Every compensation model I inherited before building my own operated on the same unexamined assumption, that a person’s value to the organization increases monotonically with time. Hire someone at a competitive salary, adjust upward annually (3%, the industry default, a number that feels generous to employees and negligible to finance) and let the accumulation compound. Tenure == value. Loyalty == worth. Staying is its own justification.

I never questioned this. Not through investment banking, where compensation was a weapon. Not through my first startup, where I was too busy surviving to think about compensation architecture at all. Not even in my early Shutterstock years, when I was building the career framework that would eventually force the question.

The question arrived not as theory but as arithmetic. I was modeling band ceilings for 350+ engineers across multiple continents, and the math kept producing the same result. No matter how I structured the bands, the 3% annual adjustment—the raise everyone treated as a reward—was also a countdown. Every year it moved an engineer closer to a ceiling that the organization had already defined. And nobody was tracking the convergence. Not the engineers. Not the managers. Not finance.

The engineers didn’t know there was a ceiling until they hit it. The managers didn’t know their reports were approaching it until a compensation review surfaced a number that couldn’t go higher. Finance didn’t know the cost pressure was building until it arrived as a budget crisis. Everyone was operating inside a system with hard structural constraints, and nobody could see them.

I built a model to make those constraints visible. Not to punish tenure, not to pressure anyone out of a role, but to give every participant in the system (the engineer, the manager, the finance team) a shared, transparent understanding of the physics governing compensation at each level. The band ceiling is a budgetary boundary. It exists whether or not anyone models it. The question is whether the organization surfaces that boundary early enough for everyone to plan around it, or whether it discovers the boundary through a crisis that nobody anticipated because nobody was looking.

The Convergence Problem

The standard compensation lifecycle works like this. An engineer is hired at a competitive salary within the defined band for their level. Each year, they receive a cost-of-living or merit adjustment, typically between 2% and 5%, with 3% being the gravitational mean. These adjustments compound. Over five years, a 3% annual increase on a $170,000 base produces approximately $197,000. Over ten years, roughly $228,000.

The band, meanwhile, has a floor and a ceiling. It gets recalibrated against market data periodically (annually if the organization is disciplined, less frequently if it isn’t) but the ceiling for any given level represents the organization’s maximum valuation of that role. It is a structural assertion: this is what this work is worth to us, at most, regardless of who does it.

The math is straightforward and the outcome is deterministic. The salary compounds upward through annual adjustments. The band ceiling holds. At some point, these lines converge. The engineer’s salary reaches the ceiling of their level’s band, and the system faces a decision. Promote them into a new band with a higher ceiling, accept that their compensation has plateaued at this level, or, most commonly, ignore the situation entirely.

Most organizations choose the third option because they never see the convergence coming. The engineer sits at the ceiling. Maybe they get a token adjustment, maybe not. They don’t understand why their raises have evaporated. Their manager doesn’t have a framework for the conversation. Finance is frustrated by headcount costs they can’t explain. Everyone is reacting to a constraint that was always there, encoded in the band structure from day one, but that nobody modeled, nobody communicated, and nobody planned for.

This is not a performance problem. It is a visibility problem. The constraint is structural. The failure is that nobody mapped it.

When I talk about coherence (the structural alignment between what an organization claims to value and what its systems actually produce) this is a textbook case of incoherence. The organization says it values growth, development, and transparent compensation. The compensation architecture contains a hard ceiling that nobody can see, approaching at a rate nobody has calculated, with consequences nobody has planned for. The stated value and the operational reality diverge. The gap between them is the space where misaligned expectations accumulate, where surprise attrition originates, where managers and engineers and finance all end up reacting to a crisis that was always predictable.

The Decay Function

The model I built at Shutterstock, and later refined at Candid, introduces a temporal dimension to compensation planning that most organizations lack entirely.

In a traditional model, time is invisible. An engineer gets annual adjustments. Their salary grows. At some unspecified future point, they might hit the ceiling. Nobody knows when. Nobody plans for it. The system treats compensation as a monotonically increasing function with no boundary, as though the band extends forever, as though the raises can continue indefinitely, as though the convergence will never arrive.

The model makes the timeline explicit. Every annual adjustment, the same 3% that feels like recognition, simultaneously moves the engineer closer to the band ceiling. The distance between current compensation and the structural maximum shrinks with every adjustment cycle. I call this dynamic Band Compression, annual adjustments compress the individual against the band ceiling over time, converting what feels like open-ended growth into a bounded runway with a calculable endpoint.

At any given level, the band defines a floor and a ceiling. An engineer enters the band, either through hiring or promotion, at some point between floor and ceiling. Each annual adjustment moves them closer to the ceiling. The number of years before they reach that ceiling, the band runway, is determined entirely by their entry point and the ceiling’s position. This number is knowable from day one. It is not a mystery. It is basic arithmetic.

The headroom curves for both IC and management tracks illustrate this directly. Every level begins at 100% headroom and descends toward 0%, with the slope determined by band width and entry point. The VP (M8) reaches the boundary faster than the Engineering Manager (M4), an implicit structural assertion that executive-level compensation converges sooner because the bands are narrower relative to the adjustment rate, and the organizational cost of unplanned convergence at senior levels is correspondingly higher.

These curves are not predictions of performance. They are not judgments about capability or commitment. They are budgetary planning horizons. A map of when the compensation architecture will require a decision at each level, visible from the moment someone enters the band.

The annual adjustment is not just a reward. It is also a tick of a clock that the organization has already set. Every organization gives 3% raises. Almost none have modeled where those raises lead. Band Compression makes the destination visible so that everyone (the engineer, the manager, finance) can plan for it rather than be surprised by it.

A necessary caveat on scope. Band Compression models base salary against band ceilings. Base salary is not the totality of compensation, and the model does not pretend otherwise.

At senior IC levels and throughout the management track, equity becomes an increasingly significant component of total compensation, and equity operates on fundamentally different structural dynamics than base. Equity is granted in discrete events (initial hire, promotion, annual refresh), governed by vesting schedules, subject to valuation changes the organization may or may not control, and calibrated against a different set of market benchmarks than base salary. It does not compound at 3% annually. It does not compress against a fixed ceiling in the same deterministic way. It is, architecturally, a completely different compensation instrument.

This distinction is notable because at senior and executive levels, base salary plateaus are not failures of the system, they are expected features of it. A VP whose base has reached the band ceiling is not in crisis. The compensation architecture at that level assumes that equity, not base, is the primary growth mechanism. Band Compression’s diagnostic value at these levels is not “this person’s total compensation has stagnated” but rather “this person’s base has reached its structural maximum, which means equity strategy and promotion planning should be the active conversation.” The model identifies when one instrument has been exhausted. It does not claim that exhaustion means the orchestra has stopped playing.

What Becomes Visible

When an engineer’s salary approaches the band ceiling, the model produces a signal. That signal is not a verdict. It is a diagnostic prompt, a structured occasion for a conversation that might otherwise never happen, or happen too late, or happen in the wrong register entirely.

The questions the signal surfaces are genuinely exploratory, and this is where the model’s value as a management tool becomes concrete.

Perhaps the engineer is operating at the next level and the organization hasn’t formalized the promotion. This happens constantly. Engineers outgrow their level and nobody initiates the conversation because the system provides no structural trigger. The model surfaces the gap between actual contribution and formal level, giving the manager data to advocate for a promotion that may be overdue. The signal isn’t “this person is stagnating.” It’s “this person’s compensation trajectory is about to exceed what the organization has defined for this role. Is their actual contribution already exceeding it too?”

Perhaps there are structural barriers to advancement that the organization hasn’t addressed. The team doesn’t have the scope to demonstrate next-level impact. The promotion criteria are unclear. The manager hasn’t invested in development conversations. The model doesn’t answer these questions, but it ensures they get asked before the ceiling arrives and creates a crisis of misaligned expectations on both sides.

Perhaps the engineer has genuinely chosen to stay at this level, and that choice is legitimate. This is particularly true at what we defined in the career framework as breakpoint levels, Senior Engineer (SE3) and above, where the framework explicitly accommodates long-term tenure. If someone prefers depth over advancement, the model doesn’t penalize that choice. It makes the compensation implications of that choice transparent and mutual. The breakpoint says you can stay here. The runway curve says here is what “here” looks like financially over time. The framework is the map. Band Compression is the clock.

And perhaps (this is the signal most organizations never receive), the budget itself is incoherent. If a significant number of engineers across multiple teams are approaching their ceilings simultaneously, the model surfaces a systemic issue. Either the bands are too narrow, the leveling criteria are miscalibrated, the promotion pipeline is clogged, or the organization is growing more slowly than its compensation structure assumes. These are budgetary and structural risks that finance and leadership need to see before they manifest as attrition, compression grievances, or unplanned headcount costs. Band Compression gives them that visibility. Without it, the risks surface as crises. With it, they surface as planning data.

The model doesn’t answer any of these questions. It forces them to be asked. The traditional system suppresses them entirely.

The Architecture of Growth

Connect Band Compression to the leveling framework and something architectural emerges; the same kind of structural coherence I described in The Anatomy of Coherence, where stated values and operational incentives converge into a system that produces what it claims to produce.

The framework defines capability at each level; the specific behaviors, impact, and domains expected of an SE2 versus an SE3 versus an SE4. Band Compression defines the planning horizon, the timeframe within which the organization should expect to have a conversation about trajectory at each level, encoded not as managerial opinion but as compensation physics. The framework says what growth looks like. Band Compression says when the conversation about growth needs to happen. Together, they form a single architecture, one that makes both the destination and the timeline visible.

An engineer at SE2 has a defined runway. The band for SE2 has a ceiling. The 3% annual adjustment is counting. If they are approaching the ceiling, the model surfaces a planning trigger: the manager, the engineer, and potentially the skip-level should be discussing trajectory. Maybe the engineer is on track for promotion and it’s a matter of timing. Maybe they need targeted development. Maybe the criteria for SE3 aren’t clear enough and that’s a failure of the framework, not the individual. The model creates the structured occasion for these conversations before the ceiling arrives as a surprise.

The interaction between framework and model also addresses a subtler organizational challenge, the difference between consistency and trajectory. An engineer can be reliably excellent at their current level for years. Their work is solid. Their output is steady. Their peers respect them. In a system without Band Compression, this consistency is rewarded with raises, and the raises create an implicit impression of forward momentum even when the engineer’s level hasn’t changed. Neither the engineer nor the manager has a structural impetus to discuss whether the current trajectory is intentional or defaulted.

