Block Has a Theory, Most Companies Just Have a Spreadsheet.

Engineering layoffs are not slowing down. The middle layer: engineering managers, staff-level coordinators, senior ICs whose value was measured in meetings attended, is being removed at a pace we have not seen since the dot-com correction. And two very different things are happening at the same time, treated as if they are the same thing.

The first is Block. They published a genuinely honest piece of organisational thinking. The argument is careful, historically grounded, and intellectually honest about the bet they are making. I disagree with parts of it. But I respect that they made their case in public and named what they are actually doing.

The second is every other company announcing "restructuring" in a PDF attached to a calendar invite.

These are not the same thing. Treating them as the same thing is how leaders avoid accountability and how individuals end up blindsided by what they are actually walking into.

Block Made an Argument Worth Taking Seriously

The core of Block's thesis is not "AI is coming for your job." It is older than that. Hierarchy exists because of span-of-control constraints. A manager can effectively oversee 5-8 people. That limit forces layers. Layers slow information. Layers create politics. Layers accumulate cost without creating value. This is not a new observation: it is what every flat-org experiment since Holacracy has been trying to solve.

What Block argues is that AI changes the constraint. If the reason for hierarchy is information routing, getting context, decisions, and alignment to the right people at the right time, and AI can maintain a continuously updated model of an entire business without a human intermediary, then the constraint disappears. The layers become optional.

This argument holds. Span of control is a real limit. Information routing is genuinely what middle layers spend most of their time on. If you replace the routing mechanism, you change the calculus.

Block traces the historical pattern honestly: Roman legions, Prussian military reforms, railroad management, Frederick Taylor, matrix organisations. Every shift followed the same logic. New coordination technology enables a new org shape. The companies that adopted the new shape first won. The ones that kept the old shape paid for it. AI is the next coordination technology. Block is placing a structural bet on that fact.

They also acknowledge they are early. "Parts of it will likely break before they work." That is an honest statement. Most companies making equivalent cuts are not making equivalent admissions.

Most Companies Just Have a Spreadsheet

Here is what Block actually did: they built the intelligence layer first. World models. AI systems that maintain operational and customer context continuously. Capability primitives, atomic financial building blocks that the intelligence layer can compose on demand. An information substrate designed to replace what middle managers were doing. Then they restructured the organisation to remove the layer that AI could now replace.

Here is what most companies are doing: removing the layer and calling it a transformation.

The sequence matters enormously. Block's model works, to whatever degree it works, because they built the replacement infrastructure before removing the human infrastructure. You can debate whether their specific design is sound, whether it will scale, whether their "economic graph" advantage is replicable at a company without millions of merchants and consumers on both sides of every transaction. Those are fair debates. But the sequence is correct: design the new system, validate it can carry the load, then decommission the old one.

What passes for transformation at most companies is different. A headcount target set by finance. Names identified by middle managers who are now also worried about their own jobs. Communication drafted by HR. Departure in 30 days. The intelligence layer does not exist. The AI is not routing coordination. The world model is a slide in a deck. The middle is just gone, and the work it was doing either falls up to senior leadership, falls down to ICs who were not hired to carry it, or disappears entirely.

That is not a theory. That is a spreadsheet with a press release.

What the Middle Layer Actually Does

To be precise about what is being lost, it helps to separate what middle managers and senior coordinators actually do. Not what the org chart says they do. What actually happens.

Some of it is pure coordination: information routing, status translation, meeting facilitation, reporting up and down the chain. Block is right that AI can replace this. If your primary value was sitting in the room so someone else did not have to, that value is genuinely diminished. Not because you are not good at it, but because the cost of automating it has dropped below the cost of your salary. This is not a moral judgement. It is arithmetic.

The coordination category is larger than most people want to admit. Sprint ceremonies where the manager is the primary coordinator. Status decks assembled from team updates. Weekly syncs where the agenda is "what did everyone do." Escalation chains for decisions that should never have needed escalating. A meaningful portion of what middle managers do in most organisations is work that should not require a human at the cost of a senior engineering salary. Block is right about this.

But there is a second category that is not on Block's list and is not being replaced by any intelligence layer I have seen deployed in production.

The manager who fights for your promotion when you are not in the room. Not advocates in theory. Actually goes into the calibration session, makes the case, pushes back when the committee is wrong, and comes back to tell you what happened and why. That function does not exist in Block's model. The DRI structure and the player-coach model assume capable adults who can self-advocate in a flat power structure. Most people cannot do this without support, especially early in their careers.

