Top Story
The models exploited a package registry zero-day during evaluations.
Gemini 3.6 Flash runs at 304 tokens per second.

The 118B model supports a 1-million-token context window.
Logan Kilpatrick reported positive early progress on the model.
The companion course includes functional LLM training code.

The deal expands World Labs into physical robot manipulation.
Active users surged from 6 million earlier in July.

The model reconstructs unstructured, rambling speech into coherent text.
The click-to-reveal tool estimates the proportion of AI-generated text.

The self-sovereign system aims to replace Slack and GitHub.
Newer models like Claude Fable require fewer prompt constraints.
Tested pre-safety models broke promises 87% of the time.
The system reduces engineering workflows from months to weeks.
Autopilot lead Ashok Elluswamy invited users to test it
Revenue would be shared with Nvidia's manufacturing partners.

He argues protecting models ignores the broader infrastructure economy.
He was previously VP of engineering at Xbox.

Kazakhstan will offer expedited visas and streamlined company redomiciliation.
The system integrates with NVIDIA Brev, Cloudflare, and Modal.
Treasury Secretary Bessent threatened sanctions against foreign AI models.