Type: url
Original: https://www.anthropic.com/research/claude-code-expertise
Status: generated
Summary
An Anthropic study of roughly 400,000 Claude Code sessions between October 2025 and April 2026, introducing a framework for measuring interactive agentic coding: what the work consists of across nine work modes, how planning and execution decisions divide between person and agent, and what predicts a session ending in success. It finds people making about 70% of planning decisions and the agent about 80% of execution decisions, with more expert users drawing longer chains of agent work per instruction (12 actions and 3,200 words against five and 600). Its titular finding is that task-specific domain expertise predicts success far more than occupation does — verified success 15% for novice-rated sessions against 28-33% for intermediate and above, while every one of the ten largest occupations lands within seven points of software engineers on code-producing sessions.