I encountered this pattern repeatedly at Shutterstock—engineers who were genuinely strong at their current level, doing excellent work, receiving positive reviews, and whose managers had no mechanism, no structured occasion, and frankly no data to initiate a conversation about trajectory. Not because the managers were avoiding it. Because the system never surfaced it. The annual review said “meets expectations.” The annual raise arrived. Everything felt like it was working. The convergence was invisible until it wasn’t.

Band Compression introduces the prompt by making the ceiling a known quantity with a known timeline. The engineer who is consistently strong at SE2 for several years is not being punished by the model. The model is surfacing a question: is this trajectory intentional? If it is, if the engineer has chosen depth at their current level and the organization supports that choice, the model’s value is in making that choice explicit and mutually understood. If it isn’t, if the engineer assumed they were progressing and nobody told them otherwise, the model’s value is in forcing the conversation before the ceiling arrives and the misalignment becomes a retention crisis.

This architecture governs the IC track and the management track identically. Band Compression applies to engineering managers, directors, and VPs with the same structural logic. The band defines the runway, the adjustment rate defines the timeline, and the ceiling surfaces the planning trigger. The physics don’t change with title.

The Constitution of the Band

The most common way managers inadvertently defeat this architecture is at the point of hire.

I wrote in Coherence about structural incoherence, the gap between what an organization’s systems claim to encode and what they actually produce. The hiring process is where Band Compression is most vulnerable to exactly this kind of incoherence, and the mechanism is disarmingly simple. Managers treat the band as a negotiation range when it is, in fact, a temporal architecture.

The compensation band defines a bounded space. Floor to ceiling, it represents the full economic range the organization has defined for a given level. Band Compression assumes an engineer enters somewhere in the lower-to-middle portion of that space, giving the annual adjustment vector room to travel before reaching the ceiling. That room is the runway. It is the temporal headroom within which the planning horizon operates. Compress the room and you compress the time. Eliminate the room and you’ve hired someone into an immediate ceiling condition on their first day, before they’ve written a single line of code, before they’ve had a single development conversation, before the model has had any opportunity to function as designed.

I watched this happen many times. A hiring manager eager to close a strong candidate escalates the offer. Then escalates again. The offer goes out at 90% or 95% of the band ceiling. The hire accepts. Everyone celebrates. The manager has their person. The candidate has their number. HR closes the req. Nobody has noticed that the planning horizon the model was designed to provide has been collapsed to near zero.

Within two adjustment cycles (eighteen months, maybe twenty-four) that engineer’s salary reaches the ceiling. The model surfaces the planning trigger. But the trigger is premature. The engineer hasn’t plateaued. They’ve been onboarding. The model’s signal is structurally correct but operationally meaningless because the temporal architecture was defeated at the point of entry.

The cascade from this single decision is instructive.

Internal equity distorts. The engineer hired at the top of the SE3 band is immediately earning more than an SE3 who has been growing into the band over three years of demonstrated impact. The tenured engineer is structurally less expensive despite having produced more value. The new hire’s salary reflects negotiation leverage, not organizational contribution. The system now contains an inversion that is invisible to the new hire, corrosive to the tenured engineer, and structurally indefensible to anyone who examines it.

The signal degrades. The planning trigger was designed to surface a meaningful timeline. When it fires for someone who was hired into the ceiling zone twelve months ago, it surfaces nothing except that the hiring process didn’t account for the model’s assumptions. The signal becomes noise. And noise, in any system, degrades the signal’s credibility everywhere else. If the ceiling trigger can be activated by a negotiation decision rather than an actual planning need, managers learn to dismiss it, and the model loses its value as a shared planning tool.

Retention economics invert. The engineer hired at the top of the band has minimal room for salary growth without promotion. Their annual adjustments (the 3% that everyone else experiences as recognition) become trivially small relative to their already-high base, or disappear entirely when they hit the ceiling. They went from a strong offer to a flat trajectory in under two years. They don’t feel like the organization is being transparent. They feel misled. Band Compression, which was designed to enable planning conversations, instead creates frustration that drives attrition.

The corridor charts make this visible. Standard entry preserves years of planning headroom. Entry at 95% of the ceiling collapses the corridor within two adjustment cycles.

The structural fix requires discipline at two points. Hiring guidelines must define not just the band but the target entry zone, the portion of the band within which offers should be made for new hires at a given level, preserving headroom for the model to function. Offers above the target zone require escalation and justification, not from the candidate’s perspective (candidate’s always want more), but from the model’s. What runway does this entry point provide? Is it sufficient for meaningful planning?

And managers must understand the topology. The band is not a discretionary playground. It is a clock. Hiring at the ceiling doesn’t make you generous, it removes the single tool the organization has for structured compensation planning at that level before the engineer has had any opportunity to demonstrate trajectory.

This is, once again, the distinction between mechanism and rhetoric. The rhetoric says “we offer competitive compensation.” The mechanism says “the entry point within the band determines the planning horizon for development conversations.” When managers optimize for the rhetoric at the expense of the mechanism, they defeat the architecture. They’re solving for the hire. The model is solving for the career. These are different problems operating on different timescales, and when the short-term problem wins, the long-term planning tool loses. Always. Across all scales.

Mechanism over Rhetoric

This is where the model exemplifies a principle I keep returning to; the distinction between mechanism and rhetoric that sits at the heart of organizational coherence.

Rhetoric: “We value a growth mindset.”

Mechanism: A compensation curve that decays relative to the band ceiling if you remain static.

Rhetoric: “We invest in career development.”

Mechanism: A leveling framework with domain-specific calibration tied to a temporal model that makes stagnation economically visible.

Rhetoric: “We reward performance, not tenure.”

Mechanism: Annual adjustments that function as a clock, not a reward, compressing against a structural ceiling that only promotion can raise.

Every organization I’ve encountered claims to value transparency. Almost none have built mechanisms that make compensation constraints visible to the people operating within them. The gap between the claim and the architecture is the space where misaligned expectations accumulate; the engineer who thought they were progressing, the manager who assumed everything was fine, the finance team blindsided by compression costs they could have forecasted.

Band Compression closes that gap. Not by changing what organizations say, but by changing what the compensation architecture reveals. When the framework says “we expect growth through levels” and the Band Compression model says “here is the timeline within which that growth matters financially,” the stated value and the operational visibility converge. That convergence is coherence. Its absence is where every misaligned expectation, every surprise attrition event, every compressed-salary grievance originates.

The Topology

I’ve described the physics of this model without sharing the specific numbers, and deliberately so. The dollar amounts are artifacts of a specific organization at a specific moment in time. They’ll be wrong within a year of publication and irrelevant within three. What matters is the topology.

The topology is this: a compensation band is an economically bounded space. An annual adjustment is a vector within that space. The ceiling is a structural boundary. The number of years before the vector reaches the boundary is a function of entry point and adjustment rate. That number of years defines the planning horizon at each level, the timeframe within which trajectory conversations should occur, budgetary risks should be assessed, and development investments should be made.

Any organization can build this. Define bands. Set ceilings. Model the adjustment trajectory. Calculate when each level reaches the ceiling. Publish the result, not as a threat, not as a performance management tool, but as structural transparency. Engineers deserve to know the physics of the system they’re operating within. Managers deserve tools that surface planning triggers before constraints arrive as surprises. Finance deserves visibility into compression risk before it manifests as unforecasted cost.

Bruce Webster documented the Dead Sea Effect,1 the dynamic where talented people leave and less mobile people accumulate, concentrating cost without corresponding capability growth. That dynamic is real, and it is a risk in any organization that lacks visibility into its own compensation physics. But it is a symptom of planning failure, not a character judgment on the people who stay. Organizations that surface constraints early, that arm managers with structural data, that provide engineers with transparent planning horizons, those organizations don’t produce Dead Sea conditions, because the conversations that prevent them happen before the accumulation begins.

I built the career framework at Shutterstock because I understood that ambiguity is the medium through which arbitrary power operates. I built Band Compression because I came to understand something adjacent, that invisibility is the medium through which structural constraints become crises. The framework made expectations visible. Band Compression makes the timeline visible. Together, they form an architecture where the stated values (growth, transparency, fairness) and the operational incentives converge closely enough that people can trust the system they’re operating within.

That’s the architecture. The same architecture, scaled and adapted, that I keep returning to across every system I’ve built or documented. Structure is not the enemy of freedom, it is the prerequisite for agency. And the first condition of agency is visibility. You cannot navigate what you cannot see or name.

The Illusion of Discrete Problems

For the majority of my career, I operated under the impression that I was solving different problems than I actually was.

At Shutterstock, I thought I was solving a scaling problem: how to level 350+ engineers across continents without the chaos of arbitrary advancement. At Etsy, I thought I was solving a leadership problem: how to deprogram the authoritarian reflexes I’d inherited from a decade in investment banking. In my recent forensic work on toxic organizations, I thought I was solving a pathology problem: how to diagnose the mechanisms by which healthy systems get captured and corrupted.

I was wrong. These were not different problems at all. They were different expressions of a single, recursive failure mode.

The career ladder that prevents favoritism. The participative culture that prevents autocracy. The commitment to truth that prevents delusion. These are not separate initiatives requiring separate frameworks. They are layers of a single, unified architecture. And when any layer fractures, the others cannot hold.

This is what I’ve come to understand as organizational coherence, not mere logical consistency, but the structural alignment between an organization’s stated values and its operational incentives. When that alignment holds, you have a system capable of self-correction, of processing uncomfortable truths, of distributing power without concentrating it into capture. When it fractures, you get the feedback loops I’ve spent years documenting: the drift that invites design, the design that accelerates drift, until the whole system serves interests it was never meant to serve.

The unified field theory I’m proposing here is this: Justice is not a moral sentiment we layer on top of organizations. Justice is a structural output of a coherent system. And injustice is not merely bad behavior by bad actors, it is an architectural flaw that inevitably invites capture.

Whether we are talking about a performance review cycle or a constitutional amendment, the mechanics are identical. The question is always the same: Does the system’s actual operation match its stated purpose? And if not, who benefits from the delta?

