status — [investor-stated · strong] for platform preference/real-options framing; moderate for superior application outcomes.
• Provenance and origin — Nelsen's 2026 account prefers three/four assets and calls Karuna an exception. BioVenture, 00:23:02, 01:02:22–01:08:15
• Mechanism and reasoning — Shared science/data/manufacturing lowers marginal experiment cost; independent shots diversify biology; staged capital preserves abandonment rights.
• Historical and vehicle context — Large LP funds/overage pools enable multiple assets and long trials; they also create correlated modality and sunk-cost risk. Fund X
• When to use — Use when assets share a validated engine but have distinct biological risk; avoid when “platform” is branding, assets are correlated, or one program consumes all capital.
• How to apply it — Draw shared engine and asset tree; estimate correlation; define asset-specific kills; price information per tranche; cap platform overhead; require at least two independent proof paths.
• Examples —
  • Illumina — researcher-applied illustration: enabling measurement platform compounded across applications. S-1
  • Juno — documented company/asset distinction: ROCKET stopped after deaths while company value/options survived. November update
  • Codiak — counterexample: exosome platform reached IPO then Chapter 11, showing optionality without proof. prospectus 8-K
• Evidence — BioVenture and Fund X above; BaseLaunch on killing bad drugs; issuer cases above.
• Limits, failure modes, and exception ledger — Platform correlation, overhead, dilution, weak asset gates; Karuna is the explicit product exception; Rubius/Unity are adverse platform controls.
• Evolution and contradictions — Platform preference intensified with fund size, while Karuna demonstrates evidence can dominate form.
• Modern VC translation (as of 2026-08-22) — [researcher inference · strong] Transfers: require a matrix of shared capability versus genuinely independent asset risk. Changed assumption/failure boundary: AI-generated assets trained on the same biased data may be more correlated than the portfolio diagram suggests. Signals: different targets, assays, patient populations, failure modes, and asset-specific kill gates. Likely misuse: relabeling one fragile hypothesis as many “shots.” Grounded current example: Sana's program suspensions and workforce reduction show that a broadly capitalized platform can still outrun asset evidence. Sana Q1 2026 10-Q

Make technical transparency a kill gate

• Rule — Do not invest when capable independent diligence cannot inspect the critical science/data, even if story, people, or social proof is strong.
• Evidence status — [investor-stated · moderate] from a self-reported Theranos pass; strongly consistent with filing-based cases but memo unavailable.
• Provenance and origin — Nelsen says Theranos secrecy was a blinking warning and ARCH passed. 2018 partial republication
• Mechanism and reasoning — Transparency enables replication, causal challenge, and falsifiers; secrecy prevents probability updates and lets charisma substitute for evidence.
• Historical and vehicle context — Biotech IP/confidentiality requires controlled diligence, not public disclosure; shared IC/expert networks can review under NDA.
• When to use — Always for core technical claims; adapt access controls for privacy/security. Do not confuse protected data rooms with refusal to substantiate.
• How to apply it — Identify irreducible claim; request raw methods/data; choose independent critic; reproduce key result; log unexplained exclusions; pass if access/replication fails.
• Examples —
  • Theranos — documented pass: secrecy blocked diligence; no investment/return. BaseLaunch
  • Karuna — researcher-applied illustration: prior human evidence and mechanistic side-effect mitigation were inspectable. S-1 business
  • Arivale — counterexample/boundary: rich measurement did not establish payer demand or sustainable assay economics; technical data is necessary, not sufficient. postmortem
• Evidence — Theranos source; Karuna filing; Arivale independent postmortem; 2009 testimony shows reliance on demonstrable technologies.
• Limits, failure modes, and exception ledger — A reproducible lab result may not scale, regulate, reimburse, or earn; Sapphire is the deployment counterexample.
• Evolution and contradictions — Later AI/data platforms increase both importance of access and legitimate privacy constraints.
• Modern VC translation (as of 2026-08-22) — [researcher inference · strong] Transfers: require held-out prospective experiments, data provenance, and independent technical review. Changed assumption/failure boundary: modern privacy, security, and licensing constraints can justify controlled access, not refusal to substantiate. Signals: reproducible raw methods, documented exclusions, critic access, and a prospective test. Likely misuse: demanding public disclosure or accepting polished model demos as diligence. Hypothetical current application: an AI-biology company can protect patient data in a controlled room while still permitting an independent replication; failure to permit either triggers a pass. Nelsen's current premise that biological data are sparse and fragmented is documented, while the application is analyst-created. Cambridge, printed pp. 3–5

Finance to decision-changing evidence

• Rule — Every check should buy a defined probability update or remove a named constraint; continue only when the evidence changes the decision.
• Evidence status — [researcher inference · strong] from real-options language, candidate-kill rule, vehicle design, and failures; no public universal ARCH rule.
• Provenance and origin — 2009 distinguishes invention/deployment capital; 2018 celebrates killing bad drugs; 2026 uses real options. testimony BioVenture, 01:07:48–01:10:16
• Mechanism and reasoning — Staging preserves abandonment option, controls downside, and makes abundant capital accountable to learning.
• Historical and vehicle context — Overage funds and megarounds allow follow-ons; closed-end LP opportunity cost and finite duration remain even when NPV is unreliable.
• When to use — All high-uncertainty stages; not when a lumpy non-divisible facility is the only feasible experiment—then define project gates.
• How to apply it — State current belief; name next decision; choose milestone; estimate cost/time; set kill/continue bands; assign board owner; prohibit new capital without result/exception memo.
• Examples —
  • Juno — documented: clinical deaths forced a program decision while other options remained. updates
  • Denali — researcher-applied illustration: $217M financed multiple neuro shots; value depends on independent milestones. contemporaneous formation
  • Sana — counterexample/live test: program suspensions/workforce reduction show platform capital can outrun evidence. 10-Q
• Evidence — testimony/BioVenture; Fund X structure; three cases above.
• Limits, failure modes, and exception ledger — Milestones can be gamed, binary trials are lumpy, and abundant reserves weaken refusal; public reserve/kill records are missing.
• Evolution and contradictions — Check sizes grew from seed to hundreds of millions; the need for explicit information gates increased.
• Modern VC translation (as of 2026-08-22) — [researcher inference · strong] Transfers: track dollars and months per uncertainty retired, not valuation uplift. Changed assumption/failure boundary: large private rounds extend runway but can weaken the abandonment option; some manufacturing or pivotal trials remain irreducibly lumpy. Signals: a named next decision, prewritten continue/kill bands, and a board owner. Likely misuse: treating runway or activity as learning. Grounded current example: Sana's program suspensions, workforce reduction, and liquidity disclosures make the missing information-gate question concrete; require an exception memo when a follow-on buys time rather than evidence. Sana Q1 2026 10-Q

