The numbers arrived on the show as good news: AI doing almost half of today’s cognitive tasks, economic growth reaching 15% a year, an enormous increase in output. Some panelists thought even that understated what was coming. Emad Mostaque heard a different story: people losing the paychecks that let them buy what the economy produces.
“We need to catch them,” he said in the Moonshots discussion.
The argument was not simply about whether AI could make society richer. It was about whether the route from greater production to household spending would still work—and whether familiar tools such as dividends and basic income would be enough to repair it.
A bigger economy, a smaller share for labor
The September 2026 Anthropic Institute working paper, Economic Scenarios for Transformative AI, by Anton Korinek and colleagues, models economic transitions from 2026 through 2030. Its scenarios are conditional exercises, not forecasts: different assumptions about automation, productivity, investment, research acceleration and workers’ movement between occupations produce different outcomes.
In the modest scenario, AI adds less than half a percentage point to annual growth by 2030. In the substantial scenario, output ends up about 8% above a no-AI baseline. Most of the scenarios’ divergence occurs after 2027.
The extreme scenario is the one the panel seized on. By 2030:
- AI automates about 49% of today’s cognitive tasks—the thinking, writing and analysis involved in work.
- Annual real GDP growth, meaning growth in economic output adjusted for price changes, reaches 15%.
- Output is roughly 32% above the no-AI baseline.
- Labor’s share of income falls from about 60% to 45%.
- Unemployment among cognitive workers reaches 17.9%.
Those figures describe different things. Automating 49% of tasks does not mean eliminating 49% of jobs. And 17.9% is unemployment among cognitive workers, not across the whole economy. Other occupations expand in the model.
Labor’s falling income share also does not mean the total wage bill must fall by the same proportion. Workers can receive a smaller slice of a larger economy. In the paper, aggregate wages can rise even while cognitive wages fall, with workers moving into other occupations. How readily wages adjust and how quickly displaced workers find new work strongly affect the results.
The demand objection—and what the model leaves out
Mostaque challenged the connection between that expanding economy and the people expected to spend money in it. If returns flow overwhelmingly to capital—the businesses, equipment and other productive assets people own—what happens to households that depend on wages?
He warned of “a complete collapse in aggregate demand.” Aggregate demand is total spending on goods and services, including household purchases and business investment. His concern was that production could surge while displaced workers lost purchasing power, undermining sales and the boom itself.
He also described the paper’s aggregate wage outcome as mathematically impossible. But a declining labor share does not, by itself, establish that. The model includes expanding non-cognitive occupations, and total labor income can grow alongside falling pay in cognitive work. Spending is not funded only by cognitive workers’ wages; capital income, other wages and investment also matter.
There is nevertheless a demand channel the authors explicitly leave out: feedback caused by prices failing to adjust quickly. If spending drops but prices do not fall enough to clear the market, businesses can cut production and employment, deepening the disruption. The paper also sets aside financial disruptions and political constraints.
Mostaque’s objection therefore reaches beyond the model’s boundaries, but it does not establish that a collapse must occur. The paper explores how productive capacity and labor markets could change under its assumptions; it does not settle how a demand shock or political response would alter that transition.
On the show, another panelist offered a counterpoint: humans still own the means of production. As long as that remains true, the panelist argued, societies already have ways to distribute the returns—dividends, sovereign wealth funds and universal basic income. Mostaque did not reject those mechanisms. His emphasis was on the disruption before support reaches people, and the need to prepare rather than assume a larger economy will automatically protect them.
Four ways to share the gains
The proposed remedies answer different questions: who is entitled to support, where the resources come from, and how much those resources buy.
Membership-based provision changes the entitlement. In his May 2026 theoretical paper, Intelligent Economics, Mostaque proposes connecting community provision directly to membership rather than paid employment. If machines increasingly produce goods, people would not have to justify access solely through their contribution as workers. He argues that a subsistence payment can relieve deprivation while leaving work’s roles in identity, community and dignity unresolved. The framework leaves the definition of membership and flourishing to collective decisions; it does not provide an empirical timetable for displacement or, by itself, a settled funding and ownership arrangement.
A sovereign wealth fund changes who holds assets. A publicly owned investment fund can collect returns and use them for public services or payments to residents. It gives the public an ownership-based claim on income rather than relying entirely on wages. But the fund first needs assets or money to acquire them. Citing the mechanism does not answer how a public stake in AI-generated wealth would be built.
A universal dividend changes how returns are paid out. Broadly shared investment income could reach individuals through regular dividends. A sovereign wealth fund could supply those returns, so the two ideas are complementary rather than competing. A universal basic income is a different kind of promise: an unconditional payment that could be financed through taxation, investment returns or a combination. Neither label alone specifies who bears the cost or guarantees that payments keep pace with lost earnings.
Diamandis’s universal high income depends on purchasing power. Peter Diamandis put forward $3,000 a month as his envisioned payment. His argument for calling it “high” rather than “basic” income rests on AI and robotics driving prices down, allowing the same dollars to buy more. That is a claim about the future cost of living, not a funding mechanism. The discussion did not establish how the payment would be financed or how far essential costs would fall.
The ownership condition
The panel’s counterargument about familiar redistribution tools had a boundary. It concerned an economy in which humans still own productive capital. A separate, hypothetical concern was an AI-only economy in which machines trade with one another and humans lose an effective claim on production. That would raise questions of ownership and control beyond the size of a transfer payment; it is not an outcome established by the working paper’s scenarios.
For the transition the paper actually models, the authors conclude that aggregate gains create potential resources to compensate displaced workers, without determining their distribution. Turning those resources into household income would require something concrete: a tax-funded payment, a publicly held asset, a dividend entitlement or a right to community provision. For someone whose paycheck disappears, a larger GDP is not yet any of those things.