Meta and Snap Just Did What Companies Have Spent Years Avoiding: Blamed AI for the Cuts
TexTak holds [white-collar-displacement] at 70% — the probability that a major layoff wave gets explicitly attributed to AI automation, not quietly absorbed through attrition. Today's news moves us closer to treating this as effectively resolved: Meta citing AI in cutting 8,000 roles, Snap announcing AI generates 65% of new code while laying off 16% of its workforce, and the year-to-date tech layoff counter at 95,878 workers. These aren't companies quietly reducing headcount — they're putting AI attribution in press releases and earnings communications. The 70% is starting to look like a floor, not a midpoint.
Let's be precise about what TexTak is actually forecasting here, because it matters for the probability. The question was never whether AI would displace workers — that's been happening for years in the attrition-based, non-attributable way the 'against' case describes. The forecast target is explicit public attribution: companies standing in front of investors and employees and saying 'AI did this.' That behavioral threshold is the hard one, and today it got crossed by two companies whose combined market cap exceeds $1.5 trillion.
Snap is the cleaner signal. CEO Evan Spiegel didn't hedge — he said AI generates 65% of new code and that this allows smaller teams to produce the same output, directly linking capability to headcount reduction. That's not a PR department accidentally letting something slip; that's a prepared executive communication to investors with a $500M cost savings target attached. Meta's framing is similarly direct: 8,000 roles cut with explicit citation of AI automation in content moderation, customer support, and engineering. These are not attrition numbers. These are announcements.
Our 70% was built on three factors: (1) the technical capability for substitution was already present in back-office and coding functions; (2) investor pressure for AI ROI was intensifying in a way that would eventually force companies to demonstrate the returns; and (3) a small number of high-profile attributors would break the dam for others. We're watching (3) happen in real time. New York assemblymember Alex Bores proposing an AI Dividend tax on token usage is a leading indicator — that kind of legislative response only emerges after public attribution has become undeniable enough to generate political salience.
Honestly, the counterargument that keeps us honest is partial rather than full: 'companies avoid PR risk of attribution' was always true for most companies, not all. The question was whether the economic incentive to demonstrate AI ROI to investors would eventually outweigh the reputational risk with employees and regulators. What today shows is that for companies where AI's substitution effect is measurable and large — Snap's 65% code generation figure is verifiable through internal metrics — the investor narrative wins. The PR risk is real but manageable when you have a clean number to point to. What would drop us below 50%: evidence that these announcements trigger regulatory backlash severe enough that the next wave of companies reverses course on attribution. We're watching the New York AI Dividend bill and whether similar proposals in other states create a chilling effect on explicit attribution language. If three more states introduce AI displacement tax proposals by Q3, the cost-benefit calculus for public attribution shifts again.