Assemble the IPO book at seed with stop discipline

• Rule — Recruit a syndicate capable of financing the plausible full journey at formation, while preserving price, governance, and milestone refusal.
• Evidence status — [investor-stated · strong] for coalition/capital doctrine; provisional for superior outcomes.
• Provenance and origin — Fund X core/overage made structure visible in 2020; Nelsen stated “IPO book in the seed round” in 2024. Fierce
• Mechanism and reasoning — Deep-pocketed expert co-investors reduce financing/recruiting risk and avoid desperate rounds; tranches/boards must preserve falsification.
• Historical and vehicle context — Fund XIII >$3B and billion-dollar private launches make this feasible; small funds/weak exit markets cannot replicate it. Form D
• When to use — Capital-intensive platforms/manufacturing with credible experiments; not ordinary software or unvalidated hypotheses whose primary risk is idea quality.
• How to apply it — Model full path/delay case; choose investors by expertise/reserves; set ownership/dilution; define tranches; allocate board rights/conflicts; test fund concentration.
• Examples —
  • Juno — documented illustration: successive large private rounds/public financing carried CAR-T through proof and sale. S-1/A
  • Denali — documented: $217M Series A assembled team/capital for long neuro programs. Forbes 2015
  • Sana/Lyell — counterexamples/live: large IPO capital did not prevent program stops, workforce cuts, dilution/reverse split. Sana Lyell split
• Evidence — Fierce/Form D; Juno/Denali/Sana/Lyell cases; Fund X.
• Limits, failure modes, and exception ledger — False precision, price inflation, herding, governance crowding, correlated reserves, sunk costs; Xaira/Altos/Prometheus remain unvalidated private marks.
• Evolution and contradictions — Capital scale became a strategic product, yet Nelsen says fundraising is a bad success metric. BioVenture, 00:43:56–00:45:02
• Modern VC translation (as of 2026-08-22) — [researcher inference · moderate] Transfers: stress-test who funds after a failed or late milestone, at what ownership, and who can say no. Changed assumption/failure boundary: billion-dollar launches and weak public-biotech liquidity magnify dilution, governance crowding, and sunk-cost risk; a small fund cannot copy the structure. Signals: committed reserves, tranche rights, concentration limits, a downside syndicate, and explicit refusal authority. Likely misuse: reading fund size or an “IPO book” as quality. Grounded current example: Xaira's reported $1B formation illustrates coalition capacity, not validated economics, so its milestone and refusal architecture—not the round size—is the test. Fierce, Xaira/seed-syndicate paragraphs

Pair urgent builders with governance scaffolding

• Rule — Back hungry, impatient leaders when the board/executive-chair design explicitly supplies missing experience and a safety slow-down mechanism.
• Evidence status — [investor-stated · strong] for preference/scaffolding; moderate for causal outcomes.
• Provenance and origin — Nelsen contrasts hungry builders with incentive-diffused pharma managers and names executive-chair/board support. BioVenture, 00:09:51–00:14:11
• Mechanism and reasoning — Urgency accelerates recruiting/experiments; experienced governance supplies pattern, controls, and replacement authority.
• Historical and vehicle context — Formation investing creates companies before complete management exists; investor boards and long capital make scaffolding possible.
• When to use — Technical founders with learning speed/integrity and identifiable gaps; not charismatic secrecy, unsafe speed, or boards without power/time.
• How to apply it — Map founder strengths/gaps; recruit chair/operators; define reserved matters/safety escalation; collect 360 references; set replacement/role-evolution triggers.
• Examples —
  • Juno — investor-stated application: urgent culture and Amazon HR recruiting; board context. BioVenture, 00:06:39
  • Denali — documented illustration: experienced Genentech neuroscience team paired with large board/syndicate. Forbes
  • Juno ROCKET — counterexample: urgency could not supersede lethal safety evidence; program stopped. issuer update
• Evidence — BioVenture/Denali/Juno; Founder Stories video repeats founder urgency.
• Limits, failure modes, and exception ledger — Boards may be captured, part-time, conflicted, or slow; urgency can compress controls; intervention evidence is sparse.
• Evolution and contradictions — The preference is stable; larger companies/megarounds make governance more complex than a single chair.
• Modern VC translation (as of 2026-08-22) — [researcher inference · moderate] Transfers: pair founder urgency with explicit board, executive-chair, and safety authority. Changed assumption/failure boundary: capital and AI can accelerate iteration, but patient safety, controls, and data integrity cannot be compressed equivalently. Signals: reserved matters, independent safety escalation, 360-degree references, and pre-agreed role-change triggers. Likely misuse: selecting charisma or impatience without a board able to slow or replace. Grounded current boundary: public Prometheus materials identify Bezos and Bajaj as cofounders/co-CEOs, while public board/cap-table evidence remains unavailable; do not infer Nelsen's scaffolding role from the ARCH profile. Prometheus leadership transcript ARCH profile

Separate science, company, and security

• Rule — Underwrite technical truth, enterprise survival, and the particular security/vehicle as three different propositions; never infer return from company value.
• Evidence status — [researcher inference · strong] from primary filings and deCODE/Juno/Karuna/GRAIL.
• Provenance and origin — deCODE's original equity bankruptcy and rescue-asset sale makes the separation explicit. bankruptcy Amgen
• Mechanism and reasoning — Debt, preference, dilution, follow-ons, timing, tax, and legal close determine investor cash independently from scientific merit.
• Historical and vehicle context — Shared ARCH funds, multiple vintages/overage vehicles, IPOs, strategic sales, rescues, and private marks make attribution/economics layered.
• When to use — Every deal/outcome claim; especially bankruptcies, recapitalizations, spinouts, M&A, and shared-control filings.
• How to apply it — Build dated cap-table/cash-flow chain; name vehicle; trace dilution/distributions; reconcile accounting versus cash; record legal/regulatory close; separate personal options.
• Examples —
  • deCODE — documented counterexample: correct science did not protect original common; the rescue entity later sold for $401M cash at close. Amgen 10-K
  • Juno — documented: $9.101B outside cash within $10.434B accounting; Fund VII gross is not net/personal. accounting Form 4
  • Karuna — boundary: 2020 12.9% stake does not prove holding/proceeds at 2024 close. proxy
• Evidence — cases above; GRAIL 9.35% shared stake; shared IC disclaimer.
• Limits, failure modes, and exception ledger — Private cap tables/distributions are absent; public beneficial ownership may mean voting control; exact carry/tax stays unknowable.
• Evolution and contradictions — Bigger headline transactions/private marks increase temptation to conflate value and return.
• Modern VC translation (as of 2026-08-22) — [researcher inference · strong] Transfers: attach an economics label to every number—financing mark, equity/enterprise value, gross security proceeds, fund net cash, or personal economics. Changed assumption/failure boundary: SPACs, reverse splits, spinouts, and private marks make headline-value comparisons especially unreliable. Signals: a dated security-level cap table, dilution/cash-flow chain, legal close, and distribution record. Likely misuse: multiplying a stale stake by a sale price or comparing an unadjusted IPO price with a post-split quote. Grounded current example: Lyell's 1-for-20 split makes its $17 IPO price $340 on a current split-adjusted basis; the adjustment still does not establish ARCH's return. Lyell IPO prospectus reverse-split release

