All AI news

AIZyphra derives a mathematical theory showing how local layers produce recency bias that propagates through the residual stream, norms, and MLPs into NoPE attention logits. The authors validate the theory on both randomly initialized and trained networks.

AIIn Zyphra's experiments, models with smaller context windows predicted text better while using less training compute. Zyphra believes the stronger recency bias induced by smaller windows outweighs the reduced compute benefit.

AIHybrid NoPE models combine sliding window attention or recurrent layers, which focus on nearby words, with global attention layers that use no positional encoding (NoPE). The post notes that NoPE layers receive no positional information yet can still learn long-range dependencies, and raises the question of how this works.

AIZyphra says that mixing information within overlapping neighboring windows makes nearby words carry similar information inside a model. This creates a built-in clue about how far apart words are, even before training begins, which the post calls recency bias.

AIZyphra Research explains how language models track word order without explicitly encoding position in attention. The post says local memory layers that read nearby words help global attention layers preserve sequence information. The source is a short teaser thread, so no further technical details are given.

AIMicrosoft AI announced three new models: MAI-Transcribe-2-Streaming, MAI-Voice-2.1, and MAI-Voice-2.1-Flash. The source says the streaming transcription model is accurate and aims for natural speech with less waiting between conversational turns for building voice agents.

AIPerplexity Computer can now generate interactive charts and visualizations directly within a user's thread. For financial data, it uses TradingView Lightweight Charts to display candlesticks, volume, and moving averages.

AIGoogle Flow, built with and for creatives, is announcing a series of interactive virtual workshops focused on the craft behind modern creative workflows. The post invites readers to RSVP for the workshops.

AIHugging Face shows that training LFM2.5-2.6B with RL inside the agent harnesses themselves lifted held-out task success from 42% to 54% across four harnesses. Before training, the model solved 62% of tasks in Mini-SWE-Agent but only 33% in Claude Code, so the same model behaved very differently per harness. The approach uses an OpenEnv capture proxy to record tokens and logprobs, Harbor for tasks and sandboxes, and TRL's async GRPO trainer, with 31% fewer tool calls on already-solved tasks; training in OpenCode alone mostly improved OpenCode.

AIRunway introduced Visual Thinking, a new harness for coding agents that uses a real-time video model to visualize each step and tool call as it happens. The visualization lets users stay in the loop during complex autonomous workflows.

AIRunway says its system can expand any media into an interactive, real-time world simulation. The post presents this as turning fixed experiences into immersive multimodal simulations for both people and agents.

AIRunway describes Portals as visual elements with soft, dream-like edges that mark where generated content begins. Users can drag a Portal onto the page, where it sits alongside their content.

AIRunway says its real-time video model generates animated, interactive overlays called portals that illustrate any concept, place, or simulation while users browse. The post describes the feature as a demonstration without listing pricing, availability, or technical specifications.

AIRunway says its video model recomposes a scene with spatial awareness when the interface is resized. Users can click inside the output to interact with it directly in real time.

AIRunway Labs introduced Project Continuum, an operating system research application built around real-time video interfaces. The early look shows four interaction concepts: Portals, Visual Thinking, Responsive Video Interfaces, and Interactive Worlds. The post says Interface World Models such as Solaris are reimagining what an interface can be.

AIHugging Face announced that users can now use MCPs (Model Context Protocol servers) to bring their own data into Hugging Face Chat's ml-intern mode. The post links to the feature at and gives no further details on setup or supported integrations.

AIAmir Efrati outlines three ways xAI sped up its data center construction: paying well above market rates for certain labor, prefabricating portions of the facilities elsewhere before moving them to the site, and using industrial robots.

AINVIDIA CEO Jensen Huang will join ElevenLabs co-founder Mati Staniszewski for a conversation on the future of AI at ElevenLabs Summit NYC on November 11. The post notes that NVIDIA has partnered with ElevenLabs since the company's earliest days.

AISenseTime says its SenseNova 6.8 Flash Preview turns static images into dynamic video with its Dynamic Design feature. The post shares a short panda animation as a demonstration.

AIGoogle Workspace says Google Pics can generate custom product mockups and update Slides with a single click. The post says this works inside Google's own tools without copy-pasting or third-party apps.

AIAMD says another Helios system has entered deployment, with OpenAI's Vamsi Boppana and Uday Ruddarraju touring the lab last week. The post frames the visit as joint work on pushing the frontier of AI infrastructure, but gives no specs, benchmarks, or deployment details.

AIDatabricks has introduced ai_decide, a new AI Function for fast, structured decisions over governed data. It classifies, scores, and chooses next actions in a fraction of a second, with lower latency and cost than an LLM on similar tasks. It is suited to model routing, document processing, agent evaluations, and real-time app logic.