No dickhead, all LLM output is slop
I could build a statistical model to predict what bolt to use in a mechanism. It would mostly predict M6, and this would mostly be right, but it would still be bullshit. There is a proper process to determine the fasteners you should use in an application based on fucking engineering: the stresses on the steel, the clamping force needed, environment, a bunch of other shit i don't know because i did stats not mechanical engineering
Using the statistically average fastener is bullshit. You can teach a computer to do the actual maths, that's an option, but something is doing the proper fucking maths at some point. If you care about the product you do it properly
And this is why software "engineering" is increasingly clearly a fucking joke, except for like 3 people who are all probably unemployable trans women. Statistically plausible code is a fucking hack, and the fact that is sometimes kinda works should make you all feel really embarrassed