Keep a forecast and failure ledger

• Rule — Record dated prediction, confidence, causal mechanism, decision, and later result—including misses—and update the rule without erasing original wording.
• Evidence status — [researcher inference · strong] need demonstrated by Sapphire/mRNA and Nelsen's tolerance-of-failure rhetoric; no public ARCH ledger.
• Provenance and origin — 2009 Sapphire testimony and 2020 COVID record preserve rare ex-ante beliefs. Congress STAT
• Mechanism and reasoning — Prevent outcome knowledge and reputation from rewriting prior confidence; distinguish correct macro call from wrong modality; calibrate gut.
• Historical and vehicle context — High-variance long-duration funds can hide misses for years; marketing pages overrepresent survivors, making a ledger essential.
• When to use — Thesis formation, IC, follow-ons, public policy claims, and postmortems; do not score unknowable outcomes before maturity.
• How to apply it — Timestamp belief/probability; write falsifier; link decision/size; schedule review; score direction/calibration; record process change and counterfactual.
• Examples —
  • Sapphire — documented miss: commercial fuel forecast failed; facility shifted. DOE retrospective
  • COVID spread — documented strong call: early undercount/asymptomatic/mitigation recognition. STAT
  • mRNA — documented counterexample within same source: skepticism was wrong while macro call was right. STAT
• Evidence — Congress/DOE/STAT; Arivale response shows theme retained with business-model update.
• Limits, failure modes, and exception ledger — Outcomes remain ambiguous; predictions can be cherry-picked; no full ARCH forecast corpus; counterfactual decision quality may differ from realized luck.
• Evolution and contradictions — Nelsen publicly values failure yet no standardized public ledger is visible; fundraising boast/bad-metric admission is another calibration target.
• Modern VC translation (as of 2026-08-22) — [researcher inference · strong] Transfers: store dated forecast cards and review positive and negative anti-portfolio, not only “called it right.” Changed assumption/failure boundary: private marks mature slowly and ambiguous milestones tempt narrative rescoring; Brier-like scoring works only when probabilities and outcomes are defined ex ante. Signals: timestamp, confidence, causal mechanism, falsifier, linked check, and scheduled review. Likely misuse: scoring fame, valuation, or a partially matured company as prediction accuracy. Grounded current example: Project Prometheus's reported $41B post-money is a financing mark to log, not a successful-outcome score; score later operating evidence against the original thesis. Axios, 2026-06-11

Evidence gaps and candidate models

• “Money follows impact” is a maxim, not a standalone model: Arivale/Sapphire show missing payer/commodity mechanisms. Incorporated as a boundary across models.
• “10,000:1 impact” lacks a stable measurable definition and application ledger; candidate impact-sizing model remains [investor-stated · insufficient public record].
• Regulatory decentralization/universal IRBs is a policy proposal, not sufficiently evidenced investing rule; no three grounded decisions/process applications.
• China competitiveness is a live macro view without an accessible decision/reject ledger; candidate model remains insufficient.
• Retrieval attempts: all 16 notes, five writings, 11 processed talks, 55-company union/filings, firm PDF, LP report, SEC/FTC/EC/CJEU/DOE routes, and named theme queries. Missing IC memos, passes, board minutes, and private-company data prevent promotion.
• Inventory reconciliation: eight distinct models cover supported sourcing, picking, portfolio, people/boards, risk, and economics rules. Synonyms—company creation/incubation, platform/options, reserves/evidence staging—were merged rather than inflated.

What the evidence supports

The strongest supported conclusion is institutional, not heroic: Nelsen helped turn a University of Chicago/Argonne commercialization system into an ARCH company-formation practice, then repeatedly paired laboratory access, specialist judgment, recruiting/boards, pooled LP capital, and follow-on capacity. The early institutional record says ARCH Development began as a university affiliate in 1986; Illumina's S-1 places Nelsen in new-company formation from 1987 and at the independent funds from 1994. [contemporaneous record · strong] UChicago Chronicle Illumina S-1, Nelsen biography

The repeatable-looking mechanism is: find inspectable science before a conventional company exists; recruit the missing organization; finance several decision-changing experiments; use boards and a syndicate with endurance; retain options only when evidence improves. [researcher inference · strong] 2009 testimony, printed pp. 26–34 BioVenture, 00:01:13, 00:09:51–00:14:11, 01:07:48–01:10:16 The evidence is strongest for formation/board/fund participation in Illumina, Juno, Karuna, GRAIL, Beam, Denali, and current public-board cases; it does not establish sole sourcing, sole IC approval, or personal returns.

Illumina, Juno, and Karuna show three different valid portions of the mechanism: enabling-platform formation, urgent multi-institution formation with asset-level stopping, and a high-evidence single-product exception. They do not prove a universal rule. Illumina's scientific founders retain invention credit; Juno's clinical deaths remain central; PureTech/Steve Paul originated Karuna and Nelsen's board tenure ended in 2021. [documented behavior · strong] Illumina S-1 Juno November update Karuna S-1

The adverse record is substantive rather than decorative. Sapphire falsifies commercial-deployment confidence; mRNA falsifies a technology forecast inside a correct pandemic warning; Xcyte/Codiak/Rubius/Unity test the platform doctrine; Arivale tests payer/unit economics; deCODE separates science from the security; EQRx separates cash transfer from operating success; GRAIL tests exit/regulatory governance. [documented behavior · strong] Sapphire DOE retrospective, printed p. 9-22 STAT COVID/mRNA record failure labs

