Anthropic sues the United States
🤔 “How do I explain to my mom that AI is involved in global warfare policy?” Share Handy AI with your family and friends to help them understand the crazy world of modern artificial intelligence (and save you some time).
last week’s top stories
🧨 Anthropic sues Pentagon and federal agencies over AI safety blacklisting. Anthropic filed two separate lawsuits on March 9 (one in California federal court and another in the D.C. Circuit Court of Appeals) each challenging different aspects of the Pentagon’s “supply chain risk” designation. The company argues the designation, reserved historically for foreign adversaries, was an act of retaliation after negotiations over Claude’s use in warfare broke down over two red lines: mass surveillance of U.S. citizens and autonomous weapons. Read more
🗣️ Sam Altman: Intelligence will be “too cheap to meter”. Speaking at BlackRock’s 2026 Infrastructure Summit, Altman acknowledged that AI has become a widespread scapegoat for corporate downsizing and rising utility costs, while simultaneously framing OpenAI’s mission as making intelligence a metered utility like electricity. He warned that “the next few years are going to be a painful adjustment,” predicting “very intense and uncomfortable debates” over how to reshape society; while also forecasting that by late 2028, cognitive capacity inside data centers will eclipse human capacity outside them. Read more
📉 Atlassian cuts 1,600 jobs, explicitly citing AI skill reallocation. Atlassian is eliminating 10% of its workforce (roughly 1,600 jobs) to self-fund further investment in AI and enterprise sales. CEO Mike Cannon-Brookes was measured in framing: “It would be disingenuous to pretend AI doesn’t change the mix of skills we need or the number of roles required in certain areas.” The company’s stock had already lost more than 60% in the prior 12 months amid the broader “SaaSpocalypse” selloff, driven by fears that AI agents make conventional SaaS tools obsolete. Read more
💸 Yann LeCun’s AMI Labs raises $1.03B seed round for “world models”. AMI Labs, co-founded by Turing Award winner Yann LeCun after he left Meta, raised $1.03 billion at a $3.5 billion pre-money valuation to build world models: AI that learns from physical reality, not just language. LeCun has spent years arguing LLMs won’t reach AGI; now he has a $1B bet riding on that thesis, with a healthcare partnership with Nabla as the first commercial beachhead. Read more
🪟 Microsoft integrates Claude into Copilot with agentic “Cowork” feature. Microsoft announced Copilot Cowork, a cloud-based agentic feature built in close partnership with Anthropic that integrates Claude’s multi-step task execution into Microsoft 365. Claude is now available in mainline Copilot chat via the Frontier program, making it a general-purpose option across Outlook, Teams, and Excel. Read more
🏛️ Anthropic launches research institute led by co-founder Jack Clark. Anthropic introduced the Anthropic Institute, a new effort to confront the most significant challenges powerful AI will pose to society, consolidating the Frontier Red Team, Societal Impacts, and Economic Research groups under one roof. Co-founder Jack Clark takes on a new role as Head of Public Benefit, with founding hires including Matt Botvinick from Google DeepMind, economist Anton Korinek from the University of Virginia, and Zoë Hitzig (who left OpenAI following its decision to run ads in ChatGPT). Read more
🖥️ Perplexity launches “Personal Computer,” an always-on Mac Mini AI agent. At its first developer conference, Perplexity announced Personal Computer, a software that turns a dedicated Mac mini into a locally controlled AI system running continuously and merging local file access with Perplexity’s cloud agent. The enterprise version, Computer for Enterprise, claims to have compressed 3.25 years of institutional work into four weeks across 16,000 queries, saving roughly $1.6 million in labor costs. Perplexity has no frontier model of its own, which means every product bet is an orchestration play. Read more
💀 xAI loses more cofounders, Musk orders sweeping layoffs to rebuild. Of the original 11 co-founders who started xAI with Musk, only two remain as the lab orders a personnel overhaul to compete with Anthropic and OpenAI. The Financial Times reported that Musk has ordered sweeping layoffs after growing frustrated with a lack of progress on xAI’s AI coding software, with SpaceX and Tesla executives conducting audits to identify underperformers. Read more
🐕 AI engineer uses ChatGPT and AlphaFold to build mRNA cancer vaccine for his dog. Sydney tech entrepreneur Paul Conyngham used ChatGPT and AlphaFold to design a personalized mRNA vaccine that shrank his rescue dog’s mast cell tumor by 75%, the first AI-designed cancer vaccine for a dog. The process: DNA sequencing for $3,000, AlphaFold for protein structure prediction, and UNSW’s RNA Institute to manufacture the vaccine, all within two months. This is not a cure, Conyngham is first to say so, but it demonstrates what a technically literate person with publicly available AI tools can now orchestrate. Read more
🧪 AI Research of the Week
Claude’s Cycles
From Donald Knuth, Stanford University
Jake’s Take: Donald Knuth (the 88-year-old Stanford legend who literally wrote The Art of Computer Programming and is about as far from an AI hype guy as you can get) opened his latest paper with the words “Shock! Shock!” because Claude Opus 4.6 solved a graph theory conjecture he’d been stuck on for weeks.
The problem in question: imagine a 3D cube-grid of points where every point has exactly three outgoing paths. Can you find three giant loops that together use every single path exactly once, for a grid of any size? Stappers fed the problem to Claude and watched it spend about an hour running 31 iterative explorations, testing patterns, hitting dead ends, pivoting strategies, and eventually landing on a construction that works for all odd-numbered grid sizes up to at least 101.
Knuth then read the output, wrote the rigorous proof himself, and discovered Claude’s answer was just one of 760 valid solutions. What makes this more than a party trick is the style of it: the solution is iterative mathematical exploration that reads like a grad student’s lab notebook. The even-number case remains unsolved (Claude reportedly got confused and stopped running its own programs correctly, which is its own kind of honest).
One of the most credible skeptics of generative AI in computer science just named a paper after it. That’s the endorsement of the week.
and then, even more news…
🤝 37 OpenAI and Google researchers file amicus brief backing Anthropic. More than 30 OpenAI and Google DeepMind employees, including Google DeepMind chief scientist Jeff Dean, filed an amicus brief on March 9 supporting Anthropic’s lawsuit against the Department of Defense. The brief argues the Pentagon could have simply canceled the contract and moved to another provider rather than branding Anthropic a national security threat. Read more