10:24 PM
Today in brief.
AMD buys Fei-Fei Li's World Labs for $8.2B
AMD is acquiring World Labs, Fei-Fei Li's spatial-intelligence lab, in a deal reported around $8.2 billion. World models move from research darling to silicon-company core bet.
🤖 AI
OpenAI DevDay teaser: "We have found a new thing"
OpenAI posted a cryptic "Get ready" and Sam Altman said he is excited for DevDay tomorrow because "we have found a new thing." Forward signal, not a spec sheet.
🎬 Media
Kling 4.0 locks October: 30s, 15 refs, 4K HDR
Kling 4.0 is set for an October launch with 30-second clips, up to 15 references, 4K HDR, and better audio plus lip sync. Creators are already posting one-prompt, no-cuts realism samples.
2 sources 4:27 PM
🎬 Media
ElevenLabs v4 lands on Runway for controllable speech
ElevenLabs v4 ships a new architecture for more realistic, directable speech and soundscapes, and Runway has it live today next to its image and video models.
2 sources 3:38 PM
💼 Business
AI power tools: 93% at OpenAI, 3% at a typical company
a16z put hard adoption numbers on screen: OpenAI's own team uses AI power tools weekly at 93%, the top 10% of companies at 19%, and a typical company at 3%. The gap is the story.
2 sources 9:17 AM
🤖 AI
NVIDIA OpenShell puts real walls around agents
NVIDIA ships the Open Agent Safety Platform with OpenShell runtime limits, BlueField-4 watchdog hardware, and 100+ partners. Perplexity published the hard eval: 9 models, 108 root-access runs, zero VM escapes.
+59 sources 4:21 AM
🤖 AI
Claude Sonnet 5.5 is live as the cheaper work partner
Anthropic shipped Claude Sonnet 5.5 today, framed as a significantly cheaper and faster work partner after last week's Opus 5.5 wave. The Monday expectation from Sep 27 just became a download button.
🤖 AI
Shopify opens checkout to browser AI agents
Shopify is letting browser-based AI agents complete checkout. Agents move from research toys to something that can spend money with a merchant's blessing.
Try today: MicroLLM Lab and a 4B decision model
Two local/open doors you can click: MicroLLM Lab runs 7 tiny LLMs in the browser, and ImaJev-4b claims #1 of 91 on JevBench after a 15-day fine-tune, ahead of GPT-5.6 Luna on DecisionBench.
No stories in this category for this edition.