What could falsify the thesis

The favorable edge thesis fails if complete, vintage/stage/modality-matched cash flows show no advantage after removing Illumina/Juno/Karuna and controlling for ownership, dilution, capital, and exits. It also weakens if contemporaneous memos show access rather than selection drove entry; follow-ons followed sunk cost instead of evidence; partner/scientist/founder work explains the cases assigned to Nelsen; or board involvement did not change recruiting, safety, financing, or stopping. [researcher inference · strong] searched denominator strongest skeptical case

The strongest alternative explanation is a selected narrative: elite institutional access, public research, ARCH partners, scientist/founder execution, Alaska and other LP capital, favorable biotech windows, and luck may explain the apparent personal record. The current Bob page is a curated 55-name union, includes private marks and disputed/team cases, and is not a loss or cash-flow denominator. [researcher inference · strong] current profile firm-wide portfolio warning/list

Discriminating evidence would be: audited vehicle cash flows and PME; complete deal cost/distribution chains; all passes and write-offs; contemporaneous IC memos/votes/dissent; partner-role allocation; board intervention records; forecast calibration; and a matched Flagship/Third Rock/5AM cohort. Without these, “consequential company builder” is better supported than “proved personal alpha.” [researcher inference · strong] retrieval ledger strongest skeptical case

Attribution and denominator audit

The mechanically reproducible person-linked denominator is 50 names in the current ARCH narrative, 36 in “Key Investments,” and 55 unique across the union. The December 2023 ARCH PDF contains 321 printed rows/320 unique company strings and is firm-wide only. Ten of 55 person-linked cases are sufficiently reconstructed for bounded role/outcome analysis; exact economic coverage is not computable. [contemporaneous record · strong] current profile dated PDF ledger/reconciliation

Ownership filings usually describe shared voting/disposition by ARCH managing directors/IC members and disclaim personal beneficial ownership beyond pecuniary interest. Therefore “ARCH/Nelsen reporting person” is not “Nelsen personally wrote the check or received proceeds.” [contemporaneous record · strong] current shared-IC Form 4 Juno 13G

Five material attribution controls passed underlying-source cross-checks. [researcher inference · strong]

1. Illumina: Nelsen director/Fund III investor; Walt, Stuelpnagel, and the scientific group retain founding/invention credit. S-1
2. Juno: Nelsen board/options plus shared Fund VII; Fred Hutch, MSK, Seattle Children's, scientists, executives, Crandell, and Bybee remain visible. S-1/A 13G
3. Karuna: PureTech founded it; Steve Paul supplied the asset thesis; ARCH joined; Nelsen boarded August 2018–June 2021. S-1 2020 proxy
4. Codiak: ARCH/MD Anderson formation and Gillis are supported; Bob-specific sourcing is not. prospectus
5. Receptos/Rubius: Receptos is a Burow/team case; Rubius's S-1 says Flagship-founded and contains no ARCH/Nelsen match despite the Bob page. Receptos/Burow Rubius S-1

Reported economics remain labeled. Fund VII's 10,552,390 Juno shares at $87 equal $918,057,930 gross before cost/tax, for the vehicle—not Nelsen or fund net. One North Carolina LP reports, at 2025-12-31, Fund VII–IX/overage IRR/TVPI/DPI figures; that table does not establish gross/net basis or the firm/personal record. [contemporaneous record · strong] Juno Form 4 NC report

Transfer limits for modern VCs

Transferable: direct technical access, explicit origin/credit mapping, formation around a missing capability, several genuinely independent assets, information-buying milestones, full-path reserve planning, and safety/governance stop rights. [researcher inference · strong] models dated modern translation

Non-transferable without ARCH's system: elite lab access, four decades of reputation, specialist partner/board networks, multi-billion-dollar LP pools, and tolerance for long private development. “Gut,” “platform,” “impact,” and “IPO book at seed” are dangerous cargo cults when separated from raw data, independent critics, priced ownership, tranches, and stop rules. [researcher inference · strong] Fund XIII Form D Fierce, seed-book paragraph

Changed assumptions as of 2026-08-22 include much larger private rounds, AI/data commoditization risk, complex payer/IRA/China branches, and long public-biotech financing droughts. The safe modern test is observable: proprietary experimental data, independent asset risk, dollars per uncertainty retired, a downside reserve/dilution map, and a named governance owner. Private valuation is never treated as realized contribution; Project Prometheus AI's reported $41 billion post-money mark is therefore unranked. [researcher inference · moderate] reported financing

Unresolved evidence

• Complete IC memos, votes, dissent, passes, price/ownership targets, reserve decisions, and board minutes.
• Complete deal cash-flow chains, fund audited returns/PME, fees/carry, recycling, tax, and personal pecuniary economics.
• Full 2003 TWST and original Xconomy bodies; timestamped Long Run/BioCentury/READOUT records; private-company cap tables.
• Project Prometheus/Stability role/governance allocation; current private marks and outcomes.
• A complete adverse/ordinary denominator and matched modern biotech-formation peer cohort.

These are public-record limits after named retrieval routes were exhausted, not permission to infer favorable answers. [researcher inference · strong] source saturation/access ledger

High-risk central-claim audit

Selection rule: every economic figure, causal edge statement, Pareto leader, signature-deal attribution, failure lesson, vehicle claim, peer conclusion, and current premise was re-opened. The table records the claim-changing checks; supporting audits appear in the linked sections.

Ordinary-claim audit

Reproducible spread: the first material non-central factual claim after the first substantive H2 in each core document, plus three later cross-document controls, for 11 claims total.

Twenty-point acceptance evidence

Link and integrity spot checks

Checked 2026-08-22 against the adjacent claims: Illumina S-1 (role/holding), Juno Form 4 (shares/$87), Fifth Circuit opinion (pp. 34–35), Illumina Q3 2024 10-Q (spin/funding/impairments), DOE Sapphire retrospective (printed p. 9-22), and the BMS/Karuna announcement (transaction terms). Each resolved and supported the bounded claim. Mechanical checks also confirmed 101 checked source rows (76/22/3) plus one unchecked ★ READOUT mapped-only route—102 total inventory rows (76/22/4)—six required H2s in every note, eight dated model translations, eight model inventories with twelve fields, and no taxonomy-first note label.

The package therefore supports completion at the research boundary. Completion does not convert the unresolved private record into evidence, imply audited personal alpha, or authorize a future claim to use a stale current-company fact without rechecking it.

