...this really would be the perfect use case for AI, wouldn't it?
Of course actual code changes to something like Chromium need to be driven by humans, for all sorts of reasons. And I'm not suggesting Google should hand over all bug triage to a machine, or anything like that.
But if you've already decided, for better or worse, that you're going to auto-close old issues—maybe at least have an AI look at them first? The bug was clear and straightforward; an agent should be able to confirm that it still exists, either by reproducing it or by looking at code changes since the problem was reported. It doesn't even need to be a particularly strong model.
Surely Google of all companies has the compute for that?
Who's downvoting this? People who hate AI or people who don't mind huge backlogs of high-effort human submitted bug reports that rot for all of eternity?
It costs very little to have an AI do a first pass on bugs. Is it a dupe of another bug? Was it fixed already? Can it be reproduced? What's the estimated impact and scope of the fix? Maybe even fix it autonomously if it's low-risk. These are all things humans don't enjoy doing.
ESR does precisely this; he uses AI to answer AI-generated bug reports.
In my experience, over half of AI generated (or AI assisted) security reports are not actual security issues.