Ek-Hou-Van-Braai@piefed.social to Linux@lemmy.mlEnglish · 3 days agoLinus Torvalds shares his thoughts on Anti-AI coding rulesmedia.piefed.socialimagemessage-square515fedilinkarrow-up137
arrow-up137imageLinus Torvalds shares his thoughts on Anti-AI coding rulesmedia.piefed.socialEk-Hou-Van-Braai@piefed.social to Linux@lemmy.mlEnglish · 3 days agomessage-square515fedilink
minus-squareSenal@programming.devlinkfedilinkEnglisharrow-up4·3 days agoThere’s nuance here though, LLM’s actually do a better job of code that looks like it should work. You get an unskilled junior in the mix, their code is obviously bad. You get a semi decent LLM with a competent harness in there and you get a mix of code that works and code that nearly works. There should be competent peer review either way, but the effort for each is potentially an order of magnitude different. Then add on top that the LLM can churn them out way quicker than a junior could possibly achieve and you have a recipe for a huge review burden.
minus-squareEk-Hou-Van-Braai@piefed.socialOPlinkfedilinkEnglisharrow-up2·3 days agoYea using a junior or a agent both comes with issues and risks. You need to know what you’re dealing with and shouldn’t trust either blindly. Ultimately someone has to take ownership of if shit goes wrong, and someone has to understand what the code does and why. If you don’t something has gone wrong.
There’s nuance here though, LLM’s actually do a better job of code that looks like it should work.
You get an unskilled junior in the mix, their code is obviously bad.
You get a semi decent LLM with a competent harness in there and you get a mix of code that works and code that nearly works.
There should be competent peer review either way, but the effort for each is potentially an order of magnitude different.
Then add on top that the LLM can churn them out way quicker than a junior could possibly achieve and you have a recipe for a huge review burden.
Yea using a junior or a agent both comes with issues and risks. You need to know what you’re dealing with and shouldn’t trust either blindly.
Ultimately someone has to take ownership of if shit goes wrong, and someone has to understand what the code does and why.
If you don’t something has gone wrong.