Join us this Saturday, March 7th, for a meetup on Moral Mazes, Zvi's Immoral Mazes critique, and what we in DC should do about it.

We'll meet starting at 6pm at Workshop House in Logan Circle, 1717 15th St NW. We'll place a group takeout order for dinner about 6:15 for those who'd like to join, then have a structured discussion about resisting maze dynamics starting about 7:30. See comments section for dinner discussion.

Our hook is that Jenn, of Balsa Research and Kitchener-Waterloo Rationality, will be in town. Her group has recently had two Partitioned Book Club meetups on Moral Mazes (I and II), which generated great conversations about how the dynamic applies to Rat and EA communities. Our discussion will focus on how Zvi's critique applies to us, people embedded in or at least familiar with DC's institutions. What can we do about maze dynamics, where are we falling prey to them, and where might we be able to bend our institutions towards truth. Current events, such as the Pentagon-Anthropic dispute and our ongoing Iran conflict, will certainly come up, but we'll moderate the discussion so that they don't fully take over. 

In case you don't have time this week to read both a full book and a collection of Zvi posts of equivalent word-count, our friendly neighborhood robots have written a summary tailored for this discussion. I'm pasting it below.

Immoral Mazes: A Summary for Discussion

This summary was written by Claude (Anthropic) based on Zvi Mowshowitz's Immoral Mazes sequence, Robert Jackall's book Moral Mazes, and Zvi's sequence summary. It is intended to give meetup participants enough context for informed discussion without requiring tens of thousands of words of pre-reading. It is not a substitute for the original material, and participants who have the time are strongly encouraged to read the sequence summary and Quotes from Moral Mazes at minimum.

Part I: What Robert Jackall Found

In the 1980s, sociologist Robert Jackall conducted extensive fieldwork inside several large American corporations, interviewing managers at every level about how they actually made decisions, advanced their careers, and understood their own moral lives. The resulting book, Moral Mazes: The World of Corporate Managers, is one of those works that people instinctively flinch away from, not because it is badly written, but because what it describes is genuinely difficult to look at clearly. As Zvi puts it: "one has the very strong instinct to avoid understanding it, to read without seeing, to hear without listening."

What Jackall found was a world in which the official purposes of the organization — making products, serving customers, generating profit — had become secondary to an all-consuming internal political competition among managers. The formal hierarchy created a system where what mattered was not what you accomplished, but how your accomplishments were perceived by the people above you.

The primacy of the boss. When managers described their work to Jackall, they almost always began not with what they did, but with who they reported to: "I work for Bill James" or "I'm in Joe Bell's group." This was not a quirk of speech. It reflected the fundamental reality that a manager's fate depended almost entirely on their relationship with their superior. As one former vice-president told Jackall: "What is right in the corporation is not what is right in a man's home or in his church. What is right in the corporation is what the guy above you wants from you. That's what morality is in the corporation."

This created a system with precise and inviolable rules. You never go around your boss. You tell your boss what he wants to hear, even when he claims he wants dissenting views. If your boss wants something dropped, you drop it. Your job is not to report something your boss does not want reported, but to cover it up. These were not cynical observations from disgruntled employees — they were stated matter-of-factly as the basic operating principles of organizational life.

Credit flows up, blame flows down. The hierarchy operated as a machine for separating action from consequence. Details were pushed down and credit was pulled up. One could not claim credit even when it was earned — credit had to be given as a gift, and accepting it reaffirmed one's fealty. Meanwhile, blame operated on a different system entirely. Being a "fall guy" — taking the rap for others' decisions — was probably the most common form of accountability in these organizations.

"Feeling comfortable." The most important evaluation criterion in this world was not competence but comfort. Managers constantly assessed whether they could "feel comfortable" with someone — whether the person was trustworthy, reliable, predictable, and above all, not likely to cause problems. This went far beyond mere likeability. It meant demonstrating that you understood the unwritten rules, that you could be trusted to look away when necessary, that you would never force your boss to act as boss. As one manager put it: "the biggest thing that I have going for me [is] that people feel that I can be trusted. I can keep my mouth shut."

Being brilliant was actively dangerous, because it signaled that you might publicly assert your intelligence and threaten others. Being principled was dangerous, because strong convictions of any sort suggested you might not be fully flexible. Being right was dangerous, because "that's when we really get mad — when the other guy is right." The ideal was interchangeability: someone who could be slotted into any role, who would align with whatever ideology currently held sway, and who would never make anyone uncomfortable by caring too much about anything other than the game itself.

Implicitness as a survival mechanism. One of the most striking features of this world was its reliance on indirect communication. Bosses gave vague objectives rather than explicit directions, because giving explicit directions meant sharing responsibility for failure. Important messages were communicated through hints, coded language, and deniable suggestions. The corporation was, in Jackall's words, "a place where people are not held to what they say because it is generally understood that their word is always provisional." Managers who could not read between the lines — who were "maze-dense" rather than "maze-bright" — were simply eliminated.