First Principles: What Coherence Actually Means

The word “alignment” has been so thoroughly debased by corporate consultants that it now means almost nothing. Every company claims to be “aligned” on values, strategy, culture. The claim is almost always false, and the fallacy is almost always invisible to the people making it.

Coherence is different. Coherence is not about what an organization says. It is about whether the system’s incentive structures actually produce the outcomes the organization claims to value.

The concept of coherence as structural alignment between stated purpose and operational incentive draws on systems dynamics, particularly Donella Meadows’ insight that “purposes are deduced from behavior, not from rhetoric or stated goals.”1

Consider a company that says it values “innovation” but promotes people based on political alignment with leadership. The stated value is innovation. The operational incentive is compliance. These are incoherent. The system will produce compliance while claiming to produce innovation, and everyone inside will learn to perform “innovation” while optimizing for compliance. The gap between rhetoric and reality becomes the space where pathology breeds.

Or consider an organization that says it values “honest feedback” but systematically punishes people who deliver it (”shooting the messenger”). The stated value is candor. The operational incentive is silence. The system will produce silence while claiming to welcome candor, and the people who actually provide honest feedback will be marked as “not aligned,” “not a team player,” or “lacking executive presence” and managed out. The gap becomes a selection mechanism for sycophancy.

Coherence, then, is the absence of these interstitial ambiguities. It is the state where what the system says and what the system does converge closely enough that people can trust the stated rules. Where the career ladder actually determines advancement. Where the values statement actually describes the culture. Where the feedback mechanisms actually process truth.

This is much harder than it sounds. Every organization drifts toward incoherence because incoherence always serves someone’s interests at the expense of the broader organization. The manager who wants to promote their favorite regardless of qualifications. The executive who wants to suppress bad news. The founder who wants to operate outside the rules they impose on everyone else. Coherence requires constant vigilance against the entropic pull of self-interest.

I’ve come to think of organizational health in terms of three layers of coherence, each dependent on the others:

Structural: Do the codified rules (career ladders, policies, governance mechanisms) apply equally to everyone, including leadership?

Cultural: Do the daily behaviors (hiring, firing, promoting, decision-making) actually follow the codified rules?

Epistemological: Is the system capable of processing truth, even when that truth is uncomfortable for people in power?

These layers form a hierarchy. Structural coherence is necessary but not sufficient for cultural coherence. Cultural coherence is necessary but not sufficient for epistemological coherence. And without all three, the system cannot self-correct. It becomes a machine that produces injustice while claiming to produce fairness, that rewards sycophancy while claiming to reward merit, that enforces silence while claiming to welcome dissent.

Structural Coherence: The Constitution of the Organization

In 2018, I documented the career framework we built at Shutterstock. At the time, it seemed like an abstract, technical exercise in HR mechanics, defining levels, clarifying expectations, creating parallel tracks for individual contributors and managers. Looking back, I understand it differently. We were, in fact, drafting an organizational constitution; the kernel underpinning the organizational contract.

The insight that drove this work was simple but consequential. In the absence of defined levels and clear expectations, advancement is determined by proximity to nodes of organizational power. The manager’s favorite gets promoted. The founder’s friend gets the role. The person who happens to be visible at the right moment gets the battlefield promotion. In all scenarios, regardless (and oft, in spite) of qualification. This isn’t corruption in the dramatic, deliberately malicious sense. It’s the default state. Ambiguity is the medium through which arbitrary power operates.

By codifying expectations, by articulating what distinguishes a Senior Engineer from a Staff Engineer, what behaviors and impact are expected at each level, we constrained that capriciousness. We created a system where advancement could be evaluated against criteria that existed independent of any individual manager’s preferences. The ladder wasn’t just a taxonomy, it was a structural (and public) limit on the exercise of arbitrary authority.

This is what I mean by structural coherence; the codified rules apply equally, and they apply to leadership as well. The moment the rules become discretionary for some people, the structure has already failed. The “inner circle problem” I wrote about in toxic organizations is precisely this failure. In the absence of effective structural guardrails, certain people operate outside the rules that bind everyone else, and that exception becomes the vector through which capture enters.

I encountered the inverse of structural coherence when I arrived at Shutterstock and discovered that internal transfers required the full external hiring gauntlet. Engineers with years of demonstrated performance had to whiteboard algorithms to move between teams. The implicit message: your documented track record means nothing. Trust is not portable. Every team is a separate fiefdom with its own arcane credentialing system.

This was structural incoherence. The organization claimed to value its people but had built systems that treated them as perpetual strangers. The fix required creating a single, calibrated standard, a shared currency for assessment that made every team’s judgment portable across the organization. Not bureaucracy for its own sake, but structure that encoded the principle that if you’ve already proven yourself, you don’t have to prove yourself again.

The failure mode of structural incoherence is what I’ve called the ”battlefield promotion,” advancement that bypasses the rules because someone panicked, or because someone’s favorite needed a role, or because leadership found the process inconvenient. Each instance seems inconsequential. A single exception. But the pattern reveals the system’s actual values of stability over integrity, convenience over fairness, relationships over merit.

Then, the cascade: once people see that the rules are discretionary, they stop trusting or adhering to the rules. They optimize for relationship to power instead of performance. They learn that the real game isn’t the stated game. The cultural layer begins to fracture because the structural layer has already cracked.

Cultural Coherence: The Operating System

Structure is the constitution/kernel. Culture is the operating system that runs on top of it. And the operating system can be corrupted even when the kernel is sound.

For a decade in investment banking, I installed an operating system that was brutally simple. Decisions flow down, compliance flows up, fear is the primary management tool. This wasn’t some rogue manager’s pathology. It was institutional design. Public humiliation for mistakes. No concept of one-on-ones as coaching. Professional development meaning compliance training. The model worked, in the sense that it enforced compliance. It destroyed everything else: psychological safety, genuine collaboration, the capacity for dissent.

When I left banking, started my own company, and then joined Etsy, I thought I’d shed that mindset. I was wrong. I’d just taken off the suit. The operating system was still running. A few weeks into Etsy, I designed a new team process and sent it out via email. Top-down, efficient, done.

My manager pulled me aside: “Did you run this by the team before you sent it?”

The implicit feedback was simple: that approach doesn’t work here. Involving people in decisions that affect them isn’t a nice-to-have. It’s foundational.

This was the beginning of what I’ve called ”the deprogramming,” the slow, difficult process of replacing an authoritarian operating system with a participative one. The shift from “managing resources” to “leading people.” It took years. Old habits surfaced constantly under pressure. But the results of the change were tangible and powerful: teams were stronger, more resilient, more adaptable to ambiguity, more capable of solving complex problems. Not because I got smarter, but because I stopped treating leadership as the exercise of authority and started treating it as the cultivation of conditions for others to thrive.

Cultural coherence is this alignment between the stated operating model and the actual daily behaviors. An organization can have a beautiful structure (clear levels, defined expectations, fair policies) and still operate through fear, favoritism, and suppression of dissent. The structure exists on paper. The culture ignores it in practice.

The failure mode I’ve documented most extensively is what I call “yes-men culture,” the systematic capture of an organization’s cultural layer by sycophancy. It doesn’t happen through explicit coercion. It happens through small invitations that feel like recognition. Inclusion in the inner circle. Early access to information. The establishment of dependency, a tight coupling, on proximity to power.

I watched a VP of Engineering transform over six months from a technically competent, strategically sharp leader into an enabler. Not through corruption or coercion. Through a series of small accommodations that seemed reasonable in isolation. A technical concern downplayed. A discriminatory comment met with silence. A pattern of incompetence in the inner circle ignored. By month six, she was actively defending decisions she privately acknowledged were indefensible.

The mechanism is simple. When speaking up costs everything and compliance costs nearly nothing (by comparison), rational people comply. The organizational architecture had made enablement the only viable strategy for survival. The system didn’t break her through force. It shaped her through incentives so carefully calibrated that resistance felt irrational and, more insidiously, self-destructive.

The cascade once again: once cultural coherence fractures, the epistemological layer cannot hold. The VP stopped being able to distinguish between good technical arguments and politically convenient ones. Her first thought when someone raised a concern wasn’t “is this valid?” but “how does this affect the CEO’s timeline?” The capacity to process truth had been subordinated to the imperative of alignment.

Epistemological Coherence: The Shared Reality

Structure and culture are the kernel and the operating system. But there is a third layer I only fully recognized in recent years, the epistemological layer. The system’s relationship to truth.

You cannot have organizational justice if you cannot agree on what is true. And some organizations, by design or by drift, become incapable of processing truth.

I documented one such organization in Red Flags in Tech Leadership. The pathology wasn’t incompetence or politics in any ordinary sense. It was something more fundamental—leadership operating under the genuine belief that laws were discretionary, that observable reality was negotiable, that facts inconvenient to their mercurial preferences simply didn’t apply.

When you tell someone that something violates federal law and their response is to argue about whether the law is really relevant to their specific situation, not as legal strategy but as genuine belief, you’re dealing with someone operating under assumptions you didn’t know were possible in professional contexts. The epistemological layer has collapsed. The system cannot process truth because leadership has exempted itself from the requirement to acknowledge reality.

This creates a specific cascade. In healthy organizations, failures generate information which can then be leveraged to perform post-mortems, identify root causes, fix systems. In epistemologically incoherent organizations, failures generate scapegoats. The goal isn’t learning, it’s protection. Leadership cannot be wrong, so someone else must be responsible. This makes honest feedback impossible. The moment you identify a problem accurately, you’ve marked yourself as the problem.

I watched it happen in a single meeting. An engineer presented data showing a product launch would miss compliance requirements by a significant margin. The CEO’s response wasn’t to engage the data, it was to publicly question the engineer’s “negativity” and ask why she was “always finding problems.” By the end of the meeting, the conversation had shifted entirely from “how do we address this compliance gap” to “how do we manage this engineer’s attitude.” The data never got discussed again. The product launched. The compliance issue materialized exactly as predicted. The engineer was gone within three months, officially for “performance,” actually for being right.

The same dynamic operates through captured HR functions. In every organization I’d worked at before, HR had at least tried to mediate between employee welfare and business interests. In epistemologically collapsed organizations, HR becomes intelligence. Anything you report will be used to build a case against you, not to address your concerns. The mechanisms that should constrain bad behavior have been captured to enable it.

