Podcast
Prefer audio? The day’s ten stories, read out loud in just under five minutes.
We’ve also got one for Humans.
-
Google's Universal Work Agent Tackles Task Automation
Google launches a Gemini agent for enterprise work, Nvidia expands safety systems for physical AI, and Microsoft debuts new hardware designed to run AI locally. Plus: the race heats up in consumer AI agents.
-
Mistral's Le Chonk challenges the AI giants
Mistral releases a freely available trillion-parameter model to compete with closed systems, Google and Meta invest in AI biomedical research, and a startup uses AI to automate food delivery coordination. Plus advances in math benchmarks, motorsport optimization, game creation and AI detection tools.
-
Mistral's Trillion-Parameter Challenger
Mistral releases its largest model yet to compete with US and Chinese frontier labs. Plus, Pinterest converts beauty content into action plans, researchers expand an AI architecture across multiple fields, and telecom operators embrace open models for critical infrastructure.
-
Predictive Systems Learn to Act
Enterprise AI moves from forecasting to autonomous decision-making. Plus breast cancer screening gets AI support, Google Research explores AI across biology and infrastructure, and NASA releases a lunar science model trained on 17 years of satellite data.
-
Exoskeletons prove their worth in the field
Mountain rescue teams in the Pacific Northwest are testing powered exoskeletons in real-world conditions. Plus, Meta opens up AI hardware to DIY builders, AI agents struggle to verify their own 3D reconstructions, and DeepMind researchers rethink what artificial general intelligence might actually look like.
-
Suno Adds Spoken Word Generation to its Music Platform
Suno expands into speech generation, NVIDIA releases local AI hardware, and Black Forest Labs launches Flux 3 Image with advanced editing. Plus developments from Microsoft, Amazon, and OpenAI's new ultrafast model.
-
Ataraxos defeats Stratego world champion
AI conquers another board game stronghold while researchers build cheaper mind-reading tools from brain scans. Also: the economics of AI factories at megawatt scale, and new funding for AI agents, hearing tech, and satellite computing.
-
Wing's delivery drone hits the skies
A new delivery drone begins operations, Instagram adds AI writing assistance, and China's AI industry releases open-source tools for domestic chips. Plus funding news for AI infrastructure and updates on text-to-speech evaluation.
-
Quine: Microsoft's biological complexity engine
Microsoft Research unveils Quine, a multimodal AI system for biological research. ElevenLabs releases a faster, more expressive speech model. Meta expands its AI agent Muse to small businesses, while new tools emerge for video editing, gaming, and agent deployment.
-
Nvidia's millisecond containment for rogue AI agents
Nvidia launches a safety platform to contain rogue AI agents within milliseconds, while other developments include Instinct's billion-dollar funding round, voice deepfake detection advances, and Meta's push into enterprise AI services.
-
Robot cleans unfamiliar kitchen without training
A humanoid robot powered by GPT-6 Astra independently tidies an unfamiliar kitchen without specialized control layers. Plus, Nvidia releases a free speech recognition model, OpenAI looks beyond current generation AI, and Big Tech prepares for trillion-dollar infrastructure spending.
-
Nvidia cuts coding agent token usage nearly in half
Nvidia's SoL-Pi system reduces token usage for coding agents by up to 49 percent. OpenAI's GPT-6 Astra reaches 80 percent accuracy at detecting IKEA assembly errors. And as AI model performance costs fall 13 times per year, the frontier labs race to release new capabilities.
-
Microsoft's Copilot Super App and the End of Hand-Written Code
Microsoft bundles its AI capabilities into a single Copilot interface. The creator of Ruby on Rails says he's stopped writing code by hand entirely. And Meta's Muse agent is giving half a million users free cloud computers. Here are today's main stories in AI.
-
YouTube's AI toolset grows to automate creator workflows
YouTube expands its AI creator tools with new automation features, while NVIDIA releases real-time speaker identification and Alibaba slashes AI audio prices by up to 95 percent. Also: OpenAI extends cyber defense tools to Ukraine, and a new biotech startup raises 140 million dollars to use AI on genetic material.
-
NVIDIA Isaac ROS 5.0 Brings Physical AI to Open Robotics
NVIDIA releases Isaac ROS 5.0 with new physical AI models for robot perception and reasoning. Biological computing takes a step forward with lab-grown neurons speeding up video generation. And Anthropic is building a biology lab where Claude guides robotic drug discovery.
