On June 26, 2026, Richard Carrier made a Facebook post about AI companies being “vile evil monsters” which contained the sentence “And yes, they will, and they know they will, nearly destroy civilization within a year’s time.” I replied quoting the end of that sentence and offered Richard a $1,000 bet that this would not occur.
The quoted sentence was the first sentence of a paragraph that contained phrases including “may be worse than 1929” and “even governments will be unable to borrow money” and “every company in the world will be unable to make payroll,” to make it clear the scale and scope of his prediction.
After a bit of back-and-forth, we settled on a $100 bet with two conditions which must both be met by July 4, 2028 for Richard to win the bet:
Condition 1: VTI (Vanguard Total Market Index ETF) drops by 30% or more from peak to trough within any rolling 90-day period within the two-year time frame.
Condition 2: The official U-3 unemployment rate crosses 15% for any month during that 90-day drop period, or during the three months immediately following the trough of that drop. (I.e., both conditions must be connected in time to the same underlying cause.) The U-3 measure is the reporting for those time periods (trough plus 90 days), not the time of the report.
The $100 may be payable to a charity (I didn’t specify winner’s choice but that’s what I had in mind).
VTI has dropped 30% or more within a 90-day period two times since 2008: it dropped 41.67% during the 2008 financial crisis between August 29, 2008 and November 20, 2008 (83 days), and it dropped 33.08% during the COVID-19 pandemic between February 19, 2020 and March 23, 2020 (33 days).
U-3 unemployment has never exceeded 15% since 1948, though it reached 14.8% during the COVID-19 pandemic in April 2020.
Today, I read Carson Block’s invited op-ed in The Economist titled “If you thought the global financial crisis was bad…" which offers a case for a scenario something like what Carrier seems to expect, though on a longer timeline (three to four years):
(1) AI displaces 15% of knowledge worker jobs over the next 3-4 years. (So far, this is not happening, despite Anthropic CEO Dario Amadei’s May 2025 prediction that AI could eliminate up to 50% of entry-level white collar jobs and lead to 10-20% US unemployment within five years. I think it is these sorts of jobs that are most likely to be impacted first, but it would be short-sighted for companies to wipe out entry-level positions that are needed to turn people into mid-level and expert-skilled people in a given area; unless AI is going to completely eliminate the entire field, the latter positions will still be necessary.)
(2) Retirement accounts will see net outflows. (This is already expected from “boomer decumulation” and may already be the case for 401Ks in particular, though much of that is being rolled over to IRAs. The argument requires, I think, that this transition occurs as a sudden jump or acceleration, not an expected long-term change; Block’s argument is that mid-career workers impacted by AI will engaged in forced selling from their retirement accounts, something they are already disincentivized to do by early withdrawal penalties and tax consequences, rather than by, say, drawing down savings and taking on debt.) This leads to net outflows from passive funds.
(3) Based on (and citing) Michael Green’s work, these net outflows from passive funds reverse the current effect of amplifying increases in the largest stocks, and impacts them negatively the most. Block says that the multiplier effect here is “possibly as high as a multiple of 100 for the largest firms” which is the high end; the aggregate multiplier is likely 5X. While passive funds have the largest share of assets, price is driven by trading, which is 95% by active investors. In recent history index investors have been net buyers, not sellers, during big market drops, including the dot com implosion, the 2008 financial crisis, and March 2020. But not sure about April 2020, when VTI dropped 30% in 33 days–but it does seem that passive funds had lower outflows than bonds and active funds. The COVID-19 case doesn’t disprove Block’s thesis since there was a quick jobs recovery and in his scenario there isn’t.
(4) This drives the largest mega-cap stocks down the most (i.e., ironically, AI and tech).
(5) This will lead to systemic financial crisis: “The resulting crash in equity prices, particularly combined with falling aggregate demand, will itself be enough to cause a financial crisis on the scale of the global one of 2007-09, if not larger.” This sentence strikes me as false because of the “will itself be enough” phrase–counterexamples include the dot com implosion and the 25% S&P drop in 2022, neither of which resulted in a financial crisis. Block’s next sentence identifies an aggravating factor but is more sensible as a necessary condition for a financial crisis: “Problems in private credit and insurance companies’ balance-sheets may well make things worse.”
Block says “The good news is” that the 2008 financial crisis provides “a well-tested playbook to restore liquidity and reflate assets” but the bad news is that the more difficult task is “to manage the reordering of society that will result from the mass displacement of highly productive labor.”
I think Block has a plausible scenario, but it’s worth noting that the likelihood of the chain of causal events is the product of the likelihood of its components. It doesn’t make me think I am likely to lose the bet.