Hacker News Reader

And then people turn around and hand-wave that away by saying: But youre assuming that the development of LLMs stop and they wont be able to address that problem.

Like Okay, it sounds like a valid point. The issue is the hand waving and the fact it is not grounded in reality.

View on HN

The assumption that it will just keep getting better is a big one. Maybe they will. Maybe they wont. If they get better I'll stop complaining. Right now I just don't think they are at the level people claim.

I took GP's comment to say that, even if LLMs get massively better in the next six months (which they will), the colleagues still have to wade through the slop during the review this week.

So while one person is getting a 5x increase in productivity, one or two get a corresponding 5x decrease.

Why would we expect LLMs to get “massively better” in the next 6 months instead of the incremental improvement we have seen to date?

LLMs didn't get massively better in the last 6 months, and most probably won't get this better in the next 6 months either.

I disagree. Couldn’t do what I’m doing with GPT-4 but I can with 5.

Reply on news.ycombinator.com