Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)
TL;DR
Hiring is the company strategy, not an HR function — Keith Rabois says Vinod Khosla’s line at Square still holds: “the team you build is the company you build,” and a founder who can assess talent “ruthlessly and accurately” can go very far even without many other skills.
Most startups add headcount when they actually need more “barrels” — His barrels-and-ammunition framework argues that only a tiny number of people can independently drive an initiative from idea to outcome, so piling on more support staff just increases coordination tax unless you increase the number of true owners.
AI is collapsing traditional role boundaries, especially for PMs — Rabois thinks the classic product manager role “makes no sense in the future” because roadmaps change weekly, and the enduring skill across PM, design, and engineering is basically CEO-like business judgment: what to build, why, and how it moves the business.
The best AI adopters aren’t always engineers — One of his more surprising observations is that at some top companies the biggest token consumer is the CMO, because intellectually curious operators can now ship analytics, campaigns, and drafts themselves instead of relying on layers of deputies.
Customer feedback is often worse than useless in consumer products — He says talking to customers for consumer, SMB, and micro-merchant products is frequently directionally harmful because people rationalize subconscious behavior, whereas enterprise customer development works better when there’s a real decision-maker and a defined buying process.
High-performance cultures trade comfort for clarity and speed — Rabois defends public criticism, rejects psychological safety as a core principle for elite teams, and argues that the better a company is doing, the harder the CEO should push because talented people get demoralized by coasting, not intensity.
The Breakdown
Keith opens with the most Keith detail possible: he hasn’t used a computer since 2010
Rabois starts by casually revealing that he’s run his life from an iPad, phone, and watch since joining Square in September 2010, after seeing Jack Dorsey do the same. It’s a funny flex, but also a tell: he likes lightweight tools, hates drag, and sees the phone-plus-AI future as something he’s been accidentally training for for years.
The real job of a founder is building talent density
He says the deepest lesson from Square and PayPal is simple: the team is the company. PayPal worked because Peter Thiel and Max Levchin created a ridiculous density of talent, and Rabois admits he was initially mediocre at hiring strangers — his breakthrough was realizing he could identify talent once he had context, then building systems like ruthless references to get that context faster. His favorite examples are Tony Xu doing 20 references per senior hire and David Sze’s old Greylock rule: don’t stop referencing until you hit a negative.
“Barrels and ammunition” explains why hiring sprees disappoint founders
This is one of the sharpest sections: most CEOs hire a lot after raising and then wonder why output barely moves while burn explodes. Rabois’s answer is that very few people are true “barrels” — people who can take an objective, go “over that hill,” and just get it done — while everyone else is ammunition. He makes it vivid with the Square smoothie story: an intern, Taylor Francis, solved a standing 9 p.m. cold-smoothie problem on day two when bigger teams had failed, and Rabois instantly thought, “I found a barrel.”
How to attract killers: don’t compete for the obvious ones
Yes, mission matters, but his more nuanced point is that great people join when they believe their specific superpower maps to the company’s biggest blocker. That’s how Square pulled him away from Google in two weeks: investors told him he was one of only a few people who understood both financial services and entrepreneurship. More broadly, he thinks startups should build on “undiscovered talent,” because chasing the same proven stars as everyone else is a salary-cap game startups usually lose.
Great CEOs push harder when things are going well, not badly
Rabois leans hard into the “bar raiser” reputation and says the CEO’s job is the relentless application of force against complacency. He flips the usual intuition: when companies are struggling, he becomes supportive; when they’re thriving, he gets more critical, because winning is exactly when teams start to coast. He and Lenny connect this to Brian Chesky — momentum feels good, but the best people actually get unhappy when everyone starts skating.
AI is rewarding curiosity and turning PM into a CEO-like skill set
His most talked-about AI take is that the traditional PM role breaks down when foundation-model capabilities are changing month to month and annual roadmaps become absurd. What survives is not the title but the skill: business acumen, judgment, and the ability to decide what to build and why. He says the strongest signal he sees in top organizations is often the CMO devouring the most tokens, because curious executives now produce analytics, campaigns, and work product directly rather than waiting on a chain of teams.
Design, engineering, and demos are all merging into one execution loop
Rabois thinks design and code are converging, though he’s not sure whether code absorbs design or vice versa. He points to Shopify’s rule that product presentations can’t be slides — they have to be working demos — as evidence that static specs are dying. Even while saying the functions are merging, he still believes the alpha increasingly lives in storytelling and design taste, because cutting through the clutter matters more as AI makes building easier.
His spiciest management takes: stop talking to customers, criticize in public, and forget psychological safety
He says customer conversations in consumer and SMB products are often actively harmful because people can’t accurately explain subconscious choices; DoorDash and Airbnb worked because founders had a strong underlying insight, not because surveys told them to exist. Then he doubles down on culture: criticism in public helps the whole system learn, while private feedback optimizes for the individual, and “high performance machines don’t have psychological safety — they’re about winning.” He closes with a very on-brand life motto — “no days off” — and a workout streak he tracks with almost venture-level obsession.