Eric Ries’s new book Incorruptible hit shelves this month, published by Authors Equity, and the Hacker News AMA thread accompanying it has already racked up more than 750 points and 500 comments — the kind of reception reserved for someone who genuinely shaped how a generation builds products. The book’s premise, as Ries himself put it in a Product School interview on May 26, is clear: not how to build a great company, but how to keep it that way. Structural forces, not moral failings, cause successful enterprises to betray their founding values.
The argument is well-researched and, by early accounts, compelling. Ries traces how incentive structures warp governance, how growth demands corrupt mission, and how Costco’s Jim Sinegal and FedMart’s Saul Price each cracked a different half of the longevity problem. He even spotlights Anthropic’s Long-Term Benefit Trust — a purpose trust engineered to legally lock safety commitments into corporate DNA — as one of the few modern attempts to solve this.
What nobody in the AMA thread seems willing to say out loud: Ries is trying to solve a problem his own biggest idea helped create.
The Lean Startup Did Not Invent Iteration. It Invented the Permission Structure to Kill Moonshots.
When The Lean Startup arrived in 2011, it landed in a corporate landscape that was already addicted to quarterly earnings. What Ries offered was a methodology — build, measure, learn — that made small-batch experimentation legible to finance. That was the book’s genius. It translated entrepreneurial intuition into something a CFO could put in a spreadsheet.
The downstream effect, fifteen years later, is not that more startups succeed. It’s that large companies stopped doing the kind of R&D that doesn’t validate in eight weeks. The minimum viable product framework is brilliant for testing whether customers want a feature. It is useless for testing whether a novel transistor architecture, a fundamentally different drug delivery mechanism, or a safety protocol that adds friction today but prevents catastrophe in a decade is worth pursuing. Those things don’t have MVPs. They have decade-long cash burns and uncertain endpoints.
What happened instead: corporate R&D labs — the Bell Labs, the Xerox PARCs, the IBM Research divisions — didn’t disappear all at once. They got lean-startup’d into a thousand small innovation sprints, each one optimized to deliver a validated learning report to the C-suite by end of quarter. The work that survived was the work that could demonstrate traction on a time horizon shorter than the tenure of the average VP.
One former director of a now-shuttered advanced research unit at a major semiconductor firm told me last fall, standing in the parking lot after his exit interview: “We used to pitch five-year roadmaps. After lean came in, anything over eighteen months was called ‘waterfall thinking.’ The phrase was a slur.”
We Spent a Decade Optimizing for Learning Velocity. Long-Horizon Trust Spent That Decade Atrophying.
Ries’s new book focuses on corporate governance — the long-term benefit trust structure, board composition, charter provisions. These are real levers. But they operate on the assumption that once a company’s governance is aligned around mission, the company will naturally do the difficult, long-horizon work that mission demands.
That assumption is missing a step. Even a perfectly governed company cannot do long-horizon R&D if it has systematically eliminated the institutional muscle for it. And the lean movement — not alone, but as the most influential philosophy of the past fifteen years in product development — did exactly that. It taught a generation of managers that any investment that can’t be validated in a two-week sprint is waste.
The result is visible in the numbers. U.S. corporate basic research spending as a share of GDP has been flat to declining for most of the last decade, even as R&D spending overall has climbed — a shift away from fundamental inquiry toward near-term applied development. Companies are spending more on R&D than ever and getting less foundational science out of it. The pipeline isn’t empty because governance is corrupt. It’s empty because the tools we gave managers to fight waste were also tools to kill patience.
The Book’s Real Blind Spot Is Its Silence on the Pipeline It Inherits.
There is a genuinely interesting idea in Incorruptible: that mission primacy, legally enforced through structures like purpose trusts, can outlast founders. Anthropic’s Long-Term Benefit Trust is the poster child here — an attempt to build a governance firewall around safety commitments that shareholder pressure can’t override. It is a noble experiment, and it may even work for a company whose product is a large language model, where safety is a property of the system’s design that engineers can iterate on directly.
But for most of the economy — for the semiconductor firms, the materials science companies, the pharmaceutical pipelines — the problem isn’t that governance got corrupted. It’s that the methodological revolution Ries led convinced everyone that the only legitimate way to build is in small, validated increments. The structural corruption Ries now diagnoses is downstream of a cultural shift his own book accelerated.
A junior engineer who recently left a major AI lab put it to me in a Slack DM this week: “We have a governance structure that’s supposed to protect safety. But our actual development process is still two-week sprints. Nothing in the governance doc addresses that.”
Ries deserves credit for noticing that something is broken and for doing the serious institutional work of designing solutions. But Incorruptible reads, from the excerpts and interviews available, like a book that treats the preceding fifteen years of lean management as background condition rather than contributing cause. That’s not an error of analysis. It’s an error of self-awareness.
If you spend a decade teaching the world to value speed, learning, and validated iteration above all else, you don’t get to act surprised when the institutions that emerge from that decade are bad at patience.