Remix.run Logo
rubicon33 4 hours ago

But actual progress seems to be slower. These modes are releasing more often but aren’t big leaps.

gallerdude 4 hours ago | parent | next [-]

We used to get one annual release which was 2x as good, now we get quarterly releases which are 25% better. So annually, we’re now at 2.4x better.

minimaxir 3 hours ago | parent | prev | next [-]

Due to the increasing difficulty of scaling up training, it appears the gains are instead being achieved through better model training which appears to be working well for everyone.

wahnfrieden 4 hours ago | parent | prev [-]

GPT 5.3 (/Codex) was a huge leap over 5.2 for coding