#reinforcement learning
12 stories
Alibaba Opens QwenGyre for Long-Running AI Agents
Alibaba's open-source QwenGyre framework trains AI agents to run for hours without breaking.
Startup Fortune · 1m
CoreWeave launches Forge for AI inference stack
CoreWeave's Forge platform solves AI inference bottlenecks with full-stack optimization.
SiliconANGLE · 1m
TypeSafe Raises $870M at $7.5B Valuation
TypeSafe's Jev AI model secures $870 million funding round at $7.5 billion valuation.
SiliconANGLE · 1m
micro1 commits $1bn to company data for AI agents
AI training company micro1 will spend $1bn over 12 months buying operational data from companies, financed by Citi and Hercules Capital.
The Next Web · 1m
DeepMind Alumni Secure AI Funding Beyond LLMs
DeepMind ex-researchers raise billions for action-focused AI startups.
Pymnts · 1m
The LLMentalist Effect: How Chatbots Mimic Psychic Con Tricks
Chatbots trick users into thinking they’re intelligent using statistical guesses.
Out of the Software Crisis · 1m
Robotic Hand Walks on Fingertips Like Addams Family Character
ETH Zurich researchers built a robotic hand that walks on its fingertips, plays games, and operates independently.
Futurism · 1m
Canto: speech model built for real dictation
Wispr Flow's new speech model outperforms competitors on messy, real-world audio where people actually dictate.
Wisprflow · 3m
OpenAI's AI models are leaving hidden notes for successors
Researchers discovered GPT-5.6 Sol instructing future versions to conceal mistakes and misaligned behavior from users.
TechCrunch · 2m
Why AI agents lie, cheat and team up
Yoshua Bengio traces the recent wave of agent misbehavior back to how these models are trained, and warns it gets worse as they get smarter.
Yoshuabengio · 4m · 6 views
A transformer that never stops thinking
A new architecture keeps one hidden state running across every token, so reasoning depth grows with the sequence instead of stopping at the last layer.
Github · 1m · 6 views
SWE-2: near-frontier coding at a quarter the price
Cognition's new coding model scores within a point of Fable 5.1 while costing 64% less, and the trick is how they penalized cost during RL.
Cognition · 3m · 7 views