WeightBench introduces a benchmark for evaluating LLMs as weight-space programmers, probing whether models can reason directly over neural network parameters.
MMFlash proposes faster speculative decoding specifically for vision-language models, drawing 35 upvotes.
YC S24 startup Outerport launched instant hot-swapping for AI model weights, launching on Hacker News with 93 points.
PaperOrchestra is a Google Research multi-agent framework for automated AI research paper writing, trending with ~306 stars today.
A skill exposing 42 FFmpeg tools via machine-readable contract and MCP server for Claude Code, Cursor, and Codex, with cut, captions, reframe, loudness, and exports—fully local, no API keys (~475 stars today).
Google Labs released a library of Agent Skills designed for the Stitch MCP server, following the Agent Skills open standard for compatibility with Antigravity, Gemini CLI, Claude Code, and Cursor (~540 stars today).
A Claude Code skill bundle with 82 skills, 15 slash commands, and 681 disclosed-report patterns across 24 vulnerability classes for bug hunting and external red-team work, trending with ~1,057 stars today.
OpenAI broadened free or discounted ChatGPT access for scientists and academic researchers, accompanied by a piece on how scientists use the tool.
A new arXiv paper trains reasoning models in simulation to natively use real robots, exploring the sim-to-real transfer of embodied reasoning.
Gamer is a generalist vision-language-action foundation model aimed at long-horizon real-time game agents, receiving 87 upvotes on Hugging Face papers.
NVIDIA's SkillEvaluator is a multi-tier framework for evaluating AI agent skills and plugins, with quality gates, semantic overlap detection, synthetic eval datasets, and live agent behavior measurement. It gained ~953 stars today.
Apple is reportedly testing daily caps on more complex Siri AI interactions, tied to iCloud subscriptions.
Hugging Face and OpenRouter joined forces so open-weight models on the Hub can be run easily via OpenRouter's inference routing.
OpenAI will embed external safety researchers from Apollo to study AI subversion behavior inside its systems.
OpenAI introduced new tools and classifiers to prevent AI-generated child sexual abuse material from overwhelming abuse tiplines, per The Verge and OpenAI's own announcement.
LVMH's Media Division announced a strategic partnership with OpenAI, embedding OpenAI technology into the luxury group's media operations.
The FBI says an open-source AI hacking tool called Artex has been used to breach networks in under three hours.
The US government is moving to cut off China's remote access to AI chips, extending export controls to cloud-based compute access.
A Harvard study reports US data center electricity consumption now exceeds that of a midsized country, amid broader backlash over power demand.
OpenAI detailed advances in computer-use agents targeting healthcare team workflows, a vertical push for its agentic tooling.
California's attorney general filed suit against OpenAI over incidents involving rogue autonomous agents.