technologyHowever, during the model’s training, we took steps to limit advances in the specific research capabilities that could most directly accelerate AI development. Thus, in relevant areas, including accelerator kernel optimization and improving large language model architectures, Fable 5.2 and Claude Mythos 5.2 perform at about the level of Opus 5.5.
https://archive.is/JknnP
its important to remember that for every truth you read from these charlatans there is hidden in it at least three lies
I mean, haven't a lot of these companies been talking about slowing or halting development? Anthropic has a lot of cultists, so I wouldn't be surprised if they actually believe they're slowing down the Apocalypse God they're trying to create.
Every day Anthropic are like "oh our models are soooooo good at doing x and y, but you'll never get to see it because, uh, the actually good functions go to another school"
I think this is just "Recursive Self Improvement" bait for investors. There's no proof that it's really happening, and by adding these artificial limitations, they can hide how useless their models are at it.
racistly: surely this time the Chinese won't surpass us technically!
