Speaker

Boris Cherny

Boris Cherny

Head of Claude Code at Anthropic. Creator of Claude Code, MCP (co-contributor), and the Anthropic desktop app, all developed within the ‘Anthropic Labs’ team.

Background: self-taught engineer (economics degree); wrote a widely-read TypeScript textbook; senior engineer at Meta/Instagram (responsible for code quality across Facebook, Instagram, WhatsApp); brief stint at Cursor (2 weeks) before returning to Anthropic. Born in Odessa, Ukraine; emigrated to the US in 1995; lived in rural Japan before joining Anthropic.


Talks and interviews

SourceDate
Boris Cherny on Claude CodeFeb 2026
Boris Cherny on Opus 5, Deleting the System Prompt, and Giving Models Harder ProblemsJul 2026

Recurring themes

  • Latent demand as the foundational product principle — in both its user form and model form.
  • Build for the model 6 months out, not today’s model.
  • The bitter lesson — bet on the most general model; scaffolding gains evaporate.
  • Exponential thinking — tracing the curve is more reliable than intuition.
  • Safety as purpose — returned to Anthropic from Cursor because of mission, not product opportunity.
  • Printing press analogy for the coding transition — short-term disruption, long-term democratic unlock.
  • Delete the harness at every model release — for Opus 5 his team cut 80% of Claude Code’s system prompt, and he tells users to delete their CLAUDE.md files and skills every six months and restore only what the model shows it needs (Harness Ablation).
  • Coding is solved ‘for the kind of coding that I do’ — the Startup School talk adds the caveat: deep systems code, distributed systems and pixel-level interface work still defeat the model.

Key concepts attributed to Boris


See also