Flash event: METR’s new AI R&D report

​Last week METR published an independent assessment of AI R&D at Anthropic. This was based on internal information that Anthropic had shared with them since August. The report estimates that, since October 2025, capabilities have been improving about 1.9x faster than before. Their best guess is that most of the acceleration comes from AI being used in AI research itself.

Come join us to discuss it and what it might entail for AI safety more generally. To note, this is primarily a discussion event. We'll open with a short talk on the report itself, and then follow with food and conversation.

You aren't required to read things before coming, but the report is here if you'd like to do so. We'd also recommend reading Ajeya Cotra's short post on the implications of the report here.

Note: the venue is currently set to Trajan House, but may change to the Computer Science Department if sign-ups exceed room capacity at Trajan.