NVIDIA opens preorders for RTX Spark Windows laptops, shipping Oct. 16, with compact desktops due in November. NVIDIA says the chip pairs a Blackwell GPU with a Grace CPU, up to 128GB unified memory and one petaflop of FP4 compute. Microsoft also makes Microsoft Execution Containers generally available for agents on Windows.
Liquid AI releases two open-weight decision models on Hugging Face: d1-3B and d1-omni-600M. The company says d1-3B scores 48.57 on its Decision Index v0.2.1, ahead of every model under 10B. It reports 8 ms answers on an RTX 4090 and 50 ms on a Jetson Orin Nano. d1-omni-600M handles text, image and audio.
Elon Musk says Grok Bot will route tasks to rival models, including Anthropic's Claude Opus 5.5, Midjourney and Suno, in a Wednesday post on X. He wrote that SpaceX will use "the best back end model for any given task." Musk gave no detail on which tasks go to which model.
Google launches Playground, an experimental browser-based platform where users 18 and older in the U.S. build and share custom games from text prompts, no coding needed. Creation access is tiered by Google AI subscription. Games can be private, shared by link, or published to a gallery. A Unity Spark integration is coming.
Tony Fadell, the iPod and Nest founder, says the Rabbit R1, Humane Ai pin and Limitless pendant failed because they met no real need. Speaking at MIT Future Fest, he says fewer than 0.01% of people have had an assistant, and predicts successful agents will run on-device. He names Apple as best placed.
LlamaIndex launches OpenDocRouter, a single API that converts PDFs and images to markdown using 10 frontier and open-source models, including Claude Opus 5.5 and MinerU2.5-Pro. Billing is per token with credit top-ups from $25, and layout output adds $0.2 per million tokens. ParseBench cost per 1,000 pages ranges from $0.80 to $48.82.
Scale Labs introduces Humanity's Sixth Sense (HSS), a benchmark of image and video items testing intuitive visual reasoning across temporal, spatial, social and abstract structure. Human participants reach 93.1% accuracy, while the top model, GPT-6-astra, scores 53.6% at maximum reasoning effort. An agentic setup with dynamic visual manipulation narrows but does not close the gap, the researchers say.
The Internet Watch Foundation says it has assessed 6,310 AI-generated images meeting the legal definition of child sexual abuse so far in 2026, 40% above the more than 4,500 in all of 2025. Most depict girls aged seven to 13. IWF policy head Hannah Swirsky backs binding UK legislation forcing companies to build safer models.
Apple may launch Siri AI smart home devices as soon as October 13, according to Bloomberg-style Apple watcher Mark Gurman. He reports a 6-inch square display hub with a FaceTime camera that personalizes content by voice and facial recognition, plus upgraded HomePod mini and Apple TV models. A doorbell, thermostat and cameras made with LG Electronics are also reported.
Xreal opens pre-orders for its Aura Android XR smartglasses at $1,279 for the 12GB RAM, 256GB model, with a 16GB, 512GB version at $1,499. Reservation-deposit holders order first. The 95-gram glasses tether to a puck running Android XR, with a 70-degree field of view. Shipping starts soon in the US, Canada, Japan and South Korea.