Maybe make a whole llm/ai category where showcases can be posted and questions can be discussed, while keep the possibility to simply ignore the llm stuff category for those who don’t want to read about it. Since, if I remember it correctly, this possibility of filtering llm stuff out of someone’s view, was the original reason for introducing these tags.
Yeah, I just meant maybe a complete own category so there can be discussed other llm topics you mentioned beside showcases. And anyone (like me) who is not very interested in it could mute the whole category
People basically start abusing the “LLM disclosure policy” to looking for LLM advisories. Some even with little programming experience.
And people with little experience in system programming language and coming from web stack languages are directly asking advisories about using LLM, instead of learning system programming language…
fair but I actually have been learning systems as Morphic grows, Zig also doesnt exactly have the huge ecosystem and decades of examples that C and C++ do, so ive been learning from whatever production approaches actually survive real tests and workloads. Sure, thats not the purist build every layer yourself first route, but the principles dont change just because an agent helped write the code. Memory still has to be correct, syscalls still have to behave, and real software still has to run, frankly i’d like to think whatever this project gets right or wrong, people can learn and the language itself can grow in terms of knowledge! There is nothing worse than a blank page, maybe nothing better, but when languages like rust exist we risk falling behind simply from lack of application, the goal for me, no matter the end state of this project is for zig as a language to grow and learn, and for assumptions and language issues to be brought out! Also, I don’t really appreciate the claims of my experience the post implies.
The current policy has been very successful from my point of view. We see tons of projects, many of which do use AI (and the requisite llm tag) that have fruitful discussion, without anyone arguing about the usage of AI. We occasionally do still get Vibe-Coded/slop projects, but those are removed.
I’m also finding the policy to be good as-is. I filtered out the llm tag for myself and basically don’t even know such rules are in place, apart from the occasional post that should have the tag but doesn’t. When that happens, I just add the tag[1] and move on with my life.
I have to disagree here, and push back a little, while AI can help you ship faster, and while it can be a great tool to learn if used adequately, thinking that letting it write, and only review is enough, or not expansive in the long run is imo risky. The important part is not the typing on the keyboard, rather what’s behind it, every-time you translate a thought or a plan into finger movements, is a place where you actually learn, and recent studies have shown that especially in non senior, non expert, the difference between the prompter and the guy who actually typed is huge.
Also Zig doesn’t need to compete with Rust, by shipping more things, rather like anything the quality should speak for itself, and when it comes to software, I’d rather have one FFMPEG class project, rather than 1000 of smaller less known projects. Because the former is definitely bringing more value to the end user.
I agree, but I don’t think “shipping” should be a concern at all for most cases Zig is used for (i.e hobby/recreational programming). I think it’s weird to vibecode personal projects for that exact reason; there is nothing pushing you to write code faster but laziness, and you end up having learned next to nothing.
Yes I totally agree, when I say “shipping faster” i mean within the context of professional code, if you need a quick POC, or you are a bit behind, or in a startup that pushes you to use AI, I think it’s ok you do what you have to do, to get food on the table, but yeah for personal projects, unless you are a senior, or simply don’t have the time because of personal life, I think you are missing the point of personal projects I agree.
I feel the discussion is getting a bit off topic. Whether or not it is smart, pragmatic, productive, or good to use AI or not is irrelevant to me about what should be allowed on this forum or the nature of the llm tag.
I think this discussion should be focussed on how we want to come together and how this as a platform supports that. Personally, I’ve had my experiments with AI and I want it cut out of my life. I hate that I used it for parts of my own project and that the smear of slop code will live on in my git history even after rewriting the code. I want to know upfront when other projects have used AI partially especially because of how I know, personally, the damage it can do to otherwise good projects and I don’t want to disparage my fellow devs, but I do think we need to be open and upfront about our uses of AI. That’s how we learn and grow together.
Firstly, I’m a fan of the LLM disclosure policy, and think it’s working well as-is. I also agree that having a no-LLM at all policy on a per project basis (like the zig compiler itself) is a healthy choice too.
Personally, I’m not a fan of the relentless hype side of the subject, or the way LLMs seem to be misused in general practice.
There are some negatives though that could be addressed. The LLM policy shouldn’t end up fostering an “us vs them” mentality. It’s about as useful as “vi vs emacs” debates.
The fact is, there are quality projects that use LLMs, and there are slop projects that don’t. As a developer community, we are interested in quality, and fostering everyone in the community to strive to be better programmers.
Suggestion - maybe it’s worth opening 1 thread, where members can have an open, honest and professional discussion about the positive application of LLMs specifically with using Zig.
I’m assuming that it’s not all bad news, and that there are some people out there who have found disciplined ways of using LLMs to improve their zig work ? If so, I wouldn’t mind being able to study what worked for them, and what doesn’t .. and draw my own conclusions.
If something is positive, I would like to be able to engage in talking about it freely, without being labelled a slopmeister if I agree, or a Luddite if I don’t.
I just wish there were more zig kernels out there to learn from, I admit that challenging yourself, is the main goal of any programming project, and the joy should be in learning, but I sincerely wanted zig to feel competitive in an area of which, I don’t see as much projects from excepting other languages, maybe it’s just plain arrogance, but you have to be honest, whatever may happen with this project, I hope zig programmers can learn from it
I filter out the LLM tag and I enjoy ziggit more because of it. I periodically (every few weeks) turn it back on to check on things and re-decide if I like the filter.
I really appreciate being singled out by the OP. Thanks. I got into computer programming (in the mid 90s!) because it satisfies my creativity. Not because I enjoy the labor. Computers were already meant to make things easier. If there’s a thing that can take the leg-work out of a task then I’m not going to punish myself by ignoring the option. It’s not great at what it does, but it helps me fast-forward through the bits I do not want to do.
I used to code in a text-editor mate. I ain’t about that life no more.
