Yes, this. I returned to manual coding for almost everything… Using AI for anything that isn’t a toy/prototype is truly a waste of time and money.
Some of it feels slower in the beginning but ultimately you move faster on your own… and BONUS: you don’t frequently want to smash your keyboard in frustration with the AI.
@superus3rdone@trashh_dev@trashh_dev this really is a decent suggestion: one note/doc per skill and wiki link from each skill doc to dependencies. Then look at graph view.
@NJMonsterHunter@Rainmaker1973 It wouldn’t electrocute, but transmitting tens of kilowatts of energy via radio or microwave energy wouldn’t exactly be a good thing for any creature flying near them. For reference, your microwave oven is ~1KW.
@sudobunni https://t.co/CMUyNFrwkO
I love this stool… I’ve always found the ever so slight recline on desk chairs to be killer on the neck. It’s rare for me to get neck pain now. It’s all I use.
@DGuardian_1977@_avichawla Yeah, I’m just curious. I’ve spent a lot of time using AI to write code and I find it quite unenjoyable for most of the work.
For complicated or complex codebases AI is slower.
AI is faster at writing individual pieces of code, so why wouldn’t it be faster at writing all the pieces?
Because getting a complicated system working reliably requires precision, and natural language is not precise.
By the time you specify your architecture with enough detail for the LLM, you might as well have just written the code.
After the LLM has drifted from your specification and you spend time correcting it, you have negative ROI.
@mitchellh Thank you for a refreshingly balanced review! I think the “simultaneously excellent and entirely unremarkable” reflects the jagged intelligence of LLMs perfectly, and I suspect that will never really change.