I'm just trying to earn an honest day's wages by getting this LLM to write my code for me but it just keeps cranking out buggy code! (╯°□°)╯︵ ┻━┻
Discussion
Ah, fun times. x)
I recently learned that my company has a ChatGPT sub. Might take advantage of that for some Kubernetes shenanigans with k3s and MetalLB. ^^
Its very helpful. But you can't trust it.
Personally, I treat it as both a tool and source of entertainment.
Run an LLM locally, crank up the temperature, give it a very simple prompt on purpose and ... watch it go. xD
Or, feed it a lot of context, reduce the temperature and rerun the prompt a few times to get different versions/takes and use it like a tool. It can give you some great pointers - but it can also make you run into a wall if you aren't careful x)