When Block's Jack Dorsey cut roughly 40% of staff in February and framed it as AI, Cloudflare's Matthew Prince countered that the real divide is between what he calls 'generalist proxies' — coordinators, writers, and software mediated roles — and the engineers who physically run…
The AI-jobs conversation is treating two very different kinds of knowledge workers as one headline.
When Block CEO Jack Dorsey announced roughly 40% of his company was being cut on February 26, 2026, the framing was that AI had finally arrived at the white-collar layoff. The same week, reporting on the Dorsey cuts appeared alongside coverage framing the moment as a turning point for AI-driven job losses — a narrative amplified by Dorsey's own framing.
The counter arrived from inside tech, in a Wall Street Journal op-ed by Cloudflare co-founder and CEO Matthew Prince titled "How I Choose Which Cloudflare Employees to Replace with AI." Prince did not argue AI was about to wipe out engineers. He argued something narrower: AI replaces generalist workers whose output runs through software, and raises the value of what he calls "builders": the engineers, technicians, and creators who actually produce the product.
Dorsey's cuttable workforce, in Prince's terms, were not builders in the structural sense. They were knowledge-workplace proxies: people whose primary artifacts were tickets, slides, memos, dashboards, and software-mediated workflows that an AI can increasingly draft, route, or skip. The 40% figure that Dorsey walked through in a follow-up Fortune interview is the CEO's framing of a reorg that used AI as the rationale.
Manufacturing engineers fall into a different category. As a February–March 2026 PFOnline analysis argued, the knowledge an industrial, electrical, materials, or process engineer carries is not a static file or a recurring meeting. It is physically instantiated: in a process window narrowed over years of yield work, in a fault signature tied to a specific spindle, in a corrective-action log that only makes sense inside the line that produced it. The engineer's job is not to produce text. It is to keep the production system producing.
That is the builder category in Prince's sense. The unit of account is the production system, not the document trail around it. AI can be threaded through the software that runs design (CAD), machine programming (CAM), and shop-floor execution (MES), and it almost certainly will be. It can draft the failure analysis report, write the work instruction, summarize the downtime event. None of that eliminates the engineer whose name is on the line when the line goes down.
There is a real caveat. Prince's framework is a tech CEO's framework, derived from running a company that Cloudflare describes as handling a large share of internet traffic, and that Fast Company's 2026 Most Innovative Companies profile frames as AI-era infrastructure. Applying it to factory floors is an extrapolation, not an audited fact. The manufacturing-engineer thesis is reasonable but unproven, and AI's effect on manufacturing headcount depends on which factories, which processes, and how fast automation and AI tooling advance together.
The test will come in the next round of layoff filings. If the second wave concentrates in support, coordination, and software-proximate roles while engineering, maintenance, and process-control headcount stays flat or grows, the structural argument is starting to hold. If manufacturing-engineer postings soften in tandem with software cuts, the elevated-engineer thesis is wrong, and the headline AI-jobs collapse is real.