How Anthropic, Costco, and Patagonia all build incorruptible companies | Eric Ries

Key points

• **The Harvard Law stat that should make every founder file a PBC tomorrow:** 'Among venture-backed companies that have the standard best-practice governance docs you got from your lawyer, only 20% of founders are still CEO 3 years after going public.' Standard governance is statistical extinction for founder control. The 'five month CEO' story: hot pre-IPO company ignored Eric's warning, went public, competitor got acquired, stock dropped, founder ousted in 5 months — same bankers/lawyers/VCs who said 'wait, this is too early' kept all their fees.
• **'Harder is easier' as the leadership principle.** Cloudflare's SSL-encryption-for-free decision — junior engineer asks Matthew Prince 'if a better internet is an encrypted internet, why are we charging for it?'. Prince says 'let's figure it out', team hand-rolls assembly-language software to cut costs, conversion rates briefly drop, top-of-funnel 10x's, $70B company today. **'Trustworthiness is the most underrated asset in business. Things that create trust by definition stack-rank to the bottom if you do ROI — tangible costs, intangible rewards.'** Direct continuation of the [Issue 03 software 'melting iceberg' thread](/issues/2026-05-03) — quality is the moat AI cannot strip.
• **The 'spiritual holding company' framework.** Five flavours of mission guardian: (a) **Public Benefit Corp** (Anthropic, OpenAI, xAI — bare minimum, 2-page Delaware filing). (b) **Long-Term Benefit Trust** (Anthropic — outside AI-safety trustees, no equity, appoint directors). (c) **Industrial foundation** (Novo Nordisk — nonprofit owns the for-profit). (d) **Perpetual purpose trust** (Patagonia — non-economic trust + 'purpose protector' who can sue trustees). (e) **Employee ownership trust** (John Lewis, Alibaba voting trust). 'Companies with foundation structures are **6x more likely to live to year 50** vs conventional counterparts. Superior ROIC. Better in so many ways.'
• **Anthropic Long-Term Benefit Trust as a market-tested fortress.** 'Whenever you see Anthropic do the right thing — refuse to release a model, turn down the Pentagon's $200M contract — think about how much that's costing them. People say they do it for publicity. Publicity is nice. You know what's nicer? Having the number-one model that everyone has to pay you to use.' **Cross-reference to Krishna Rao on ILTB this week** ($30B+ ARR end-Q1, $44B run-rate per Altimeter) — Eric attributes Anthropic's commercial dominance directly to the LTBT mission protection that pulled the talent from OpenAI. **'If you ask why Anthropic is winning, people cite lower inference cost, faster velocity, focus — keep asking why and it ends at: people want to work for the good guys.'**
• **Vectura cautionary tale.** UK inhaler-therapeutics company (asthma, COPD), spun out of University of Bath, public on London Stock Exchange. **Actual Philip Morris** bid 165p/share. American PE firm bid 155p/share. They could have stayed independent. The board met twice, decided 'our hands are tied — fiduciary duty to accept the highest bid'. **PM paid £1.1B, took a $900M writedown 3 years later, disposed of the carcass for parts.** Eric: 'I only know this story because I'd been using the Philip Morris hypothetical for years and I asked Claude if it had ever actually happened. It had.' Delaware charters are public record — pull them.
• **The 'don't be evil vs the quarterly report' framing.** 13-year ex-Googler describes the loss of mission. Eric: 'What's the probability Google files its next quarterly report on time?' '100%.' 'Probability they accidentally kill someone and cover it up?' 'Well… 90%?' Why is the quarterly more certain than no-harm? **'Because there is a massive unbelievably expensive apparatus to make sure the quarterly happens every time. Don't-be-evil was a slogan. If someone says they're serious about something, show me the apparatus.'** This is the most ruthless line in the book — and the test every founder can run on their own company today.
• **Patagonia and Yvon Chouinard as the quality-zealot archetype.** 'He believed quality was an objective function — every product has a quality level it deserves. Most people think that's insane. That's why Patagonia is such a success.' Echoes Steve Jobs fighting over the layout of wires inside computers customers would never open. **Parallel to Dana White on Founders ('I don't have a Plan B', refuse-to-give-up-equity-in-Lower-Merrion-gym moment) and Costco's Saul Price ('customers first, employees second, shareholders last') as the through-line of founder-as-mission-guardian.** Three different founders, one shared trait: they would rather die than betray the principle.
• **The Vital Farms / BlackRock anecdote** as the in-flight example of financial gravity. 'It was on Twitter — everyone complaining the eggs got worse, full of toxins. Turns out BlackRock bought them.' **The pattern is so universal we don't have a name for it** — Eric calls it corruption (in the old sense, not embezzlement). 'You can taste the ownership structure of a company in the food.' Direct parallel to [Lemkin's 'terminal state of decay' on 20vc this week](/issues/2026-05-17) — but Eric's frame is structural (governance docs), not market (AI displacement).
• **The 3 things to do this week if you're an early-stage founder:** (1) **File as PBC** in Delaware — write a mission you'll feel good about, adversarially prompt with co-founder: 'can I make money while violating this?' If yes, write the constraint in. (2) **Director's oath** — write into the corporate charter that every board member takes an oath as a precondition of joining. 'Nurses have a hippocratic oath. Directors make far more consequential decisions than nurses.' (3) **Mission-protective provisions on top of founder-preferred shares** — 'Anthropic-style: 10% of equity pledged to a non-profit foundation, 1% of future revenue, foundation gets a board seat. Write it into the charter now. Fire and forget. Boot it up later.'
• **Vatican AI-governance panel anecdote.** Eric on a panel with Anthropic, OpenAI, Google, Cohere, Palantir. 'I looked down the row and realised **not a single one of these companies has standard governance.** That's how bad it is. Nobody making this technology would say standard governance is fine — it's just too dangerous.' Footnoted with the concept of 'emergent intelligence' — Conway's law (org chart visible in architecture) + ant-colony piano-puzzle experiment (more ants → faster solution; more humans → worse, unless aligned). **'If we can't agree on the human values to align AI to, we're already cooked.'** The strongest meta-frame on this week's tape for the AI-governance crowd.
• **The Mary Parker Follett 'invisible leader' final.** Follett (1920) — 'the leader's job is to create more leaders'; 'we focus on power-with, not power-over'; 'superior and subordinate together obey the law of the situation'. **'The most consequential decisions in any organisation are by definition made when no manager is present.'** If you don't cultivate common purpose, your promises are worthless because you're not the one in the room when the designer chooses rounded vs straight corners. **'Why is trust collapsing in our institutions across the board? We are teaching a leadership philosophy that is anti-trustworthy.'** Pair with the Issue 04 Drucker-ish thread on management discipline.

Notable quotes

Themes

• Standard VC governance = statistical extinction (20% founder retention 3yrs post-IPO)
• Public Benefit Corp + Long-Term Benefit Trust as Anthropic's structural moat
• Harder is easier — trust is the unpriced asset all ROI-thinking destroys
• Mission guardian as the design pattern for incorruptible companies
• Vectura as the cautionary tale of fiduciary duty forcing value destruction