Alex4d ago

It's slight disheartening to see the debate about #LLM usage in open source bifurcate into two mutually opposed camps of never-AIers vs AGI believers. Engineering is all about trade-offs including making the best use of your #tools.

It doesn't make sense to burn a few million tokens when your editors search and replace would do. But if an LLM allows you to finish up that long standing re-factor that reduces your overall CI usage is it still unadulterated evil?

Show threadpenguin424d ago
@stsquad Yeh, and a lot of people seem to be stuck in what they saw how badly a chatbot did 3 years ago without seeing how much better they've got; they're still very fallible but they're managing some useful stuff if used carefully.
Show threadAlex4d ago
@penguin42 how we separate the artisanal patches crafted with #LLM assistance from the YOLO'd vibe coded slop is going to be the challenge.
Show threadpenguin424d ago
@stsquad It's not that different from separating the random rust crates/npm packages/etc that someone through together in an evening.
Show threadAlex
@penguin42 heh, I actively avoid anything with #npm requirements. I did hope crates.io would be better but I thought they had had some supply chain attacks themselves recently.
Feb 26 at 5:46pmWeb
Show threadpenguin424d ago
@stsquad It's not so much it's the attacks I mean, but people who have thrown together a little project in an evening and uploaded it as a crate.