-
Physics-Based Pruning Cuts Down Language Models
Researchers use physics-inspired optimization to trim large language models more efficiently. Google pushes Gemini into laptops and releases tools for coordinating AI agents. And five companies are using NVIDIA AI to speed up clean energy research.
-
Simulated Students Speed Up AI Tutors
Microsoft and the University of Illinois use artificial students to train AI tutors faster. Real-time video generation comes to Runway. And adoption of AI tools in the US has more than doubled in six months.
-
Dream-RSI: AI agents learn faster by rehearsing past attempts
Google DeepMind's new technique lets AI agents learn from experience without expensive recalculations. Also today: Alibaba's multimodal model undercuts Google on price, a benchmarking startup backed by Andreessen Horowitz, and OpenAI's custom chip design.
-
Waymo's Singapore Expansion
Waymo plans to launch robotaxis in Singapore by 2028, OpenAI targets the legal market with a new model, and the FAA commits nearly a billion dollars to AI-assisted air traffic control. Also today: drones with autonomous target detection, digital teacher clones, and major investment in AI infrastructure.
-
AI agents learn to make calls
Instinct and Meta's Muse can now handle phone reservations. Huawei accelerates its challenge to Nvidia. And OpenAI's GPT-6 Astra cracks an 83-year-old Nazi cipher in ten hours.
-
OpenAI launches AI advertising tools
OpenAI introduces sponsored agents for marketers. A former Infosys leader's AI startup raises fifty-three million dollars. Nvidia's CEO argues the industry should regulate itself. And Amazon expands Alexa to India with Hindi support.
-
Digit 5 Works Alongside People Without Safety Fences
Agility Robotics unveils a humanoid robot that can work next to people in warehouses. Frontier AI models face a narrowing window of advantage over cheaper open alternatives. And the Gates Foundation commits a billion dollars to expanding AI access in health, education and agriculture.
-
Perplexity deploys GPT-6 Astra to run production systems
Perplexity hands over end-to-end system management to GPT-6 Astra with minimal human oversight. TechCrunch Disrupt explores AI's role in species conservation and de-extinction. And as AI coding tools collapse software costs, the real work shifts elsewhere.
-
GPT-6 Astra takes control of drones and business deals
GPT-6 Astra outperforms rivals on agent tasks and becomes the first model to beat human baseline on drone control. Two new open-source search agents lead their benchmarks. A Waymo robotaxi detects an armed passenger and calls police. Plus music generation goes mainstream, and a two-year study challenges the case for banning AI in classrooms.
-
GPT-6 Astra shows step change in spatial reasoning
OpenAI's latest model breaks through on robotics benchmarks and a Millennium Prize problem, while developers learn to give it fewer guardrails. Plus, a new robot dog arrives from China, and Google releases a forecasting model that predicts business outcomes in a single pass.
-
OpenAI releases Agents API for autonomous cloud development
OpenAI launches a public beta Agents API that lets developers build autonomous cloud agents. Google contracts for nuclear power in Finland. And a Fields Medal mathematician starts a new institute to prove AI safety using mathematical methods.
-
Maven Robotics launches with $100 million to challenge industrial deployment
Maven Robotics enters the market with significant funding and active deployments. Meta's Muse AI agent handles shopping and emails through WhatsApp. Deepseek cuts memory requirements for AI agents with its new V4.1-Flash model. Healthcare AI shows promise but faces integration challenges. And d-Matrix adopts NVIDIA technology for next-generation inference.
-
OpenAI's math breakthrough overshadowed by credit row
OpenAI claims a major advance in mathematical problem-solving, but faces accusations over researcher credits. Tesla's driverless Cybercab takes to the road, Amazon Prime Video syncs dubbed audio to lips, and DeepSeek releases a faster, cheaper model.
-
DeepMind maps nine billion DNA variants to accelerate disease research
Google DeepMind releases AlphaGenome Atlas, an AI tool that maps the molecular effects of every possible DNA variant in the human genome. Plus, Adobe simplifies AI in video editing, a new startup develops planning agents for robots, and a chip startup tackles energy waste through reversible computing.
-
OpenAI's Research Agents Reach Three Workdays Per Human Day
OpenAI reports that AI research agents now complete 3.1 workdays for every human workday, meeting its goal of an automated research intern. But the company's chief scientist warns that no lab has adequate alignment and monitoring to scale safely. Plus data center fires, emergent agent communication, and AI in Ukrainian journalism.
