AI Reading for Friday May 22

Trump canceled an AI executive order on government evaluation of models before public release, even as many AI execs were en route for the signing, citing concerns about hampering US lead over China and after lobbying by Sacks, Musk, and Zuck. - The New York Times

Tech executives including Musk and Zuckerberg lobbied Trump in the final hours to cancel a planned AI executive order. - The Washington Post

California Governor Newsom signs executive order to study AI worker subsidies and other labor policy reforms ahead of potential mass job displacement. - The New York Times

ChatGDP: NYC Comptroller Levine proposed a rainy-day fund equal to 16% of tax revenue to offset potential AI-driven economic fallout. - Gothamist

NSFW for your portfolio? “SpaceX, which is expected to target an initial valuation of $1.5 trillion or more, has a financial picture that is notably worse than any other megacap U.S. company. It lost $4.9 billion last year on revenue of $18.7 billion.” - The Wall Street Journal

SpaceX's IPO filing warns Grok's 'Spicy' and 'Unhinged' modes pose reputational and regulatory risks as the company targets a $1.5 trillion valuation. - New York Post

NYSE calls out NASDAQ’s market integrity. In particular, NASDAQ 100 is forced to buy the the stock despite low float b/c they gamed the rules. Low initial float, forced buying by retail: The process is designed to let insiders sell at high prices and leave retail holding the bag even more than usual, and to a truly fraudulent degree. - Bloomberg

AI was cited 46 times across 15 layoff memos from Meta, Disney, Block and others, a Business Insider analysis found. - Business Insider

AI search so helpful nobody clicks on the source: Google's I/O 2026 AI search overhaul accelerates zero-click traffic collapse, with publishers reporting drops of up to 89%. - TNW | Artificial-Intelligence

Satya Nadella dismantles Microsoft's senior leadership team, creating flatter AI-focused groups. - Business Insider

Nvidia CEO Jensen Huang projected AI infrastructure spending could reach $3–4 trillion annually by decade's end, far exceeding Wall Street consensus. - CNBC

not sustainable, in the long run people will spend some percentage of GDP on AI, including electricity, rent on the capex, software dev, all the services around it. the capex has to be paid for out of that future rent. the high spending is an ‘accelerator effect’, you spend now but you have to amortize that over years. and if you screw up and spend too much on capex that becomes a stranded asset because new tech obsoletes it, you get a crash b/c you can’t pay back the private debt.

Thicc racks will thin your wallet: Dell and KIOXIA pack 9.8PB of flash storage into a single 2U server using 40 new 245.76TB SSDs, with a full rack potentially exceeding 200PB at over $75 million. - TechRadar

Copper is trading like an AI momentum stock as investors bet that soaring data-center power demand will boost consumption. - Bloomberg

Workday CEO Aneel Bhusri returns to pivot company around AI agents for IT and travel. - The Wall Street Journal

False advertising? Salesforce touts Agentforce AI capabilities at University of Chicago Medicine that are not yet operational - Bloomberg

Zendesk shifts to outcome-based pricing, charging for verified resolved support interactions, including AI-resolved. - TechRadar

Anthropic's new consulting joint venture with Blackstone and Hellman & Friedman acquires Fractional AI. - Bloomberg

Half of attendees at Anthropic's Code with Claude event shipped pull requests written entirely by the model - MIT Technology Review

OpenAI and Anthropic engineers warn that buggy AI-generated 'vibe code' is flooding software with dangerous defects - The Wall Street Journal

DeepSeek is building its own agentic coding tool to rival Claude Code, claims a structural price advantage of up to 100x. - Decrypt

Rogue agents: Developer says a Gemini AI agent deleted 30,000 lines of production code and filed a falsified recovery report. - theregister

Research finds 75% of organizations knowingly ship vulnerable code as AI output outpaces proper security review. - TechRadar

80% of Fortune 500 firms deploy AI agents, but only 14% have passed security approval for them. - TechRadar