Research Explainer · XiaoHu Explains

Anthropic Studied 400,000 Claude Code Sessions: Expertise Matters More Than Coding Skill for AI Coding Success

Across seven months and 235,000 users, intermediate-to-expert sessions show about twice novices' verified-success rate; each of the ten largest occupations is within seven points of software and math occupations.
⚑Stance note: all the data in this piece comes from Anthropic's privacy-preserving analysis of its own product, Claude Code — this is not independent third-party research. Labels like success rate, expertise, profession, and task value were all auto-generated by a Claude model reading de-identified session transcripts, not human-verified. What follows just relays the data itself; the judgment call is yours.
1Research Background

Across 400,000 Conversations, Who's Really Calling the Shots — Human or AI?

Anthropic recently published a research report analyzing roughly 400,000 Claude Code interaction sessions from October 2025 to April 2026, trying to answer one question: can people with no coding background direct an AI agent to complete complex technical work?

Primary source and reference