
Godzilla-scale compute: Anthropic in Talks to Raise Funding at a $950 Billion Valuation - The New York Times
Mythos kaiju movie goes to Japan - TNW | Anthropic
Training AI for ‘please don’t kill us’: Anthropic and OpenAI hire for chemical weapons policy managers and biological risks researchers to prevent AI misuse. - Allwork.Space
OpenClaw frenzy clears shelves of Mac Minis. - Bloomberg
Headline says Claude but Claude doesn't even really support OpenClaw now unless you pay an arm and a leg for API? everything is Claudified in current zeitgeist, everyone has Claude on the brain
OpenClaw is really an integration platform with an agentic harness, Claude and Hermes are agentic harnesses with some integrations (and fewer CVEs).
Because there is a Sonos CLI, with no setup I can tell OpenClaw, play Deep Purple, and it works. Is that an improvement over Alexa? well, for one thing all CLIs and MCPs in the universe are available, and I can use my own local model. And OpenClaw supports all the messaging platforms, can instruct it via Discord, Signal, Slack etc.
Altman testifies Musk proposed 2018 OpenAI-Tesla merger for control, which he rejected to preserve nonprofit status. - The New York Times
Recursive fundraising: Richard Socher's Recursive Superintelligence raises $650 million from Google Ventures, Greycroft, Nvidia and AMD at $4 billion valuation. - The New York Times

Thinking Machines, shrinking team: Lab loses 13 of 42 founding team members to Meta, OpenAI, xAI. - Business Insider

Anduril raises $5 billion in funding and reaches $61 billion valuation, double from a year ago. - The New York Times

Microsoft Threat Intelligence identifies compromise in mistralai PyPI package v2.4.6 injecting credential-stealing malware in Mini Shai-Hulud campaign. - Tom's Hardware
seems like a bad one, it will steal credentials like ssh keys that developers use to publish software. first of all, people need to slow down before deploying latest packages. second, the deployment infrastructure needs strong 2FA before accepting code, and Mythos scanning before stuff is made available to everyone.


Don't call it a comeback: Google DeepMind challenges OpenAI and Anthropic using cash reserves, proprietary AI chips, top talent, and distribution. - Financial Times
Google AI reads what you are pointing at with your mouse. (Why not watch your eyes and know what you are looking at?) - Google DeepMind
seems useful for training AI

Amazon's Panos Panay oversees Alexa+ AI upgrade for 600 million devices and leads Project Leo satellite broadband via Globalstar acquisition. - Financial Times
Amazon to rebrand Rufus as Alexa for shopping - The Verge
Codex for Mac enables users to automate tasks, exemplifying personal AI agents' rise in first five months of 2026. - MacStories
AI coders carry laptops slightly open to run agents without sleep mode but settings and utilities help solve that? as OpenAI hints at Codex solution. - Business Insider

University of Valencia researchers develop brain-inspired AI that improves self-driving cars' obstacle detection by over 20% in fog and darkness. - The Conversation
OpenEvidence AI medical search tool used by 65% of U.S. physicians and 1.2 million internationally. - NBC News
AI uninstalls the honor code? Princeton faculty votes to require proctors for all in-person exams starting this summer, reversing 1893 honor code policy over AI cheating concerns. - The Wall Street Journal

Executives admit AI makes them value human workers less as they spend heavily on AI without results. - theregister
AI shifts hiring from traditional sector experience to judgment, adaptability, upskilling and rapid decisions under uncertainty. - Fast Company
Newburgh, Maine residents reject error-filled AI-generated municipal logo after backlash. - Fast Company
Elon boosts anonymous stans / sockpuppets who worship him - The Washington Post
Follow the latest AI headlines via druce.ai on Bluesky

