Categories
AI News

Google’s Agent Teams Wrote 1.8M Lines of Proof. Humans Still Pick the Question

Google’s new /teamwork has Gemini agents check each other’s work. It formalized the Weil conjectures in Lean. Humans still vet the question.

Categories
iRaluca

Give Everyone Their Own AI Agent and the Group Does Worse, Study Finds

Researchers gave each user their own AI agent. In all four tests the group did worse than with one shared agent. Simulated, but telling.

Categories
AI News

DeepMind’s Agents Proved Signal’s Compiled Code Secure, Within Limits

DeepMind’s agents wrote ~6.5M lines of Lean proving Signal’s compiled code keeps key security promises. Humans still trust 26,000 lines.

Categories
AI News

An AI Edited 1,023 Weights in a Small Model. Its Math Score Quadrupled

GPT-6 Astra edited 1,023 weights in Qwen3.5-0.8B, with no training, and its MATH-500 score rose from 8% to 38%. It doesn’t work every time.

Categories
AI News

DeepMind’s Agents Made Old RSA Cheaper to Attack. The Human Sent 20 Messages

DeepMind’s AI agents found two ways to weaken legacy 1024-bit RSA. The papers say 2048-bit keys are safe. A human mostly clicked approve.

Categories
AI News

Anthropic Wants Claude Agents to Dream. Mostly, That Means Editing Notes

Anthropic says Claude agents remember better by “dreaming”: rereading logs to fix notes. The gains are real but uneven, and self-reported.

Categories
AI News

Ai2’s ToyBlocks Teaches Robots on a Tabletop. Spelling Is Still Hard

Ai2’s open ToyBlocks is a simulated arm-and-blocks world with 85,336 demos. Its robots spell words about half the time. The scores are self-reported.

Categories
AI News

METR Says AI Roughly Doubled Anthropic’s Pace. Most Evidence Is Redacted

METR spent 11 weeks in Anthropic’s data and says AI R&D now runs about 2x faster, mostly thanks to AI. Much of the key evidence is redacted.

Categories
AI News

x = y Is Where Even Frontier Models Lose Track, Says Ai2’s StateBench

Ai2’s open StateBench tests how models track changing values in long text. Frontier models fall to 13% or lower on the hardest level.

Categories
AI News

FlashProve Can Prove Which Model Answered You. Slowly, for 1,024 Tokens

Anthropic’s open FlashProve proves which model wrote a reply without revealing its weights. It costs ~70x the compute and stops at 1,024 tokens.