How AI Coding Agents Modify Code: A Large-Scale Study of GitHub Pull Requests Conference: 23rd International Conference on Mining Software Repositories; April 13–14, 2026; Rio de Janeiro, Brazil 23rd International Conference on Mining Software Repositories (MSR ’26), April 13–14, 2026, Rio de Janeir
Daniel Ogenrwot, John Businge · 2026
Read How AI Coding Agents Modify Code: A Large-Scale Study of GitHub Pull Requests Conference: 23rd International Conference on Mining Software Repositories; April 13–14, 2026; Rio de Janeiro, Brazil 23rd International Conference on Mining Software Repositories (MSR ’26), April 13–14, 2026, Rio de Janeir by Daniel Ogenrwot, John Businge (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-17581.txt.
Description
Daniel Ogenrwot and John Businge study how AI coding agents change code on GitHub. Agents now submit pull requests as autonomous contributors. Little proof exists on how their changes differ from human ones. The team runs a large-scale study of agent pull requests. They compare how agents modify code. They also compare how agents describe their changes.
Such proof is key to judge agent reliability. Reviewers need to know where agents excel or slip. Description quality shapes how fast humans can check work. The study gives teams data for policy on agent code. It moves debate from anecdote to record. Open-source projects can set rules with proof in hand.
Catalog record
| CREATOR | Daniel Ogenrwot, John Businge |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-17581