First-person and authored material

• ★★★ | House Climate for Innovation testimony | https://www.congress.gov/event/111th-congress/house-event/LC20982/text | 2009-07-29 | official full record, printed pp. 26–34; lab formation/IP/deployment and Sapphire forecast.
• ★★★ | Official hearing PDF | https://www.congress.gov/111/chrg/CHRG-111hhrg62451/CHRG-111hhrg62451.pdf | 2009-07-29 | primary pagination control; same underlying work as HTML, not independent corroboration.
• ★★ | Email on Yellen biotech comments | https://www.forbes.com/sites/matthewherper/2014/07/17/an-unsolicited-comment-on-janet-yellen/ | 2014-07-17 | direct email reproduced by journalist; full; no valuation model.
• ★★★ | COVID posts/interview | https://www.statnews.com/2020/03/30/biotech-vc-bob-nelsen-called-it-right-on-the-coronavirus-now-he-has-thoughts-on-therapeutics-and-masks/ | 2020-03-30 | direct/reproduced posts; spread insight plus mRNA miss.
• ★★ | Fund X and overage announcement | https://www.archventure.com/arch-venture-partners-announces-1-46-billion-raised-in-two-new-funds-to-invest-in-transformative-biotechnology-companies/ | 2020-04-02 | official; amount/structure primary, thesis promotional.
• ★★ | Fund XIII announcement | https://www.archventure.com/arch-venture-partners-announces-new-fund-xiii/ | 2024-09-26 | official >$3B release; promotional thesis bounded by Form D.

Talks, interviews, and transcript routes

• ★★★ | TWST early-stage interview | https://www.twst.com/interview/early-stage-technology-venture-capital-robert-nelsen-arch-venture-partners | 2003-02-03 | opening/metadata accessible, body partial/paywalled; earliest formation snapshot.
• ★★★ | Xconomy interview ARCH route | https://www.archventure.com/co-founder-and-managing-director-robert-nelsen-speaks-with-xconomy-on-biotech-and-big-pharma/ | original interview, mirrored 2018 | partial; IPO, drug kills, Theranos, AI.
• ★★ | BaseLaunch Xconomy republication | https://baselaunch.ch/blog-post/biotech-pharma/ | 2018-05-15 | partial/edited; same underlying appearance as prior row.
• ★★★ | STAT coronavirus interview | https://www.statnews.com/2020/03/30/biotech-vc-bob-nelsen-called-it-right-on-the-coronavirus-now-he-has-thoughts-on-therapeutics-and-masks/ | 2020-03-30 | full article; same underlying work as authored-post row.
• ★ | Axios Re:Cap on AstraZeneca/Resilience | https://www.axios.com/2020/11/23/bob-nelsen-astrazeneca | 2020-11-23 | metadata/context only after transcript search; no public transcript or content-grade body, so not processed.
• ★★ | The Long Run ep. 87 | https://timmermanreport.com/2021/03/investing-in-manufacturing-mental-health-more-bob-nelsen-on-the-long-run/ | 2021-03-01 | 49:10 audio/show notes; no public full transcript or quote-grade timestamps.
• ★★★ | C&EN aging episode | https://cen.acs.org/biological-chemistry/biochemistry/Podcast-research-aging-keep-us/99/i11 | 2021-03-24 | edited full transcript/audio; motivation, not efficacy.
• ★★ | WMIF official day-three recap | https://www.massgeneralbrigham.org/en/about/newsroom/articles/2021-world-medical-innovation-forum-day-3-recap | event 2021-05-21 | official recap; gene/cell manufacturing/scalability.
• ★★ | WMIF transcript mirror | https://ceointerviews.ai/quote/resilience-wants-to-be-a-proactive-investor-well-OTI4MzA3/ | upload 2021-12-14 | timestamped mirror; not event-owner custody.
• ★ | GIC healthcare panel summary | https://www.gic.com.sg/uploads/2022/01/GIC-Insights-2021-A-Sustainable-Future-for-Healthcare.pdf | 2021 | edited summary, not full transcript; context only.
• ★ | SeedCon resilience keynote metadata | https://polsky.uchicago.edu/2021/02/26/seedcon-2021-resilience-and-opportunity/ | 2021 | no transcript; route mapped, no content claim.
• ★★ | Axial Founder Stories video | https://www.youtube.com/watch?v=VD1sPcLp0CY | 2023-06-30 | full video/auto captions; exact ASR quotes excluded.
• ★★★ | BioCentury AI/China/IRA article | https://www.biocentury.com/article/651502/bob-nelsen-on-ai-china-and-the-ira | recorded/published 2024-02-22 | specialist excerpts; no full reliable transcript.
• ★★ | BioCentury audio reissue | https://podbay.fm/p/the-biocentury-show/e/1724360400 | 2024-08-22 | 35:33 audio; same underlying appearance.
• ★★★ | Fierce megarounds interview | https://www.fiercebiotech.com/biotech/bob-nelsen-arch-led-megarounds-we-want-ipo-book-seed-round | 2024-05-01 | full edited interview; “IPO book” doctrine.
• ★★★ | Cambridge AI/healthcare landing | https://www.cambridgeassociates.com/en-eu/podcasts/ai-healthcare/ | 2024-03-01; updated 2026-01-29 | multi-guest episode.
• ★★★ | Cambridge AI/healthcare transcript PDF | https://www.cambridgeassociates.com/wp-content/uploads/2024/10/CA-UU-Epi-6_Healthcare_Script_FINAL.pdf | 2024 | full edited transcript, printed pp. 3–5, 13–14; same underlying appearance.
• ★★★ | BioVenture VoiCes primary video/page | https://www.biotechtv.com/post/bioventure-voices-bob-nelsen | 2026-01-11 | primary chaptered video.
• ★★★ | BioVenture VoiCes timestamped transcript | https://www.bishoprockcap.com/assets/library/bob-nelsen/January%202026%20Episode%2032%20ARCH%20Venture%20Partners%27%20Bob%20Nelsen%20%28transcript%29 | 2026-01-11 | ASR average confidence 0.77; cited spans spot-checked; same underlying appearance.
• ★ | READOUT LOUD live from JPM | https://www.statnews.com/2026/01/14/readout-loud-podcast-jpm-2026-bob-nelsen-mike-doustdar/ | 2026-01-14 | audio/video, no public transcript; mapped only.