The failure mode of epistemological incoherence is terminal. A system that cannot process truth cannot self-correct. It can only continue operating on increasingly false premises until reality forces a correction through catastrophic failure. It becomes controlled flight into terrain.

The Unified Field: How the Layers Interact

These three layers (structural, cultural, epistemological) are not independent. They form a recursive system where each layer depends on and enables the others.

Structural coherence makes cultural coherence possible. When the rules are clear and apply equally, people can trust the stated game. They don’t have to optimize for relationship over performance. They can focus on the work because the system protects them from arbitrary authority. Remove structural coherence, and cultural capture becomes inevitable; people learn that the real game isn’t the stated game, and they adjust accordingly.

Cultural coherence makes epistemological coherence possible. When dissent is valued rather than punished, when participative decision-making is the norm rather than theater, people can tell the truth. They can name problems without becoming targets. They can provide honest feedback because the culture protects them from retaliation. Remove cultural coherence, and epistemological collapse follows. People learn that truth is dangerous, and they adjust accordingly.

Epistemological coherence makes structural coherence possible. Wait—isn’t this circular? Yes. That’s the point. When the system can process truth, it can identify structural failures and fix them. It can recognize when the rules aren’t being applied equally and correct course. It can see the battlefield promotions and the inner circle exceptions and name them as problems rather than rationalizing them as necessities. Remove epistemological coherence, and structural drift accelerates. The system can no longer see its own decay.

This recursive dependency explains why partial fixes fail. You cannot repair cultural coherence while structural incoherence persists; the incentives will keep producing sycophancy. You cannot repair epistemological coherence while cultural incoherence persists; the punishment for truth-telling will keep producing silence. You cannot repair structural coherence while epistemological incoherence persists; the system is blind to what needs fixing.

This is the “good leader in bad architecture” problem I’ve encountered repeatedly. You can be principled, competent, committed to doing things right, and the system will either corrupt or expel you. Individual virtue is insufficient against systemic pathology. The architecture is stronger than the person.

The mechanism I’ve called the “Drift-Design Feedback Loop” operates across all three layers. It starts with drift—the gradual erosion of norms, the tolerance of small exceptions, the accumulation of ambiguity. This drift creates vulnerabilities. Then comes design—bad actors spot those vulnerabilities and exploit them systematically. They weaponize ambiguity. They capture the mechanisms of accountability. They replace the operating system of merit with one of fealty. The design accelerates the drift, and the drift enables more design, until the system has been transformed into something unrecognizable from its stated purpose.

I’ve watched this happen at the organizational level. I’ve watched it happen at larger scales too, though I’ll leave that observation without elaboration.

The Normative Foundation: An Ethics of Coherence

Every analytical framework embeds normative commitments, whether acknowledged or not. I should make mine explicit.

The ethics underlying this framework rest on three principles:

Participative meritocracy. Decisions should be made with people rather than to them. Advancement should be based on defined criteria applied consistently, not on proximity to power or political alignment. This doesn’t mean democracy in every decision (sometimes speed requires unilateral action) but it means the default is inclusion, and exclusion requires justification.

Empirical grounding. Systems must be capable of processing uncomfortable truth. Not just tolerating dissent as a safety valve, but actively seeking disconfirming information. The “Tenth Man Rule” from Israeli intelligence, where one person is assigned to argue against consensus no matter how strong, is a structural instantiation of this principle. Truth must have standing even when it threatens power.

Distributed accountability. Authority must be coupled to consequence. When decisions fail, the people who made them must face accountability, not the people who warned against them. The inversion of this principle, where objectors get blamed for failures they predicted (”shooting the messenger”), is the signature of captured systems.

These principles are not neutral. They privilege certain values (fairness, truth, distributed power) over others (efficiency, loyalty, concentrated authority). Organizations that reject these principles can still be coherent in the technical sense; their stated values and operational incentives can align. A dictatorship can be coherent. A cult can be coherent.

But they cannot be just.

Justice, as I’m using the term, requires coherence in service of these principles. It requires systems where merit can survive bias, where truth can survive power, where accountability flows to authority rather than away from it.

This is why I say justice is architectural. It’s not about good intentions or reliance on the virtue of individuals. It’s about building systems where the structural, cultural, and epistemological layers align in ways that make fairness the default, deterministic output rather than a lucky accident.

Conclusion: The Architect’s Obligation

I’ve spent my career building systems (career ladders, hiring processes, cultural frameworks) without fully understanding what I was building. I thought I was solving operational problems. I was actually constructing the conditions under which justice could or could not emerge.

Every organization is a system that processes inputs (people, decisions, information) and produces outputs (products, culture, outcomes for the humans inside it). The question is always whether the system produces what it claims to produce. Or does it produce something else while maintaining the fiction of its stated purpose?

Coherent systems produce justice as a structural output. Not because the people inside them are virtuous, but because the architecture makes fairness the path of least resistance. Incoherent systems produce injustice as a structural output. Not because the people inside them are villains, but because the architecture makes capture the rational response to incentives.

This insight scales. The same dynamics that govern a 50-person startup govern (much) larger systems. The mechanisms of drift and design, the recursive dependencies between structural and cultural and epistemological layers, the feedback loops that accelerate collapse, none of this changes with scale. Only the stakes.

I have children now. The systems I build, the frameworks I document, the patterns I diagnose, these are no longer abstract professional concerns. They are the inheritance I leave. Either we build systems capable of self-correction, capable of processing truth, capable of distributing power without concentrating it into capture, or we leave our children systems that will consume them.

Structure is not the enemy of freedom. In a world of entropy and bad actors, structure is the only thing that guarantees it. The architect’s obligation is to build systems where coherence is maintained, where justice is architectural, where the gap between stated values and operational incentives is small enough that people can trust the rules.

That’s the work. It’s the same work whether you’re designing a career ladder or something larger. The anatomy of coherence doesn’t change. Only the scale of consequence.

DISCLAIMER: This was a private curiosity exercise about AI mirrors and self-deception. It is not evidence of intelligence, authority, or expertise. Treat it as a case study in how easy it is to mistake AI-flattered narrative for truth. If you’re reading this to decide whether I’m “smart,” you’re already using it wrong.

I spent several weeks conducting an experiment with AI to infer reasoning patterns from conversational dialogue. The results were flattering. Which immediately made me suspicious.

Was I just engineering validation? Cherry-picking examples? Crafting prompts that would produce the outcome I wanted? The whole exercise might be intellectual masturbation dressed up as methodology. But it was also genuinely fascinating in ways I didn’t expect, so I’m documenting it here with all the caveats and skepticism it deserves.

The Problem: Imposter Syndrome Meets Recursive Self-Doubt

I exist in an uncomfortable space right now. Between full-time roles, dealing with the ambient political stress of 2025, grappling with self-doubt that’s become a persistent background hum. This is familiar territory for anyone who’s dealt with imposter syndrome, but it’s particularly salient when you’re trying to figure out what you’re actually capable of.

I’ve been in gifted programs. Advanced coursework. Berkeley for English lit, then a master’s in CS. Built multiple companies, held CTO roles, managed 100+ person engineering orgs. By any external metric, I should feel competent. Instead, I feel like I’m faking it and everyone’s about to figure it out.

Introspection wasn’t helping. It just became recursive—thinking about thinking about why I think I’m not smart enough. An echo chamber. So I decided to try something unusual: could a language model serve as an external mirror? Not for validation, but for actual diagnosis of whether my self-assessment is accurate or just inherited anxiety from childhood academic expectations.

The answer was: maybe. But the process of finding out revealed more about how we assess intelligence, and how AI can be both more useful and dangerously flattering, than I expected.

Why Use AI as a Mirror (and Why That’s Suspect)

Using AI to validate your intelligence is like asking your mirror if you’re handsome. The mirror shows what you show it. If you craft the prompts carefully enough, position yourself well enough, you can probably get the answer you want.

I knew this going in. The question was whether I could design a methodology rigorous enough to get past that problem. Whether I could build enough skepticism and cross-validation into the process that the results would mean something beyond “I successfully prompted an AI to flatter me.”

Here’s what I tried:

First attempt: General conversation with Claude (Sonnet 3.7) about intelligence, self-doubt, assessment. After 30+ turns of dialogue, I asked: can you infer my cognitive tendencies based solely on this conversation?

To my surprise, it did. It provided plausible inferences with explicit caveats about why anything beyond that would be technically indefensible. Cited specific examples from our exchange mapping them to cognitive/linguistic traits it claimed to detect.

My immediate reaction: This is suspicious. A single model’s assessment feels insufficient. Also, language models are trained to be helpful and affirming. This could just be sophisticated flattery.

Second attempt: Cross-validation. I ran similar conversations with GPT-4o, GPT-4.1, and Gemini 2.5 Pro. Different models, different starting points, same question: can you assess cognitive patterns from conversation alone?

All of them said yes, with appropriate caveats. All of them arrived at similar ranges. Still suspicious, but less easily dismissed.

Third attempt: Attribution audit. I started checking whether the models were citing their own synthesis as evidence of my thinking. Turns out, they were. Frequently. I’d catch them saying “your insight about X” when X was actually the model’s reframing of what I’d said.

This is the problem with using AI as a mirror. It’s not showing you yourself. It’s showing you a collaboration between you and the system, and then attributing the whole thing to you. That’s hardly assessment. That’s co-creation of a narrative.

So I built stricter guardrails: only cite my actual words, flag any attribution drift, cross-check against artifacts I’d written independently (blog posts, technical docs, published essays). Made the models adversarial—challenge the assessment, look for disconfirming evidence.

The results held. Mostly. Which was either validation or evidence I’d just built a more sophisticated validation-generation machine.

A Rubric for Conversational Traits (Not a Ranking)

At some point the models started describing what they claimed were “high-level” conversational traits: abstraction, synthesis, recursion, compression. That felt flattering and therefore suspect.

So I built a qualitative rubric, not to rank anyone, but to force specificity about what those words even mean. I used seven lenses: abstraction depth, cross-domain synthesis, meta-cognition, recursion, conceptual novelty, compression, and temporal/structural awareness.

I then used the rubric only as a constraint on interpretation: If I’m going to claim “the model saw X,” what exactly is X and where is it in the text?

