Hacker News Reader

The same way we have text classification for the spam in our mail inbox, couldn't someone train a model to classify issues as actionable bugs vs. noise for large projects like the one mentioned in the OP? Data would come from closed issues:

- if the issue lead to a commit, or a merge is mentioned in the thread, then it is actionable,

- if the issue was closed without any code change, then it is noise.

View on HN

maybe. but i think this process really helps develop a deeper understanding of the evolution of the project and doesn't take that much time.

and I bet your AI isn't going to be able to say 'oh yeah, thats just that thing we fixed in 2.1'

seems useful in a second-order capacity though - process introspection

You do see a decent number of GitHub bots that will trash issues if they don’t comply to some sort of “expected/observed/steps/specs” format, which I think more or less accomplishes the same thing. The only issue is you’re potentially losing out on issues from less-tech-savvy users, but I guess simply using GitHub filters many of those people anyways.

Reply on news.ycombinator.com