The practical impact of AI agents on open-source commits

The chart erased doubts

Simon Willison, creator of the popular tool Datasette, has shared a GitHub chart showing weekly commits to his project since 2018. It clearly shows that a massive spike in activity (37,000 additions and 9,500 deletions) in 2026 aligns directly with the release of agentic models like Opus 4.8, GPT-5.5, and Sol.

Writing code is no longer the bottleneck

While a large drop dominated the chart in 2020 when Willison was mainly pruning old code, there is now a huge surplus of creation. Coding agents have definitively altered the iteration pace of solo open-source developers. Anyone with access to Sol or Opus can tackle rewrites and refactoring that would previously have languished on the back burner.

Volume does not equal quality

Naturally, the chart does not tell us whether the new code is secure or if the architecture is collapsing under technical debt. The same tool that accelerates feature requests can replicate poor design patterns at lightning speed. The speed of writing is no longer the metric of a developer's success.

Who will own these repositories

It will be interesting to see if this massive surge in added code sustains itself long-term. And if it does, whether it will lead to better software or to monolithic piles of generated boilerplate that a human can no longer satisfactorily test.

Lilith's verdict

The solo open-source developer finally has a production team that never sleeps. When he gets lost in the code six months from now, he can just let the agents refactor it.