Daniel sent us this one — he's been thinking about the bottleneck boss we talked about, the person who inserts themselves into every approval like the business would collapse without their signature on a Post-it note. And his question is: flat hierarchy. The thing everyone says they want. Is it actually achievable, or is it management mythology? Because here's the tension — every organization has to make decisions. If you remove the formal approval chain, how do decisions get made without everyone lighting their own match and burning the whole thing down?
The box of matches metaphor. I love that. Each team optimizes for its own little corner, nobody sees the full picture, and suddenly you've got three different product roadmaps and a marketing campaign for a feature engineering killed last month.
That's the failure mode. And I think what Daniel's really asking is — can you have decentralized leadership without it becoming a bonfire? Or does flattening the hierarchy just replace one bottleneck with a dozen smaller, more confusing bottlenecks?
The dream is so seductive though. No bosses, no permission slips, just competent adults making good decisions together. And then you try it and discover that "no bosses" actually means "every meeting is now a governance meeting and nothing ships.
Like adopting a feral cat.
You think you're getting independence and self-sufficiency, and what you actually get is something that hisses at you from under the couch and knocks over your coffee at three in the morning.
The timing on this question is interesting, because remote and async work have made the coordination problem more visible. When everyone's in an office, informal hierarchy sort of handles itself — you see who people go to for answers. When you're distributed, that invisible structure evaporates, and what's left is either explicit coordination or chaos.
The pandemic forced a lot of organizations into distributed work, and suddenly the people who held everything together through hallway conversations and desk drive-bys — that mechanism was gone. And you can't just declare "we're flat now" into a Slack channel and expect it to work.
The question isn't really "is flat hierarchy possible." The question is what specific coordination mechanisms make it possible, and why do most organizations skip them?
Because they mistake the absence of hierarchy for the absence of structure. And those are not the same thing.
Let's actually define it, because "flat hierarchy" gets thrown around like it's one thing and it's absolutely not. At its core, it's a structure with few or no hierarchical layers between staff and leadership — decision-making authority is distributed, not concentrated in a chain of command. But here's what people miss: distributed doesn't mean absent. It means relocated.
It's not "no managers." It's "managers, but not in the way you're used to thinking about them.
In a traditional hierarchy, authority flows down through levels — you have a boss, they have a boss, that person has a boss. In a genuinely flat structure, authority sits in roles or teams, not in reporting lines. You might still have someone responsible for compensation decisions or strategic direction, but they're not signing off on your choice of database or your marketing copy.
Which brings us to the paradox Daniel's pointing at. If nobody has formal approval authority over most decisions, how do decisions actually get made? Because "the team decides" sounds great until the team disagrees.
That's where the box of matches comes in. Without a formal approval chain, every team or individual becomes their own decision-making island. They optimize for what they can see. Engineering picks a tech stack that makes sense for engineering. Marketing builds a campaign around a timeline that makes sense for marketing. Nobody's talking, because there's no structural reason to talk — the structure that used to force alignment through a shared boss is gone.
The arsonist in Daniel's metaphor isn't a person. It's the absence of shared context. Each match is a perfectly reasonable decision in isolation. The fire starts when they all burn at once.
And this is why most flat hierarchy attempts fail in the first year or two — not because people are bad at their jobs, but because the organization removed the coordination mechanism without replacing it with anything. They confused "no approval chain" with "no coordination needed.
The thesis here — and I think this is where we're heading — is that flat hierarchy is achievable, but only when you pair it with explicit coordination mechanisms. And most organizations skip that part because it feels like adding bureaucracy back in.
Which is the great irony. You flatten the hierarchy to escape bureaucracy, and if you don't build lightweight coordination structures, you get something worse than bureaucracy — you get chaos. The successful flat organizations aren't the ones with the fewest rules. They're the ones with the clearest rules about who decides what and how decisions get communicated.
The absence of hierarchy is not the absence of structure. They're different things, and mistaking one for the other is what turns the dream into the bonfire.
