@Retreival9096 I think you'll find that all actual software developers who have tried using LLMs as aid the last ~6 months agree that they are quite useful.
The people who claim differently are either 1) not actual developers or 2) haven't checked if their gut feelings agree with reality.
That said, I would most definitely not have an LLM write me a cryptographic library, but I'm also very certain Proton aren't either.
@troed @Retreival9096 I think the problem is the "wo wo" crowd, raising LLM:s to the skies. That's just ridiculous.
Yes, they do some things well, yes, they might save me time on boilerplate or easy stuff, but no, they will not make me 10x faster.
They might make me 5%-15% more productive or so. But yes... I am not a programmer, but a devops guy.
I do know programmers though, and the most brilliant of them also are in the 5%-15% range.
@h4890 Exactly this.
An interesting thing is that it allows someone unskilled to perform a task they wouldn't have been able to do at all otherwise. As long as they don't think the solution is as good as if someone senior in the subject had made it I think it's all good.
Myself I use the regular 80/20 rule of thumb. The LLM helps a lot with 80% but it's my input that decides whether the final 20% have good quality or not.
@troed @Retreival9096 Amen!
I wonder if vibe coding by juniors programmers will give us a lot of security holes in the year to come?
I've seen some of my students vibe code assignments, and they have no clue what the code does. The code also tends to be convoluted at times.
