Question, I check HN often, and noticed the trend towards "it's obvious that was written by AI so I didn't read it", or couldn't handle it, or hate the way the prose it's written.
And a similar behavior towards AI generated code, also mirrored by multiple AI policies around AI submitted patches/PRs.
Now question, say you have an idea about x product, or y improvement as a patch to a OSS, or whatever where AI can help you, then you
a) outsource to this team of people, local to you, same language, you can go and say hi to the person working on it, then don't review the output and put on github b) team is in your country, more difficult to go say hi, but you don't review the output and put on github c) same but the team is in another country d) same but the work is done by a contractor of a contractor, again you don't review and put it on github e) ask it to the free chatgpt version, don't review, post to github f) ask it to haiku on your 20$ claude account, don't review, post it to github g) ask it to claude mythos... the version that hacks stuff as has to be contained in a magnetic field air gapped... don't review the output and post it to github...
where does it start being slop and where is not? now, if instead of that, I review the output, read it, understand it, dogfeed it to myself to make sure it works, test it for 3 weeks, go through 100 rounds of polishing, and then post it to github... is it the same or not?
if not then AI is not the issue... the issue is hastily done stuff that's not reviewed, that you didn't guide, that you didn't test, that you didn't check EVEN if it has "the load bearing decision" 5 times.
IF the AI style writting is the issue, then polishing, ai or not, is not the problem.
NOT A PRO-AI DISCOURSE, if the thing dissapeared tonight, thanks GOD we will be able to buy RAM and gpus again at sane prices and stop the insane wave of quickly,hasted,PRs making it or not to the software we use.
Comments URL: [https://news.ycombinator.com/item?id=49289341](https://news.ycombinator.com/item?id=49289341)
Points: 1