> How do you know we aren't already through the exponential part of the S-curve

From a recent METR report, benchmarks point towards an acceleration, LLMs are getting better at a faster rate. There's no sign that we are about to reach the top of the S curve.

> increasingly high costs of training frontier models

Somewhat surprisingly the gains do not seems to come from more compute, we have on the METR report:

> We are confident that the acceleration was not due to an acceleration in the scaling of training compute.

Link to said report: https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleratio...

> Somewhat surprisingly the gains do not seems to come from more compute

I think this could form part of an argument that acceleration is slowing. It's only because we are not making as much improvement out of hardware gains that we're having to find the improvements in more complex architecture and more tricky to create training data.

The fact that the gains are not coming from the traditional place gains come from in technology might be a sign that a high percentage of the low hanging fruit has now been picked.

>From a recent METR report, benchmarks point towards an acceleration

On an internal benchmark with internal models, if you look at the public benchmark with public models there is no acceleration. That report also has a big graph labeled "fake data".