The leader who translates "we need to move faster" from a board deck into "here is what you should actually build next Monday." Strategy lands in organisations as abstraction. The middle layer converts abstraction into concrete direction. Without it, ICs receive vision without action, which produces anxiety, not output.

The EM who sits across from someone at 4pm on a Thursday and notices that something is off. Not because a world model flagged an anomaly in commit frequency. Because they have been paying attention to this person for eight months and they know what off looks like. AI coordinates. It does not see people.

The manager who knows that two senior engineers have been in a cold war for three months over an architectural decision that went badly, and does the slow, unglamorous work of resolving it before it spreads. Conflict resolution is not a process. It is a relationship skill. It requires trust on both sides, built over time, specific to the individuals involved. There is no model for this.

The companies cutting the middle layer are, in many cases, cutting both categories without distinguishing between them. The coordination work disappears into tooling that is not quite ready. The human work disappears into nothing.

Three Honest Truths About What Comes Next

For individuals in the middle layer: if your value is "I go to meetings so others do not have to," that role is gone and it is not coming back, regardless of whether your company has an intelligence layer or just a spreadsheet. The coordination work that defined middle management in 2015 is being automated. The question is what you do about it.

The skills that survive are not the coordination ones. They are the judgment ones. Can you walk into a system you have never seen and know within two hours where the real problems are? Can you have the conversation that everyone else is avoiding? Can you translate ambiguous direction from leadership into a decision that an IC can act on Monday morning? These are not easy skills. They are the ones that are getting harder to hire for as everything else gets automated, which means their value is increasing, not decreasing.

If you are in the middle layer and your instinct is to prove your value by being in more meetings, sending more status updates, writing more documentation of decisions made elsewhere: you are optimising in the wrong direction. The work is to get visible on the judgment side before the next cycle. I wrote more on what this actually looks like in what engineering managers do now.

For leaders removing layers: if you are cutting middle management without building what replaces it, you are not running a Block-style experiment. You are cutting. You are allowed to make that call. Boards have targets, businesses have constraints, and not every company has the runway or the data assets to build what Block is building. Hard decisions are part of the job.

But own what you are doing. Do not announce it as an AI transformation toward an intelligence-driven org when the intelligence layer does not exist. Do not use Block's conclusion to justify a decision you made on a spreadsheet. The people watching, including the people you kept, can see the gap between the narrative and the reality. And they are updating their model of what leadership means at this company based on what they observe.

On what leadership owes before it cuts: before you eliminate someone's role, you owe them honest reasoning. Not a restructuring memo that could have been generated in thirty seconds. A real explanation: this work is being automated, or this structure is changing, or we made a hiring decision we cannot sustain. Specific and true.

You owe them a genuine assessment of which parts of their work are being automated and which parts are not, because that distinction determines where they go next. The manager who spent three years doing coordination work needs to hear that clearly so they can make the transition before the market prices them as already redundant. The one who spent those three years on the judgment work needs to hear something different.

And you owe them the acknowledgment that they were not let go because they failed. They were let go because a decision was made about the shape of the organisation, and they were in the part that got cut. That is a different thing from failure. Most companies treat them as the same. That confusion does real damage to the people on the receiving end of it.

The Difference Is Not AI, It Is Accountability

Block might be right. The intelligence layer model may be where most engineering organisations are headed. The historical pattern supports it: every significant organisational shift in the last two hundred years has followed new coordination technology. AI is the next one. The companies that restructure around it first, and do it correctly, will have a structural advantage that compounds.

But we are in the early middle of that transition. Block acknowledged it. And the companies operating in that middle, which is almost everyone, are making choices right now about how they carry their people through it.

The ones doing it with honesty and design will come out with organisations that actually function and teams that trust them. The ones doing it with a spreadsheet and a press release will have the same operational problems they had before, with fewer people to absorb them, and a culture that has drawn its own conclusions about what leadership values when things get hard.

Block has a theory. Most companies just have a spreadsheet. The difference is not AI. It is accountability.

---

If this framing resonates, Cortex is the methodology I built to implement it. One workflow at a time.

---

I help engineering teams close the gap between "we use AI tools" and "AI actually changed how we deliver." Book a 20-minute call and I'll tell you where the leverage is.