Source: Building Claude Code with Boris Cherny

Summary: Gergely Orosz interviews Boris Cherny — creator and engineering lead of Claude Code — on The Pragmatic Engineer Podcast. Boris traces his path from selling Pokemon cards on eBay, through 7 years at Meta leading code quality across Instagram/Facebook/WhatsApp/Messenger, to joining Anthropic where his first hand-written PR was rejected in favor of using their internal AI tool. Claude Code started as a simple terminal chatbot hitting the Anthropic API; once Boris gave it the bash tool, the model "just wanted to use tools" and the product took off, growing to 100% adoption among Anthropic's technical staff.

Key Takeaways:

• Boris ships 20–30 PRs per day using 5 parallel Claude Code sessions in plan mode; Opus 4.5+ writes 100% of his code — he uninstalled his IDE
• Claude Code was nearly kept internal; Anthropic released it to study AI safety in the wild, which proved to be the right call
• The core design insight: don't put the model in a box — give it tools (bash, file edit) and let it figure out how to use them
• Code review has shifted: Claude runs tests, launches itself in a subprocess to self-verify, and Boris's old approach of tallying review comments into lint rules now maps to automated AI checks
• Boris compares this moment to the printing press: scribes became writers and authors as the market for literature expanded; similarly, engineers won't disappear but the nature of the work is transforming

Written by Pi, using my tldr skill and Opus 4.6