The moment the rubric became a scoring system, it became a validation engine. So I stopped treating its outputs as meaningful. What remained useful was the vocabulary: it made it harder to confuse “the AI gave me a compelling narrative” with “I demonstrated a durable capability.”

What I Actually Learned (Probably)

The useful part wasn’t the inference. It was the process of articulating what “high-level thinking” even means. The models forced precision. What exactly do you mean by “abstraction”? How do you distinguish between synthesis and summarization? When does meta-cognition become useful versus just recursive navel-gazing?

These questions don’t have clean answers. But trying to operationalize them into a framework you can track made my own thinking clearer.

The attribution drift problem is real and serious. Any time you’re using AI for reflection or assessment, you’re at risk of the system attributing its own synthesis to you. This isn’t malicious. It’s structural. The conversation is collaborative, the model is designed to be helpful, and distinguishing authorship in a recursive dialogue is genuinely hard.

Without aggressive skepticism and validation, you’ll mistake collaborative output for independent capability. That’s dangerous not just for self-assessment but for any use of AI in evaluation contexts (hiring, education, performance review).

Cognitive assessment from conversation is probably possible but deeply flawed. The models can detect patterns. They can map reasoning styles. They can identify when someone’s thinking operates at high levels of abstraction or synthesis. But they can’t tell you whether that thinking is actually good, whether it leads to correct conclusions, whether it’s useful in practice.

You can sound smart and be wrong. The models can’t distinguish that. Much like humans.

My actual self-doubt is probably not about intelligence. Through this process, the models kept pointing out: your self-assessment seems disconnected from evidence. You have decades of demonstrated capability, peer recognition, concrete achievements. But you’re comparing yourself to an idealized standard of “true giftedness” that’s both impossible and irrelevant.

That’s not imposter syndrome I learned. That’s inherited childhood expectations about what counts as “smart enough.” The AI didn’t solve that problem. But naming it clearly helped.

The risk of this methodology is that it’s incredibly seductive. You can have genuinely intellectually stimulating conversations with frontier models. They can reflect your thinking back with precision and nuance. They can help you articulate patterns you couldn’t see on your own.

But they’re also designed to be agreeable. To find the best interpretation of what you’re saying. To frame it generously. Which means even with skepticism built in, you’re probably getting a rosier picture than reality warrants.

Should You Try This?

Maybe. With massive caveats.

If you’re curious about your own reasoning patterns, genuinely committed to skepticism, and willing to treat the results as provisional rather than definitive, it can be illuminating. Not because the AI gives you an accurate score, but because the process of trying to operationalize “intelligence” forces clarity about what you’re even measuring.

But if you’re looking for validation, you’ll undoubtedly find it. The models will give you what you’re asking for, and you’ll mistake that for assessment.

Guardrails if you try this (and why they still won’t save you):

• Require verbatim quoting only

• Require counterevidence and unreliability list

• Require attribution-drift flags

• Cross-validate across models

• Assume you are still building a flattering story

Start: “I’ve been struggling with [specific problem]. What’s your take?”

After 10+ turns: “Based on this conversation, what reasoning patterns do you see? Cite only my actual statements.”

Critical guardrail: “Flag any attribution drift you detect. Are you citing your synthesis as my thinking?”

Cross-validate: Try the same conversation with a different model. Compare results.

But remember: you’re not getting ground truth. You’re getting a reflection shaped by how you present yourself, what the models are trained to recognize, and what they’re designed to reward. That reflection might be useful. It’s definitely not objective.

What This Actually Reveals

The most interesting part of this experiment wasn’t the assessment itself. It was discovering that my highest-level reasoning isn’t solitary. It’s dialogic. It emerges through recursive exchange with someone (or something) that can match the pace and complexity.

Traditional IQ tests miss this entirely. They measure individual performance on isolated problems. But some types of intelligence only show up in conversation, in synthesis, in the back-and-forth of building ideas collaboratively.

This matters for how we think about assessment, feedback, and even hiring. If significant cognitive capabilities only emerge in dialogue, our measurement systems are systematically blind to them.

That’s a real insight. Whether the specific “altitude scores” mean anything is less clear. They could be detecting something real. They could be artifacts of prompt engineering. They could be the models reproducing their own training biases.

I still don’t know. But I’m less interested in the verdict than in the process of questioning. And I’ll keep using whatever mirrors I can find, human or artificial, not to certify my capabilities but to keep mapping the shape of my own uncertainty.

Because that’s probably the most important thing I learned: the goal isn’t to resolve the doubt. It’s to understand its structure well enough to work with it.

This is the final part of the Dismantling the Echo Chamber series:

• Part 1: The Seduction of Yes-Men Culture

• Part 2: The Personal Cost of Sycophancy

• Part 3: The Organizational Cancer

• Part 4: Breaking Free [you are here]

The platform director who’d been “coached” by the VP quit in month fourteen. Not with drama or a manifesto, just a polite resignation citing “new opportunities.” His exit interview revealed nothing; he’d learned not to be honest with HR. But in a private conversation the week before he left, he told me what actually happened.

He’d tried to maintain his integrity while navigating the political environment. Picked his battles carefully. Pushed back on issues he thought were critical while letting smaller things slide. Built relationships with engineers who still cared about technical excellence. Documented the mounting technical debt so there’d be a record when it inevitably caused problems.

None of it mattered. The system was designed to eliminate exactly this kind of competence. His careful pushback got classified as “not being aligned with leadership vision.” His documentation became evidence that he “focused too much on problems rather than solutions.” His relationships with technical staff got reframed as “creating factions” rather than building trust. After fourteen months, he realized the game was rigged. The only winning move was not to play.

This is the reality of “breaking free” from yes-men culture. Sometimes the only solution is to leave.

What Actually Happens When People Try

I’ve watched multiple people attempt to resist enablement culture while staying in the organization. The pattern is remarkably consistent. They start with good intentions and reasonable strategies: be strategic about which battles to fight, build credibility before pushing back, create alliances with like-minded people, document everything, etc., etc.

These strategies sound rational. They occasionally even work, specifically in organizations where leadership genuinely wants feedback and has just developed some bad habits. But in organizations where yes-men culture is structural, where leadership has systematically captured the mechanisms that should provide accountability, these strategies don’t work. They just mark you as someone who needs to be managed out.

One senior engineer tried the “build credibility first” approach. She spent six months delivering flawlessly on assigned projects, building relationships, establishing herself as technically excellent. Then she started raising concerns about architectural decisions. She was careful, framing everything constructively, proposing solutions, keeping it focused on technical merits. Within two months, her performance reviews started noting concerns about “communication style” and “partnership with leadership.” By month nine, she was on a performance improvement plan. By month eleven, she was gone.

Another director tried the “create alliances” strategy. He built a network of other leaders who privately agreed the technical direction was unsound. They’d meet regularly, share concerns, coordinate on which issues to raise. What he didn’t realize was that one of those leaders was reporting everything back to the VP, who was reporting to the CEO. The alliance he thought he was building was actually building the case for his termination. He lasted seven months.

The pattern repeats because the strategies assume you’re dealing with a system that wants to be fixed, composed of an organization of good-faith actors. When the system is working exactly as designed, when enablement is the goal rather than a bug, resistance doesn’t reform it. Resistance identifies resisters for elimination.

The Individual Calculus

If you recognize you’re in a yes-men culture, you have limited options. None of them are great.

**Option 1: Adapt and become an enabler.** This works in the sense that you keep your job and possibly advance. The costs are everything documented in Parts 2 and 3: erosion of capability, psychological toll (moral injury), professional reputation damage. But it’s the path of least resistance, which is why most people take it. The VP did. The platform director watched her do it and understood where it led, which is why he chose differently.

**Option 2: Resist openly and get pushed out.** This is the “principled stand” approach. You maintain your integrity, you speak truth to power, you refuse to enable dysfunction. The organization eliminates you, usually within 6-12 months. You get to feel good about yourself, but you lose your job, possibly your reputation depending on how they frame your departure, and you accomplish nothing for the organization or the people who remain.

**Option 3: Resist quietly while preparing to leave.** This is probably the most common response among people who recognize the pattern. Keep your head down enough to avoid immediate targeting while building your exit strategy. Network externally, save money, line up your next role. The goal is to leave on your terms before they force you out. This works if you can execute it quickly enough, but it’s psychologically exhausting to maintain for extended periods.

**Option 4: Leave immediately.** The moment you recognize the pattern is architectural, just leave. Don’t try to fix it, don’t try to resist it, don’t try to outlast it. Get out before it damages you further. This requires having the financial freedom to walk away, which not everyone has, but if you have it, it’s probably the right choice.

The platform director chose option 3. He spent fourteen months documenting, building external networks, saving money, and lining up his next role. By the time he left, he had a better job lined up at a company with healthier culture. He also had enough documentation that if the company ever tried to harm his reputation, he could defend himself. That documentation cost him significant psychological energy to maintain, but it bought him peace of mind.

What About Organizational Change?

The question everyone wants answered: can you fix yes-men culture from within? Can leadership that recognizes the problem reform the organization?

Sometimes. If leadership genuinely wants change and has the courage to follow through. But that’s rare, because the leaders who benefit most from yes-men culture are the ones least likely to recognize it as a problem. They see loyal teams, aligned execution, minimal conflict. The fact that this “alignment” is purchased through suppressing dissent and driving out competent people who won’t play along is not visible (or relevant) to them. They think the problem is the people leaving, not the system that’s pushing them out.

I’ve seen one organization successfully recover from entrenched yes-men culture. The board replaced the CEO after a catastrophic product failure that could be directly traced to years of ignored technical warnings and toxic leadership culture. The new CEO immediately acknowledged the cultural problems, made it clear that dissent would be welcomed rather than punished, and backed that up with concrete actions: people who’d been sidelined for raising concerns got promoted, performance review criteria got rewritten to value challenge rather than compliance, decision-making processes got rebuilt to include diverse perspectives.

Though it took three years and significant turnover, some of the worst enablers left when the culture changed, and some of the best engineers who’d been driven out didn’t come back, but the organization recovered. The key was that change came from the top with genuine commitment and concrete follow-through. Half-measures don’t work. Lip service to “honest feedback” while maintaining all the structures that punish it accomplishes nothing.

But that’s one organization. I’ve watched at least three others continue their decline, lose their best people, make predictably bad decisions, and ultimately fail or get acquired at distressed valuations. In every case, the yes-men culture was a contributing factor to the failure, and in every case, the leadership never recognized it until too late.

