Zoe Hitzig is a Research Scientist at Anthropic. Her published work on AI usage in practice examines how real users of different backgrounds engage with agentic coding tools at scale. The June 2026 study she co-led analyzed approximately 400,000 Claude Code sessions from 235,000 users to map how domain expertise, occupation, and skill level affect success rates and output volume when working with coding agents.
Operating themes
- Domain expertise as the key predictor: Research consistently shows that what a user knows about the problem domain matters more than what they know about coding when using an AI coding agent.
- Large-scale empirical AI usage research: Studies are grounded in real session data across diverse populations, not in lab settings or synthetic tasks.
- Cross-occupational parity findings: The finding that all major occupations achieve within 7 percentage points of software engineers contradicts the common assumption that coding agents primarily serve technical users.