The Head of Claude Code Says Coding Is Solved. Is Knowledge Work Next?

In Brief

The head of Claude Code at Anthropic says he hasn't written a single line of code by hand since November 2025. The implication is that all of knowledge work will be next.

What Happned

Boris Cherny, the head of Claude Code at Anthropic, sat down for a wide-ranging interview on Lenny's Podcast this week and made a blunt claim: coding is effectively solved.

Cherny said he hasn't edited a single line of code by hand since November 2025. Every line has been written by AI. He built Claude Code as a side project at Anthropic in September 2024. Within five days of releasing it internally, half of Anthropic's engineering team was using it.

Engineering output at Anthropic has increased 200% per engineer. The product is now generating a billion dollars in annual run rate revenue. Four percent of all public GitHub commits, or changes in a file, are authored by Claude Code, and Cherny predicts that number will hit 20% by end of 2026. 

But the part that matters most isn't about engineering. On the Claude Code team, everyone codes, including the product manager, the engineering manager, the designer, the finance person, the data scientist. Cherny predicts that by end of year, the title "software engineer" will start to go away. He said it's going to be painful for a lot of people.

The Key Numbers

$1B annual run rate — Claude Code's current revenue, up from a side project built 18 months ago.

4% of public GitHub commits — Number of commits authored by Claude Code today. Cherny predicts 20% by end of 2026.

200% productivity increase — Productivity increase per engineer at Anthropic since adopting Claude Code.

8 months — The time it took Cherny to go from 20% AI-assisted work to 100%.

Why This Is Really About Knowledge Work

Coding is only part of the story. "Replace code with whatever you do," says Marketing AI and SmarterX founder and CEO Paul Roetzer on Episode 200 of The Artificial Intelligence Show. "Every time he says code, replace it with the things you do, the tasks that you do for your living. Because that's what he's implying. Whatever we just did with coding is now going to come to the rest of knowledge work."

The speed should get your attention. "This is not like a multi-year thing," Roetzer says. "Within eight months it went from 20% of his work to 100% of his work."

The trajectory is the roadmap. Anthropic's development path is explicit: coding, then tool use, then computer use. Roetzer breaks it down: "Coding means it's good at writing code. With tool use, it doesn't just rely on the language model itself for the outputs. It accesses other things like the internet, like search. And then computer use is seeing everything on your screen digitally and then being able to act in a digital way. Agents that can do things on the screen just like you and I would."

That's the sequence that brings AI from a coder's tool to something that touches every desk job.

The models themselves are changing the game. "I'm increasingly seeing people within the labs talking about the fact that these models are functioning as innovation partners," says Roetzer. "They're actually starting to come up with and solve problems." Claude isn't just generating code to specification. It's also analyzing feedback, bug reports, and telemetry to suggest what to build next. That's product strategy, not coding.

What's my token budget? Cherny's advice to engineering leaders is to give teams as many tokens as possible and optimize after, not before, you've proven value. Roetzer says Cherny talks about coders who start to work at labs asking: "What's my token budget? How much access to intelligence do I have?"

"I could see in other industries by the end of 2026 that becomes a regular conversation: 'What is my agent budget?'"  

— Paul Roetzer, founder and CEO of SmarterX

Cherny compared the current moment to the printing press. Before printing, scribes spent their time copying books. As literacy spread, their core skills remained relevant but their daily tasks transformed entirely.

He predicts a similar shift where "software engineer" gives way to "builder," and the people who thrive are generalists who can combine technical fluency with design, business, and strategic thinking. 

Despite this dramatic shift, Anthropic is still hiring. "But these people are 4x more productive than they used to be," says Roetzer. The demand has simply expanded to match the new capacity.

SmarterX Take

The pattern here is clear: AI masters a domain, the people in that domain resist, and then the transition happens faster than anyone expected. Coding just happens to be first because code is the most structured, measurable, verifiable form of knowledge work. It won't be the last.

The real question for every professional isn't whether their work will be affected. It's how much and how soon. Will AI assist their work by 20% or 100% eight months from now? And is that a choice they'll make or will it be made for them.  

What to Watch

• Cherny's 20% GitHub prediction as the near-term benchmark. If Claude Code hits that by the end of 2026, it validates the exponential adoption curve and signals the tool use and computer use phases are next.

• "Token budget" and "agent budget" to enter mainstream corporate vocabulary. When CFOs start asking about per-employee AI spend the way they ask about software licenses, the shift from experimentation to infrastructure is complete.

• Job titles. If "builder" starts replacing "software engineer" at major companies this year, the role consolidation Cherny predicts is underway.

Resources

Lenny's Podcast: Head of Claude Code Interview → podcasts.apple.com

What Happens After Coding Is Solved? — Lenny's Newsletter → lennysnewsletter.com

Will Claude Destroy Software Engineer Jobs? → fortune.com

What Happens After Coding Is Solved — Roger Wong Summary → rogerwong.me

Claude Code Creator's 3 Principles for Every Team → dnyuz.com

8 Game-Changing Insights from Boris Cherny → waydev.co

Heard on The Artificial Intelligence Show, Episode 200
Boris Cherny, head of Claude Code at Anthropic, claims coding is solved, and the implications extend far beyond engineering to all knowledge work. Listen →