Ghost GDP and the Day a Hypothetical Moved Markets
Yesterday, a memo from the future slipped into the present. Fortune amplified a viral Substack “pre‑mortem” by Citrini Research that reads like a June 2028 post‑mortem on a white‑collar recession we haven’t had—yet. In that imagined endpoint, U.S. unemployment touches 10.2% and the S&P 500 sits 38% below a 2026 peak. The culprit isn’t a banking crisis or an oil shock. It’s the moment agentic AI graduates from quiet assistant to tireless operator, producing so much work without wages that the economy becomes top‑heavy on output and light on paychecks. Citrini gives the imbalance a blunt name: ghost GDP.
A Memo That Time‑Traveled
The piece is not a forecast, and its authors are careful to underline that. It’s a stress test aimed at a collective blind spot: the easy assumption that “more AI productivity” equals “more prosperity” across the board. By writing from a near future where that assumption breaks, the memo treats left‑tail risk as a scene, not a statistic. We’re invited to inhabit the downstream logic of extremely fast adoption and to ask whether the institutions that convert innovation into incomes can keep pace.
When Copilots Become Agents
In the memo’s world, the tipping point isn’t model accuracy alone; it’s reliability across planning, execution, and verification. Once agents can chain steps and check their own work, they stop being a gloss on human workflows and start being the workflow. Firms thread them through coding, research, customer support, procurement, and finance. Output stays shiny. Dashboards glow. But the wage bill thins. Mid‑skill knowledge roles—large enough to anchor household spending and mortgage payments—compress. The result is a paradox the memo names and circles: measured productivity climbs while the people who buy the output fall behind it.
The Intelligence Displacement Spiral
That paradox doesn’t sit still. As incomes soften, consumption cools. With targets unchanged and costs scrutinized, managers double down on agents to make the numbers. Each turn of that wheel adds slack to the labor market and subtracts confidence from households. White‑collar job losses leak into credit and housing, not because workers are unproductive but because the productivity no longer lands in their wallets. In this telling, the feedback loop is the story: automation begets layoffs, layoffs beget weaker demand, weaker demand justifies deeper automation.
Where Agent Behavior Shreds Moats
The scenario is concrete about where this pressure shows up first. In software, enterprise buyers use coding and IT agents to replicate features they once licensed, dulling growth at the exact moment valuations assume compounding seat expansion. Headcount follows revenue down. In payments and delivery, autonomous buyers—software shopping on behalf of humans—optimize for latency and cost, not for habit or brand. The fee‑laden detours that supported card networks and delivery platforms become expensive routes agents refuse to take. It isn’t disruption by a better product; it’s starvation by a better optimizer.
The Household Balance Sheet Takes the Hit
Once prime earners wobble, risk migrates. Mortgage books look different when white‑collar income volatility rises. Private‑credit portfolios that rolled up software vendors suddenly face churn they didn’t underwrite, as customers self‑provision with agents or simply need fewer seats. The memo lingers on this plumbing not to be dramatic, but because the plumbing is how macro pain propagates. Wages fund spending; spending validates revenue; revenue justifies payrolls and debt. Break the first link and the rest starts creaking.
Hypothetical, Meet the Tape
Because employment is the transmission channel, the story was covered as a jobs piece with macro consequences. What stands out is not only the thought experiment itself, but how quickly a thought experiment changed behavior. By the next trading day, names the memo framed as exposed sold off. Markets reacted to a scenario precisely labeled as a scenario. That’s new. It suggests investor psychology has shifted from asking whether agents will arrive to asking which cash flows look mispriced if they arrive fast.
Why This Matters Now
The uncomfortable idea at the center of “ghost GDP” is that a surge in measured productivity can coexist with stagnating household demand if the surplus accrues to machines and the owners of the machines. Companies, acting rationally, take labor savings and buy more compute instead of more labor. The spreadsheet loves it. The mall does not. If that wedge widens faster than new, well‑paid human work appears, the macro math stops balancing—even as output hits records.
Citrini’s point is not to claim the future but to price it. If agents scale on enterprise timelines while institutions adjust on political ones, the gap is where the risk lives. Yesterday’s reaction was a tell: a vocabulary—ghost GDP, displacement spiral—escaped a Substack and briefly set the cadence of the market. Whether the scenario comes true is secondary to the reminder it delivers. We’ve built systems that are excellent at converting intelligence into compute. We have not yet proved we are as good at converting compute into incomes.
