Definitely would be good to see more objective figures but benchmarks arent always great samples either.

If there was a silent degradation then the recent chatter about suspected nerfing is what you would expect to see. Of course that's not evidence, but I dont think we can say it's far more likely that its all subjective.

I mean the basic intuition doesn't seem unfouded. Maybe people can confirm/deny this contention for compute. But Anthropic is compute constrained. Inference takes compute. Model development takes compute. Isn't it plausible that there will be periods of contention? Or do we have information that suggests inference compute is affected by model development?

It feels more like a crowded casino where someone asks "Is it just me, or is this table cursed?" and 50 people all run to agree.

Please stop gaslighting people. I'm tired of people claiming that my subjective experience has changed and not the observable behavior of the models.

Anthropic has confirmed that they make changes to the models, for example with the recent update to limit reverse engineering.

I think it's similar to how an artist's eye is ahead of their hand [1]. As you get more experience with the models you start to notice their shortcomings more and more. When a new model comes out, everything seems crazy good for a while, until you start to realize the weak points again. Benchmarks aren't everything, but people have created numerous benchmark trackers because of this phenomena and no one has caught degraded performance yet, other than this post catching that something changed (no connection as of yet to performance though)

[1] https://www.rensoart.com/why-your-eye-is-ahead-of-your-hand-...