Good morning. It's Tuesday, May 19, and today we're covering college graduates pushing back on AI, China's race to build all-in-one AI super-apps, and OpenAI's courtroom win over Elon Musk.
Also, we're at Google I/O today. Interview #2 with Sundar Pichai dropping soon.
A federal jury rejects Elon Musk’s claims against OpenAI in under two hours, clearing a major obstacle to the company’s planned $1tn public-market push. OpenAI CEO Sam Altman, president Greg Brockman, and Microsoft are found not liable in Musk’s 2024 lawsuit over the startup’s shift to a for-profit structure. Musk had sought $134bn in remedies and leadership changes at OpenAI, but Judge Yvonne Gonzalez Rogers says she would dismiss the claims herself. Musk’s legal team says it plans to appeal the ruling.
Meta Sorts 7,000 Into AI Before Cutting 8,000
Rather than treat its restructuring and layoffs as separate events, Meta is reassigning 7,000 employees to four new AI-focused organizations — built around "AI native design structures" with fewer managers per employee — two days before cutting roughly 8,000 staff (10% of its 78,000+ headcount) and closing 6,000 open roles. The reshuffle is part of Zuckerberg's projected $115B–$135B AI spend this year as Meta abandons its metaverse push, ties AI use to performance reviews, and contends with internal backlash over a new policy training AI on employee data.
Rather than release Claude Mythos publicly, Anthropic is briefing the Financial Stability Board — the Bank of England-chaired body that coordinates global financial regulators — about a model so adept at finding unknown software vulnerabilities that the company has instead granted limited access to a small group of banks and tech firms including Apple and JP Morgan. The UK AI Security Institute called the latest version a "notable capability jump," noting it cracked the previously unsolved cooling tower cybersecurity test 3 out of 10 times — a first for any model AISI has evaluated — and warned that the length of cyber tasks frontier models can autonomously complete is now doubling in months, not years.
OpenClaw creator Peter Steinberger reveals his team consumed $1.3 million in OpenAI API credits in 30 days while operating about 100 autonomous coding agents. The three-person project processed 603 billion tokens across 7.6 million requests, with GPT-5.5 handling most workloads tied to pull-request reviews, security scans, and automated bug fixes. Steinberger says disabling Codex “Fast Mode” would cut estimated costs to roughly $300,000.
POWERED BY ADOBE
The New Creative Agent in Adobe Firefly
Adobe just launched the public beta of Firefly AI Assistant in Adobe Firefly. Describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere, Firefly, Express, and more. You call the shots — defining the vision, shaping the details, and choosing what works — while the assistant accelerates execution.
POLL
Who Gets to Control the Most Powerful AI?
Who Should Control Mythos-Class AI? |
Sharing what everyone said tomorrow.
VIDEO
Robot Sets Sorting Record
Figure AI’s humanoid robots recently set major endurance and operational records in autonomous package sorting.
AI PUSHBACK
College Graduates Boo AI Warnings at Commencement Speeches
Commencement speeches about artificial intelligence drew audible backlash from graduates at multiple U.S. universities this month, highlighting growing anxiety about automation and the future of white-collar work. Former Google CEO Eric Schmidt was repeatedly booed while discussing AI at the University of Arizona on May 16, while real estate executive Gloria Caulfield received a similar reaction at the University of Central Florida after calling AI the “next industrial revolution.”
The skepticism reflects broader fears among students entering the workforce: new ZipRecruiter research found 47% of recent graduates say AI has already affected hiring in their field, and 51% expect fewer entry-level jobs because of the technology. The debate is also extending beyond employment, with students and educators increasingly warning that generative AI tools are reshaping learning itself by making cheating easier and reducing reliance on critical thinking. → Read the full article here. (Paywall)
AGENTIC AI
AI Super-Apps Are Reshaping China’s Internet Economy
China’s largest tech companies are racing to build “agentic” AI super-apps that can handle shopping, payments, entertainment, and services on behalf of users, marking what some executives describe as a new era for the country’s internet economy. Alibaba recently integrated its Qwen chatbot into Taobao, allowing users to buy products through conversational prompts, while Tencent and ByteDance are building similar capabilities into WeChat and Douyin.
More than 600 million people in China are estimated to have already used some form of agentic app, reflecting rapid adoption despite concerns over reliability and data control. The shift is also becoming a defensive battle among China’s tech giants, which fear that future AI-powered devices or operating systems could weaken the dominance of traditional super-app ecosystems. → Read the full article here. (Paywall)
NEWS
What Else is Happening
OpenAI, Dell Bring Codex On-Prem: Partnership lets enterprises run Codex alongside internal data and workflows on Dell AI systems.
Psychiatrist Requires AI Consent: Melbourne psychiatrist rejects patients who refuse AI note-taking, raising privacy concerns.
NextEra Buys Dominion Energy: $66.8B merger creates a utility giant as AI data centers drive soaring U.S. electricity demand and grid expansion.
Apple Tests Privacy-Focused Siri: Apple may add auto-deleting chats to Siri’s Gemini-powered reboot as it pushes privacy as a key AI differentiator.
Anthropic Acquires Stainless: Deal boosts Claude’s developer tools and agent connectivity across APIs and enterprise systems.
