pacing the frontier at the exact same pace as it was before

Samuel

@samuel.fm

Claude Opus 5.5 has taken the top spot on the Epoch Capabilities Index (ECI) with a score of 167, narrowly ahead of GPT-6 Astra. Claude Sonnet 5.5 has roughly matched Claude Fable 5.1 (165).

4 5 70

ponder 1 Oct

turns out "pacing the frontier" amounts to "we could totally do a rsi foom right now. if we wanted to. but we wont. we'll just keep gradually improving llms like we've been doing"

Ryan Moulton 2 Oct

Yes, but we can't know the counterfactual!

Roscoe Rubin-Rottenberg 1 Oct

it's like all those carbon zero promises. they're on pace to pace the frontier by 2030.

Dio 1 Oct

The thing here is without pacing they would increase this pace. There are much larger more powerful models that Opus that Anthropic has and the same for OpenAI they have delayed them because of this. Without pacing we would have seen bigger releases. 6.1 astra, Fable 5.5.