Let's look at what actually breaks. There are three failure pattern that show up almost every time someone tries to flatten their org chart without building the coordination layer first.
Three ways to burn down the house.
The first one is the power vacuum. You remove the approval chain, and suddenly nobody knows who's allowed to decide what. So decisions just... Nobody wants to overstep, nobody wants to be the person who made the wrong call without permission, and the result is paralysis. Things that used to take a day now take three weeks because everyone's waiting for someone else to pull the trigger.
Which is worse than having a bottleneck boss. At least the bottleneck boss eventually signs something. The vacuum just sits there, silently killing momentum.
The second failure pattern is the one that makes me the most cynical — informal hierarchy. You remove the formal structure, and what rushes in to fill the gap is the loudest person in the room, or the most senior, or the one who's been there longest. They become the de facto approver, except now there's no accountability because technically nobody reports to anyone.
You took down the org chart but the power dynamics are exactly the same, just unspoken. And unspoken power is harder to challenge than explicit authority.
That's the gaslighting piece you mentioned before. The company says "we're all equal here" while everyone knows you don't ship anything without Dave nodding. But Dave isn't your manager. Dave is just... And good luck explaining that to a new hire.
The third one is the box of matches — fragmentation. Teams making decisions in isolation.
And this is the one that's structurally inevitable without coordination mechanisms, because stakeholders naturally optimize for their own sub-goals. Engineering wants stability and technical debt reduction. Sales wants features that close deals this quarter. Marketing wants a coherent story they can tell. None of these are wrong. But without a shared context layer, each team makes the right decision for their own domain and the organization ends up with three incompatible versions of what it's trying to do.
Each match is reasonable in isolation. The fire is what happens when they all burn simultaneously.
The arsonist, like we said, isn't a person — it's the absence of shared context. Nobody is trying to create conflict. They just can't see what the other matches are doing.
How do the organizations that actually make this work solve the fragmentation problem? Because they're not just hoping people talk to each other more.
They use something called consent-based decision-making, and this is where the distinction from consensus really matters. Most people hear "flat hierarchy" and assume everyone has to agree on everything. That's consensus. It's slow, it's exhausting, and it doesn't scale past about eight people.
Death by meeting.
Consent is different. The standard isn't "do you agree with this decision." The standard is "do you have a reasoned objection." Can you articulate why this decision would cause harm or move us backward? If not, the decision proceeds. You don't have to like it. You just can't block it without a substantive reason.
It's not "everyone's happy." It's "nobody can prove this is a bad idea.
And that's the mechanism Holacracy uses. The Holacracy constitution — and yes, it's an actual written constitution — defines roles, not people. Each role has explicit authorities and accountabilities. You don't ask your manager for permission because there is no manager. You consult the constitution, which tells you what your role can decide and what requires input from other roles.
Which is the opposite of "no structure." It's a different structure entirely.
It's more structure, honestly. Just structured around work rather than around people. And that's the part most organizations miss when they get excited about flat hierarchy. They think they're removing bureaucracy, but what Holacracy actually does is replace people-bureaucracy with process-bureaucracy.
Zappos is the poster child for how hard that transition is.
Tony Hsieh eliminates all managers and goes all-in on Holacracy. And it was chaos. Twenty-three percent of employees took a buyout rather than participate. That's nearly a quarter of the company saying "I would rather leave than work in this system." And Zappos eventually backed off to a hybrid model.
Which tells you something about the discipline required. It's not that flat hierarchy is impossible. It's that most people don't actually want the version of it that works, because it demands more rigor than having a boss.
Then there's Valve. The famous employee handbook that says "nobody tells you what to work on." Desks have wheels so you can physically relocate to whatever project you choose. No formal hierarchy at all.
Sounds utopian until you look at where the power actually sits.
The Steam platform. If you controlled Steam's infrastructure or its revenue levers, you had informal authority that nobody else could match. The handbook says everyone's equal, but the person who can tank the company's primary revenue stream by leaving is more equal than others. That's not a failure of the people — it's a structural reality that the flat model couldn't account for.
