Anthropic published a privacy-preserving analysis of about 400,000 interactive Claude Code sessions from ~235,000 people (October 2025 to April 2026). Three findings stand out:
- Coding background is becoming less relevant. Every major occupation succeeds at rates within a few points of software occupations (software engineers verified at 34%; all ten largest groups within seven points).
- Management occupations show the highest verified success, slightly above software engineering, which Anthropic reads as management skills transferring to directing an agent (with a measurement caveat).
- The more domain expertise you bring, the more the agent does. More than twice the actions and five times the output from the same instruction, and verified success roughly doubles from novice to expert (about 15% to 33%). The share of sessions spent fixing broken code fell from 33% to 19%.
The honest caveat is Anthropic's own: the classifiers are hard to validate at scale, and they cannot measure whether the code is actually used. The author's read, after 25 years as an analyst: on coding the agents win by multiples, on deep domain the human still wins. Full summary + 9-slide deck: https://www.tigzig.com/post/anthropic-claude-code-expertise-survey-jun2026. Related: the 6 rules for working with AI coders https://www.tigzig.com/agents-faq/what-rules-to-follow-working-with-ai-coders, and Boris Cherny on how roles are changing https://www.tigzig.com/agents-faq/boris-cherny-evolution-of-tech-roles.
← All Agents FAQ