Post #3875391
2026-07-16 22:12 UTC
@fell@ma.fellr.net @torvalds@social.kernel.org The main reason I'm reluctant to use these is the "power dynamics" with having to go through a big tech giant to use the actually useful models.
Also the intrinsic error rate, but that seems to slowly be going down.
Replies (1)
-
@fell@ma.fellr.net 2026-07-17 03:46
@sjb@mstdn.io They seem to get pretty good at doing the right thing correctly, that's for sure. I had exactly the same feeling feeding mine and others works through Microsoft or Anthropic, but that goes away as soon as you hear that your coworkers are already using it heavily. At that point you might as well. I got curious, and every second attempt yields a successful useful change. The other times the model misunderstands me or barks up the wrong tree. Not really a time saver to be honest. 🤷🏻♂️