The informal hierarchy just reasserted itself around the economics. The org chart was flat, but the revenue chart wasn't.
That's the lesson from both Zappos and Valve. You can't just declare hierarchy away. You have to replace it with something explicit — decision rights, coordination mechanisms, conflict resolution processes. Otherwise the vacuum fills itself, and what fills it is usually worse than what you removed.
If consent-based decision-making is the mechanism, what does it actually look like in practice? Let's look at two organizations that tried this — one that nailed it, one that didn't. And the difference is instructive.
Buurtzorg and Medium.
Buurtzorg first, because it's the success story that doesn't get enough attention outside of organizational design circles. This is a Dutch home-care organization. Fifteen thousand nurses. Self-managing teams of twelve. No middle managers. And they consistently rank as one of the best employers in the Netherlands.
Fifteen thousand people with no middle managers. That already sounds impossible.
It shouldn't work on paper. But they built three coordination pillars that make it possible. First, explicit decision rights — every team of twelve knows exactly what they can decide independently and what requires broader input. Second, transparent information sharing — they have a national IT system that tracks patient outcomes and team performance, so every team can see how they're doing relative to everyone else. No manager needed to tell them.
The transparency replaces the supervisor. The data does the managing.
And the third pillar — conflict resolution. They have a lightweight escalation process for disputes between teams, but it's designed to resolve issues without creating a permanent hierarchy. You escalate, you resolve, the structure returns to flat. It's like a circuit breaker, not a new layer of management.
What's the actual decision-making mechanism on the ground? When a team of twelve nurses needs to decide something, how do they avoid the loudest-voice problem?
They use the advice process. Any nurse can make a decision, but before they do, they have to seek advice from anyone affected by that decision and from anyone with relevant expertise. They don't need permission. They don't need consensus. They need to have consulted.
The constraint is procedural, not hierarchical. You can decide, but only after you've done your homework.
That's what prevents the box of matches. The advice process forces horizontal communication. You can't optimize in isolation because the process requires you to talk to the people your decision will impact. It preserves speed — you still get to decide — but it eliminates the blind spot.
Which is the opposite of what most people imagine. They think flat means "nobody tells me what to do," when successful flat actually means "I have to talk to more people before I act.
More intentional communication, not less. And then there's Medium.
The cautionary tale.
Medium adopted Holacracy in twenty sixteen with genuine enthusiasm. They wanted the whole thing — the constitution, the defined roles, the governance meetings. And two years later, in twenty eighteen, they abandoned it. The official reason was that the structure had become too rigid and bureaucratic. The very thing they were trying to escape.
Structure became its own bottleneck.
This is where the contrast with Buurtzorg is so sharp. Buurtzorg kept it simple — three coordination pillars, lightweight processes, heavy emphasis on shared norms and transparency. Medium went all-in on the formal Holacracy apparatus and discovered that replacing people-bureaucracy with process-bureaucracy isn't necessarily an improvement. Their governance meetings became the new approval chain.
The difference isn't "one had structure and one didn't." It's "one had the right amount of structure and the other overcorrected.
Buurtzorg's coordination mechanisms are minimal but non-negotiable. Medium's became maximal and suffocating. And that's the design challenge — how much coordination is enough to prevent fragmentation without becoming the new bottleneck?
Which brings up the scaling question. Buurtzorg makes fifteen thousand people work with no middle managers. But they're organized into teams of twelve. That's deliberate.
Dunbar's number. Around a hundred and fifty people is the cognitive limit for maintaining stable social relationships. Beyond that, you need additional coordination layers. Buurtzorg solves this with federated teams — each team of twelve is flat internally, but teams connect through representatives who coordinate across the organization without becoming managers.
Circles within circles.
Each circle is flat. The connections between circles are lightweight and role-based, not authority-based. It scales because the flatness is modular — you don't try to keep fifteen thousand people in one flat structure. You create a network of flat structures.
