Some of those bugs had beaten a 30-year veteran for over a week. Claude Code ran a /loop around the clock on a prototype compiler, fixing over 500 crashing test cases while they glanced in now and then.
Result Including bugs that beat a 30-year veteran
“I've had Claude Code running a /loop for the last week driving down complex crashing bugs in a prototype compiler entirely unilaterally.”
Spotted: found in the wild. Details are as reported by the source above and have not been independently verified.
Set Claude Code running a /loop for a week to drive down complex crashing bugs in a prototype compiler entirely on its own, glancing over its work occasionally.
Working 24/7, it has fixed over 500 crashing test cases so far, including ones the developer had spent more than a week trying to track down. They expect some technical debt, but nothing close to the cost of fixing them all by hand.
Work through the failing test cases in [test directory] one at a time. For each one, reproduce the crash, find the root cause, fix it with the smallest change, keep a regression test, run the full suite and commit with a message explaining the cause. After [N] failed attempts on a case, log it in SKIPPED.md and move on. Keep going until none are left.
A starting prompt written by Agents in the Wild. Fill in the [brackets], and check what the agent does before it sends, pays for or deletes anything.