Anthropic: domain expertise, not coding skill, drives Claude Code success
An Anthropic study of ~400,000 Claude Code sessions finds that domain expertise — not coding proficiency — is the strongest predictor of success, while task complexity and value have risen roughly 25% over seven months.
Score breakdown
The findings show that agentic coding tools reward domain understanding over formal programming training, with non-engineers succeeding at roughly the same rate as software engineers — a direct signal about how these tools may reshape the labor market for knowledge workers.
- 01Study covers ~400,000 Claude Code sessions from ~235,000 people, October 2025–April 2026
- 02In a typical session, people make most planning decisions; Claude makes most execution decisions
- 03Domain expertise — not coding proficiency — is the strongest predictor of session success
Anthropic's research team conducted a privacy-preserving analysis of approximately 400,000 interactive Claude Code sessions from roughly 235,000 people between October 2025 and April 2026. The study introduces a framework for evaluating agentic coding usage across nine work modes — including building new code, fixing bugs, testing, orchestrating agents, operating software, understanding existing systems, and planning changes — and assesses task composition, human-AI collaboration patterns, and session success rates.
The central finding is that domain expertise, rather than coding proficiency, is the primary driver of success.
The central finding is that domain expertise, rather than coding proficiency, is the primary driver of success. The greater the domain expertise a person brings to a session, the more work Claude does per instruction, and the more frequently the session ends in success — defined as accomplishing what the person set out to do, with verifiable evidence such as passing tests or committed work. Importantly, the gap between intermediate and expert users is described as modest, suggesting that solid domain understanding is sufficient to use the tool nearly as effectively as deep mastery. Every major occupation succeeds at approximately the same rate as software engineers on average.
Over the seven months studied, the nature of Claude Code usage shifted meaningfully: the share of sessions spent debugging fell by nearly half, and usage moved toward more end-to-end agentic tasks such as deploying and running code, analyzing data, and writing non-code documents. The estimated value of the typical task — benchmarked against freelance job postings — rose approximately 25% on average across almost every category of work. The report also notes that the share of GitHub projects with coding agent activity more than doubled since late 2025, and that Claude Code users now average 20 hours per week using the tool.
Key facts
- 01Study covers ~400,000 Claude Code sessions from ~235,000 people, October 2025–April 2026
- 02In a typical session, people make most planning decisions; Claude makes most execution decisions
- 03Domain expertise — not coding proficiency — is the strongest predictor of session success
- 04Every major occupation succeeds at nearly the same rate as software engineers on average
- 05The gap between intermediate and expert users in success rate is described as modest
- 06The share of sessions spent debugging fell by nearly half over the seven-month period
- 07Estimated value of the typical task rose about 25% on average, benchmarked against freelance job postings
- 08Claude Code users now average 20 hours per week using the tool
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 17, 2026 · 10:39 UTC. How this works →