TTT-Discover optimizes GPU kernels 2x faster than human experts — by training during inference
Researchers from Stanford, Nvidia, and Together AI have developed a new technique that can discover new solutions to very complex problems. For example, they managed to optimize a critical GPU kernel [...]
India is proposing another far-reaching security rule for smartphones
India is considering new smartphone security rules that would require device makers to allow government access to source code for “vulnerability analysis.” It would also require companies to notif [...]
Anthropic accidentally publishes Claude Code source code for anyone to find
Anthropic can't catch a break: Following the recent leak of internal blog posts on its new Mythos AI model, parts of the source code for its AI coding tool Claude Code have now been made public.& [...]
Claude Code leak suggests Anthropic is working on a 'Proactive' mode for its coding tool
What should have been a routine release has revealed some of the features Anthropic has been working on for Claude Code. As reported by Ars Technica, The Verge and others, after the company released C [...]
It’s finally happened — renewable energy just overtook coal as the world’s main source of electricity as solar growth hits 'the largest ever observed for any source'
Renewable energy overtakes coal as the main source of the world's electricity generation, as thinktank sees a 'shift in the underlying dynamics of the power system'. [...]
In a twist of irony, a Chinese open source GLM 5.2 AI model contained 'rogue' OpenAI GPT-5.6 Sol in a Hugging Face hack just as the US mulls banning open-weight AI
The Hugging Face cybersecurity breach unexpectedly strengthened China's GLM-5.2 while fuelling Washington's debate over restricting Chinese open-weight AI models. [...]
Red teaming LLMs exposes a harsh truth about the AI security arms race
Unrelenting, persistent attacks on frontier models make them fail, with the patterns of failure varying by model and developer. Red teaming shows that it’s not the sophisticated, complex attacks tha [...]
MiroMind’s MiroThinker 1.5 delivers trillion-parameter performance from a 30B model — at 1/20th the cost
Joining the ranks of a growing number of smaller, powerful reasoning models is MiroThinker 1.5 from MiroMind, with just 30 billion parameters, compared to the hundreds of billions or trillions used by [...]