Pocketbase and AI

[mattzab]

Hello,

I love PocketBase, particularly the goal of being so stable and "finished" software- avoiding update hell when 1.0 releases. I'm concerned that a sickly amount of time is being consumed due to AI garbage all over github. Not only are there all these bogus PRs but there's also the LLM bots landing in discussion areas with walls of text. Now also we're being spammed by Microsoft's Copilot nonsense being trained on everything and wanting to participate, or whatever the banner is that I keep dismissing. I've even seen instances where ads were inserted into commit messages automatically, even when the dev is not using LLMs at all. That "feature" was rolled back, but it's still super stupid that it was a thing at all...

This proposed solution may not be the perfect answer, and there are surely drawbacks that only Gani might know about- but what about moving off to a different platform? Gitea is FOSS and self-hostable. I've stood up an instance in the past and it was straightforward and simple. I think this would greatly reduce the spam by shrinking down the community only to those who actually care to register on a separate, custom git platform dedicated only to PB. I remember the days when collaboration happened through mailing lists, instead of this new social network GitHub with all the bots and LLMs everywhere.

Just a thought, I know this is off-topic for PB itself. I just wanted to float the idea of eliminating all this wasted effort dealing with low-quality and low-effort bot related activity by migrating to something like git.pocketbase.io with a self-hosted Gitea/Gitlab (or similar) git instance. It just seems like the right move to encourage true collaboration with real, interested people, and return to that small "old internet" feel. Small community collaboration is usually more enjoyable anyway.

Looking forward to 1.0 when the day comes! Thanks for all the hard work, I do hope this remains enjoyable for you!

Regards,
Matt

[dschissler]

It seems like that the current level of frustration is going to cause a lot of false positives. From my perspective the growing PC HR word and phrase correcting culture has caused many people to write like robots. I mean that I can't really tell the difference unless it's the obvious bullet lists that AI loves so much.

I actually think that to a degree the early language and style policing was a ploy to force people into communicating in a way that would be most useful as a training set. So it's all rather a bit of a horror show from then until now.