Biography, firm history, denominator, and vehicles

• ★★★ | ARCH current Nelsen profile | https://www.archventure.com/team/robert-nelsen/ | current; accessed 2026-08-22 | official current roles/self-attribution; 50 narrative, 36 key, 55-name union; promotional, not audited.
• ★★★ | ARCH 2023 comprehensive portfolio PDF | https://www.archventure.com/wp-content/uploads/2024/03/for-website-comprehensive-portfolio-list-12-2023.pdf | 2023-12-31 | 321 printed rows/320 unique strings; firm-only and explicit profitability warning.
• ★★★ | ARCH current portfolio | https://www.archventure.com/portfolio/ | current | firm search control, not Bob denominator.
• ★★★ | UChicago Chronicle ARCH history | https://chronicle.uchicago.edu/950119/arch.shtml | 1995-01 | contemporaneous 1986 university-affiliate origin and 16-company claim.
• ★★★ | ARCH on Steven Lazarus | https://www.archventure.com/steven-lazarus-1931-2018/ | 2018 | institutional origin/Lazarus role/1992 separation; prevents lone-founder myth.
• ★★★ | Illumina S-1 biography | https://www.sec.gov/Archives/edgar/data/1110803/000101287000003639/0001012870-00-003639-0001.txt | 2000 | Nelsen ARCH employment/education/board and Fund III ownership.
• ★★★ | Hong Kong exchange Nelsen biography | https://www1.hkexnews.hk/listedco/listconews/sehk/2024/0425/2024042500752.pdf | 2024 | education and board chronology; primary issuer custody.
• ★★★ | Fund XIII Form D | https://www.sec.gov/Archives/edgar/data/2006772/000101297524000050/xslFormDX01/primary_doc.xml | 2024 | $3B offering; fee paid quarterly in advance but rate absent.
• ★★★ | Current shared-IC Form 4 | https://www.sec.gov/Archives/edgar/data/1806952/000101297526000235/xslF345X03/form4.xml | 2026 | five IC members/share-control disclaimers; antidote to personal ownership.
• ★★★ | NC Investment Authority CY2025 performance report | https://www.nctreasurer.gov/documents/files/ncia/performance-report-calendar-year-2025/download?attachment= | figures at 2025-12-31 | one LP's Fund VII–IX/overage exposures; table basis caveat.
• ★★ | Fund VIII close report | https://www.prnewswire.com/news-releases/latest-arch-venture-partners-fund-closes-above-400-million-272850081.html | 2014 | >$400M versus $250M target; release distribution.
• ★★ | Fund XI announcement | https://www.archventure.com/arch-venture-partners-announces-new-1-85-billion-fund-to-create-and-fund-early-stage-biotechnology-companies/ | 2021 | vehicle control only; >$1.85B and $50K–$250M check range; no qualifying Nelsen-authored statement for the writing corpus.
• ★★ | Fund XII announcement | https://www.archventure.com/arch-venture-partners-announces-2-975-billion-fund-xii-to-create-and-fund-early-stage-biotechnology-companies/ | 2022 | vehicle control only; official $2.975B mandate; no qualifying Nelsen-authored statement for the writing corpus.

