This is not an April Fools’ joke — even though the timeline reads like one.

On February 26, 2026, Block laid off ~4,000 people. Forty percent of its workforce. Dorsey told shareholders that most companies would reach the same conclusion within a year.

On March 31 — 33 days later — Dorsey and Roelof Botha published “From Hierarchy to Intelligence” on Sequoia Capital’s site. An essay on how AI can replace hierarchical coordination. Five thousand words about the future of organizations. Zero mentions of the 4,000 who had just lost their jobs.

Is this visionary organizational design, or the most sophisticated layoff rationalization ever published?

I’ve been doing this for 35 years. I’ve seen re-engineering, lean, agile, holacracy — every decade brings its version of “this time we’ll reorganize everything.” What makes this case different isn’t the answer. It’s that almost nobody is asking the right question.

What the essay says

Credit where it’s due: the essay is intellectually serious. It’s not a vague X thread. It’s a 5,000-word piece tracing organizational coordination from the Roman legion to the modern corporation — arguing that AI is the first technology capable of replacing what hierarchy does.

The central argument: hierarchy exists because humans are bad at transmitting information. The Roman army discovered a leader can effectively manage three to eight people. Since then — through McCallum’s railroads in 1850, Taylor’s scientific management, McKinsey’s matrix org — every organizational innovation has tried to work around that limitation without breaking it. More layers mean slower information. The price of coordination is speed.

Block proposes breaking that tradeoff with two “world models”:

Company world model. Block is remote-first. Everything they do produces artifacts: decisions, code, designs, discussions. An AI system can monitor those artifacts continuously and maintain a live picture of projects, blockers, resources, and outcomes — what a middle manager used to do by walking hallways or reading status reports.

Customer world model. Cash App sees the buyer. Square sees the seller. Both sides of every transaction, every day, millions of times. People lie on surveys, abandon carts, and ignore ads. But when they spend, save, lend, and repay — that’s the truth. Block argues this signal is more honest than any product roadmap.

On that foundation, they propose building four things: capabilities (payments, lending, payroll — modular financial primitives, not products), a world model (operational + customer), an intelligence layer (that composes capabilities into solutions for specific customers at specific moments), and interfaces (Square, Cash App, Afterpay — delivery surfaces, not where value is created).

The organizational outcome: three roles. Individual contributors (ICs) who build. Directly Responsible Individuals (DRIs) who own cross-cutting problems with authority to pull resources. Player-coaches who combine building with developing people. No permanent middle management layer. Everything the hierarchy did, the system now coordinates.

The ideas that carry weight

The timeline is uncomfortable, but dismissing the ideas because of it would be lazy. There’s real substance here:

Middle managers are information routers. That’s not an insult — it’s a functional description. A significant portion of what middle management does is knowing what’s happening across their team and relaying that context up and down. In large organizations, that function consumes most of the workday. Alignment meetings, status sessions, priority negotiations. If a system can do that better and faster, there’s a legitimate case to examine it.

Remote companies generate machine-readable work artifacts. In an office, decisions happen in hallways. In a remote company, they live in Slack, Notion, Jira, commits, PRs. The material is already machine-readable. You don’t need to digitize it because it was born digital.

Transaction data is the most honest signal. NPS, satisfaction surveys, focus groups — all have well-documented biases. Real financial behavior is a more reliable proxy for what people actually need. Block’s access to both sides of the transaction is genuinely rare.

I wrote two weeks ago that AI is infrastructure, not a tool. Block’s essay takes that idea further: not just infrastructure, but the coordination layer that replaces hierarchical structure itself. It’s a big leap. But the direction is consistent.

What you won’t find in the essay

The timeline isn’t a detail. It’s the context that redefines everything above.

The 4,000

The essay talks about “normalizing down to three roles” and “no need for a permanent middle management layer” as if it were a design exercise. It doesn’t mention that 40% of the company was eliminated a month earlier. It doesn’t mention that teams went from eight engineers to one. It doesn’t mention the human cost.

Publishing a theory about why your company doesn’t need middle management — after laying off middle management — doesn’t invalidate the theory, but it changes how it reads. Darden (UVA) is already asking: was AI the strategy or the scapegoat? Josh Bersin questions whether it’s a bellwether or an anomaly.

The generalization problem

Block has advantages most companies don’t. It sees both sides of every transaction — buyer and seller. It operates fully remote, generating digital artifacts by default. It has millions of real financial signals, not surveys or NPS scores.

The essay reads as if the model transfers to any company. It doesn’t. Most B2B companies in Latin America — the ones we work with every day at IQ Source — aren’t remote. Their decisions don’t live in documents; they live in WhatsApp groups and in-person meetings. They don’t see both sides of the transaction. They don’t have the data density to feed a “world model” that replaces human coordination.

The transition cost

“No permanent middle management layer” — sounds clean in an essay. In practice: who coordinates while the AI learns? Who handles the cases the model hasn’t seen? Who makes the decisions that the essay itself acknowledges require human judgment — ethics, novel situations, high-stakes moments?

The essay is silent on the transition period. On the cost of getting it wrong. On what happens when the “intelligence layer” isn’t intelligent enough yet.