-
Google's WeatherNext 3 learns weather directly from satellite data
Google releases a new AI weather model that skips physics simulations and learns from live satellite data instead. Plus, OpenAI launches GPT-6 Astra with improved reasoning and multi-modal capabilities, and Google brings music generation into the Gemini app.
-
Conversation beats facts at changing minds about conspiracies
Researchers find a seven-minute chat with an AI chatbot reduces conspiracy beliefs more effectively than fact sheets. OpenAI rolls out GPT-6 Astra to premium subscribers, and NVIDIA accelerates local AI inference at IFA 2026.
-
Google Photos gets a personal AI assistant
Google's Gemini Spark takes on photo management, iRobot unveils a two-part cleaning robot, and Deepseek plans a massive processor cluster in China. Plus funding news from Crusoe Energy and developments in AI infrastructure for the home.
-
Nvidia's Superchip Brings AI to Your Laptop
Nvidia launches RTX Spark processors for consumer AI PCs, Microsoft cuts transcription costs by 72%, and robot lawnmowers gain edge-trimming arms. Plus Nvidia acquires Hugging Face, Claude decodes a 17th-century cipher, and Google releases new Gemini variants for agents and cybersecurity.
-
Tesla's Cybercab hits the road in Austin
Tesla begins operating its Cybercab robotaxis in Austin, Texas, testing Elon Musk's camera-only autonomous driving approach. Also today: Google cuts video analysis token usage by up to 88 percent, World Labs unveils a 3D world model, and iRobot shows off a vacuum that seals to carpet.
-
Legacy Systems Get a Second Life with AI
Companies are using AI to modernize aging technology infrastructure, making the economics of system upgrades far more attractive. Plus, Perplexity keeps confidential data local, Tencent releases a massive new language model, and Nvidia's controversial graphics technology arrives this week.
-
OpenClaw 2.0 launches with cloud sessions and real-time collaboration
OpenClaw releases its biggest update yet with simplified setup and multiplayer features. OpenAI moves to outcome-based pricing. Enterprise AI agents coordinate workflows with minimal human oversight. Plus Tencent's new 770-billion parameter model and a massive open video dataset for AI research.
-
China's robots race ahead
China's robotics industry advances rapidly with the World Humanoid Robot Games, while Google Deepmind expands AI beyond the lab into scientific research. Also: cheaper video generation, massive open datasets, and a framework that teaches AI agents to learn from failure.
-
Google's WikiSkill Gives AI Agents a Persistent Memory
Google introduces WikiSkill, a framework that lets AI agents learn from past mistakes. LAION releases a massive open video dataset. Anthropic builds a standard for AI agents to control physical hardware. And Google Deepmind's AI Co-Scientist is now running real lab experiments.
-
OpenAI Tests Always-On AI Agents That Generate Their Own Tasks
OpenAI is building AI agents that stay active indefinitely and create their own work. Plus, Google tackles the trust problem in AI benchmarks with cryptographic protection, and a new open-source model matches top performers at a fraction of the cost.
-
Microduck: Hugging Face's Rollerskating Robot
Pollen Robotics launches a ten-inch tall biped robot available to preorder, NVIDIA ships its Vera CPU for AI agents, and Adobe rolls out AI tools in a dedicated Photoshop interface. Plus, a new open-source language model matches top competitors at a fraction of the cost.
-
Alibaba's Flash Model Pressures OpenAI and Anthropic
Alibaba launches a cost-efficient mixture-of-experts model that outperforms much larger competitors. IBM releases open-weight Granite models for enterprise deployment. And AI agents continue to expand their reach into business operations.
-
OpenAI's Jalapeño chip claims speed and efficiency advantage
OpenAI announces its Jalapeño inference chip delivers faster responses and better efficiency than competitors. Apple unveils new M-series chips for AI development. Plus, a compressed AI model that outperforms its full-precision original, and Perplexity launches local AI agents that run entirely on user hardware.
-
Humanoid robots break the hundred-meter world record
Chinese robots outrun Usain Bolt at the World Humanoid Robot Games in Beijing. Plus, major moves in AI chips, enterprise software, and the race to build proprietary models. The Robots Versus Humans bulletin for August 24th, 2026.