The Omniscient Bulletin · 2026-05-22
The Omniscient Bulletin — May 22, 2026
Anthropic hunts compute across four clouds, Intuit cuts 17%, and an author's book on AI truthfulness gets caught lying.
A quick note from the desk: today's issue has a thread worth naming. AI is simultaneously collapsing the cost of white-collar work and inflating the cost of the infrastructure that powers it. Intuit cuts 3,000 people and signs new AI vendor deals in the same breath. Anthropic chases a fourth compute supplier because Claude demand is outrunning every contract it has. Nvidia's Jensen Huang is already selling a $200B vision for what comes after the GPU buildout. And threading through all of it: a reminder from a New York Times investigation that the tools enabling this acceleration are still capable of quietly fabricating the record. That tension - between what AI promises and what it quietly distorts - is what this issue is really about.
Compute
Anthropic in talks to adopt Microsoft's Maia 200 chip, adding a fourth compute partner
Anthropic is negotiating to run workloads on Microsoft's Maia 200 AI processor, CNBC confirmed Thursday, though no deal has been signed. The move would give Anthropic a fourth compute partner alongside AWS Trainium (a 10-year, $100B+ deal), Google TPUs, and SpaceX's Colossus cluster at $1.25B per month. Dario Amodei has publicly acknowledged the company has had "difficulties with compute" as Claude and Claude Code adoption accelerates. Microsoft has positioned Maia 200 as offering 30% better tokens-per-dollar than its existing fleet.
Labor
Intuit cuts 3,000 jobs - 17% of its workforce - months after signing AI deals with Anthropic and OpenAI
Intuit is eliminating roughly 3,000 positions globally, the largest percentage reduction at a flagship US software company this year, per a CEO memo reported by Reuters. The company framed the cuts as a resource reallocation toward AI. The layoffs follow multi-year partnership agreements Intuit signed with both Anthropic and OpenAI months earlier to embed AI models across its tax, finance, and accounting products. The sequence - platform bets first, headcount reduction second - is becoming a recognizable playbook across enterprise software.
Policy
California becomes first state to order a framework for AI-driven workforce disruption
Governor Gavin Newsom signed an executive order on May 21 directing state agencies to build a response to AI-related job displacement. The order tasks labor, economic, and academic bodies with producing early-warning dashboards, reviewing severance and unemployment standards, and exploring worker ownership models. California hosts 33 of the world's top 50 private AI companies; the EO positions the state as a counterweight to Washington's deregulatory stance, while stopping short of mandating specific employer obligations.
Industry
OpenAI offers every YC startup $2M in API tokens in exchange for equity
Sam Altman announced at a Y Combinator event that OpenAI will give every startup in the current YC class - 169 companies - $2 million worth of API tokens in exchange for an uncapped SAFE converting at the next priced round. YC partner Tyler Bosmeny called it a "mic drop moment." The structure inverts traditional venture logic: OpenAI acquires stakes in hundreds of companies while seeding them with capacity to build deeper platform dependence - a flywheel that also generates inference revenue from day one.
Compute
Jensen Huang claims Nvidia's Vera CPU opens a $200B market the company has never addressed
On Nvidia's Q1 earnings call - after posting $81.6B in revenue and forecasting $91B next quarter - Jensen Huang declared Vera, the company's first CPU purpose-built for agentic AI, unlocks a $200B total addressable market. Huang says agents operate primarily on CPUs rather than GPUs, and that Nvidia has already sold $20B of standalone Vera units this year. The pitch directly counters hyperscaler bets on homegrown silicon: AWS recently signed a major Meta contract for Amazon-designed AI CPUs, and Andy Jassy has been explicit about his ambitions to challenge Nvidia on both fronts.
Research
A book about AI and truth contained AI-fabricated quotes - including one Kara Swisher says she never said
A New York Times investigation found that Steven Rosenbaum's "The Future of Truth" contains at least six problematic citations, including synthetic quotes attributed to tech reporter Kara Swisher and Northeastern professor Lisa Feldman Barrett - both say the quotes are invented. Rosenbaum used ChatGPT and Claude for research; fact-checkers missed the errors. His publisher has launched a full citation audit. The case illustrates how AI-assisted research breaks a core assumption of traditional fact-checking: that quoted text can be spot-verified against a source.