Dyna-2.1: A Robot Agent Built on a Million Hours of Human Video
Dyna Robotics says its world-action model and custom semi-humanoid hardware can sustain real-world physical work for hours at a time.
AI & ML
Frontier models, research results, agents, and developer tooling.
Why Model Routing Is Harder Than It Looks
A new IBM Research post explains why picking the right LLM per query is deceptively complex despite its simple premise.
Open Models Face a Six-Month Viability Test
A new analysis argues that the coming months are the most serious test yet of whether open-source AI can survive.
AI labs buy dead startups' data to train models
A growing end-of-life startup industry is selling defunct companies' code and workspace data to AI labs, raising legal and ethical questions.
Google and AIM Launch Gemini-Powered ATL Saathi for Indian Educators
A new AI assistant aims to help teachers run robotics labs in Indian schools.
Cars24's OpenAI Agents Handle 1M+ Conversation Minutes a Month
A case study from OpenAI shows how Cars24's voice and chat agents recover 12% of lost leads while handling over a million conversation minutes each month.
NVIDIA Opens Vera Rubin Rack Designs to OCP for AI Factories
At the OCP Global Summit, NVIDIA contributed Vera Rubin rack and compute tray designs to the open MGX architecture to help standardize AI factory infrastructure.
Vélez and Vince Join OpenAI's Foundation and Group Boards
The appointments add senior finance and governance experience to OpenAI's oversight as the company scales.
How OpenAI's creative team uses Codex to build custom tools
OpenAI's creative team treats Codex as a context-aware collaborator for faster ideation and prototyping.
Hugging Face Releases Grabette, an Open Robot Data Recording System
Grabette is a new open system from Hugging Face designed to record robot-manipulation data.
OpenAI makes case for teen access to safer ChatGPT
OpenAI argues that teens should have access to AI tools, and is pairing that case with new safeguards such as parental controls and age-appropriate protections in ChatGPT.
Newer Models Keep Old Advantages, Says Hugging Face Blog
A Hugging Face post argues that recent AI models retain the same benefits as earlier versions.
NTT DATA cuts incident analysis to 30 minutes with OpenAI Codex
A vendor-published case study reports that NTT DATA Group deployed ChatGPT Enterprise and Codex to 9,000 employees, reducing incident analysis time to half an hour.
OpenAI CFO Proposes Scorecard for Measuring AI ROI
A new framework from OpenAI aims to help businesses evaluate AI investments beyond raw model performance.
Kimi K3 and the Open-Weights Escalation
Nathan Lambert argues that Kimi K3 marks a new phase in open-weights AI with significant global consequences.
Google DeepMind Unveils Gemini 3.6 Flash, 3.5 Flash-Lite, and Cyber
The announcement expands Google's Gemini family with three new Flash-tier models, though technical details remain sparse.
OpenAI to Work With US Labs on AI-Driven Science
OpenAI is partnering with the Department of Energy and national laboratories to apply frontier AI to accelerate scientific discovery.
OpenAI Launches Health Data Integration in ChatGPT for U.S. Users
The feature lets eligible U.S. users securely connect medical records and Apple Health to receive more personalized health insights.