The precedents

Dorsey and Botha cite three previous attempts to move beyond hierarchy. Zappos adopted Holacracy — a system that eliminates manager titles and distributes authority into self-governing “circles” — and lost 30% of its workforce in a single year; Tony Hsieh himself admitted he regretted how he implemented it. Valve operated without formal hierarchy, but former employees revealed a “hidden layer of powerful management” — the flat structure existed on paper, not in practice. And Spotify stopped using its own squad model as it grew, quietly reverting to managers with real authority. Dorsey and Botha dismiss all three because “they lacked a technology capable of actually performing the coordination functions that hierarchy exists to provide.”

That’s the bet: AI is the missing technology. Maybe. But these three attempts didn’t stumble only because of missing tech. Organizational coordination is more complex than routing information. It includes trust, cultural context, power dynamics, misaligned incentives — all the things the essay delegates to “the edge” without explaining how to solve without structure.

Bayer tried something similar with “Dynamic Shared Ownership” — cutting management positions by ~50%. Amazon, under Andy Jassy, mandated a 15% increase in the IC-to-manager ratio. Neither has declared victory yet.

What this means for your company

Every time a company with unique advantages publishes a framework, a hundred companies without those advantages copy it. Google’s 20% time turned out to be more myth than reality — but that didn’t stop every startup from 2006 to 2012 from claiming they did it. Facebook’s “move fast and break things” worked at Facebook’s scale and cratered at everyone else’s. OKRs were designed for Intel and got shoved into companies where the problem wasn’t goal alignment but basic execution.

Block’s model will get the same treatment. Some CEO is going to read this on Sequoia’s site next week, call the leadership team, and say “we need to eliminate middle management and build a company world model.” Without transactional data. Without remote operations. Without digital artifacts an AI system can read.

Before you reorganize around AI coordination, there’s a question almost nobody asks: does your company generate material that an AI system can actually read?

If your decisions happen in unrecorded meetings, WhatsApp calls nobody documents, or hallway conversations that evaporate before reaching a document — you have nothing to feed a world model. The first step isn’t AI. It’s operational discipline.

Then comes data infrastructure. Not spreadsheets emailed monthly. Not Power BI reports nobody updates. Clean, accessible, real-time data. It’s the difference between speed and velocity we’ve already described — without foundations, AI is velocity without direction.

And the hardest part: an honest assessment of which coordination functions are information routing and which require human judgment. Block’s own essay acknowledges that people are needed for “intuition, opinionated direction, cultural context, trust dynamics, the feeling in a room.” That’s not the margin of coordination work — in many organizations, that’s the center.

At IQ Source, our AI assessment helps you map exactly this: which functions in your organization are real candidates for automated coordination, and which require human structure you shouldn’t eliminate because an essay on Sequoia sounds convincing.

The real test

The manifesto gets validated or debunked over the next 12 months — and the metrics won’t be subtle.

Revenue per employee is the obvious one. If 6,000 people can do what 10,000 used to do, the number should jump. But product velocity matters more: does the “intelligence layer” actually compose financial capabilities into customer solutions, or do Block’s products stall because nobody connects engineering to business anymore? Merchants and Cash App users didn’t read the Sequoia essay. They care if Square works and if loans arrive on time.

The signal I’m watching for is quieter. If Block starts rehiring coordinators under new titles — “program managers” rebranded as “DRI enablers,” project managers disguised as “intelligence architects” — we’ll know the intelligence layer didn’t replace what hierarchy did. It repackaged it.

Deloitte already documented that 60% of companies have AI but only 34% transform anything with it. Block is betting it can be in that 34% — and that the reason will be having eliminated hierarchy, not despite having eliminated it.

Oracle’s 30,000 and what nobody is measuring

Block isn’t alone. Today, as I write this, Oracle laid off up to 30,000 people — a 6 AM email, no prior warning, no conversation with their manager. The reason: free up $8-10 billion in cash flow to build AI data centers. Two companies, same day, same logic: AI replaces human functions, whether in coordination (Block) or operations (Oracle).

I asked Eytan Starkman — co-founder of IQ Source and someone who’s spent decades evaluating technology risk — what he thought about the thousands walking out of Block and Oracle. His answer was direct: “They’re going to create startups. They have deep knowledge of systems that very few people in the world understand.”

Not everyone will start a company. Not everyone wants to or is wired for it. But these people have something to contribute to the ecosystem — as principal engineers at companies that are building, as advisors, as technical co-founders. The institutional knowledge that Oracle and Block just put on the street doesn’t vanish. It redistributes.

I say this from experience: I’ve been an entrepreneur my entire life. I started at 15. And every wave of corporate cuts I’ve lived through — from IBM’s layoffs in the ’90s to the dot-com bust — ended up producing a generation of founders who wouldn’t have started if they hadn’t been pushed out of corporate structure. That’s no comfort to whoever got the 6 AM email. But the ecosystem has memory, and talent finds ways to create value when companies decide they don’t need it.

Block’s bet is legitimate. But publishing it one month after a 40% cut — on your most prominent VC’s website, the same day Oracle cuts 30,000 more — is a PR decision as much as a strategic statement. The difference between the two matters if you’re thinking about copying the model.