GPT-6 Astra saturates benchmarks, but Anthropic's Fable 5.1 retakes crown 48 hours later
PeterDiamandis · x · 2026-09-08
Peter Diamandis rounds up a frenetic AI week: GPT-6 Astra hit 99.9% on ARC AGI 3, 98% on FrontierMath and 100% on Exploitbench, yet Anthropic's Fable 5.1 retook the broad-capability crown within 48 hours and formalized Fermat's 358-year-old proof across 13 million lines, proving 29,000 new theorems en route. Tesla also launched Cybercab robotaxis at $30K vs Waymo's $100K+ per vehicle, and GLP-1 drugs extended mouse lifespans by 100 days.
More from Embodied
- Meta Remotely Disables Recording on Thousands of Tampered Smart Glasses Hiding LED Indicators — Polymarket · 2026-09-08
- Cybercab's entire outer body is plastic: Tesla's RIM process cuts emissions 35% — MatthewChang · 2026-09-08
- NVIDIA's VoLo lands at CoRL 2026: a VLM agent that orchestrates robots through long-horizon tasks — erwincoumans · 2026-09-08
- AI designs a full flat-six engine from a text prompt — tool-free assembly, FDM-printable — ycombinator · 2026-09-08
- Dev builds macOS writerdeck on a 60Hz hi-dpi E Ink display, demos typing and browsing — Bedrovelsen · 2026-09-08
- MIT's Phillip Isola on the advent of 'robot-use agents': an inflection for robotics — phillip_isola · 2026-09-08