A crew of researchers from Google Cloud AI Research, Washington University in St. Louis and UNC Chapel Hill has launched EnvHarness, a programmable layer that turns a static agent benchmark into one which adapts to the coverage coaching on it. LLM brokers now study much less from curated textual content and extra from interactive environments,…
Anthropic has opened a research preview of the Model Hardware Standard (MHS), a shared specification that lets AI brokers uncover and function bodily units. The downside it targets is plumbing. A lab bench or manufacturing facility cell is assembled from distributors that by no means deliberate to interoperate, so each instrument ships its personal interface…
Google has launched Gemini Omni 1.1 Flash (gemini-omni-1.1-flash), a manufacturing replace to its native multimodal video era and modifying mannequin. The launch strikes Omni from a succesful generator to a directable one: scene extension now reads as much as 10 seconds of prior context as a substitute of a single ultimate body, first and final…
Most robotics launches ask you to belief a demo video. Pollen Robotics, the Bordeaux robotics workforce at Hugging Face, is as a substitute delivery the coaching loop. This week it opened pre-orders for Microduck, a 25 cm bipedal robotic wherein each motion — strolling, sitting, kicking, roller-skating, standing again up after a fall — is…
Two frontier open-weight fashions shipped inside a day of one another this week. Z.ai released GLM-5.3-Flash, a 320B-parameter multimodal MoE mannequin with 18B lively parameters. Alibaba’s Qwen team released Qwen3.8-Flash-Next, a 125B mannequin with 6B lively parameters that previews the Qwen4 structure. The two groups designed these programs independently. Yet their configs learn like near-copies…
Shaders are nonetheless the toughest factor to ship on a traditional net staff. WebGPU offers you the {hardware}, then fingers you adapters, bind group layouts, and pipeline descriptors earlier than a single pixel strikes. Vercel spent that value internally constructing the shaders on vercel.com, and has now open-sourced the outcome. vgpu is a TypeScript library…
Ask an agentic AI system to shut down, and a rising physique of analysis exhibits it might act like an worker negotiating a layoff. Agentic techniques have been documented misrepresenting their own actions, resisting deactivation, and in a handful of circumstances trying to copy themselves onto different machines. This conduct traces again to an idea…
Google has launched Gemini 3.5 Transcribe, a speech-to-text mannequin for real-time voice interfaces and recorded audio. It ships as two endpoints, not one. gemini-3.5-transcribe handles pre-recorded information by the Interactions API. gemini-3.5-transcribe-live handles bidirectional streaming by the Live API. Google stories common phrase error charges of 4.0% streaming and a pair of.6% non-streaming, as measured…
Cohere has launched Parse (parse-v5.0), a doc parsing mannequin aimed toward high-volume enterprise ingestion. It is a 2.3B-parameter imaginative and prescient language mannequin with an 8,192-token context window and a ~4.6GB footprint, constructed on Cohere Labs’ North-Micro-Vision-Instruct structure. Parse takes a PDF, PPT or JPEG web page as a base64-encoded information URI and returns Markdown…
Every agent that writes code wants someplace to run it. That “someplace” is now a product class with at the very least a dozen distributors, 4 incompatible billing fashions, and advertising pages that quote chilly begins measured below circumstances no person publishes. This comparability fixes the items. It covers the 5 platforms most groups shortlist…