@zeenix@toot.cat I don't think the jury is back yet on any of this. In the next few years, we'll be able to see the state of projects that have gone full-bore pro-LLM, made modest use of LLMs, and those that have either rejected LLMs or severely constrained their usage. We'll be able to look at the code and communities. We'll be able to see what's still maintainable, which communities have grown, shrunk, or stayed the same. At that time we *might* be able to draw some real conclusions. I say *might* because we don't exactly have a control group for any of this. There's no Kernel B, for example, that we could say "OK, it didn't adopt LLMs and here's what happened versus the Linux kernel." There's no equivalent project. (No, the BSDs don't count because they're far from the same baseline.) But I strongly disagree that it's an "adopt or die" scenario. I don't think, for example, projects can just ignore when LLM-discovered security reports come their way. But that doesn't mean that they have to adopt generative AI to create new code. I don't buy it. I definitely don't think projects have to start accepting LLM-generated PRs/MRs from new contributors. If I want LLM-generated code in my whatever, then I'll do the prompting. Having to filter through the good, bad, and ugly LLM-generated stuff is not good for maintainers. Anyway, we'll see. I don't think it's cut and dried. I am prepared to be wrong, but the verdict is not in yet.