Except ai is actually good at improving performance and finding bugs. Literally. And every time an author denies this I feel like they or I are smoking crack.
Some combination of ego or loss of self seems to make people totally unable to admit that an llm can produce code as good as (i would say better) than them.
Except ai is actually good at improving performance and finding bugs. Literally. And every time an author denies this I feel like they or I are smoking crack.
AI can find bugs, and it can find targets for performance improvements. If you tell it to vibe code something that's fast, you're not going to get the results you want. Like the time Anthropic mostly but not quite created a compiler: https://www.anthropic.com/engineering/building-c-compiler. And if you want it fix things that are complex, you're going to have a bad time: https://1password.com/blog/why-ai-generated-patches-still-re...
Some combination of ego or loss of self seems to make people totally unable to admit that an llm can produce code as good as (i would say better) than them.