Lenny's Podcast

Claude Code head Boris Cherny: Why he ships 30 PRs a day

Through hundred-percent AI-written code and parallel running agents on autopilot; 'clodify everything,' unlimited tokens, and latent demand make the builder.

BCBoris ChernyguestLRLenny Rachitskyhost

Feb 18, 2026·1h 27m·Watch on YouTube ↗

🙋 FREQUENTLY ASKED QUESTIONS

Direct answers grounded in the episode transcript. Tap any timestamp to verify against the source.

1. 1

   Why is Claude Code a terminal app instead of an IDE?

   Claude Code stayed terminal-first because that was the fastest form factor Boris Cherny could keep improving. The first version was a solo prototype called Claude CLI, built in the terminal because it was the easiest way to ship when the project was just him. Anthropic later considered other form factors, but Boris says the model was improving so quickly that a heavier interface could not keep up. The terminal also put the tool where engineers already worked, while still feeling unusual enough that users had to be open-minded. Early internal reaction was tiny, then daily active users at Anthropic went vertical, and the product eventually expanded into iOS, Android, desktop, the website, IDE extensions, Slack, and GitHub. The lesson he draws is to under-resource early products a bit and keep the interface flexible enough for the model's rapid progress.

   ▸ 6:42 in transcript
2. 2

   How does Claude Code come up with PR ideas?

   Claude Code gets PR ideas by reading the same feedback streams Boris Cherny used to watch manually. He says the simple workflow is to open Claude Code or Cowork and point it at a Slack thread, especially Anthropic's internal Claude Code feedback channel. In the early days, Boris personally fixed feedback within a minute or five to make people feel heard and keep the feedback loop alive. Now he still uses that loop, but Claude does much of the work. It reads the channel, identifies a few concrete things it can do, puts up PRs, and asks Boris to review them. Elsewhere in the conversation, he adds that Claude is also looking at bug reports and telemetry to suggest bug fixes and things to ship, moving it closer to a coworker than a passive assistant.

   ▸ 19:30 in transcript
3. 3

   Why does Boris Cherny tell companies to give engineers unlimited tokens?

   Boris Cherny treats loose token budgets as fuel for AI experimentation, not as waste. His advice to CTOs is to avoid cost cutting at the beginning and give engineers as many tokens as possible, because that freedom lets them try ideas that would otherwise sound too crazy. At small scale, he says an individual engineer's token cost is usually low compared with salary or other business costs. Once a useful idea works and starts consuming many tokens, that is the time to optimize, perhaps by moving work from Opus to Sonnet or Haiku. He also says Anthropic is already seeing some engineers spend hundreds of thousands a month in tokens, with similar patterns beginning at other companies. The important sequence is experiment first, prove value, then optimize.

   ▸ 25:54 in transcript
4. 4

   Why does Boris Cherny compare AI coding to the printing press?

   The printing press analogy is Boris Cherny's way of explaining coding becoming broadly accessible. He says today's agent users still need to understand the layer underneath, but in a year or two that may matter much less. His historical model is Europe in the mid-1400s, when literacy was below one percent and scribes handled reading and writing for lords and kings. After Gutenberg's printing press, printed material exploded, costs dropped roughly one hundred times over fifty years, and literacy later rose toward seventy percent globally. Boris sees AI coding as a similar democratizing shift: a skill once concentrated in a small group becomes available to many more people. He also compares himself to a scribe who no longer has to copy books, because the tedious part of coding can disappear while the interesting work of deciding what to build remains.

   ▸ 32:27 in transcript
5. 5

   What does Boris Cherny mean by replacing software engineer with builder?

   Boris Cherny uses builder to describe a future where job boundaries blur and everyone codes. Near the end of this section, he says the title software engineer may start to go away in some places and be replaced by builder, or by a world where everyone is a product manager and everyone codes. Just before that, he says engineering, design, and product management still persist in the short term, but he already sees about fifty percent overlap across roles. His examples are practical: Kat, the team's PM, does more coordination, planning, and forecasting, while Boris codes more, but both contribute across the same space. Earlier he says Claude Code's team has a product manager, engineering manager, designer, finance person, and data scientist who all code. Builder means broad ownership over problems, not only writing software.

   ▸ 43:06 in transcript

Answers are AI-generated from the transcript and may contain errors. Tap a question to verify against the source.