成功把公司变成靶子：Eric Ries 讲怎样做成不可腐蚀的企业

Host Lenny Rachitsky opens by noting that today's top AI labs ship the way The Lean Startup prescribed: a small research preview, then iteration. Ries is glad to hear it but says the point was never the vocabulary. The deeper lesson is that nobody can predict the future, and that ChatGPT, Claude Code and similar products began as small experiments whose popularity surprised their own makers.

His new book, Incorruptible, is about the other half of the job: protecting what you have built. Ries describes a force "that no one controls but everyone obeys" that drags organizations toward mediocrity until their founders lose control, either by being ousted or by watching the company turn bureaucratic or malign. He says he has watched this for years while helping people build companies.

To make it concrete, he describes tasting private-equity ownership in a restaurant meal, and hearing the same pattern in a natural-foods brand whose founder was pushed out for growth and margins. Quality falls, customers and employees get angry, and market share shrinks. He calls this pattern corruption in the old-fashioned sense, comparable to a bridge whose bolts have rusted, and argues the answer is the organizational equivalent of stainless steel rather than a shrug at "financial gravity."

Ries makes what he calls a radical claim: if you get this wrong, no later decision matters because you will not be the one making it. He cites Harvard Law School data that among venture-backed companies with standard best-practice documents, only about 20% of founders are still CEO three years after going public. He recounts advising a hot company before its IPO, which ignored his warnings after its bankers, lawyers and investors called him a downer, and whose founder was ousted five months after listing.

He also pushes back on the idea that thriving companies without these protections prove they are unnecessary. He says Cloudflare and Costco, which people assume are ordinary, in fact use at least one of the governance structures he describes. His advice is to check the math before concluding you can be the exception.

To the objection that he should first find product-market fit, he answers that the key question is when protections are put in place, not which ones. Every adviser along the way, from lawyer to VC to growth investor to banker, says it is too early, until the CFO tells the founder at IPO time that it is too late. Success, he says, does not protect you; it makes you a target.

Ries's blueprint is ethos plus integrity: internal alignment of character and choices, plus a structure that keeps the company aligned with human flourishing. He illustrates it with 1920s Denmark, where August and Marie Krogh brought insulin technology back from Canada. Worried that investors might one day exploit patients who depended on the drug, they set up the company as a for-profit owned and governed by a nonprofit foundation.

That company became Novo Nordisk. Ries says the structure has protected its scientific ethos for more than a hundred years, and that on one occasion the foundation's trustees intervened against a sale in a way that ultimately created more than $500 billion of shareholder value. The German optics company Zeiss, he adds, had a similar structure in 1885.

He cites research showing that companies with these structures are six times more likely to reach year 50 than conventional peers, with superior return on invested capital. His point is that financial gravity is not inevitable, since there is an actual dataset of exceptions, and that today's "best practices" are younger than the trees in a local park.

Ries asks founders to name the most evil company they can think of, using Philip Morris as his example, and to imagine it offering to buy their company for slightly more than it is worth. Most refuse instantly, he says, yet the standard charter they signed imposes a fiduciary duty to say yes. Founders are outraged, and when they call their lawyers, the lawyers confirm it.

The real case was Vectura, a UK inhaler company spun out of the University of Bath. Philip Morris bid 165 pence per share against a private-equity bid of 155 pence, and the public and the British Thoracic Society begged the board to refuse. The board said its hands were tied by its duty to accept the highest bid.

Philip Morris paid £1.1 billion, took a $900 million write-down within three years and disposed of the company for parts. Ries tells founders that Delaware charters are public records and anyone can read what a company has really committed to. He says the stakes are rising as technology becomes powerful enough to have planet-scale consequences.

Ries's first leadership principle, available to anyone with any influence, is that business may be hard partly because nobody trusts you. Employees who trust you need less communication and align more easily, and customers who trust you cost less to acquire and forgive mistakes. He calls trustworthiness the most underrated asset in business, but says it only works if you pursue the principle for its own sake, not for the reward.

His example is Cloudflare, whose founders disliked consulting-speak and had no mission statement for years. Engineers kept saying they were trying to "make a better internet," and that eventually became the mission, with "be principled" as the top value. He stresses that the mission statement is not the mission; the mission emerges from how the company behaves, and quality has to be built in from the inside.

Then a junior engineer asked why Cloudflare charged for web encryption, its top driver of upgrades, if a better internet would be an encrypted one. CEO Matthew Prince answered with "let's figure it out," and the team found ways to give it away sustainably even as premium conversion dipped. Ries says that trust is a large part of why Cloudflare is worth about $70 billion today. By contrast he describes Groupon, where Andrew Mason was ground down from one daily email to two and eventually eight by ROI arguments.

Ries avoids consultant vocabulary and asks a plain question: what is your purpose, and who would you rather die than betray? Examples include Steve Jobs obsessing over unseen wiring, Patagonia's Yvon Chouinard treating quality as an objective function, and Devoted Health telling staff to treat every customer like their own parents. He warns that purpose matters at the product-manager level too, citing Johnson & Johnson's baby powder, where the stated mission of patient health gave way to growth and quarterly targets.

He distinguishes mission-driven companies from merely mission-hopeful ones. A mission-driven company cannot profit except by achieving its mission, so you should ask whether anyone on a team could hit bonus targets or OKRs by betraying a principle. Safety, performance, quality and design are the things cut first because trust means a betrayal can go unnoticed, and he notes AI now offers new ways to audit for them.