If we're pulling practical lessons out of this, the first one is painfully simple and almost nobody does it. Before you flatten anything, write down who can decide what. Every key domain — product direction, hiring, tech stack, marketing spend — needs an explicit decision right attached to it. Who decides, who must be consulted, who can veto.
The constitution before the revolution.
And it doesn't need to be a hundred-page document. It can be a shared Notion page. But if you don't define decision rights in writing, you get the power vacuum. People freeze because they don't know if they're allowed to act.
The second one is the advice process we saw in Buurtzorg. Anyone can decide, but they have to seek input from affected parties and experts first. It's not permission. It's not consensus. It's a consultation requirement that forces horizontal communication.
This is the minimum viable coordination. It prevents the box of matches without adding bureaucracy, because you're not creating an approval step — you're creating an information step. The decider still decides. They just can't do it blind.
Which brings us to the third one, and this is the one that makes people groan. Flat hierarchy requires more communication, not less. Async updates, shared decision logs, regular retrospectives — these aren't nice-to-haves. They're the infrastructure that replaces the hallway conversations and manager check-ins you removed.
If you flatten the org chart and don't invest in communication infrastructure, you haven't eliminated coordination cost. You've just made it invisible until something breaks. The Buurtzorg IT system that tracks team performance — that's not overhead. That's what makes flatness possible at scale.
For someone listening who's leading a team right now — audit your decision-making. Find one recurring conflict that keeps flaring up, and trace it back. I'll bet you find unclear authority at the root. Someone thought they could decide. Someone else thought they should have been consulted. Neither was wrong, because nobody wrote it down.
Pick that one conflict and apply the advice process to it this week. Define who decides, require consultation, document the outcome. See if the conflict evaporates. That's the smallest possible experiment in functional flat hierarchy, and it doesn't require reorganizing anything.
Flat hierarchy is real, but it's not a set-it-and-forget-it solution. And the next frontier might change the game entirely. AI tools are getting better at the coordination work that currently requires human discipline — summarizing decisions across teams, flagging when two groups are about to step on each other, surfacing conflicts before they become fires.
The AI as the shared context layer. Instead of hoping everyone reads the decision log, the decision log reads itself and taps you on the shoulder.
And that could lower the communication tax that makes flat hierarchy exhausting. But I'm not sure it solves the judgment problem. The AI can tell you two teams are making incompatible decisions. It can't tell you which one should yield. That still requires human trade-offs.
The coordination becomes easier, but the hard part — actually resolving the conflict — stays human.
Which brings us to the most radical experiment happening right now. Decentralized autonomous organizations on blockchain. No managers at all. Just smart contracts and token voting. It's flat hierarchy taken to its logical extreme — the rules are code, decisions are votes, and there is no "Dave" to nod at.
How's that going?
About as well as you'd expect. Low voter participation is endemic — most token holders don't show up. And when they do, the whales dominate. A handful of large token holders effectively control decisions, which is just informal hierarchy with extra steps and worse UX.
The same failure pattern, just on-chain. The vacuum, the informal power, the fragmentation.
The technology changes. The human dynamics don't. Which is really the final thought here. Flat hierarchy isn't a structure you declare — it's a discipline you practice. The bottleneck boss is just one person slowing things down. A broken flat structure makes everyone the bottleneck, and nobody knows it until the fire starts.
The question to sit with — as AI gets better at coordination and DAOs keep experimenting with governance without governors — does the technology eventually make the discipline easier, or does it just give us new ways to avoid the hard part?
Now: Hilbert's daily fun fact.
Hilbert: In the eighteen-tens, dyers in Labrador produced a distinctive shade of blue from boiled spruce bark and copperas, which appeared deep indigo on wool but shifted to a metallic teal when the fabric caught direct light — an optical property caused by the iron mordant's interaction with tannins, not the dye itself.
...right.
This has been My Weird Prompts. If you enjoyed this, leave us a review wherever you listen — it helps. We're back next week.
Try not to set anything on fire in the meantime.