Short-termism as structural necessity. Managers thought in the short run because they were evaluated on short-term results, and because they changed jobs frequently enough to outrun the consequences of their decisions. "Our horizon is today's lunch," said one manager. "Outrunning mistakes" was described as the real meaning of being on the fast track. Managers routinely "milked" their plants — deferring maintenance, cutting inventory to the bone, reducing quality — to boost numbers during their tenure, leaving the costs for their successors.

The hollowness of success. Perhaps most devastating was Jackall's observation about what happened to those who actually won the game. The few who rose through the hierarchy and became CEO still lost. Their success did not bring happiness or meaning. The fruits were hollow. Yet even for those who realized this, stopping felt worse than continuing. The self-modifications required to succeed — abandoning outside interests, choosing friends for strategic value, learning to speak out of both sides of one's mouth without missing a step — had become the person. There was nothing left to go back to.

Part II: Zvi's Framework — Why Mazes Form, and Why the World Isn't Worse

The Immoral Mazes sequence takes Jackall's empirical observations and asks two questions: Why does this happen? and Why hasn't it consumed everything?

Moloch hasn't won. Zvi's starting point is a response to Scott Alexander's "Meditations on Moloch," which argued that competitive selection pressure, given enough time, inevitably destroys all values and all slack. Zvi's crucial intervention is to observe that this clearly hasn't happened yet. Most of the world still has slack — the absence of binding constraints on behavior. Most markets feature highly imperfect competition. Most people still have genuine relationships, real hobbies, and authentic values. Most organizations, despite their flaws, still accomplish real things. The question is not "why is everything terrible?" but rather "why are some things terrible while most things are merely mediocre?"

The mechanism: super-perfect competition. The answer is that the Molochian nightmare arises specifically in conditions of perfect competition — where participants are interchangeable, the playing field is level, and the only variable is effort. Under perfect competition, all constraints bind and all slack is eliminated. Worse, Zvi identifies what he calls super-perfect competition: when exit is costly (and all exit has some cost), competitors don't just break even — they actively lose, eating their seed corn in a race that leaves everyone worse off.

Middle management in large corporations is one of the purest examples of super-perfect competition in modern life. Too many managers seek too few promotions. Past a certain level, everyone is assumed to have equivalent competence at the actual work. The differentiating factor becomes entirely political — who can best navigate the social dynamics, who makes others feel most comfortable, who is best at claiming credit and avoiding blame. The protections that most people have against this kind of competition — differentiated skills, outside options, genuine metrics of performance — have been stripped away.

Why mazes spread. Once maze dynamics take hold, they are self-reinforcing. Managers who have internalized the political game instinctively reward and ally with others who play the same way, and punish those who don't. Caring about anything other than the competition — morality, family, technical excellence, the actual mission of the organization — makes you a bad ally, because it suggests you might be distracted, and distracted people lose. People who lose are dangerous to be associated with. So the system selects ever more strongly for people whose entire identity is the game.

This explains why things in any given organization tend to get worse over time rather than better. The maze-people accumulate power and reshape the institution in their image, driving out anyone with different values.

Why things get better anyway. The good news — and this is central to Zvi's argument — is that maze-ridden organizations eventually collapse. They over-optimize, eat their seed corn, and lose to competitors who still have some connection to reality. The cycle of decay and renewal is what keeps the world from being consumed. Civilizations rise and fall. Corporations are born, grow, ossify, and die. The replacement of the old by the new is not a tragedy but a necessary part of how the system self-corrects.

The bad news is that our society has increasingly chosen to prevent this renewal. "Too Big to Fail" is not just a banking policy — it reflects a broader instinct to preserve existing institutions at all costs. We protect organizations from disruption, especially in times of crisis. Our laws and regulations favor large incumbents over new entrants. Our educational system trains people for the maze. And the people running the mazes have gotten powerful enough to cause a vicious cycle, using their influence to structure society in ways that favor more mazes.

This is why, Zvi argues, people increasingly have the instinct that Moloch wins everywhere — not because it's true, but because we are systematically disarming the forces that keep Moloch in check.

Part III: What Makes an Organization More or Less Mazey?

Not all large organizations are equally maze-like, and Zvi spends significant effort identifying the factors that push an organization toward or away from mazedom. This matters because it determines where to focus resistance — and where resistance is probably futile.

Factors that increase mazedom. The single strongest predictor is the number of management layers. The more levels of hierarchy between the people doing object-level work and the people making strategic decisions, the more room there is for the political game to dominate. Closely related: when the number of people competing for each promotion is high and the playing field is perceived as level, competition shifts from substance to politics. When outcomes are difficult to attribute to individuals — when it's genuinely unclear whether a business unit succeeded because of its manager or despite them — evaluation becomes subjective, and subjective evaluation is the native habitat of maze dynamics. When exit is costly (because of golden handcuffs, specialized skills that don't transfer, or dependence on employer-provided healthcare and status), people tolerate worse conditions and the competitive pressure intensifies. And when an organization is insulated from external consequences — through regulation, monopoly position, government bailouts, or sheer size — the feedback loop that would normally punish dysfunction is severed.

