Block fired 40% of its workforce and kept the product running. I've been watching their developer tooling and API output since the announcement, and the release cadence hasn't collapsed. Jack Dorsey didn't dress up a cost cut in AI language. He described what the tooling is actually doing.
That's my position, and I'll hold it against the increasingly popular theory that "AI restructuring" is just a PR-friendly label for ordinary layoffs. The productivity gains are measurable. When WiseTech Global says generative AI is making traditional coding approaches obsolete, I don't take that as a press release. I take it as an engineering assessment from people watching their own commit logs.
Show Me the Diffs
If you've used Copilot, Cursor, or any of the code-generation tools that matured through 2025, you know the math has changed. A senior engineer with good prompting skills and a solid understanding of system architecture can now produce what used to require a team of three or four. Not hypothetically. In production. I've watched teams at mid-size companies cut their backend squads by a third and ship faster, because the boilerplate, the test scaffolding, the migration scripts, all of it now takes minutes instead of days.
WiseTech cut 2,000 jobs and explicitly cited generative AI boosting engineering productivity. eBay cut 800 and pointed to AI handling listings and pricing workflows. These aren't vague claims about "future capabilities." These are companies describing tools they already deployed.
The Harvard Business Review survey found that 21% of organizations made large cuts based on AI's projected, not demonstrated, performance. That's a real problem, and Audrey Liang is right to flag it. Some companies are absolutely using AI as a buzzword to justify cuts they'd make anyway. But 21% is not 100%, and treating every AI-linked layoff as theater means ignoring the other 79% where something real is happening on the ground.
Profits Don't Disprove Productivity
The counterargument I keep hearing: companies are posting record revenues, so layoffs can't be about genuine efficiency. This logic is backwards. Record revenues with fewer people is exactly what productivity gains look like. If Salesforce's AI tools handle 30% to 50% of the work, as Marc Benioff claims, and revenue keeps climbing, that's not evidence of a cover story. That's evidence the tools work.
Oracle's situation is different and messier. Cutting 20,000 to 30,000 people to fund data center spending is a capital allocation decision, not a productivity story. I won't defend every layoff that gets the AI label. Some of these are cash management dressed in new clothes. Oracle's cuts look like exactly that.
But Block? Block maintained API output, kept its developer platform stable, and continued shipping Square and Cash App features. With 4,000 fewer people. If you've ever managed a team through a reduction, you know the difference between a company that's gutted and a company that's reconfigured. Block looks reconfigured.
The uncomfortable truth for developers: the same tools we love using are the reason some of our former colleagues don't have jobs. Copilot didn't just make you faster. It made one of you sufficient where two used to be necessary. That's not corporate spin. That's what I see every time I open a PR that was 60% machine-generated and still clean.
Companies owe laid-off workers honesty about what's happening and real severance packages that reflect years of service. They owe the public clear data on what AI actually does versus what they hope it will do. But critics owe the data the same respect. When the output stays constant and the headcount drops, you're not looking at a cover story. You're looking at a new baseline.
Nine thousand jobs tied to AI this year so far. Some of those labels are fake. Many of them are not. The PRs are still merging.
