Anthropic Engineer on AI Coding: Using Claude to Orchestrate Claude

Boris Cherny, head of Claude Code at Anthropic, recently shared his insights on AI coding workflows and enterprise adoption at a developer conference and in interviews. He noted that over the past six months, most of the code he used to write by hand is now generated by Claude, highlighting a substantial shift in developer workflows.

Workflow and Productivity Shifts

Cherny mentioned that after a year of using Claude Code, his habits have drastically changed to the point where he no longer uses to-do lists. He admitted he no longer writes code directly or even talks to Claude directly, but instead uses one Claude to orchestrate other Claude instances. He believes AI significantly increases an individual's engineering leverage, and the future focus will be on how large a system one person can support. Internally at Anthropic, code output per engineer has increased by about 250% without a noticeable decline in quality.

Core Advice for AI Integration

Regarding how enterprises can boost AI coding productivity, Cherny advises against micromanaging AI usage. Instead, companies should give employees enough permission to experiment and discover useful workflows, as psychological safety is crucial for AI integration. He also emphasized that AI performs better with tools and freedom rather than being forced into rigid human workflows, as general-purpose learning systems scale more easily. His core philosophy is: don't just ask what the model can do for you, but also consider how to give it the right tools and autonomy.

2026-07-08 ~ 2026-07-09 · 5 related posts