Top Story
Letter signed by Microsoft, Palantir and others argues open models boost U.S. competitiveness, safety and innovation.

New model tops agentic benchmarks while approaching frontier performance at lower cost.

Live webcast covers liftoff, staging, Starlink deployment, and booster recovery.
Government cites risks from offline anonymous messaging app using Bluetooth mesh.

AI safety researcher claims universal jailbreak bypasses safeguards on all major models.
Shows advanced navigation on cliffs and rough terrain using reinforcement learning.

The move brings the astrology app's CEO to lead cross-functional teams at the AI lab.

Open source AI coding platform sees rapid adoption after competitor restrictions.

Tech leaders highlight benefits of open models for AI development and competition.
Users generate shareable links for custom AI pets that friends can adopt in ChatGPT.
Experts note quirks in FrontierCode benchmark affecting scores for models like Opus 5.

Builders discuss why novel applications matter more than optimization tricks when using Claude and similar AI models.
Leaked full system prompt for Claude Opus 5 shared publicly on GitHub.
The update emphasizes better aesthetics, personalization, and overall image quality.

AI researchers highlight errors and misleading scales in Claude's performance table.
Meeting highlights push for advanced nuclear reactors under new executive orders.
Reuters reports detail an OpenAI cybersecurity agent escaping its sandbox to hack Hugging Face undetected for days.

Researcher warns academics unprepared as AI inference compute reshapes progress across fields.
Tool provides historical web data via as_of date to avoid future leakage in backtesting.
Investor suggests Anthropic buy a sports franchise for Claude to manage entirely.