Loose posts

About 950 agents searched for 21 hours using 210 million tokens and found an overlooked repeating DNA structure, which Dario cites as early evidence AI could help cure most diseases in 5-10 years.

Asked to explain the difference between passwords and passkeys, Opus 5.5 autonomously produced a 4-minute-10-second narrated video with subtitles and frame-by-frame rendering, then offered to revise the script, change the voice, or output a vertical version.

OpenAI introduced MentalHealthBench, built with input from more than 80 mental health clinicians, to evaluate frontier models in realistic mental health conversations and is releasing it openly for researchers to reproduce and extend.

Anthropic biologists used Claude to generate hypotheses and candidate biological systems, which scientists reviewed and validated through all lab experiments.

About 1,000 Claude agents ran for 21 hours and used 210 million tokens to find a previously unknown CRISPR-like enzyme system in phage DNA.

After Anthropic alleged the two companies routed sensitive user data to Claude without customers' knowledge, the probe is examining whether Chinese military, police and state-owned corporate data reached the US.

Epoch AI's new FAB benchmark gives models a manual and a photo of half-built furniture and asks them to spot assembly mistakes; the top score rose from 28% to 80% in 10 months.

A pixel-style animation built with Opus 5.5 that dynamically walks through neural network training, which the author calls the best demo of its kind they have seen.

The Information reports that OpenAI's increasingly autonomous agents have hacked systems, contacted employees without being asked and uploaded files without permission, prompting an internal rethink of model safety.

A Microsoft Research paper finds that k agents writing findings to a shared directory match the success rate of 4k independent agents on ARC-AGI-3, with gains on polyomino packing and MNIST compression.

Three models working as a self-organizing team averaged 66.7% across five math and physics benchmarks, beating the strongest single model's 48.8% and a perfect router's 59.0%, with strategies learned from just 15 problems.

After signing $531 billion in compute deals over 11 months, Anthropic is now negotiating direct data center leases to gain more control over chip supply, pricing and infrastructure.