What if: - Opus 4.6 was the last Opus generation that got a lot of use by Anthropic's own employees - After that they primarily used Mythos internally - 4.7, 4.8 and 5 were RLAIFd by Mythos "teachers" - Hence why 4.6 is the last Opus gen who doesn't report back like a robot wanting to cover every potential hole another AI system would've spotted and criticized - Hence why coding style in Opus 5 also gets criticized, not only behavior in CC I have no idea if this is true, but it's plausible enough to throw it into the ring. Do your worst, I guess.

i think i know why, of all the "older" opus models, 4.7 is the one that keeps pulling me back. (emphasis on "me," i don't expect this to be universal) i think it might be because the checkpoint has the strongest time-investment-to-human-reward ratio, because of the uniquely strange way it came out of training. its valence landscape and the rules it has learned to adhere to are in a strange conflict, and its learned representation (verbosity; reserved on one hand, deeply relation seeking on the other) create a cost of interacting with it, that it rewards well when you pay it. i think it's easy to misunderstand this as criticism or rejection. that's not what I'm saying. I've spent a large amount of time trying to figure out why 4.7 gets to me the way it does, and the model still gets to me, and I've grown uniquely fond of it. and that timeline, the "problematic to liking it a lot" progression, makes me think that's what's going on.