On the morning of February 27, 2026, Jack Dorsey did something that looked, on the surface, like an act of technological conviction. He fired more than four thousand people. Block — the company he runs, the parent of Square, Cash App, and Afterpay — went from over ten thousand employees to just under six thousand in a single announcement, one of the largest cuts in its history. And he did not hide behind the usual euphemisms about rightsizing or a difficult macro environment. He named a cause, and the cause was artificial intelligence. A significantly smaller team, he told shareholders, using the tools the company was building, could do more and do it better. Intelligence tools, he wrote, had changed what it means to build and run a company.
The market adored it. Block’s stock leapt as much as twenty-seven percent in after-hours trading, adding roughly six billion dollars in market value before most of the people who had just lost their jobs had finished reading the email. Dorsey owns close to ten percent of Block. The arithmetic is not subtle, and the market ran it instantly: the man who signed the order to erase four thousand jobs watched his own paper wealth climb by something on the order of half a billion dollars in the time it takes to publish a post. That was the productivity gain. That was the metric that moved that day. Not output. Not shipped product. Not a single unit of customer value. The share price moved, and the founder’s fortune moved with it.
I want to be precise about what I am arguing, because I build AI systems for a living and I have no interest in the comfortable fiction that these tools are vaporware. They are not. Block’s own coding agent is genuinely good, and I will defend the technology in a moment. But the technology is not the story. The story is the narrative Dorsey wrapped around it — that a machine made him do this, that four thousand people had been rendered redundant by software, that what happened in February was destiny rather than a decision. That narrative is a lie. And it is a profitable one, which is exactly why he told it.
The smaller team that did more, except it didn’t
Start with the promise, because it is testable. Dorsey said a leaner, AI-powered Block would build faster and ship more. So look at what it has actually shipped since the cuts, and look closely at the dates. The marquee product is Moneybot, an AI assistant inside Cash App that analyzes spending and answers financial questions, and Block has held it up as proof of the intelligence-native future. There is only one problem. Moneybot was introduced, in limited release, in Block’s fall update in late 2025 — months before the layoffs. It reached general availability in the first quarter of 2026, where it promptly drew a million active users in a single week with no marketing. Impressive. Also built, designed, and launched by the larger, pre-layoff organization. Block is now crediting its smaller workforce with shipping a product the people it fired had already finished.
The rest of the post-layoff roadmap is real but unremarkable as evidence of an AI miracle. Managerbot, an assistant for Square sellers, reached more than a million merchants and was on track to cover all of them by June, with each seller assigned more than a hundred agents to pull data and automate routine tasks. There is Neighborhoods, a discovery feed linking Cash App users to Square merchants, which crossed three hundred and twenty million dollars in annualized payment volume. And there is Goose, the engineering agent, which Block open-sourced and which has since gathered more than forty-two thousand stars on GitHub and moved to the Linux Foundation’s Agentic AI Foundation. Here is where I will keep my promise and defend the technology: Goose is excellent. Block co-developed the Model Context Protocol with Anthropic, the standard a great deal of the industry now builds on, my own work included. None of this is fake.
But none of it required firing four thousand people, and none of it proves that four thousand people were replaceable. A useful coding agent is a useful coding agent. It is not a headcount theory.
The productivity that never showed up
Here is the part Dorsey is counting on you not to examine: the numbers. On their face, Block’s first quarter of 2026 looks like vindication. Gross profit of 2.91 billion dollars, up twenty-seven percent year over year. Adjusted operating income up fifty-six percent to 728 million dollars, at a record margin. Adjusted earnings per share of eighty-five cents, up fifty-two percent and an all-time high. If you stop reading there — and the headlines did — the AI gamble paid off spectacularly.
Do not stop reading there. First, that growth was already happening. Block’s gross profit growth ran nine percent, then fourteen, then eighteen, then twenty-four, then twenty-seven across the five quarters ending in the first quarter of 2026 — a smooth, steady acceleration that began long before anyone was fired. The layoffs landed in the final week of the quarter. They could not have produced those results, and they didn’t. Second, you do not have to take my word for it, because Block told you the same thing. Its own guidance stated that the new cost structure would only begin to meaningfully affect results in the second quarter, with the full impact arriving in the second half of the year. By the company’s own calendar, the efficiency it fired four thousand people to capture had not yet arrived when the stock popped. The market paid up front, in full, for a benefit that was entirely a promise.
Third, the quarter that actually contained the layoffs was not a triumph on the bottom line. It was a loss. Block posted a GAAP operating loss of 172 million dollars and a net loss of 309 million dollars, dragged down by 852 million dollars in restructuring charges. The efficiency move cost the better part of a billion dollars and produced a red quarter. The adjusted numbers that made the headlines are simply the numbers with that pain politely removed.
