Google’s multi-agent framework keeps long-form AI video coherent
Google Research describes a unified multi-agent system that plans visual continuity across shots to reduce drift and cascading errors in long-form AI video generation.
AI & ML
Frontier models, research results, agents, and developer tooling.
NVIDIA Nemotron 3 Diarization Targets Real-Time Speaker ID
The new model helps developers build systems that know who spoke when in live, multi-speaker audio.
Gemini 3.8 Live Adds Real-Time Avatars for Enterprise Agents
DeepMind's new Live Avatar feature pairs Gemini 3.8's live dialogue with streaming video and lip-synced speech, letting enterprise agents converse with a visible, responsive persona.
Faulty Firmware Update Bricks Samsung Bespoke AI Fridges
A botched firmware update disabled Samsung’s Bespoke AI fridges during the Chuseok holiday, forcing owners to throw out food.
Ringg says OpenAI agents resolve up to 65% of customer calls
Ringg says its OpenAI-powered agents now resolve up to 65% of customer calls and handle over 7 million calls a month.
Gemini 3.8 TTS models add custom voices and line-by-line direction
Google DeepMind's new text-to-speech models let creators design original voices and direct performances line by line, with built-in watermarking.
AI Arms Race Heats Up in Bio-Defense, Says Radical Numerics CEO
A startup is applying chain-of-thought AI to biology, aiming to outpace adversaries in the race to design and defend against new genomes.
XPRIZE Wildfire AI Spots Blazes in Minutes but Can't Stop Them
The XPRIZE Wildfire competition showed AI can detect fires within minutes, but autonomous suppression remains out of reach.
NVIDIA Warp and MJWarp Bring GPU-Scale MuJoCo to Robotics
A new workflow moves MuJoCo models to GPU-accelerated parallel environments, scaling to thousands of simulated worlds.
Google Beam expands to six countries with new hardware and workspace partners
Google's immersive video-call technology reports internal gains in connection and meeting efficiency as it launches in new markets.
NVIDIA's Sakeena Fiza: Validation Is Detective Work for AI Hardware
A profile of NVIDIA validation engineer Sakeena Fiza reveals how AI hardware reliability depends on engineers who deliberately try to break systems before customers do.
OpenAI Expands Academy with Role-Based Learning Paths
OpenAI's expanded Academy now offers learning paths tailored to five audience groups, focusing on practical AI skill-building.
Google DeepMind Adds Private Persistent Memory to Cloud AI
A new server-side memory layer aims to give cloud-based assistants long-term context while keeping decryption keys on users' devices.
Xiaomi MiMo-V2.6-Pro tops open-weights models with $3M training run
Latent Space reports that Xiaomi's MiMo-V2.6-Pro 1T-A42B is now the leading open-weights model, trained for $3 million.
Parallel Cuts Research Time and Cost in Half with GPT-6 Astra
OpenAI reports that Parallel's agents using GPT-6 Astra halved both time and cost for labor-market research and synthesis.
GPT-6 Prompt Caching Boosts Hit Rates, Adds Diagnostics
OpenAI's improved prompt caching for GPT-6 promises higher cache hit rates, lower costs, and new tools for developers to monitor and optimize cache performance.
Google’s ERA Uses LLM-Guided Search to Automate Science, John Platt Says
In a Latent Space podcast, Google researcher John Platt describes an “auto-Kaggle” system that turns scientific problems into score-maximization tasks and has already produced at least ten papers.
UK AISI and EvalEval Aim to Make AI Benchmarks Reproducible
A new collaboration focuses on standardising evaluation practices so benchmark results can be trusted and repeated.