[RFC] Surviving the AI PR Flood: The Macro X-Ray & Asymmetric TDD

This topic was created three months ago, but I feel that the concern renich raised was very prescient and correct.

Looking back at it now, I find one point especially interesting. A human correctly expected the problem, but the proposal generated by AI (probably Gemini) missed the point. Because of that, I feel that the importance of this topic may have been overlooked.

Perhaps renich did not predict the problem only by logical reasoning. Maybe he noticed it through his experience in the larger Fedora ecosystem. I may be wrong about this.This suggests to me that, as of March 2026, humans still had an advantage over AI in noticing future problems.

@renich, what do you think about the topic now?
Can we learn anything from recent trends in Fedora?

As you observed, the proposal was indeed generated by Gemini. The curious thing is that, when one says that, one imagines the situation going something like:

— “Hey Gemini. Generate a proposal for an RFC in crystal. Make no mistakes.”
— “OK… … Done!”

And, as you might’ve surmised as you read that allegory, it’s not at all that simple.

Fedora, at the time, hadn’t been hit with this problem. In fact, if it was happening, it was largely unnoticed. Fast-forward to today and Fedora just had a huge, multi-month fight about AI in Fedora. The community is divided, not exactly by the use of AI, but by how the proposal was presented (it favored NVidia and it ignored some of the Fedora core Tenants).

The proposal evolved. That’s the good side of Fedora’s heated discussions. People respect each other’s opinions an listen.

More to the point, I expected that to happen in most projects. Look at the kernel right now. AI-assisted code has been steadily growing. They don’t oppose it and Linus has even said something along the lines of “actually, the code’s not that bad anymore” and he even coined an AI code policy: AI Coding Assistants — The Linux Kernel documentation

If my proposal at least planted the seed of the idea that lead to something like a start of a policy about AI-assisted code, hey, it was a win.

I am still worried about supply-chain attacks of the sort that I mentioned in my RFC (which, btw, shouldn’t been an RFC). I think my PoC faintly proved an attack of sorts. Someone with better skills could take advantage of something similar. My intention was just to illustrate this.

And thank you for the acknowledgement and insight, @kojix2.

@renich Hi, thanks for your reply.
I see that the Linux kernel community had already established an official policy back in April. I had missed that.
Personally, I feel that Assisted-by is more appropriate than Co-authored-by.