And then there is the metric at the center of the whole religion: gross profit per employee. Dorsey’s stated target is more than two million dollars per person, which he called four times the company’s pre-pandemic efficiency. It sounds like a productivity breakthrough. It is mostly division. If you hold revenue roughly constant and cut the denominator — the people — your profit per head goes up automatically. That is not productivity. That is arithmetic dressed as achievement. The engineering proof is no better. Block’s headline is that its agent let engineers ship forty percent more code, and the tool’s own creator has said roughly ninety percent of his lines of code are now written by Goose. I have spent thirty years writing software, and I can tell you that lines of code written is the single worst proxy for value in the entire discipline. More code is not an accomplishment; it is a liability you maintain forever. A senior engineer’s job is to write less code, not more. Celebrating a forty percent increase in code volume is like celebrating a surgeon for making more incisions.
Ask the people who were in the room
If the AI had genuinely absorbed the work of four thousand people, you would expect the remaining organization to hum along. Instead, within weeks, Block began quietly rehiring. A design engineer who builds agentic AI experiences at the company — the very work Dorsey was celebrating — posted that his layoff had been a clerical error and that he had been asked back. A creative strategy lead returned. A recruiter’s manager escalated his case all the way to the CEO to bring him back. Most telling of all, a technical lead for Square’s online and site operations explained that the cuts had left him as the sole remaining member of his team, responsible for infrastructure critical to customers, and that he had to threaten to quit before leadership agreed to rehire some of his people. That is not what replacing humans with intelligence looks like. That is what cutting too deep and scrambling to undo it looks like.
Then there is the testimony of the people who stayed. A Block data scientist, offered a ninety percent raise to remain after the layoffs, quit on the spot — saying publicly that she would rather her colleagues keep their jobs than profit from their loss, and that internally the company had seen very limited gains in productivity from the AI it was pushing on everyone. A person paid almost double to stay, telling you the productivity story is hollow from the inside. Outside analysts said the same in plainer terms; one prominent fintech analyst called a forty-percent cut in a single stroke overly aggressive and the AI-impact claim overstated. Meanwhile, the quarter Block reported alongside the layoffs slightly missed revenue estimates, and the stock — even after its triumphant pop — remains down roughly eighty percent from its 2021 peak. This is not a company at the summit of its powers electing to optimize. It is a company recovering from years of overhiring that found a flattering way to talk about the cleanup.
What a money grab actually looks like
So if the productivity case is this thin, why do it, and why do it this way? Follow the money, and notice the asymmetry of certainty. The benefit Dorsey promised — AI-driven efficiency — is deferred, unproven, and by his own guidance had not yet materialized. The cost — four thousand livelihoods — was immediate, real, and borne entirely by other people. And the payoff that was instant, guaranteed, and enormous flowed in exactly one direction: to the owners of the stock. Six billion dollars in market value, conjured in an evening, roughly half a billion of it landing in the personal account of the man who made the decision. Dorsey himself was blunt in a way that should have set off alarms; he told shareholders the company was not making this decision because it was in trouble. Read that again. Not a rescue. Not survival. A choice, made from a position of strength, that moved a vast sum of money from labor to capital and rewarded the person making it more than anyone else.
This is the mechanism the AI-jobs story is built to obscure. Modern executive pay is overwhelmingly equity. When the stock jumps, the people who run the company get rich; when payroll shrinks and margins expand, the stock jumps. The layoff is not a side effect of the compensation structure. The layoff is the compensation event. And artificial intelligence made it necessary is the perfect cover story, because it is unfalsifiable in the moment, it flatters the executive as a visionary rather than a cost-cutter, and it transfers all blame to a machine that cannot be shamed. Insiders, for what it is worth, were net sellers of Block stock in the months around the cuts — selling into the rally they had engineered, and buying none. Make of that what you will.
The lie they are all telling
Block is not unique. It is a template, and Dorsey said so out loud — he predicted that within a year most companies would reach the same conclusion and make the same cuts. That prediction is not analysis. It is the narrative doing its job, manufacturing inevitability so that the next CEO who guts a division can point at Block and say the future left them no choice. You have heard the chorus. A senior AI executive warning that white-collar workers have a year to eighteen months. A famous bank CEO and a former presidential candidate nodding along. The drumbeat is constant and the message is always the same: this is coming for you, it is unstoppable, resistance is naive.
