The Feedback Loop Is Collapsing

Block cut 4,000 employees in February 2026 — nearly half its workforce. The stock jumped 24% overnight.

Jack Dorsey's explanation was clean and confident: "intelligence tools" have fundamentally changed what it means to build and run a company. Smaller, highly talented teams. AI automating more work. The future arrived, and Block was choosing to act on it now rather than drag it out over months or years of slow cuts.

Wall Street loved it. The narrative wrote itself. AI makes companies leaner. Leaner companies move faster. Faster companies deserve higher valuations. Buy.

Fewer people is the speed gain

A company with 10,000 employees cut to roughly 6,000. That's 4,000 fewer people in Slack channels. Fewer people in meetings. Fewer people reviewing pull requests. Fewer people who need to be aligned before a decision ships.

Anyone who's worked on a team that went from fifteen engineers to six knows the feeling. Suddenly decisions happen in hours, not weeks. Code reviews come back the same day. Nobody schedules a "sync" to prepare for the other "sync." You stop spending half your time explaining your work to people who need to know about it but can't actually change it.

The speed gain is real. But it's a coordination cost story, not a technology story.

AI tooling helps. A senior engineer with Copilot or Cursor can scaffold faster, write tests faster, churn through boilerplate faster. That's a multiplier on individual output. The order-of-magnitude improvement in organizational speed? That comes from removing the coordination overhead that 4,000 people were generating just by being part of the system.

Brooks figured this out in 1975. The Mythical Man-Month. Adding people to a project makes it slower because communication pathways scale quadratically. Block rediscovered what software engineering has known for fifty years.

The narrative the market rewards

Here's where the feedback loop starts breaking.

Dorsey cuts half the workforce and says "AI." The stock soars. Every board in the country watches that happen. The signal is unmistakable: if you cut headcount and say the magic word, your valuation goes up.

So now Meta is planning sweeping layoffs to offset AI costs. Other companies will follow. Not because AI has actually replaced those workers' output — but because the market has created an incentive loop where saying it has gets rewarded immediately and any consequences show up quarters later.

That's a collapsing feedback loop. The signal reaching executives and boards isn't "we produced better outcomes." It's "cutting people produced a stock bump." Those are wildly different signals, and they lead to wildly different decisions.

A healthy feedback loop looks like this: you change how your team works, you measure whether output quality and velocity actually improved, you adjust based on evidence. Slow, boring, reliable.

A collapsing feedback loop looks like this: you cut headcount, the stock goes up, you declare victory, and you don't check whether the remaining team can actually sustain the workload until the cracks show up in production six months later.

What gets lost when you optimize for the wrong signal

The 4,000 people Block cut weren't all dead weight. Some of them were. Every company at 10,000 employees has slack in the system. But a significant portion of those people were doing work that someone still needs to do — customer support, risk review, compliance, the kind of operational work that doesn't show up in a product demo but keeps the business from catching fire.

When you cut that work and the stock goes up, the feedback loop tells you the cut was right. But the feedback loop is measuring market sentiment, not operational health. Market sentiment is a leading indicator of investor excitement and a lagging indicator of whether the company can actually function.

There's a version of this that plays out on engineering teams specifically. You cut from twelve engineers to six. For three months, velocity increases — less coordination, faster decisions, the senior engineers stop being blocked by review queues. Leadership looks at the velocity charts and sees confirmation.

Then month four hits. The six remaining engineers are carrying twelve people's worth of institutional knowledge, on-call rotations, and maintenance burden. Technical debt that was manageable with twelve people starts compounding. Incident response gets slower because fewer people know the systems. The people who are left start burning out, and the best ones — the ones with options — leave first.

But by month four, the narrative has already been set. The stock price has baked in the new cost structure. Admitting the cuts went too deep means admitting the story was wrong.

The real question nobody's asking

Dorsey told investors that the majority of businesses will reach the same conclusion within a year. He might be right — not because AI will have replaced those jobs, but because market incentives will have made the cuts inevitable regardless of whether AI can backfill the work.

That's the feedback loop collapsing. The signal from the market is "cut and say AI." The signal from operations is "we're starting to strain." When those two signals diverge and leadership only listens to one, you get a system optimizing for a metric that doesn't measure what matters.

Fewer people is faster. That part is true, up to a point. The dangerous part is attributing the speed to AI and using that attribution to justify cuts beyond what the organization can absorb. AI is good at generating code, summarizing documents, and automating repetitive workflows. It is not good at replacing the judgment, institutional knowledge, and cross-functional awareness that experienced employees carry.

Cutting 4,000 people and calling it a strategy is like removing the brakes from a car and calling it a speed upgrade. Technically accurate. Strategically reckless.

The companies that will actually get faster aren't the ones making headlines with mass layoffs. They're the ones quietly giving their existing teams better tools, watching what actually changes, and making staffing decisions based on evidence instead of market narratives. No press release. No 24% stock jump. Just teams that are genuinely faster because the work got easier — not because half the people disappeared.

Nobody's going to write that CNBC headline, though. It doesn't move the stock price.