GitHub spam

Stream: general

[Yaël Dillies]

Today I've seen three separate clearly AI-powered accounts spam on PRs to mathlib: ayestore, eyestoreoye, eyestorefivemstore. See eg #24747

[Yaël Dillies]

What do we do? Is it reasonable to restrict interaction with our repos to people who have push access? That makes up a lot of the people who could reasonably be commenting on a PR, but probably not all

[Bryan Gin-ge Chen]

Yeah, it seems to be happening across GitHub: https://github.com/leanprover-community/queueboard/issues/99

I've been blocking the accounts from leanprover-community as they pop up

[Robin Carlier]

-1 for restricting repos to people with write access. Spams like this are annoying but opening issues/PR and discussing them should remain open to everyone, regardless of push access.

[Kevin Buzzard]

#24743 was another one (thanks Yael for removing it)

[Junyan Xu]

Any screenshot / archived page of the bots' behavior?

[Kevin Buzzard]

It was generic posts of the form "this is an insightful PR" by various bot accounts whose home pages were just ads

[Yaël Dillies]

All my PRs are insightful and my LinkedIn profiles pristine :sparkles:

[Edward van de Meent]

Apparently my performance optimizations are quite clever, and I did nice engineering work :sparkles: too bad it was basically a copy of one of Yaëls PRs to try to debug the import summary issue...