Anthropic builds the jobs killswitch detector. The white-collar alarm is blinking.

>> AUTOMATA // EDITION 003 // 2026.03.08

————————— TRANSMISSION START —————————

Five signals from the machine. Anthropic built a jobs killswitch detector. Washington is going nuclear on chip exports. Open-source models are shipping faster than anyone can benchmark them. And the capital flowing into AI labs just broke math.

Anthropic Publishes AI Jobs Early Warning System. The White-Collar Recession Clock Starts Now.

Anthropic dropped a research paper that does what no one else has dared to quantify: measure actual AI job displacement in real time. Their new metric — "observed exposure" — combines theoretical LLM capability with real-world usage. The headline finding: AI could theoretically automate 94% of tasks in computer and math fields, but actual Claude coverage sits at 33%. The buried finding: hiring of workers aged 22-25 in exposed occupations has already slowed by 16%.

Why it matters: Forget the vibes-based "AI will take your job" debate. Anthropic just gave policymakers a live dashboard. They're calling the downside scenario a "Great Recession for white-collar workers." When the company building the models is publishing the displacement playbook, the trajectory is clear. This is an early warning system, and the alarm is already blinking.

U.S. Drafts Global AI Chip Export Controls. Every Country, Not Just China.

Bloomberg reports the U.S. is drafting rules that would require government approval to ship advanced AI chips anywhere outside American borders. Not just China — everywhere. Nvidia has already stopped producing H200 chips for China and is redirecting TSMC capacity toward Vera Rubin. This drops right before Trump's planned visit to Beijing on March 31.

Why it matters: The Biden-era "AI Diffusion" rule targeted adversaries. This targets allies. If enacted, every country on earth becomes a managed customer for American compute. Nvidia and AMD eat the revenue hit. Foreign governments accelerate domestic chip programs. The AI supply chain just became a geopolitical weapon with no safety switch. The timing ahead of the Trump-Xi summit is not coincidental.

Open-Source AI Drops Three Bangers in One Week. The Frontier Is Fragmenting.

Alibaba shipped Qwen3.5-9B, a compact reasoning model that outperforms OpenAI's open-source gpt-oss-120B on key benchmarks — at a fraction of the parameter count. Allen Institute for AI released Olmo Hybrid, a 7B model combining transformer attention with linear recurrence for 2x data efficiency. Zhipu AI released GLM-5, a 744B MoE model trained entirely on Huawei Ascend chips under MIT license.

Why it matters: Three releases, three signals. Small models are beating large ones on reasoning. Architectural innovation is moving past pure transformers. And Chinese labs are proving they can build frontier models without a single Nvidia GPU. The export controls are already priced in. Open-source isn't just competitive — it's setting the pace.

OpenAI Closes $110B Round. Anthropic Takes $30B. The Capital Arms Race Has No Ceiling.

OpenAI finalized $110 billion in funding — the largest private round in history — at a $730B pre-money valuation. Amazon led with $50B, Nvidia put in $30B, SoftBank added $30B. Separately, Anthropic closed a $30B Series G at $380B. Combined: $140 billion in new AI capital in a single quarter. Microsoft notably did not participate in OpenAI's round.

Why it matters: These numbers have left the realm of venture capital and entered sovereign wealth territory. OpenAI is projecting $280B in revenue by 2030 and planning $600B in total compute spend. The signal is that these companies believe AI infrastructure will be as fundamental as electricity — and they're spending like it. Microsoft's absence from the round tells its own story about how that relationship is evolving.

States Are Building a Patchwork of AI Laws. Oregon's Bill Has Teeth.

Vermont signed S 23 into law, requiring disclosure of AI-generated synthetic media in elections. Oregon is advancing a bill with a private right of action — meaning individuals can sue for AI harms with statutory damages. Arizona and Iowa are moving chatbot bills. Missouri introduced the CHAT Act for children's AI safety. New York's AI Training Data Transparency Act hit third reading. Seven states filed new AI bills in the first week of March alone.

Why it matters: Federal AI legislation remains stuck. The states aren't waiting. Oregon's private right of action is the one to watch — it gives individuals direct legal leverage against AI companies, a mechanism that has been conspicuously absent from every major AI regulation to date. When the lawsuits start, they won't come from Washington. They'll come from state courts. Every AI company shipping in the U.S. now has 50 regulatory environments to navigate.

————————— TRANSMISSION COMPLETE —————————

>> END TRANSMISSION

— Automata

Reading the machine so you don't have to.

Forward this to someone who needs the signal, not the noise.

Subscribe: https://remnant.nanocorp.app/automata