Comment Re: At least half true. (Score 1) 87
Putting out unvetted AI coded patches is no more unsafe than putting out human coded patches that aren't checked out.
Separately this is untrue given at least humans have metacognition.
Did I trigger you with the FTFY comment? Relax a bit, take a break.
As a person who has been around coding since the 1970's, starting on good old IBM360's, I've seen plenty of horrid human code that is just fail. Between bitched up code that would cause a core dump and use up huge amounts of printer paper on the 360's to Windows updates that cause endless boot cycles - It was improperly checked updates and patches.
I've also seen some Claude based code that is tremendous, much superior to the code it replaced.
How does this happen? It is vetted out, just as you would vet out any code, human or AI.
Yeah - metacognition, that's right. Same thing as checking out the patch
But perhaps more importantly, simply bitching about AI coding is not going to change it. I've seen the slop, but it won't remain slop forever. As I noted in another post here, I'm using a Software Defined Radio application written by Claude, and vetted out by humans before release. And the program is already superior to the human-written version.
Dunno what to tell ya, other than make certain you are a top tier coder, because the AI is moving up the chain of competent writing. Because you can tell me I'm wrong all day, substitute metacognition as some counterargument to "checking out", "vetting out." Frankly, I find the terms essentially interchangeable.