For the most part I don’t think a community as slow as this should have much in the way of moderation. The fewer restrictions, the better. We are all adults here (well, probably).
Content-free clutter like OS playground fight memes and “Look! I, too, have answered the Steam survey!” posts is an exception.
And if somebody’s talking to us via an LLM, that’s (in my own dramatising-prone opinion) disavowing what makes us special among animals (apart from our infinite capacity for conscious cruelty): Language, reasoning, imagination, creativity, self-expression, personal flair and style.
AI-generated code is unknown territory for me. The people using it sometimes seem to know pretty damn well what they’re doing, in principle, even if they’re not quite doing it themselves. So I’m not in favour of banning it outright. But, as with the parent reddit community, I have set a “two months of verifiable development time” barrier.
What do you think? Thumbs up? Thumbs down? Too lenient? Too strict? Just vote or make alternative suggestions. (I haven’t got the energy to discuss this again…)
PS: I do miss the old banner but it was AI-generated back when that was still fresh & weird & interesting to play with… now it’s more like a plague
I feel like it should be a bit stricter, as it is so easy for LLM generated code to hide errors and present the output as something that seems valid (when in fact there are a lot of pitfalls involved). LLMs and the chatbot features they possess often act in tandem to deskill the user of them, they become disconnected from their craft, gradually become more confident in the outputs of prompt slot machines. There is a political movement behind the push for the use and normalization of LLM use to cause a mass disenfranchisement event for skilled workers and creatives.
It’s just a very uncomfortable reality to face. Calling it “AI” feels incorrect in my opinion, it is just a prompt slot machine remixing the training data that humans either willingly or unwillingly (collected through scraping attacks) contributed to the model. Which can often produce very weird results. Just recently, Linus caused an issue by using an LLM to vibecode something and went on vacation dumping the problem in the laps of contributors. Mind you, this is a person who is supposed to know better but caused a massive problem. This is a scary time to be alive.
I’m guessing over time it will lead to people becoming more dependent paying for tokens among those who can’t afford to own the hardware needed to run AI locally. And those who don’t review what is being generated could accidentally publish something that is malicious as the manual review process becomes less common. Could be a good attack vector among those who are vibe coding everything, and not be diligent double checking things. Especially with how people will run random LLMs to use.
linus got old and got something else to hold his marbles.