The Trick Behind "Free Shipping," "Just a Post," and "We're Profitable"
Amazon called a price hike free shipping. Facebook called an ad a post. AI companies call a product pitch an extinction warning. Anthropic's own IPO numbers just handed the same trick its newest costume.
What was checked, and what could not be.
Amazon Buy Box claim: anchored to the 2016 ProPublica investigation, carried here via a Retail Dive summary link rather than the original ProPublica piece directly; the roughly 80% figure and the shipping-cost-exclusion mechanism both come from that investigation, and Amazon's own rebuttal, that nine in ten orders already ship free, is included, which is the fairer way to present a contested industry finding.
Facebook mobile-ad history: anchored to two dated sources, TechCrunch's October 2012 earnings report and MIT Technology Review's March 2013 followup. The engagement-dip figure and the mobile ad pricing figure are both attributed to Facebook's own internal testing, as reported by MIT Technology Review, not independently audited by a third party. That is disclosed correctly as coming from the company's own internal testing.
The FLI and CAIS correction is the piece's most careful piece of sourcing: both letters are linked directly to their publishing organizations (futureoflife.org, safe.ai), and the claim that Musk did not sign the CAIS statement and Altman and Amodei did not sign the FLI letter is checkable against each org's own signatory list. This correction is presented as fact, not opinion, and it holds up against the primary sources given.
Anthropic's profitability claims: sourced to a single wire chain, Financial Times reporting, carried by Irish Times, with Yahoo Finance also cited. The piece is explicit that other reporting could not independently verify the figures and that Anthropic declined to comment, which is an important caveat carried into the piece itself rather than omitted. The Chanos quote is sourced directly to an X post. The Asay quote (that the companies "weren't and aren't profitable" and calling it a "made-up accounting fiction") has no direct source link in this piece's dossier, unlike the Chanos quote, so that attribution is thinner than the others.
The central interpretive claim, that doom messaging and profitability messaging are the "same trick" as Amazon's shipping math and Facebook's ad design, is an analogy, not a factual claim, and the piece is careful to frame it that way as its own read of the incentives, not a proven case of coordination.