5,838-developer study: Claude Code adoption adds 41 commits and 0.83 languages per month
Economist Alexander Quispe tracked 5,838 GitHub developers over 28 months and found Claude Code adoption correlates with +41 monthly commits, +0.83 programming languages, and growing language diversity over time. The doubly robust design controls for selection bias, making this the most credible large-sample estimate of AI coding-assistant productivity to date — quasi-experimental evidence, not a vendor benchmark.
Source: arxiv.org ↗
AI provides free signals about unfamiliar technologies and lowers the switching barrier.
Quispe et al., arxiv study
Why this matters
- → Claude Code adoption correlates with developers writing 41 more commits monthly and learning 0.83 additional programming languages.
- → Doubly robust causal design controls for selection bias, providing the most credible large-sample productivity estimate for AI coding assistants to date.
- → Effect grows over time, suggesting AI lowers switching costs and provides signals about unfamiliar technologies.
AI expands developer range