Factors that decrease mazedom. Organizations that maintain real contact with reality — where products must actually work, where customers can leave, where failure is visible and attributable — are structurally more resistant. Fewer management layers help enormously. Genuine skill differentiation (where people are not assumed interchangeable) protects against the leveling dynamic that drives political competition. Skin in the game, where decision-makers bear personal consequences for their decisions, changes behavior dramatically. And perhaps most importantly, organizations whose leadership is aware of maze dynamics and actively, continuously pays costs to fight them — firing politically skilled but mission-indifferent people, tolerating inefficiency to preserve slack, refusing to add management layers even when it would be convenient — can hold the line for a time.

But Zvi is not optimistic about the long run. His view is that mazedom is the default trajectory for any organization above Dunbar's number (roughly 150 people). Below that scale, personal relationships and direct observation can substitute for formal hierarchy. Above it, you need structure, and structure creates the conditions for political competition. Organizations can fight this — and should — but they are fighting gravity. The question is not whether a large organization will eventually succumb, but how long it can hold out, and whether it can accomplish its mission before it does. This is why organizational death and replacement is so important to the overall health of the system: it is the mechanism by which the inevitable decay of any particular organization is prevented from becoming the decay of everything.

Part IV: What To Do About It

The final section of the sequence is prescriptive, and it operates at several scales: personal, organizational, and systemic. These are presented as Zvi's arguments — reasonable people may disagree about the feasibility or desirability of specific recommendations, and that disagreement is part of what makes this productive discussion material.

Personal: Recognize and escape. Zvi's most urgent recommendation is individual. If you find yourself trapped in a maze, get out. Quit and do something else. He argues that people systematically overestimate the costs of leaving and underestimate the costs of staying. The costs of staying are not just financial — they include the progressive erosion of your values, your relationships, your capacity for independent thought, and your connection to anything you actually care about. The costs of leaving, meanwhile, are usually lower than they appear from the inside, because the maze has trained you to see the outside world as more dangerous than it is.

This is easier said than done, but Zvi argues it is easier done than those in a position to do it think. "You're already doing something hard, and you have the skills to do a different hard thing that won't make you miserable, even if the pay starts out lower." If you genuinely cannot leave, the next best option is to stop caring about advancement and instead take bold risks in pursuit of things that actually matter.

Identifying mazes. Before you can escape or resist, you need to recognize what you're in. The diagnostic criteria are covered in Part III above — hierarchy depth, skin in the game, skill differentiation, slack, and above all, paying attention to how things are actually done rather than what the org chart says.

Organizational: Design against mazes. For anyone building or running an organization, Zvi's advice centers on several principles. First, minimize hierarchy — keep the organization as flat as possible, with as few management layers as you can manage. Second, maintain skin in the game — ensure that the people making decisions bear real consequences for those decisions. Third, maintain "soul in the game" — hire and promote people who genuinely care about the mission, not just about advancement. Fourth, be extremely vigilant about personnel. Anyone who is making the organization more maze-like needs to go, regardless of how painful that is. This is the hardest recommendation, because maze-people are often charming, politically skilled, and well-connected — which is precisely what makes them dangerous.

Do fewer things and do them better. Minimize interaction with other mazes, because maze dynamics are contagious. Be smaller than you think you need to be. Every additional layer of hierarchy, every additional degree of separation between decisions and consequences, is a step toward mazedom.

Systemic: Policy and culture. At the broadest level, Zvi identifies several policy levers that could reduce the prevalence and power of mazes: regulatory reform that reduces the advantages of scale; healthcare reform that reduces the hold employers have over employees (making exit less costly); tort reform; ending corporate welfare; and in some cases, forcibly breaking up organizations that have become too large. He also argues for cultural change: educating people about what mazes are, lowering the social status of maze-participation, and reducing demand for the "illusion of security" that mazes are particularly good at providing.

Perhaps his most ambitious proposal is to "create a full alternative stack" — a comprehensive set of institutions that would allow people to meet their needs (employment, healthcare, housing, social connection, meaning) without having to participate in maze-like organizations at all. This would require someone with several hundred million dollars and an unusual amount of vision, but Zvi argues it would be the single highest-leverage intervention available.

Why this matters for us specifically. One thread running through the sequence is a warning directed at the rationalist and EA communities. As these communities grow, professionalize, and attract more resources, they face exactly the pressures that produce mazes. Funding concentrates in a few large organizations. Status hierarchies form. People begin optimizing for advancement within the community rather than for the community's stated goals. The same competitive dynamics that consumed Jackall's corporations can consume a movement just as thoroughly.

This is not inevitable. But avoiding it requires the same vigilance Zvi recommends for any organization: keep things small, keep them flat, maintain real accountability, and be willing to bear costs — in money, in status, in social comfort — to preserve the things that actually matter.