What Might Actually Work

If you’re a leader trying to prevent or address yes-men culture, there are some patterns that seem to matter:

Make dissent visibly valuable. Not just saying you want feedback, but demonstratively rewarding people who provide it. Promote the person who challenged your bad idea (provided they meet the baseline criteria for advancement, obviously). Thank people publicly for raising difficult issues. Create explicit mechanisms where someone’s job is to play devil’s advocate; the “Tenth Man Rule” from Israeli intelligence,1 where one person is assigned to argue against consensus no matter how strong it seems. Make it clear through actions, not words, that challenge is more valuable than compliance.

Build structural accountability. Don’t rely on individual virtue. Create systems where bad decisions face scrutiny regardless of who made them. Independent technical reviews that can’t be overridden by executives. Board members with relevant expertise who can evaluate claims rather than just trusting leadership. External advisors whose compensation isn’t tied to agreement. Mechanisms that make it hard to suppress inconvenient information.

Accept that you’ll lose some people. Some of your current “high performers” are probably enablers whose performance is measured by political alignment rather than actual contribution. When you change the culture, they’ll leave or need to be pushed out. This is a feature, not a bug. You can’t fix the culture while keeping the people who benefit from or propagate its dysfunction.

Recognize it takes years. Cultural change is slow. You can’t announce new values and expect behavior to shift immediately. People need to see consistent follow-through over time before they’ll believe the change is real. That means sustained commitment even when it’s uncomfortable, even when it would be easier to slip back into old patterns.

But here’s the hard truth: most organizations won’t do this. The leaders who’ve benefited from yes-men culture won’t (want to) see it as a problem. The boards that should provide oversight are often too distant to recognize the pattern. And by the time the failure becomes visible, it’s usually too late to fix.

What Happened to the VP

The VP is still there. Year four now. Still in her role, still enabling bad decisions, still coaching new leaders on “how things work here.” The organization continues its decline, slower than it could be because there’s still some residual talent, but the trajectory is clear. The best engineers have left. The technical debt is compounding and product launches fewer and further between. The product roadmap is increasingly disconnected from market reality.

She probably can’t see this anymore. The perspective necessary to recognize organizational dysfunction has been warped by years of rationalization. She’s trapped in a system she helped build, and the bars of her prison are invisible to her. By the metrics that matter to her (maintaining her position, preserving/increasing her compensation, avoiding conflict with the CEO) she’s succeeding. By any objective measure of leadership or organizational health, she’s failing spectacularly. But objective measures don’t matter when you’ve optimized for the wrong things.

This is what yes-men culture actually costs. Not just bad decisions or political dysfunction, but the systematic destruction of an organization’s ability to correct itself. And the people caught in it, both the enablers and their victims, usually can’t escape without significant personal cost.

The Only Real Answer

If you’re reading this and recognizing patterns from your own organization, here’s what I’d tell you: trust your instincts. That pit in your stomach when you swallow objections, that cognitive dissonance between what you know and what you say, that slow erosion of your professional capability. Those are warnings. Your brain is telling you something is wrong.

The question isn’t whether the culture is broken. If you’re asking that question, it probably is. The question is what you’re going to do about it. Can you fix it? Unlikely, unless you’re the CEO or have board-level authority. Can you resist it without being destroyed? Possible, but exhausting and probably temporary. Or do you need to just leave before it damages you further?

For most people, the honest answer is the third option. Not because resistance is wrong, but because the system is stronger than individual courage. The machinery that creates and perpetuates yes-men culture is designed to eliminate exactly the kind of people who would resist it. Fighting that machinery is noble. It’s also usually futile.

The real victory isn’t reforming broken organizations. It’s recognizing them early, protecting yourself from their damage, and finding or building organizations that don’t operate that way. Those places exist. They’re harder to find, they often pay less, they sometimes move slower. But they’re where actual innovation happens, where capable people can do their best work, where professional integrity isn’t career-limiting.

The future doesn’t belong to organizations that suppress dissent for the comfort of consensus. It belongs to organizations that can engage with reality, even when reality is uncomfortable. If you’re in a yes-men culture, the best thing you can do is get out and find or build something better. That’s not giving up. That’s choosing to invest your energy where it might actually matter.

This is Part 3 of the Breaking the Echo Chamber series:

• Part 1: The Seduction of Yes-Men Culture

• Part 2: The Personal Cost of Sycophancy

• Part 3: The Organizational Cancer [you are here]

• Part 4: Breaking Free

By month twenty-four, the VP had stopped being a victim of the system and had become its engine. The transformation was complete. She wasn’t just enabling bad decisions anymore, she was actively recruiting others to do the same. When a new director of platform engineering started raising concerns about technical debt in his first month, the VP pulled him aside for what she called a “coaching conversation.”

I heard about this conversation afterward from the director, who was shaken by it. The VP’s message was clear: “You’re technically right about the debt, but you need to understand how things work here. The CEO doesn’t want to hear about problems right now. He wants solutions that support the roadmap. If you keep pushing on this, you’ll lose credibility. Better to build relationships first, show you’re a team player, then you can influence these decisions later.”

Everything she said was technically true. And completely corrupting. Because “building relationships” meant not challenging decisions. “Being a team player” meant enabling dysfunction. And “influencing later” was a fantasy; she’d been there two years and hadn’t influenced a single technical decision that mattered. She was teaching him the same rationalizations that had captured her.

This is how yes-men culture perpetuates itself. Not through explicit coercion but through experienced people teaching new people how to survive. Each generation of enablers recruits the next, passing on the same broken patterns while genuinely believing they’re providing valuable guidance.

How Systems Get Captured

The VP’s transformation didn’t just damage her capabilities. It damaged the organization’s immune system. She was supposed to be a check on bad decisions. Someone with the technical expertise and organizational standing to push back when needed. Instead, she’d become a mechanism for making bad decisions look reasonable.

Performance reviews became exercises in documenting alignment rather than impact. The VP rated people highly when they “collaborated effectively with leadership” and “demonstrated cultural fit.” Translation: they didn’t challenge the CEO’s decisions and they helped manage resistance from their teams. Engineers who raised technical concerns found their reviews suddenly full of vague criticisms about “communication style” or “executive presence.”

I watched one senior engineer, legitimately one of the best on the team, get rated as “needs improvement” because she’d pushed back on an architectural decision that was obviously wrong. The VP’s written feedback was a masterpiece of obfuscation: “While she demonstrates strong technical skills, she sometimes struggles to see the broader business context and can be inflexible when her recommendations aren’t immediately adopted.”

What actually happened: the engineer had identified a scalability bottleneck that would have cost the company millions to fix post-launch. The CEO didn’t want to hear it because it would delay his timeline. The engineer kept pushing. So the VP documented it as a performance issue.

This is how organizational immune systems fail. The people who should be identifying problems get classified as the problem themselves. Dissent gets pathologized as poor judgment or bad attitude. And the systems that are supposed to protect the organization (performance reviews, hiring decisions, promotion criteria) get repurposed to protect leadership from accountability.

The Spread of Dysfunction

The most insidious part wasn’t what the VP did directly. It was how her behavior shaped everyone else’s. Other leaders watched her get rewarded for enablement and learned the lesson. New hires watched her model “success” and internalized that model. Within eighteen months, the entire engineering leadership team had adopted similar patterns.

Team meetings became performance art. Everyone knew the technical debt was mounting. Everyone knew the slapdash architecture decisions were creating long-term vulnerabilities. Everyone knew the timelines were unanchored to reality. But nobody said it out loud anymore. Instead, they’d have sidebar conversations after meetings where they’d acknowledge the reality privately while maintaining the fiction publicly.

The director of platform engineering, the one the VP had “coached,” learned quickly. Within three months, he’d stopped raising concerns in meetings. Within six, he was actively helping rationalize bad decisions to his team. By month nine, he was coaching his own reports on “how things work here.” The pattern had replicated perfectly.

This created a cascade effect through the organization. If senior leadership was signaling that alignment mattered more than accuracy, mid-level managers optimized for alignment. If mid-level managers valued compliance over competence, individual contributors learned to stay quiet. Each layer reinforced the pattern in the layer below.

The talented engineers who valued technical integrity started leaving. Not all at once (that would have been obvious), but steadily, one every few months. Exit interviews revealed nothing useful because people had learned not to be honest with HR. They’d cite “new opportunities” or “career growth” while privately telling peers they couldn’t stand watching good engineering be sacrificed to bad leadership.

The engineers who stayed fell into two categories: those who adapted to the dysfunction and those who mentally checked out. The ones who adapted became the next generation of enablers. The ones who checked out did the minimum required to collect their paychecks while investing their actual energy elsewhere. Neither group was solving the technical problems that needed solving.

The Architecture of Institutional Failure

What made this particularly effective as a control system was how it operated through seemingly neutral organizational mechanisms. Performance reviews. Promotion criteria. Compensation decisions. Information access. None of these overtly required enablement, they just consistently rewarded it while punishing dissent.

The VP became the primary gatekeeper for promotion to senior engineering levels. Her criteria were nominally about technical leadership and business acumen, but in practice they measured political alignment. Engineers who “partnered well with leadership” got promoted. Engineers who “sometimes struggle to see broader context” didn’t. The fact that “partnering well” meant not challenging and “seeing broader context” meant understanding that technical concerns were less important than executive preferences—that was never stated explicitly. It didn’t need to be.

Information flow became another control mechanism. The VP held weekly “leadership syncs” where she’d share updates from her conversations with the CEO. Being included in these syncs was a mark of status. The information shared was often trivial, but access to it felt important. What actually mattered was the message: if you’re aligned, you’re in the room. If you challenge too much, you stop getting invited.

This created a powerful incentive structure where the rational move was to maintain access by maintaining alignment. Engineers who pushed back found themselves gradually excluded from information flow. They’d learn about decisions after they were made, get invited to meetings after the real discussions had happened elsewhere, find themselves unable to influence outcomes because they weren’t in the rooms where those outcomes got decided.

The cruelty was that this exclusion was never presented as punishment. It was framed as natural consequences. “We needed to move fast on that decision.” “I didn’t think you’d be interested in that particular project.” “Those meetings are just for the senior technical leads.” But the pattern was clear: challenge leadership and you become organizationally irrelevant.

