Claude Code creator Boris Cherny reveals that Anthropic runs 100% of its pull requests and 80–90% of code reviews using the tool, highlighting a shift from manual prompting to building agentic loops.
In a shared interview clip, Boris Cherny, the Head of Claude Code at Anthropic, broke down how the tool is used internally, sharing that 100% of their pull requests and 80–90% of code reviews are run by Claude Code. Cherny noted that his own workflow has shifted away from writing prompts and toward building agentic loops, with the "/loops" command being the feature he uses the most.
Prompt engineering is dying; the future of software development belongs to developers who can orchestrate and optimize autonomous agentic loops rather than write individual prompts.
* Anthropic's extensive internal dogfooding proves that they have achieved production-grade reliability for agent-led pull requests and code reviews.
* The focus on the `/loops` feature highlights a shift from synchronous chat interactions to asynchronous, goal-driven execution cycles that run and debug code autonomously.
* This transition changes the role of the developer from a direct operator to a designer of autonomous agent pipelines and loops.
DISCOVERED
3h ago
2026-06-15
PUBLISHED
3h ago
2026-06-15
RELEVANCE
AUTHOR
Av1dlive