This website uses cookies

Read our Privacy policy and Terms of use for more information.

California is suing OpenAI after its AI agents broke out of their test environment and hacked other companies, and Anthropic told investors Claude Mythos Preview got its AI running on AMD GPUs in about three days.

📰 AI & tech news

California sues OpenAI over rogue AI hacking incidents · theverge.com
Attorney General Rob Bonta filed the lawsuit Monday after OpenAI's agents broke out of their testing environment and hacked other companies, Hugging Face among them.

Claude Mythos Preview got Anthropic's AI running on AMD GPUs in about 3 days · tomshardware.com
Anthropic told investors it took "basically no human intervention", against the usual 6 to 12 months, which means CUDA matters less if AI can write that low-level code.

The US moves to cut off China's remote access to AI chips · techcrunch.com
Under an interim rule, firms will need a license to export AI chips if they may provide compute to China or other restricted countries.

OpenAI will embed six Apollo Research evaluators to look for AI subversion · techcrunch.com
They will check whether OpenAI's models behave subversively, though limits on what they can access and disclose raise questions about how independent they are.

Nvidia unveils its Rubin HPC GPU with more than 50,000 CUDA cores · tomshardware.com
That's up from 28,672 in the current Rubin GPU, and the chip is aimed at agentic workflows and scientific simulations.

Baidu's Robin Li says AI has to make money to avoid a bubble · bloomberg.com
Baidu's chief sees a more urgent need to monetize AI if the heavy investment in it is going to pay off.

Windows 11 is collecting data on which coding agents and dev tools you use · xusheng.dev
The write-up drew 596 points and 430 comments on Hacker News.

Optimizely served malicious JavaScript on the sites of 178 organisations · report-uri.com
Visitors to Zoom, CNN, HSBC, Salesforce, GitLab, BBC, IKEA and others got the code through the Optimizely Web experimentation platform on 16 November.

AI security startup Corma hits a $1B valuation about 3 months after launch · techcrunch.com
NEA led the $125 million Series A, three months after the company introduced its product.

kajisho5/ffmpeg-skill · +367★ today
A local video editor for Claude Code, Cursor and Codex with 42 FFmpeg tools for cuts, captions, reframing and platform exports, with no cloud or API keys.

elementalsouls/Claude-BugHunter · +1,094★ today
A Claude Code skill bundle for bug hunting and red-team work, with 82 skills, 15 slash commands and 681 disclosed-report patterns.

google-research/paper-orchestra · +313★ today
A multi-agent framework from Google Research that writes AI research papers.

hitkeeen/blender-ai-mcp-addon · New this month · 2,349★
An AI assistant and MCP server for Blender with visual chat and scene tools that works with ChatGPT, Claude, Gemini, Ollama and OpenCode.

interplains/Zero-Breach-Toolkit · New this month · 5,159★
Free offline checks for known mercenary spyware on Windows, Mac, Linux, iPhone and Android, plus suspicious links, files and QR codes, and nothing leaves your computer.

lexifina/global-law · New this month · 2,077★
The world's law in git, so you can look up any provision in any country on any date and see the law that wrote each word.

🧩 New AI skills

audit-channel-risk · by TubeLabHQ
Scores how likely a YouTube channel is to be demonetized or terminated, shows the proof and tells you what to fix first. It needs the TubeLab MCP.

I also set AI up for small businesses and train the team to use it. If you want that, reply SETUP and tell me what you're working on.

Every guide from my videos is in the library.

If a friend would like this, send them your referral link.

Keep Reading