AIpocalypse.Now
Today's doom 3.4
Doom 6/10 · 8 stories

Wall Street Prices In Your Replacement, Calls It Upside

Coinbase trims headcount for the automation era while investors model returns on mass unemployment and OpenAI's founders fight in court.

Published · By · Story-level doom average 3.8/10

The Quiet Part, Said Into A Microphone

Coinbase laid off 14% of its workforce and did something unusual. It told the truth. The press release blamed crypto volatility, sure, but it also named the automation era as a reason to get lean now. For years, every layoff memo blamed macro headwinds and strategic realignment. Today a public company looked at its remaining staff and effectively said, fewer of you will be needed soon, and we are getting ahead of it.

The Financial Times made the subtext text. Wall Street is actively modeling returns on mass unemployment. Funds are positioning around the thesis that companies displacing workers at scale will deliver outsized margins, and that the displaced will not, in any timeframe that matters to a quarterly report, organize a meaningful response. This is not a rumor or a hot take. It is a strategy deck. The labor share of income is now an arbitrage opportunity with a ticker symbol attached.

If you were waiting for the moment when the automation conversation stopped pretending to be about productivity gains shared with workers, that moment was this morning, on the second page of the FT.

OpenAI's Diary Goes Public

The Musk versus Altman litigation continues to generate the most useful primary source material in the industry, mostly by accident. Greg Brockman's personal communications and diary entries are now being entered into the record, which is a sentence I did not expect to type but which feels inevitable in retrospect. The Guardian and The Verge are both circling the same question. Did OpenAI abandon its founding mission to benefit humanity in favor of maximizing shareholder returns?

The honest answer is that the mission was always load-bearing marketing, and the for-profit cap was always a temporary accommodation to investors who do not believe in caps. Musk's complaint is less a moral grievance than a contract dispute dressed in robes. He wanted to own the lifeboat. Altman built the lifeboat and sold tickets. The trial is going to keep producing documents that nobody at OpenAI wanted produced, and the cumulative effect is corrosive even if Musk loses every motion.

Meanwhile JPMorgan and BlackRock spent the morning telling anyone who would listen that there is no AI valuation bubble. Demand is durable, capex is justified, multiples are fine. Take this for what it is worth, which is what large institutions always say about the assets they hold heavy positions in. The bubble question will be settled by cash flows in eighteen months, not by a CNBC panel today.

Safety Theater And The Copyright Backlog

The Commerce Department announced expanded AI safety testing agreements with Google, Microsoft, and xAI. The framing is reassuring. The substance is voluntary testing of frontier systems by the companies that build them, coordinated by an agency with no enforcement authority over model deployment. This is useful. It is not regulation. Treating it as regulation is how you end up surprised in two years.

The copyright pile keeps growing. Major publishers sued Meta over Llama, alleging millions of books and articles were ingested without permission. The legal theory is the same one being tested against every frontier lab, and the eventual answer will probably be a licensing regime that locks in incumbents, because only incumbents can afford to license a library. The lawsuits feel like resistance. They are actually the negotiation phase of a settlement that will make training data a permission economy controlled by the same five companies already winning.

And in the small-but-telling category, NPR's analysis suggests FBI Director Kash Patel tweeted AI-generated content built from frames of a Beastie Boys music video. The Beastie Boys are famously, aggressively litigious about unauthorized use of their work. The director of federal law enforcement allegedly running copyrighted material through a generator and posting it publicly is the kind of detail that future historians will use to date the period. Not because it is the worst thing happening, but because it captures the texture. Everyone is doing it, nobody is checking, and the people in charge of checking are doing it too.

The through-line at midday is that the polite fictions are eroding faster than the replacements are being built. Layoffs cite AI on the record. Funds price unemployment as alpha. Founders' diaries become exhibits. Cabinet officials remix copyrighted videos. The accounting is getting honest. The accounting is not encouraging.

Sources cited in this digest
  1. Coinbase Cuts Staff, Blames Crypto Volatility and AI Inevitability · New York Times · 5/10
  2. Wall Street Bets on Profits From Mass Unemployment · Financial Times · 6/10
  3. Musk's Lawsuit Examines OpenAI President's Personal Communications · The Guardian · 3/10
  4. Musk and Altman Battle Over OpenAI's Profit-First Direction · The Verge · 4/10
  5. JPMorgan and BlackRock Dismiss Concerns About AI Valuation Bubble · Financial Times · 2/10
  6. US Government Expands AI Safety Testing With Major Tech Companies · BBC News · 3/10
  7. FBI Director Allegedly Generated Content From Copyrighted Music Video · NPR · 2/10
  8. Publishers Sue Meta for Unauthorized AI Training on Copyrighted Works · The Guardian · 5/10
Today's doom Weekly column