Even though I am not a contributor, I do have several places I've used (and use) TCC with modifications for both commercial and non-commercial projects. And, I am a maintainer of several other open source projects.
I have been more and more frustrated with the changes that people have been using AI for. And I've been finding projects I've allowed more AI-aided commits in to become increasingly difficult for humans to reason about and maintain. (For instance cilbox https://github.com/cnlohr/cilbox) I _think_ it's two fold - (1) that traditionally people have to put significant thought into what they are doing, and thus the changes become much more surgical, and do not change the vibe of the original codebase. And, more importantly, things that shouldn't be up-streamed aren't pushed as much. (2) the type of code AI produces, even when well prompted and documented, is harder for humans to read and work through the logic of. I don't have any explanation for why this happens, the individual commits are fine. But the code slowly morphs away from human-readable-ness. This phenomenon seems to happen regardless of what AI tools or models are used. I think TinyCC is a great starting place for lots of more unusual applications (for instance the changes someone a few weeks back made with the ;'s is a great thing to have in a branch) but, I don't this to be what happens to upstream TinyCC. A friend who works on clang commented on this: > even with code that on its surface looks okay i start having a much harder time grasping the intent behind it - it's weird and an unnerving feeling There's very few? no? other single-pass C compilers that can just compile-to-RAM like TinyCC can, and it's relatively hackable. TinyCC is a precious piece of software to me, and I don't want to go off in a corner and live on my own fork. I don't know about a true policy change or anything, but I would strongly prefer if the things that are pushed upstream are things that continue to keep TinyCC in the same ethos it's in. Charles On Sun, Apr 26, 2026 at 2:50 PM grischka via Tinycc-devel < [email protected]> wrote: > On April 26, 2026 1:07:43 PM UTC, Reini Urban <[email protected]> > wrote: > >I love LLM work, but only with good LLM's. Quality is much better, not > >worse. > >If you work with free models, like MinMax of course non-sense will appear, > >but with opus, gpt-5.x or kimi-2.6 it will definitely be better > > > >Why else banning? > > The unwritten but basically only prerequisite for people pushing on tinycc > mob is that they know what they're doing. > > IMO this however excludes AI agents because AI does not "know" (as far as > I know). > > It does not exclude people using AI though. > > But, for example, we should not end up with a nonsense test made by AI > added to .gitignore by an human. Humans must still be able and willing to > question the AI and to remove nonsense stuff rather than stabilize its > existence in our code ;) > > >Copyright? This is already solved. See the linux policy. > > > >On Fri, Apr 24, 2026 at 3:08 AM Ellie <[email protected]> wrote: > > > >> Dear TCC mailing list, > >> > >> I was wondering if the TCC project has any sort of guidelines on > >> commits, and specifically regarding LLM code. > >> > >> A couple of high profile projects, e.g. Qemu and SDL, and the > >> programming language Zig, have banned LLM code from contributions > >> entirely due to what seems to be in part copyright concerns. > >> > >> I've also found this clip that I find concerning: > >> > https://github.com/mastodon/mastodon/issues/38072#issuecomment-4105681567 > >> > >> As a result I guess TinyCC, since it doesn't seem to be focused on > >> implementing a huge feature set with pressure to compete with "big" > >> compilers, may want to consider banning LLMs. While it probably won't > >> stop everyone, it might reduce it and shift the moral responsibility of > >> ignoring that and ensuing licensing questions to the contributor. > >> > >> (This isn't legal advice, for sure, since I'm not at all a lawyer.) > >> > >> I guess a natural place to put any such thoughts and guidelines, > >> regarding committed code and LLMs, would be on tinycc.org somewhere. So > >> this might be another reason to update the website at some point. > >> > >> Sorry if this question is too off-topic. > >> > >> Regards, > >> > >> Ellie > >> > >> _______________________________________________ > >> Tinycc-devel mailing list > >> [email protected] > >> https://lists.nongnu.org/mailman/listinfo/tinycc-devel > >> > > > > > > _______________________________________________ > Tinycc-devel mailing list > [email protected] > https://lists.nongnu.org/mailman/listinfo/tinycc-devel >
_______________________________________________ Tinycc-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/tinycc-devel