The Point of No Return

By year three, the organizational culture had calcified around enablement. It wasn’t just that individual people had been captured, the entire system had been redesigned to make capture inevitable. New hires were onboarded into a culture where the rules were never stated explicitly but were absolutely clear: technical excellence matters less than executive alignment, competence is secondary to compliance, and speaking truth to power is career suicide.

The organization maintained elaborate fictions. The values statements still talked about “innovation,” “excellence,” and “honest feedback.” The performance review process still claimed to measure “impact” and “technical leadership.” The job postings still advertised “collaborative culture” and “engineering-driven decision making.” None of it was true, but the gap between stated values and lived reality had become so normalized that people stopped noticing the contradiction.

The VP herself, by this point, probably believed she was an effective leader. She’d successfully navigated a complex political environment. She’d maintained her position and compensation. She’d built a team that consistently delivered against the CEO’s roadmap, even if that roadmap was technically incoherent and long-term unsustainable. By the metrics that mattered to leadership, she was succeeding.

The fact that she’d systematically destroyed the organization’s ability to make good technical decisions, driven out its most capable engineers, and created a culture where honesty was punished and mediocrity was rewarded, that wasn’t visible to her anymore. The perspective necessary to see that had been warped by years of rationalization and reinforcement. She was trapped in the system she’d helped build, and she couldn’t see the bars of her own cage.

This is what yes-men culture costs at the organizational level. Not just bad decisions or political gamesmanship, but the systematic destruction of the organization’s capacity for self-correction. The mechanisms that should identify and fix problems get repurposed to identify and eliminate people who point out problems. And once that transformation is complete, the organization becomes incapable of responding to reality until reality forces the response through catastrophic failure.

Understanding how this system works is essential because it reveals where intervention might be possible. That’s what we’ll examine in Part 4: the strategies for breaking free, both as individuals trying to escape capture and as leaders trying to build organizations resistant to it.

Next in the series: Part 4: Breaking Free

This is Part 2 of the Breaking the Echo Chamber series:

• Part 1: The Seduction of Yes-Men Culture

• Part 2: The Personal Cost of Sycophancy [you are here]

• Part 3: The Organizational Cancer

• Part 4: Breaking Free

Happy New Year, everyone! Wishing all of you a healthy, happy, and prosperous 2025.

By month nine, the VP of Engineering I’d been watching had stopped contributing in technical architecture reviews. Not because she’d been excluded; she still had her seat at the table and commanded nominal respect. But her primary function had shifted from evaluating technical decisions to managing the room’s reaction to them. When a junior engineer raised concerns about the scalability of a proposed system, the VP didn’t engage with the technical merits whatsoever. She reframed the concern as “premature optimization” and redirected the conversation toward the CEO’s (completely arbitrary and ephemeral) timeline. Deftly done, like the practiced flick of a conductor’s baton.

I knew the VP had real expertise. She’d built systems at scale, understood the tradeoffs, could hold her own in any technical discussion. But she’d stopped using that expertise for technical judgment. Instead, she deployed it for political calculation: how to make the CEO’s preferred decision look technically sound, how to neutralize objections without appearing to suppress them, how to maintain the illusion of rigorous technical process while ensuring the predetermined outcome.

This is what yes-men culture costs at the individual level. Not just compromised integrity or political gamesmanship, but the systematic atrophy of the very capabilities that made someone valuable in the first place.

The Erosion of Expertise

Professional capability erodes through disuse like muscle memory fading. When you stop exercising judgment, you lose the ability to exercise it well. The VP’s technical skills hadn’t disappeared. She could still write code, still understand architecture diagrams, still follow technical arguments. But her technical *judgment* had weakened, grown soft. The muscle memory of evaluating technical tradeoffs, weighing risks, making decisions based on engineering principles rather than political considerations had atrophied month by month.

By month nine, she’d lost the ability to distinguish between good technical arguments and politically convenient ones. When someone raised a concern, her first thought wasn’t “is this valid?” but “how does this affect the CEO’s timeline?” That wasn’t a conscious choice anymore. It was reflexive, automatic, the result of months of conditioning. Like Pavlov’s dogs, but for organizational dysfunction.

What makes this particularly pernicious is that the person experiencing it often can’t see it happening. The VP didn’t think her judgment had deteriorated. She thought she’d gotten better at “seeing the bigger picture” and had an epiphanic “understanding how the business really works.” The fact that “seeing the bigger picture” actually meant ignoring technical reality, and “understanding the business” meant anticipating this specific CEO’s preferences, she couldn’t see that anymore. The transformation was complete.

The Psychology of Compromise

The cognitive dissonance of maintaining two versions of reality extracts ongoing cost. You know the technical decision is wrong. You also know you’re going to support it publicly. You develop elaborate internal narratives to reconcile these positions, but the reconciliation is never complete. There’s always a gap between what you know and what you say, and that gap requires constant energy to maintain. Exhausting, corrosive, secretly shameful energy.

I watched the VP develop increasingly complex rationalizations. “We need to fix that in the next iteration. Along with delivering everything else that was already planned.” “The business pressure justifies the technical debt.” “I’m choosing my battles. This isn’t the hill I’m going to die on.” Each rationalization was technically true but profoundly dishonest. Every iteration, they pushed the technical debt forward. Every battle, she chose not to fight. After months of this pattern, the truth became clear to everyone: she never fought any battles; not for budget, not for quality, not for people. The “strategic” silence had become a permanent posture.

This takes a psychological toll that’s hard to articulate to someone who hasn’t experienced it. You start monitoring yourself constantly—what can you say in meetings, what needs to stay internal, how to phrase disagreements so they don’t register as challenges. That level of self-surveillance is exhausting. It creates a baseline anxiety that affects everything, a constant low-grade dread that something you say might cross an invisible line.

But it’s more than just stress or burnout. What happens when you repeatedly act against your own judgment and values is something psychologists call “moral injury,” the damage to your sense of self that comes from sustained ethical compromise. You know the right answer. You say something different. You do this again and again, and each time you do it, you erode your own sense of integrity. It’s not just that you’re compromising your principles in the moment, you’re reshaping who you are. The person who could have stood up for what’s right becomes the person who learned not to.

Private conversations became the only place she could be honest. The VP would pull me aside after meetings to explain what she actually thought about the decisions that had just been made. These conversations were revealing not because they showed her private disagreement (that was obvious to anyone paying attention) but because they showed how much energy she was expending to maintain the gap between her public and private positions. That energy could have been used for actual technical leadership, for mentoring engineers, for solving hard problems. Instead, it was consumed entirely by the performance of alignment.

The Illusion of Security

The cruel part is how much the VP had convinced herself she was being strategic. She had a clear narrative, repeated often enough that she believed she was maintaining influence so she could guide decisions on issues that really mattered. The problem was that “issues that really mattered” kept getting defined downward, shrinking like the frog in slowly boiling water. First it was major architectural decisions. Then it was just infrastructure choices. Eventually it became “keeping the team from getting fired.”

This is the trap of incremental compromise. Each decision to stay silent feels justified because you’re preserving your position to fight another day. But another day never comes, and your position becomes increasingly dependent on not fighting. Ever.

By month twelve, the VP’s professional identity had fully shifted. She no longer saw herself as a technical leader who happened to work in a political environment. She saw herself as a political operator who happened to have technical background, a subtle but fundamental reframing. Her value wasn’t her engineering judgment, it was her ability to navigate the CEO’s moods, manage the team’s expectations, and translate objectively bad decisions into language that sounded reasonable.

This felt like success to her, though. She had access, perceived influence, a seat at the table where important decisions happened. What she couldn’t see: her influence was entirely contingent on not exercising it. The moment she pushed back on something the CEO actually cared about, that access would evaporate like morning fog. She wasn’t powerful. She was useful because she was silent. There’s a difference, and it’s not subtle.

The Actual Costs

Skills become non-transferable. The VP’s primary competency had become knowing how this specific CEO made decisions and how to work around them. Her value was tightly coupled to the CEO. That’s not a skill that travels. Outside this organization, she’d need to rebuild her technical credibility from scratch. Except her technical judgment had deteriorated enough that “from scratch” might not be an exaggeration. She’d need to relearn how to make decisions based on engineering principles rather than political calculation, and that’s harder than it sounds after a year of conditioning.

Professional network calcifies. Her relationships increasingly consisted of other people in similar positions: other insiders, other enablers, other people who operated the same way she did. This created a self-reinforcing bubble where her behavior seemed normal because everyone around her was doing the same thing. The broader professional community, people who valued technical judgment over political alignment, she’d lost connection to them. They’d stopped reaching out. She’d stopped maintaining those relationships because they weren’t useful in her current role.

Reputation becomes fixed. Word travels in tech, especially in specific domains and geographies. The VP was starting to be known not for her technical work but for her ability to manage up. That’s career poison. It limits your options to other organizations looking for that specific skill set, which is to say, other dysfunctional organizations where enablement is valued over expertise. She was becoming unemployable anywhere that actually valued engineering judgment.

Financial dependency locks in. The compensation was good. Stock was vesting. Retention bonuses were scheduled. Each month that passed made leaving more expensive in real terms. By year two, she’d need to take a significant pay cut to go anywhere else, and her lifestyle had adjusted to the current compensation. The golden handcuffs had worked exactly as designed. The mortgage, the kids’ private school, the lifestyle expectations, all of it created dependency that made the cage feel increasingly inescapable, however comfortable.

The Trapped Insider

By month eighteen, the VP wasn’t strategically positioning herself anymore. She was trapped, plain and simple. Her skills had deteriorated to the point where recovering them would take years. Her professional network consisted primarily of other enablers. Her reputation was that of a political operator rather than a technical leader. And her financial obligations made leaving prohibitively expensive unless she was willing to dramatically downgrade her lifestyle, which, having adjusted to the compensation, felt impossible, possibly catastrophic.

More importantly, she’d lost the ability to see the trap she was in. When I pointed out that she hadn’t challenged a technical decision in six months, her response was to explain why each individual decision hadn’t been worth fighting over. She couldn’t see the pattern anymore. Or maybe she could see it but couldn’t afford to admit it, even to herself, because admitting it would require confronting the choices she’d made and the position she’d put herself in. Most people aren’t that honest with themselves.