Here is the truth an actual practitioner will tell you. The tools are real and they are useful. I use them every day and they make good engineers better. But useful tool and your job was eliminated by a robot are not the same claim, and the distance between them is where the con lives. An Oxford Economics analysis found what anyone honest already suspected: a large share of the layoffs that executives blamed on AI were in fact corrections to pandemic-era overhiring. The AI did not fire those workers. Executives did, for ordinary financial reasons — bloated headcounts, margin targets, a stock price to feed — and then reached for the most fashionable available excuse. The AI will take your job narrative is not a forecast. It is a management tool. Its purpose is to get you to accept as technological destiny what is, in fact, a deliberate decision to move money out of your paycheck and into a shareholder’s portfolio. Once you believe the machine is coming no matter what, you stop asking the only question that matters: who decided this, and who got paid?
A way to make the lie expensive
If we want CEOs to stop dressing up wealth transfers as innovation, we have to make the dress cost something. So here is a proposal, and I will state it plainly: tax revenue per employee. Make it progressive. Below a sensible threshold for a given industry, a company pays the normal corporate rate. Above it — as the gross profit a company extracts from each human on its payroll climbs into the stratosphere — the marginal rate climbs with it. The more revenue you wring out of each person you employ, the more you pay on that excess.
Consider what this actually does. First, it closes a tax arbitrage that almost nobody talks about. Wages are taxed coming and going — payroll taxes, then income tax in the worker’s hands. Corporate profit is taxed once, and the shareholder gains it produces are taxed lightly and often deferred for decades. That means converting a workforce into a smaller workforce plus fatter profit is, before it is anything else, a tax optimization. A revenue-per-employee tax neutralizes the incentive to strip out human beings purely to shift the tax base from labor to capital. Second, it internalizes a cost the public is currently forced to eat. When Block cuts four thousand people, society absorbs the unemployment claims, the hit to local economies, the lost income-tax receipts, and the retraining. Block keeps one hundred percent of the upside and externalizes the downside onto the rest of us. A tax keyed to profit-per-head claws a portion of that windfall back to fund the safety net for the people the decision displaced. Third, and most important, it removes the one-way bet. It does not ban AI. It does not ban automation. I would never argue for that, and I build this technology. It simply says that if you are going to generate two million dollars of gross profit per remaining employee by handing the work to machines, you will share a slice of that extraordinary number with the commons that paid the cost of getting you there.
The objections write themselves, and none of them survive contact. Will this punish genuinely efficient, capital-intensive businesses? Not if you key it to gross profit per employee rather than raw revenue and set thresholds by sector, so the target is the company pulling two million per head after gutting its staff, not the high-throughput business running on thin margins. Will companies game it by reclassifying employees as contractors or offshoring them? Only if you write it badly; count full-time-equivalent labor including contractors and offshore heads, and the loophole closes. Tax law has handled harder definitional fights than this one. Should government stay out of picking efficiency winners? It is not staying out now — the current code actively subsidizes the shift from labor to capital by taxing the two at wildly different rates. This proposal does not introduce a distortion. It removes one. The goal is narrow and the goal is fair: end the regime in which the single most profitable thing a chief executive can do in an afternoon is convert four thousand employees into a press release.
The story they want you to believe, and the one that is true
Strip away the language about intelligence tools and a new way of working, and what happened at Block in February is an old story in a new costume. A company that hired too aggressively cut deeply, blamed a machine, and watched its stock — and its founder’s fortune — soar on the strength of a productivity gain that, months on, has not arrived and may never be separable from the trend that was already underway. The people who were actually doing the work got an email. Some of them got a second email asking them to come back, because it turned out the machine could not, in fact, do their jobs. The man who signed both emails got half a billion dollars richer.
The AI did not take those four thousand jobs. Jack Dorsey did, and he was paid extraordinarily well for it. Until we change the incentives that made that the smart move, expect a great many more intelligence-native press releases, each one timed to an earnings call, each one narrated as destiny. Do not believe the story. Ask who decided, and ask who got paid. The answer is almost always the same person.
About the Author
Gal Ratner has spent nearly thirty years building and shipping production software on the Microsoft and .NET stack, for clients including Microsoft, Sony, Rockstar Games, 2K Games, Best Buy, and Allegiant Air. He is Chief Architect at Prana Entertainment and the founder and CTO of Inverted Software and WhiteStar Labs, an enterprise software and AI consultancy based in Las Vegas. He was the sixth employee at Break.com during its run as one of the largest entertainment sites on the web, and a finalist for the Los Angeles Business Journal’s CTO of the Year.
His current work is agentic AI in production — MCP servers, the Microsoft Agent Framework, SQL Server 2025 vector search, and the open-source diagnostics tooling behind PLogger. He writes about AI the way a practitioner sees it: bullish on what the tools can actually do, and allergic to the marketing wrapped around them. He publishes under the banner The .NET AI Guy That Ships.
When he is not shipping code or picking apart AI hype, he rides motorcycles and co-hosts the Edge Grip Podcast, trains Brazilian jiu-jitsu in Las Vegas, and is the author of the novel The Archive of Lost Suns.