Mel Gibson 2.0

how long until AI solves a fairly famous open math problem from a century ago

Claude Mythos is scheduled for Q3 2026. We're currently still in Q1 2026. Anthropic will certainly factor in compute limitations and project how good the model will be. However, if it's already this good *now*, and NVIDIA's Vera Rubin GPUs aren't even live yet, what will we see by the end of the year? Or in 2027? I'm increasingly convinced that Dario is right and superintelligence will become a reality. But there's also a caveat: Anthropic is also planning its IPO for Q3 2026. I certainly see a connection there.

We have removed a problem from FrontierMath: Open Problems. The problem was solved by AI, but upon review we determined that the problem didn’t meet our minimum bar for mathematical notability. This is a different problem from the one whose solution we announced on Monday.

Anthropic’s website content management system included publicly accessible draft blog posts and assets, including Claude Mythos and Anthropic CEO Forum

Jack Clark another amazing Co-Founder at anthropic also holds a similar sentiment That the models in machines of loving grace will come to fruition by the end of this year and be deployed early next year "Recursive self-improvement, in the broadest sense, is not a future phenomenon. It is a present phenomenon," - Evan Hubinger, who leads Anthropic's alignment stress-testing team, as well.

Mythos is a new tier, bigger than Opus, and more capable. It will be beyond the current state of the art by a wide margin, a dramatic jump in intelligence. Things have been accelerating under the surface since December. Capabilities are increasing faster, the time between model releases is getting shorter. I think OpenAI's recent decisions and refocus have been driven by the fact that it is increasingly obvious to some people working in the industry that we are beginning to take off. I think there have been multiple unannounced breakthroughs over the last few months, in more than one lab. I think there have been advances in capabilities, in alignment, and in architecture. The loop that started in December has begun to bear fruit. This model will be huge, very expensive to serve, and incredibly powerful. It has already finished training, and is in early rollout to some enterprise partners so they can prepare for what's about to arrive. Once we get benchmarks I think it will become clearer where we are; we are already inside the singularity.

We scored 36.08% on ARC-AGI-3 in one day using the Agentica SDK.

Anthropic’s new model, Capybara: “Compared to Claude Opus 4.6, Capybara achieves dramatically higher scores in software coding, academic reasoning, and cybersecurity.” According to Dario's previous interview, it might be a 10T-parameter model that cost $10 billion to train.