The new self‑evolving language model improves the Cosmos‑3 AI's benchmark results, surpassing Veo‑3.1.
NoteOossa1 min read
Researchers from NVIDIA, MIT and the University of Oxford released Physis‑Lang, a self‑evolving physical language for AI. The system lets a video‑world model reason about physics by using language instead of extra visual or numeric cues. In tests, Physis‑Lang lifted the Cosmos‑3 model’s score above the previous best Veo‑3.1 on standard physics benchmarks. The work was posted on Sep 30, 2026.
Better physical reasoning can make AI‑generated videos look more realistic and reliable.
Research
Former OpenAI, DeepMind and Anthropic scientists say there may be a 10‑50% chance AI could cause human extinction.
Research
The Blue Cross Blue Shield Association says hospitals’ use of AI tools to prepare insurance claims was linked to $942 million in extra healthcare spending over two years. Its analysis found more patients coded as having complex conditions, without evidence of matching changes in care.
Research
A research team reported that their AI system, built from a large language model and a rule‑based playbook, outperformed 17 practising audiologists on 58 simulated cases in English and Chinese.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
| # | Source | Outlet | Date | Key takeaway |
|---|---|---|---|---|
| 1 | NVIDIA Researchers Introduce Physis-Lang: Self-Evolving Physical Language That Lifts Cosmos 3 Past Veo 3.1 on Physics Benchmarks ↗ | MarkTechPost | Sep 30, 2026 | Video world models can render convincing clips that still break physics. |
1 sources
