Fireworks AI M3 Blackwell kernel hits ~980 TFLOP/s

Salesforce's Dreamforce 2026 session pairs Ethan Mollick with Paula Goldman to discuss which human skills, judgment, and instinct stay irreplaceable as AI becomes a daily collaborator.

Speaking as guest of honour at the IDEAS 4 INDIA Indian American Tech CEOs Forum in Seattle, Microsoft CEO Satya Nadella highlighted India's growing role in global technology and AI innovation.

Suleyman, Microsoft's AI CEO, said AI models need to stay aligned with humanity's interests and called that the most responsible path for the tech industry. The CNBC Squawk Pod conversation closed a week of AI warnings from the show's guests.

Tao's keynote at the Science x AI Summit at Caltech on September 11, 2026, co-organized by SAIR, Caltech, and the Merkin Center for Pure and Applied Mathematics, previews the Open Math Model initiative. He notes AI now generates and verifies formal proofs faster than mathematics can absorb them.

Grok Voice Transcribe 2.0 hits 2.7% WER at 0.49s after end of speech on AA-WER Streaming, ranking first among 32 streaming models. SpaceXAI claims 2x the accuracy of version 1.0 at the same $0.10 per hour price.

Claude Code v2.1.278 switches auto mode for Claude API and Enterprise users, plus Bedrock, Vertex, Foundry and gateways, to the server-side classifier, which does not charge for classifier overhead. CLAUDE_CODE_AUTO_MODE_SERVER=0 opts out on Bedrock, Vertex, Foundry and gateways; a new Auto mode server row in /status shows where the classifier runs.