AI doomsday Substack scenario rattles US markets

The S&P 500 slid more than 1% on Monday as a speculative “AI doomsday” scenario circulated on Substack, and a handful of named consumer and payments stocks fell 4% to 6% in sympathy. According to the Guardian, the post—published by a little-known firm, Citrini Research—described a 2026–2028 timeline in which AI “agents” strip friction out of the US economy, pushing unemployment above 10% and triggering an “Occupy Silicon Valley” protest outside OpenAI and Anthropic.

The mechanism in the scenario is not a new model release so much as a new habit: software that can shop, book, bargain and build workflows without a human in the loop. Citrini argues that once an autonomous agent can complete tasks end-to-end, it does not “loyally” return to DoorDash or Uber; it routes around them, fragments demand into countless bespoke alternatives, and compresses margins across the middleman economy. It extends that logic to payments, imagining agents choosing crypto rails to minimise fees—an explicit hit to Visa and Mastercard—and to enterprise software, where tools such as OpenAI’s Codex and Anthropic’s Claude Code are presented as substitutes for subscription platforms.

Investors reacted to the narrative because it offers a coherent story for why a seemingly healthy economy could tip: revenue models built on habit, bundling and contractual lock-in become vulnerable when the user is no longer a user but a machine optimising cost. The Guardian quotes Saxo analyst Neil Wilson calling it “doomsday porn”, but the market move shows why the genre persists: it provides a ready-made causal chain that can be traded, hedged and repeated.

That feedback loop runs into a second market: regulation. If “agents” are framed as systemic risk—capable of detonating employment, credit and consumer demand—then the obvious political response is to demand licensed deployment, audits, and controlled access to frontier systems. Those requirements are expensive, and they naturally favour incumbents with compliance teams, legal budgets and privileged relationships with regulators.

OpenAI’s own structure points to how much power concentrates when the product is treated as critical infrastructure. Business Insider’s look at the company’s internal org chart shows CEO Sam Altman with 10 direct reports, while Fidji Simo—installed as CEO of Applications—runs consumer and enterprise products with 13 direct reports and, according to a source cited by Insider, manages nearly two-thirds of the company. In other words, the part of OpenAI that touches the public is already being organised like a scaled operating business, while the research and safety decisions remain in a small executive circle.

When markets swing on a scenario labelled “not a prediction”, the winners are not only the traders who timed the move. The firms best positioned to sell “safety” are the same ones large enough to survive the paperwork.

Citrini’s scenario ends in June 2028 with tents outside AI labs. For now, the more immediate sign of influence is that a Substack post can move blue-chip shares in a single morning.