Already rich, already successful, why the last wave of tech winners is grinding again

Tom Blomfield co-founded Monzo. He spent four and a half years at Y Combinator picking winners. On Monday he announced he's taking a leave of absence to join Anthropic's compute team — not as an executive, not as an advisor, but as a member of technical staff. The title is deliberate. It's the same flat label Anthropic and OpenAI slap on nearly everyone who writes code, regardless of whether they've built a billion-dollar bank or just graduated Stanford.

Mike Krieger co-founded Instagram. He's now Anthropic's Chief Product Officer. Andrej Karpathy helped found OpenAI, ran AI at Tesla, started his own lab, then folded it to join Anthropic's pre-training team. Peter Bailis spent less than a year as Workday's CTO, overseeing AI strategy across an eight-billion-dollar revenue business, before trading the corner office for the same technical staff badge. Chamath Palihapitiya, who hasn't held an operating role since leaving Facebook in 2011, just took the CEO seat at his own AI coding startup, 8090 Labs, with a hundred thirty-five million dollar Series A led by Salesforce Ventures. Eric Wu ran Opendoor for a decade, stepped back in 2023, and launched NavigateAI with twenty-five million in seed funding. He told me directly: "I knew if I looked back in ten years and didn't do something related to it, I would probably regret that."

Regret is the operative word. Not greed. These people have more money than they can spend. They have status, networks, the luxury of saying no to anything. They're choosing the grind because the alternative — watching from the stands while the defining technology of their lifetime gets built by someone else — has become intolerable.

The hierarchy collapse is real. At Anthropic and OpenAI, a founding member of OpenAI and a twenty-six-year-old PhD share the same business card. The title "member of technical staff" strips away the signaling machinery that Silicon Valley has spent decades perfecting. No managing director. No vice president. No principal engineer. Just the work. Blomfield isn't slumming it. He's signaling that the only credential that matters now is proximity to the model.

This cuts against the valley's instincts. The standard playbook for successful founders: become an LP, join boards, advise, angel invest, maybe start a fund. Stay close to the action without getting your hands dirty. The new cohort is rejecting that playbook. They're not allocating capital. They're allocating themselves.

Palihapitiya's move is the outlier that proves the pattern. He's not joining someone else's lab. He's building his own, but he's doing it as a full-time operator — the first time in over a decade. His language: "there was no decision to make except to be all in." The SPAC king, the boardroom philosopher, the guy who made his name financial-engineering outcomes, is writing code reviews.

The fear is specific. Not AI doom. Not AGI timelines. The fear is that the next three to five years will settle the architecture of intelligence for decades, and the people who-trained the foundational models will write the rules everyone else follows. Karpathy put it plainly: "the next few years at the frontier of LLMs will be especially formative." He didn't say profitable. He didn't say impactful. He said formative. The word choice matters.

Critics will call this FOMO dressed in mission language. They'll note that Anthropic and OpenAI pay extraordinarily well, that "member of technical staff" at these labs commands compensation packages that would make a Series B founder weep. They'll argue that risk profiles change when you have generational wealth — that Blomfield can afford a leave of absence because failure costs him nothing. True. But the easier path for a wealthy founder is the board seat. The speaking circuit. The memoir. The harder path is the daily standup, the code review, the on-call rotation, the anonymity of a flat title. They're choosing the harder path.

Wu's construction copilot is not a vanity project. Construction is a trillion-dollar industry that runs on clipboards and phone cameras. An AI that actually works there — that reads blueprints, catches rebar errors, sequences pours — changes how the physical world gets built. That's not a toy. That's infrastructure. Wu spent a decade digitizing home buying. He knows what adoption curves look like. He's betting the next curve is steeper.

The pattern suggests a shift in how the valley's elite perceive their own agency. For fifteen years, the winning move was to build a platform, capture a network effect, extract rent. The AI wave doesn't settle. It accelerates. The models improve monthly. The context windows expand. The reasoning deepens. There is no moat to capture. There is only the frontier. And the frontier moves at the speed of the next training run.

Blomfield's leave of absence is temporary on paper. Y Combinator says he'll return. But the signal is permanent. When the people who built the last generation of winners decide the only place worth being is the compute team at the lab training the next generation, the rest of the ecosystem should pay attention. They're not investing in the wave. They're swimming in it.