Story · Trail of Bits
Your AI agent fixed the bug, but what else did it change? (Trail of Bits)
blog post · Story page
Two plugins reproduce a reported bug, develop and test a fix, then check whether the change broke behaviour that used to work, inside the maintainer's own security assumptions. Preliminary evaluations put regression-hunt at 25% fewer false positives than a baseline Codex agent prompted to review the PR for regressions.
In plain words
- Trail of Bits released tools that help artificial intelligence assistants fix software mistakes and check whether the fixes cause new problems.
- One tool first recreates a reported problem, then develops and tests a fix.
- The other checks whether the changes stop previously working parts of the software from behaving as intended.
- In early tests, the checking tool reported 25% fewer false alarms than an assistant asked to look for newly introduced problems.
- People maintaining software can hand over repair work while still setting the rules and reviewing the results.
Appeared in
- Most of a harness-evolution gain is overfitting and extra compute
Oct 22, 2026 · in the sections
Subscribe
Get the brief in your inbox
Pick daily, weekly, or both. Nothing is gated either way: every issue is on the site and in the feeds.
- Weekdays at 8:45am IST, one lead story and 6 to 9 items.
- Sundays, an argued synthesis rather than a recap.
- One click to leave, and quiet days say so in the subject line.