alt Hacker News> heh apt username
?
> Not worse than humans have been doing on their own since forever
If it's not reviewed well (some people stopped even looking at the code) it can be immensely worse.
And most people end up producing much more code with LLMs, quickly reducing how much of it they review; a lot have become fine with turning the interface upside down often just because it's easy, even if the new interface is worse and a lot buggier.
> If it's not reviewed well
When was the last time you ..or anyone.. reviewed the assembly language or binary output of your compiler?
...
What if the CODE DOESN'T MATTER anymore?
No really, wait:
What if all that matters is the result?
First, you ask the AI to make what looks correct.
Then you ask it to make it work correctly.
Then you ask it (or another adversarial AI) to do a security review pass.
Then a performance benchmark.
Then a few more iterations of whatever.
If all pass, does any human really NEED to look at the code?