To make the point, he describes talking to a longtime Google employee who put the chance Google files its next quarterly report on time at 100% but gave high odds that it might accidentally harm someone and cover it up. The difference, Ries says, is an enormously expensive apparatus guaranteeing the report. "Don't be evil" was only a slogan, and when someone says they are serious, you should ask to see the apparatus.

Ries notes that integrity has two meanings, being reliable and being structurally sound, and in an organization they become one. A weak organization cannot keep promises because the person who made them will not be there. The goal is to resist temptation from inside, betrayal at board level and pressure from outside, as Costco's governance fortress does when critics complain that it spends shareholders' money on customers.

The starting point is your charter, whose standard line says a company is formed to pursue any lawful act. Under shareholder primacy, he says, that effectively means maximizing shareholder returns, an idea only about 40 years old. In the 19th century a company had to declare a public benefit, and taking one over to change its purpose to maximizing value could have cost it its charter.

The fix he calls the easiest on the whole podcast is the public benefit corporation, a two-page Delaware filing that states a specific purpose. All the major AI labs use it, Anthropic most famously. It does not make you the good guys, but it answers a lawsuit alleging breach of duty to investors, because investors agreed to the purpose.

Asked about downsides, Ries says the only cost is people wasting your time trying to talk you out of it, and the book includes scripts for talking to investors and lawyers. Lawyers advise keeping your options open, which he mocks by asking about the option of turning customers into Soylent Green. Anthropic's success, he adds, shows that a public benefit structure does not stop you raising money or growing fast.

He traces the usual explanations for Anthropic's lead, such as lower inference costs, faster product velocity and focus, back through repeated whys to talent that wants to work for the good guys. That mission only lasts because they paired ethos with integrity. Lenny adds that the head of product for Claude Code told him mission alignment makes decisions fast.

Ries describes mission-aligned companies as the organizational equivalent of a flow state, where nobody needs to build a spreadsheet to defend doing the right thing. He also teaches the culture bank, a rule he credits to Todd Park, who learned it from Howard Schultz: whenever someone sacrifices for the company's values, that is a deposit, while a greedy move is a withdrawal. The rule is to only make deposits, with the H-E-B manager who let customers take groceries home during an ice storm as his example.

Ries says his role in the Anthropic story was minor, and credits Dario and Daniela Amodei and their team. When they left OpenAI, Dario was a first-time founder, no top venture fund wanted in, and generative AI had not yet boomed, yet they wanted to protect their safety mission. They became a public benefit corporation from the start and wrote into their charter the right to add reforms, which they defended for about two years until the Long-Term Benefit Trust arrived with their Series C.

Today Anthropic has directors appointed by, and accountable to, outside trustees who are AI safety experts without equity. Ries argues that refusing to release a model, or turning down a $200 million contract at the risk of angering the government, is costly and made possible partly by a structure investors cannot easily override. He says the structure is better than founder control, and that every AI company on a Vatican panel he joined has non-standard governance.

The critical need, he says, is a mission guardian, either a person or a thing. Founder control can be a temporary bridge but wears founders down, while permanent options include employee ownership trusts, cooperatives, nonprofit foundations like Novo Nordisk's, and perpetual purpose trusts such as Patagonia's, which also add a purpose protector who can sue trustees. He groups all of these as "spiritual holding companies" and says the evidence shows they are more durable and better for shareholders.

Ries recounts attending a celebration for a friend who had been ousted as founder, a company-wide event that felt like a wake. His point is that if a company can be decapitated at any time, nobody can trust its promises, and that this philosophy erodes trust in institutions broadly. He says he writes books slowly, only about practices tested across hundreds of companies, and has nothing to sell.

For early-stage founders, he says the moment before raising a priced round is precious because you can do anything. Become a PBC and write a mission by brainstorming adversarially with your co-founder: can you think of ways to make money while violating it, and would you be sad? He also suggests a director's oath, like the Hippocratic oath, written into the charter and required of board members.

Third, use founders' preferred shares and mission-protective provisions to get founder control, extra board votes or similar rights, and talk to a lawyer about them. He mentions that he helped start a law firm called Virgil that does not charge by the hour. He wants companies to be mission-controlled rather than investor-controlled or founder-controlled, and says you can pledge equity and a board seat to a future nonprofit in the charter now and set it up later.

Lenny points out the parallel with AI alignment, and Ries says it is no coincidence: the unsolved problem is who aligns the aligners. Through Conway's law, an organization's imprint shows up in the software it builds. He calls corporations the oldest form of artificial intelligence, emergent super-organisms.

He illustrates with a research video of ants solving a piano-movers puzzle: one ant cannot, but a thousand together appear to reason, pausing and reorienting like a human. Adding ants speeds the solution, while adding humans makes it worse unless they are carefully aligned. His lesson is that organizations you do not tend develop emergent traits you do not like, and founder mode cannot fix what is in the DNA.

His closing recommendation is Mary Parker Follett, a contemporary of Frederick Winslow Taylor, whose 1920s ideas about power with rather than power over and leaders creating more leaders sound current. Her concept of the invisible leader holds that the most consequential decisions are made when no manager is present, so a common purpose has to be the leader. Without it, he says, your promises are worthless.