Eric Schmidt Says AI Coding Tools Now Beat Him at the Job That Made Him Rich

Eric Schmidt spent decades being one of the best programmers in Silicon Valley before he ran Google. Now, at 71, he says the machines have passed him.

"I was a very, very good programmer when I was 22, right? I was one of the best, according to myself," Schmidt told Blackstone's "Inside Blackstone" podcast, hosted by Christine Anderson. "My field is over, right? I look at the code it writes now. I can't write that code. I couldn't when I was 22, right? So, I got beaten, right? Hard."

That's a blunt admission from a guy who built his reputation on technical chops. But Schmidt isn't mourning it. He's describing a programmer at one of his startups who, according to Schmidt, defines a task every evening at 7 PM, writes a test function to judge whether the AI's output is any good, hits run, and goes to bed. The AI grinds through the night. By morning, the work is done.

"This stuff would've taken me six months and 10 programmers at Google," Schmidt said. "This poor guy's sleeping."

The best engineers get MORE valuable, not less

Schmidt's actual argument runs against the obvious doom-and-gloom take. He says top-tier engineers have always been worth roughly 10 times more than the next tier down, and that gap is about to widen, not shrink. The grunt work of actually typing code is disappearing. What's left is framing the problem correctly, setting the right constraints, and knowing when an output is good enough to ship.

"If you can define the evaluation function and you can let it run, and if you have enough hardware, you're inventing worlds," Schmidt said.

He still tells young people to learn to code. Not so they can grind out boilerplate, but so they can give AI systems precise, accurate instructions. His analogy: he used to have to build the house himself. Now he architects it and tells Claude Code or OpenAI Codex to do the construction. "The next generation of people like me will be the architects," Schmidt said.

Schmidt isn't alone in making this claim, and that's worth taking seriously even though the sources making it have skin in the game. Ivan Burazin, CEO of a 30-person AI startup called Daytona, told Business Insider that each of his 16 engineers manages an average of five AI coding agents, and literally nobody at the company writes code by hand anymore.

Anthropic's own CFO, Krishna Rao, made a similar claim on the "Invest Like the Best" podcast, saying AI now handles much of the execution layer at his company, including some financial statement production. "90 plus percent of our code is actually written by Claude Code," Rao said. "Everyone kind of becomes a manager."

Rao runs the finance side of a company that sells the exact AI coding tool he's praising. Schmidt has venture investments across the AI sector. Burazin's entire business model depends on AI agents replacing human coders. None of that makes their claims false, but none of these are neutral bystanders describing a trend from the outside. They're interested parties with money riding on the narrative that AI coding tools are this good.

Individual anecdotes from named executives at named companies describe their own internal workflows. What remains unverified is whether a six-month, 10-engineer Google project genuinely compresses into one overnight run with the same quality, security, and reliability a large company like Google would demand before shipping to a billion users. Schmidt didn't name the startup, the project, or provide code for inspection. The claim stands on his word alone.

The broader jobs question is still unresolved. If the elite 10% of engineers become 10 times more valuable while AI eats the routine work, what happens to the other 90% of coders who aren't architecting evaluation functions for a living? Neither Schmidt nor Rao addressed what happens to the junior programmer who used to cut his teeth writing the boilerplate that AI now writes in minutes. That's the question hiring managers, computer science departments, and junior engineers themselves are going to have to answer in the next few years, not the AI companies selling the tools.