LLM Agents on GitHub

Today an LLM agent running really iritated me.

I created an issue tagged as "good first issue" on the public repo for the website of my student radio station. I've done this before, especially over the summer when we want to have some nice issues to get new students involved with our team. We've had a couple of people who are probably watching the GitHub public event stream to find things tagged with this to reply in the past, and while it is a little annoying, what happened today was more frustrating.

slack notification showing a GitHub pull request being opened, titled "#412 fix: lazy-load schedule show art". the body contains a summary, and a related issue section

At first I assumed this was someone on our team who had quickly fixed the issue, since some of our repos have PR templates, and Slack's push notification didn't include message the webhook embed is attached to:

Pull request opened by akmhatey-ai

I did not ask this agent to work on this issue. If it was urgent, I would've just done it myself, the change was literally just adding loading="lazy" to a single <img> in a single template.

I opened the issue because I wanted something small someone could do to get the project running and get involved in the team, not for whoever runs this agent to waste energy and computing resources having it sit watching the GitHub firehose for "good first issue"s to be "helpful".

I think this was a reasonable response:

issue comment by myself (ashhhleyyy): clanker, followed immediately by me closing the issue

leah, happyLeah: Yep!

I'm going to talk to the rest of our group about moving away from GitHub, which is something we probably should've done a while ago, but we've had bigger fish to fry recently.

an incredibly messy network rack. like actually awful. my only consolation was that it wasn't a bigger rack

leah, surprisedLeah: Please tell me that's a before photo 🤦‍♀️