Deal, ownership, outcome, and attribution records

• ★★★ | Illumina S-1 | https://www.sec.gov/Archives/edgar/data/1110803/000101287000003639/0001012870-00-003639-0001.txt | 2000 | Fund III $5.05359M, 3,615,299 shares, 14.2% pre/11.8% post; shared-control disclaimer.
• ★★★ | Beam 2021 proxy | https://www.sec.gov/Archives/edgar/data/1745999/000156459021020335/beam-def14a_20210609.htm | 2021 | Nelsen/Burow board chronology and ARCH ownership control.
• ★★★ | Juno S-1/A | https://www.sec.gov/Archives/edgar/data/1594864/000119312514436942/d772541ds1a.htm | 2014 | formation rounds, syndicate, Nelsen board, ARCH vehicle.
• ★★★ | Juno ARCH 13G | https://www.sec.gov/Archives/edgar/data/1402439/000107261318000076/archvii-juno_13g18160.htm | holdings 2017-12-31 | 10,552,390 Fund VII shares; shared Crandell/Nelsen/Bybee power.
• ★★★ | Juno tender Form 4 | https://www.sec.gov/Archives/edgar/data/1219042/000120919118017824/xslF345X03/doc4.xml | 2018 | 10,552,390 Fund VII shares at $87; vehicle gross, not personal/net.
• ★★★ | Celgene/BMS Juno accounting | https://www.sec.gov/Archives/edgar/data/816284/000081628419000006/exhibit995.htm | 2018 | $9.101B outside-holder cash/$10.434B total accounting consideration.
• ★★★ | Juno July clinical-hold release | https://www.sec.gov/Archives/edgar/data/1594864/000119312516643572/d201119dex992.htm | 2016-07 | two deaths/hold; contemporaneous.
• ★★★ | Juno November ROCKET update | https://www.sec.gov/Archives/edgar/data/1594864/000159486417000003/exhibit991.htm | 2016-11 | five treatment-related deaths/program decision.
• ★★★ | Juno 14D-9 | https://www.sec.gov/Archives/edgar/data/1594864/000119312518030824/d514862dsc14d9.htm | 2018 | merger process/options control.
• ★★★ | Karuna S-1/A | https://www.sec.gov/Archives/edgar/data/1771917/000119312519174168/d709743ds1a.htm | 2019 | PureTech/Steve Paul provenance; ARCH Fund IX/overage rounds.
• ★★★ | Karuna 2020 proxy | https://www.sec.gov/Archives/edgar/data/1771917/000119312520126055/d923844ddef14a.htm | 2020 | ARCH 3,362,872/12.9%; Nelsen 16,250 direct/options; time-bounded.
• ★★★ | BMS/Karuna acquisition announcement | https://investors.bms.com/iframes/press-releases/press-release-details/2023/Bristol-Myers-Squibb-Strengthens-Neuroscience-Portfolio-with-Acquisition-of-Karuna-Therapeutics/ | 2023-12-22 | $330 cash per share; $14B equity/$12.7B net of estimated acquired cash; announced terms, not ARCH proceeds.
• ★★★ | BMS Karuna completion | https://investors.bms.com/iframes/press-releases/press-release-details/2024/Bristol-Myers-Squibb-Completes-Acquisition-of-Karuna-Therapeutics-Strengthening-Neuroscience-Portfolio/default.aspx | 2024 | $330/share; does not establish ARCH closing holding.
• ★★★ | GRAIL formation | https://grail.com/press-releases/illumina-forms-new-company-to-enable-early-cancer-detection-via-blood-based-screening/ | 2016-01-10 | $100M+ Series A, ARCH/Illumina/others, Nelsen director.
• ★★★ | GRAIL S-1 | https://www.sec.gov/Archives/edgar/data/1699031/000162827920000227/grails-1.htm | 2020 | business/risks/ownership.
• ★★★ | Illumina/GRAIL S-4 | https://www.sec.gov/Archives/edgar/data/1110803/000119312520302773/d801214ds4.htm | 2020 | ARCH 63,710,240 as-converted/9.35%, shared control.
• ★★★ | Illumina 2021 10-K | https://www.sec.gov/Archives/edgar/data/1110803/000111080322000013/ilmn-20220102.htm | filed 2022-02-18 | GRAIL closing purchase accounting: $9.745B total across cash, stock, contingent and prior interests.
• ★★★ | FTC final opinion | https://www.ftc.gov/system/files/ftc_gov/pdf/d09401commissionfinalopinion.pdf | 2023 | US vertical-harm finding/divestiture route.
• ★★★ | Fifth Circuit Illumina v. FTC | https://www.ca5.uscourts.gov/opinions/pub/23/23-60167-CV0.pdf | 2023-12-15 | substantial-evidence holdings plus Open Offer legal-standard vacatur/remand, pp. 34–35.
• ★★★ | FTC 2024 dismissal after divestiture | https://www.ftc.gov/system/files/ftc_gov/pdf/d09401_commission_order_dismissal_unsigned.pdf | 2024 | proceeding endpoint after spin.
• ★★★ | CJEU jurisdiction release | https://curia.europa.eu/jcms/upload/docs/application/pdf/2024-09/cp240127en.pdf | 2024-09 | EC lacked jurisdiction.
• ★★★ | EC withdrawal | https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A52024M10188 | 2024-09-06 | prohibition/divestiture/€432M fine decisions withdrawn.
• ★★★ | Illumina GRAIL spin-off completion | https://www.illumina.com/company/news-center/press-releases/press-release-details.html?newsid=7881 | 2024-06 | completed separation.
• ★★★ | Illumina Q3 2024 10-Q | https://www.sec.gov/Archives/edgar/data/1110803/000111080324000052/ilmn-20240929.htm | filed 2024-11-05 | 85.5% distribution/14.5% retention, $974M disposal funding, impairments, retained-stake loss, FTC/EC endpoints.
• ★★★ | deCODE pre-bankruptcy 10-K | https://www.sec.gov/Archives/edgar/data/1022974/000104746909003539/a2191980z10-k.htm | 2009 | debt/liquidity/business control.
• ★★ | deCODE bankruptcy report | https://www.theguardian.com/business/2009/nov/17/decode-genetics-bankruptcy | 2009 | $69.9M assets/$313.9M debt; original security distress.
• ★★★ | Amgen deCODE announcement | https://www.amgen.com/newsroom/press-releases/2012/12/amgen-to-acquire-decode-genetics-a-global-leader-in-human-genetics | 2012 | $415M all-cash subject to closing adjustments.
• ★★★ | Amgen deCODE 10-K | https://investors.amgen.com/static-files/2c053867-736d-4413-9fbd-46295e97337b | 2012 | $401M total cash consideration at close.
• ★★★ | Denali formation report | https://www.forbes.com/sites/stevedickman/2015/05/21/why-denali-raised-a-217-million-financing-round-not-just-because-it-could/ | 2015 | contemporaneous $217M round/team/risk.
• ★★★ | Denali proxy | https://investors.denalitherapeutics.com/static-files/f0fb4a5c-6980-43b0-858b-e59448662261 | 2019 | ARCH ownership/shared control.
• ★★ | UChicago Polsky Resilience formation report | https://polsky.uchicago.edu/2020/12/04/new-biopharma-manufacturing-company-founded-by-booth-alum-robert-nelsen-raises-800-million/ | 2020 | Nelsen formation attribution and reported $800M launch; institutional promotional context.
• ★★★ | Pfizer Array acquisition announcement | https://www.pfizer.com/news/press-release/press-release-detail/pfizer_to_acquire_array_biopharma | 2019 | reported $11.4B enterprise value; does not establish ARCH proceeds.
• ★★★ | EQRx 2022 10-K | https://www.sec.gov/Archives/edgar/data/1843762/000155837023001902/eqrx-20221231x10k.htm | 2023 | ARCH 36,335,375/7.4%; financing exposure.
• ★★★ | EQRx/Revolution final proxy | https://www.sec.gov/Archives/edgar/data/1843762/000119312523247498/d508095ddefm14a.htm | 2023 | alternatives/liquidation analysis/winddown; capital-transfer outcome.
• ★★★ | Codiak IPO prospectus | https://www.sec.gov/Archives/edgar/data/1659352/000119312521041423/d100259d424b4.htm | 2021 | ARCH/Steve Gillis/MD Anderson/Flagship formation and ownership.
• ★★★ | Codiak bankruptcy 8-K | https://www.sec.gov/Archives/edgar/data/1659352/000119312523082411/d470562d8k.htm | 2023-03 | Chapter 11/delisting.
• ★★★ | Codiak dissolution order | https://cases.stretto.com/public/x239/12139/PLEADINGS/1213906282380000000044.pdf | 2023-06-28 | court-approved dissolution and cancellation endpoint.
• ★★ | Fortune Codiak/ARCH loss report | https://fortune.com/2023/08/10/moderna-covid-vaccine-flagship-pioneering-biotech-investing-tessera-gene-editing/ | 2023 | later reported Nelsen estimate of $30M–$40M ARCH loss; unaudited, vehicle allocation unknown.
• ★★★ | Xcyte sale proxy | https://www.sec.gov/Archives/edgar/data/1130166/000119312506023924/d424b3.htm | 2006 | regulatory/program/financing history and $5M asset-sale endpoint.
• ★★★ | Receptos/Burow merger biography | https://www.sec.gov/Archives/edgar/data/1463729/000119312515266171/d51890dex99e21.htm | 2015 | Burow cofounder/ARCH role and Receptos attribution boundary.
• ★★ | Arivale postmortem | https://www.geekwire.com/2019/arivale-shut-doors-inside-surprise-closure-ambitious-scientific-wellness-startup/ | 2019-04-26 | >$50M, ~5K customers, CAC/assay/pricing/timing failure.
• ★★ | ARCH after Arivale closure | https://www.biocentury.com/article/301964/arch-to-continue-investing-in-science-driven-consumer-wellness-even-as-arivale-folds | 2019 | direct Nelsen business-model response.
• ★★★ | Rubius S-1 | https://www.sec.gov/Archives/edgar/data/1709401/000104746918004710/a2236053zs-1.htm | 2018 | Flagship founded/57.3%; no ARCH/Nelsen match despite profile listing.
• ★★★ | Rubius liquidation 8-K | https://www.sec.gov/Archives/edgar/data/1709401/000110465923024013/tm237246d1_8k.htm | 2023 | board-approved liquidation/dissolution.
• ★★★ | Unity dissolution 8-K/certificate | https://www.sec.gov/Archives/edgar/data/1463361/000119312525220778/d190494d8k.htm | 2025-09-26 | effective dissolution after clinical stop/all-employee reduction.
• ★★★ | Homology/Q32 merger filing | https://www.sec.gov/Archives/edgar/data/1661998/000119312524122748/d773213d424b3.htm | 2024 | program stop, 86% workforce cut, reverse-merger context.
• ★★★ | bluebird distressed tender material | https://www.sec.gov/Archives/edgar/data/1293971/000119312525116496/d941692dex99a5y.htm | 2025 | default/bankruptcy/liquidation warning if take-private failed.
• ★★★ | Sana 2024 10-K | https://www.sec.gov/Archives/edgar/data/1770121/000095017025040254/sana-20241231.htm | 2025 | program suspension, workforce reduction, liquidity risks.
• ★★★ | Sana Q1 2026 10-Q | https://www.sec.gov/Archives/edgar/data/1770121/000119312526216769/sana-20260331.htm | 2026 | current liquidity/litigation; allegations not Nelsen finding.
• ★★★ | Lyell IPO prospectus | https://www.sec.gov/Archives/edgar/data/1806952/000119312521193519/d168165d424b4.htm | 2021 | $17 IPO and ownership.
• ★★★ | Lyell reverse-split release | https://ir.lyell.com/static-files/3b301a87-ed3f-4ace-aea4-00aecbffd44a | 2025 | 1-for-20; $17 becomes $340 split-adjusted.
• ★★★ | Project Prometheus financing | https://www.axios.com/newsletters/axios-pro-rata-c007d2f6-d2d2-41cc-96e2-ea8c5b760f11 | 2026-06-11 | reported $12B Series B/$41B post; ARCH investor; private mark.
• ★★★ | Prometheus operating leadership transcript | https://pressroom.versantmedia.com/cnbc/press-releases/cnbc-exclusive-transcript-prometheus-co-founders-and-co-ceos-jeff-bezos-and-vik | 2026 | Bezos/Bajaj co-founder/co-CEO control; role conflict with ARCH self-claim.
• ★★★ | Prometheus Biosciences identity control | https://www.sec.gov/Archives/edgar/data/1718852/000119312521076931/d14529ds1a.htm | 2021 | no Nelsen/ARCH; RXDX/Merck excluded.

