Who is using AI to code? Global diffusion and impact of generative AI (arxiv.org)

Researchers trained a neural classifier to detect AI-generated Python functions across 80 million GitHub commits from 200,000 developers (2018–2024) and mapped how generative coding tools have spread globally. By December 2024 they estimate AI authored 30.1% of Python functions from U.S. contributors, versus 24.3% in Germany, 23.2% in France, 21.6% in India, 15.4% in Russia and 11.7% in China. Adoption is higher among newer GitHub users, with similar rates for men and women. Using within-developer fixed-effects models, the authors find that shifting a developer’s workflow to 30% AI-assisted coding increases quarterly commits by 2.4%. Beyond raw adoption, the paper ties usage intensity to measurable economic and innovation effects: coupling productivity estimates with occupational task and wage data values U.S. AI-assisted coding at $9.6–$14.4 billion annually, rising to $64–$96 billion if larger productivity gains from randomized trials are used. Generative AI also spurs learning and experimentation—developers try more new libraries and novel library combinations. The study signals broad but uneven diffusion: access alone isn’t enough—intensity of use drives output and exploration—raising concerns that uneven uptake could widen skill and income gaps and suggesting priorities for training, tooling and policy.