Snap Cuts 1,000 Jobs. AI Writes 65% of Their Code. Stock Goes Up 8%.

Snap CEO Evan Spiegel announced 1,000 layoffs and closed 300+ open roles, explicitly citing that AI now generates over 65% of the company's new code. The stock rose 8%. Snap expects $500M+ in annualized savings. This is the clearest case yet of a public company quantifying AI's direct displacement of engineering headcount, with Wall Street rewarding the decision immediately.

Read that again alongside the 43% Denominator story. If three independent studies show AI code needs significantly more debugging, rework, and post-merge fixes, what does "65% AI-generated" actually mean? Is Snap measuring lines written? PRs merged? Commits pushed? And are they tracking the rework that follows? I don't know. But the market clearly doesn't care about the denominator. It cares about headcount reduction.

That tension is the story of 2026. Companies are cutting engineers based on input metrics (code generated) while the output metrics (working, stable, secure software) might be getting worse. Snap isn't alone. 96,000+ tech workers have been laid off in 2026 so far according to Trueup. In April alone, roughly 19,000 confirmed cuts came from Meta (8,000), Salesforce (~1,000), and Snap (~1,000), all citing AI as the primary driver.

Here's what I think. The companies cutting headcount based on AI code generation metrics are making a bet that hasn't been validated yet. They're extrapolating from "AI can write code" to "AI can replace the person who writes code," and those aren't the same claim. Writing code is maybe 30% of what a senior engineer does. The rest is debugging, architecture, reviewing, communicating, and making judgment calls about what not to build.

But I also think it doesn't matter what I think. The stock went up 8%. The incentive structure is set. Every public tech company CFO just watched Snap get rewarded for cutting engineers and citing AI. More layoffs are coming. If you're a working engineer, the best defense is being the person who catches what AI misses. Review harder. Test deeper. Be the denominator.