Forecast, adverse, legal, and comparison controls

• ★★★ | DOE Sapphire 2014 milestone story | https://www.energy.gov/cmei/success-stories/articles/eere-success-story-california-and-new-mexico-sapphire-energy-advances | 2014 | official ex-ante demo/5,000-bbl-day projection context.
• ★★★ | DOE commissioned Sapphire retrospective | https://www.energy.gov/sites/default/files/2021-06/2019%20-%20Meeting%20the%20Dual%20Challenge%20Vol%20III%20Chapter%209.pdf | 2019 | entity ceased by 2017/facility sold for pennies; primary commissioned outcome.
• ★★★ | ORNL BioEnergy KDF Sapphire record | https://bioenergykdf.ornl.gov/taxonomy/term/3 | current archive | facility shifted to algae nutrition/feed.
• ★★★ | Juno shareholder complaint | https://www.courthousenews.com/wp-content/uploads/2018/02/Juno.pdf | 2018 | allegations only; no finding inferred.
• ★★ | Forbes “Rx for Success” | https://www.forbes.com/sites/matthewherper/2016/05/11/rx-for-success/ | 2016 | deep profile, remembered outcomes/vintage figures; direct plus reported, not audited.
• ★★ | Peter Crisp completed peer | ../008-peter-crisp/sources.md | repo accessed 2026-08-22 | biotech/company-construction comparator.
• ★★ | Laurance Rockefeller completed peer | ../003-laurance-rockefeller/sources.md | repo accessed 2026-08-22 | patient science capital; family-capital difference.
• ★★ | Georges Doriot completed peer | ../001-georges-doriot/sources.md | repo accessed 2026-08-22 | institutional construction/outlier-control comparator.
• ★★ | Arthur Rock completed peer | ../006-arthur-rock/sources.md | repo accessed 2026-08-22 | concentrated governance versus proactive lab formation.

Coverage assessment

Inventory and access

• Rows: 102 source rows: 76 ★★★ / 22 ★★ / 4 ★. Of these, 101 are checked (76/22/3) and one ★ READOUT route remains mapped-only and unchecked. Mirrors and repeated HTML/PDF routes are rows for custody/access control but are not independent works.
• Priority parity: every accessible ★★★/★★ authored item in the bounded five-item writing corpus has a notes/w-*.md note and deep section in writings.md; every substantive ★★★/★★ appearance has a notes/t-*.md note and deep section in talks.md. Axios Re:Cap and READOUT LOUD are ★ mapped-only routes without a public content-grade transcript.
• Access gaps: complete TWST and Xconomy bodies, Long Run/BioCentury transcripts, Axios Re:Cap and READOUT content-grade transcripts, private-company cap tables, and deleted social archives remain unavailable. Partial sources are never represented as fully transcribed.

Denominators and attribution

1. Firm PDF: 321 printed rows/320 unique names, firm-wide, not Bob.
2. Current Bob page: 50 narrative names, 36 key names, union 55; self-attributed and curated.
3. Ownership: fund securities and shared voting/disposition are not personal checks/proceeds; reporting persons disclaim ownership beyond pecuniary interests.
4. Returns: transaction values, financing marks, and shares×price are not deal/fund/personal net returns.
5. Adverse: complaints are allegations; issuer failures stay company/team-level unless a source specifies Bob conduct.

Search saturation and unresolved record

• Queries covered Bob/Robert T. Nelsen, ARCH entities/partners, each current-union and deep-lab company, writings/letters/blog/books/social, every decade of interviews, podcasts/video/transcripts, fund/LP/IRR/DPI/TVPI, SEC ownership, lawsuits/controversy/criticism, mistakes/losses, and identity variants.
• Official firm, university, SEC/FTC/EC/CJEU/DOE, company, LP, major specialist press, podcast, video/caption, and completed-peer routes were exhausted. Final repeated searches produced duplicates, low-grade aggregators, or private-data gaps rather than a new claim-changing fact. Bounded public-web saturation was reached 2026-08-22.
• Still unavailable: IC memos/votes, complete opportunity/pass and loss ledger, full vehicle cash flows/fees/carry/PME, deal-level distributions, partner attribution, rejected companies, board minutes, and private cap tables.

Source notes