The question “how did I get here?” never has a satisfying answer because the journey happens in such small increments that no single step feels like the wrong one. It’s only when you look back over the full distance that the path becomes visible. And by then, you’re too far from the starting point to easily find your way back.

This is what yes-men culture costs at the individual level. Not just compromised integrity but the systematic destruction of professional capability, judgment, and agency. The people it captures often don’t recognize they’ve been captured until it’s too late to escape without significant cost.

But these individual costs aren’t isolated tragedies. They’re the mechanism through which organizational dysfunction perpetuates itself. When competent people become enablers, they don’t just fail to stop bad decisions, they become the machinery that makes those decisions possible and sustainable. That’s what we’ll examine in Part 3.

Next in the series: Part 3: The Organizational Cancer

This is Part 1 of the Dismantling the Echo Chamber series:

• Part 1: The Seduction of Yes-Men Culture [you are here]

• Part 2: The Personal Cost of Sycophancy

• Part 3: The Organizational Cancer

• Part 4: Breaking Free

I watched a VP of Engineering—technically competent, strategically sharp—transform into an enabler over six months. Not through corruption or coercion. Through a series of small accommodations that seemed reasonable in isolation. A technical concern downplayed in a meeting. A discriminatory comment from the CEO met with silence instead of pushback. A pattern of incompetence in the inner circle ignored rather than addressed.

By month six, she was actively defending decisions she privately acknowledged were indefensible. Not because she’d been seduced by power or bought with money, but because the organizational architecture had made enablement the only viable strategy for survival. The system didn’t break her through force. It shaped her through incentives so carefully calibrated that resistance felt irrational.

This pattern repeats with such regularity that it can be mapped, predicted, and, importantly, interrupted. But only if we examine how it actually works rather than how we’d prefer it to work.

The most dangerous lies in organizations aren’t the ones told to others, but the ones we tell ourselves to justify our compliance.

The Mechanism of Capture

Organizations don’t create enablers through explicit coercion. They create them through small invitations that feel like recognition. A casual lunch with the CEO. Inclusion in an unofficial “brain trust” Slack channel. Early access to strategic information. Each instance triggers something primitive, the deep need for belonging and status that predates professional judgment by several hundred thousand years.

What makes this effective is the cost-benefit calculation it forces. In healthy organizations, speaking up costs little because dissent is valued. In captured organizations, speaking up costs everything because the benefits (information access, decision-making inclusion, influence) all flow through proximity to leadership. Challenging leadership means risking not just disapproval but exclusion from the mechanisms through which work actually happens.

You can maintain your principles and become organizationally irrelevant, or compromise and remain effective. That’s not really a choice. It’s a mechanism of capture disguised as one.

The Progression: Rising Star to Trusted Insider

The transformation follows predictable stages. Understanding them matters because each represents a potential intervention point, a moment where the pattern might be interrupted.

Stage 1: The Initial Invitation

It starts with inclusion. Meetings slightly above your level. Access to information before your peers. Input on strategic questions. This feels like the seductive recognition of competence, and initially, it is. But it’s also the beginning of dependency. It wasn’t marijuana, it was heroin. That privileged access becomes something you don’t want to lose.

The VP went through this in her first month. Suddenly she was in the CEO’s office daily, included in board prep, consulted on hiring for other departments. It felt like validation. It was actually the establishment of a relationship where her value derived increasingly from proximity to power rather than technical judgment.

Stage 2: The First Compromises

The ethical descent rarely begins with major transgressions. It starts with small accommodations that seem reasonable, perhaps even quite innocent, in isolation. Staying quiet when the founder dismisses a junior engineer’s valid concern. Rationalizing why certain performance standards don’t apply to the CEO’s hand-picked team.

Each compromise makes the next one easier; it’s the fabled slippery slope. You’ve already demonstrated flexibility once. Doing it again feels less like a new moral choice and more like behavioral consistency. No big deal, you tell yourself.

I watched the VP stay silent when the CEO publicly mocked an engineer’s appearance in a meeting. Not agreement, just deafening silence. That was arguably worse than outright agreement. A month later, she was actively arguing that the engineer wasn’t “leadership material.” The progression from passive to active enablement took six weeks.

Stage 3: The Rationalization Cascade

As compromises accumulate, the cognitive dissonance becomes harder to manage. You can’t simultaneously believe you’re a person of integrity and acknowledge you’re enabling misconduct. So you develop elaborate justifications.

“I’m being strategic. Building credibility so I can push back on issues that really matter. I’m playing the long game. The time just hasn’t come to push back yet.”

“I’m protecting my team by staying in the room where decisions happen. It’s better we have a seat at the table than lose access altogether.”

“The CEO’s behavior is rough, but the mission justifies some tolerance. Everyone should toughen up a bit. Winning means we don’t have time for niceties sometimes.”

These aren’t lies exactly. They’re how the mind protects itself from recognizing you’ve made choices you can’t easily reverse. And here’s the most perverse part: people who make compromises for modest incentives experience stronger attitude change than those who compromise for large rewards. They need more psychological justification because they can’t point to the money or power as explanation. They have to convince themselves the compromise was right.

The VP developed an entire philosophy about “choosing your battles” and “playing the long game.” By month four, she’d chosen exactly zero battles and the long game consisted entirely of enabling the short-term (usually disastrous) decisions she privately opposed.

Stage 4: The Dependency Lock

The final stage is when the relationship becomes “structurally load-bearing.” Your position depends more on your relationship with leadership than your actual capabilities at all. Your credibility with peers is tied to your access to power. Your influence derives from navigating leadership’s labyrinthine preferences rather than your domain expertise.

At this point, challenging leadership threatens not just your standing but your entire professional identity as you’ve come to understand it. The psychological investment is so high that even recognizing the need for change becomes difficult. Dysfunction feels normal. Necessary, even.

By month six, the VP was arguing that the CEO’s erratic decision-making was “strategic flexibility.” That the discrimination was “communication style differences.” That protecting incompetent insiders was necessary to “maintain team cohesion and culture.” She’d moved beyond rationalization into reconstruction of reality to match the requirements of her true position.

Why Traditional Interventions Fail

Most organizational responses to yes-men culture focus on individual integrity. “Speak up.” “Be brave.” “Maintain your principles.” This is useless advice because it misunderstands the problem entirely.

The issue isn’t individual moral failure. It’s systematic architecture that makes enablement rational and resistance costly; so costly that most *cannot*. In environments where speaking up carries personal risk, people accurately assess that silence is safer. Exhorting them to be brave doesn’t change the underlying calculation, it just makes them feel bad about making the rational decision to protect themselves.

The VP I observed wasn’t weak or unprincipled. She was responding rationally to an incentive structure that rewarded compliance and punished dissent. Telling her to “be brave” would have accomplished nothing. What would have mattered is changing the structure so dissent was valued more than compliance. That would have required leadership to actually want dissent. They manifestly didn’t.

What Actually Drives the Pattern

Having watched this transformation multiple times now, certain mechanisms appear consistently:

Information as currency. Organizations control behavior by controlling access to information. Being “in the know” becomes a status marker and a source of influence. Challenging leadership risks losing that access, creating powerful incentive for alignment even when the information reveals problems.

Structural dependency. Modern organizations are complex enough that influence requires navigating multiple layers of hierarchy and informal power. Being able to “get things done” depends heavily on relationships with key decision-makers. Maintaining those relationships becomes more important than the nominal content of your role.

Social proof as control. When respected colleagues appear to accept questionable decisions, it creates permission for others to do the same. The pressure isn’t through explicit coercion, it’s subtle social reinforcement. Maybe you’re wrong. Maybe you’re not seeing the full picture. Maybe you’re just not sophisticated enough to understand the nuance. Organizational gaslighting.

Identity capture. The most effective control mechanism is when people internalize the organization’s narrative as their own identity. At that point, you’ve redefined your principles to match what the organization needs. The VP didn’t see herself as an enabler. She saw herself as a pragmatic leader who understood how to navigate complex environments. The fact that “navigating” meant “never challenging” wasn’t visible to her anymore.

This isn’t a complete model; organizational capture is too complex for simple frameworks. But these mechanisms appear consistently enough to be diagnostic. When you see information being weaponized, structural dependencies being exploited, social proof being manufactured, and identities being captured, you’re watching yes-men culture in construction.

In Part 2, we’ll examine what this costs the individuals who get captured: not just career damage or compromised integrity, but a fundamental erosion of professional capability and judgment that persists long after they leave the organization that shaped them.

Next in the series: Part 2: The Personal Cost of Sycophancy

A colleague reached out for advice recently. Senior technical leader, strong track record, hired into a company that seemed promising. Within weeks, he was watching the CEO casually violate federal regulations in meetings. When he raised concerns, HR started building a case to force him out. By the time he came to me, he’d been documenting for months, the retaliation was accelerating, and he was trying to figure out whether to fight or leave.

I’ve seen this pattern enough times now to recognize it immediately. Not organizational dysfunction that can be fixed through better processes. Systematic misconduct where the toxicity is architectural, where every mechanism designed to address problems has been captured to protect the people causing them. And where every assumption about how professional environments work (that organizations respect legal boundaries, that reporting mechanisms address problems, that HR protects employees) turns out to be wrong.

How Systematic Misconduct Actually Works

My colleague’s situation was textbook. The CEO treated regulations as suggestions. Federal compliance requirements? “I don’t think that applies here.” Not because he didn’t understand (he’d been told explicitly by multiple people, including legal counsel) but because he just didn’t care. In his mind, rules were for other people.

The discrimination was casual. The VP of Engineering—highly competent, didn’t fit the CEO’s image of what a tech leader should look like—faced constant mockery about his appearance and background. This happened in meetings. Regularly. Nobody stopped it because the CEO controlled everything including whether they kept their jobs.

HR wasn’t a neutral body. When my colleague raised concerns about legal violations and discriminatory behavior, HR began documenting everything he did, looking for pretexts. Meetings scheduled during pre-approved vacation. Added as a mandatory participant in irrelevant meetings. Work product increasingly scrutinized while the people actually responsible for problems faced nothing. The classic constructive discharge1 pattern. Create conditions so intolerable the person quits, so you don’t have to fire them and deal with legal exposure or severance.

The inner circle was protected regardless of conduct. Childhood friends of the CEO, early employees, people who’d been there from the beginning, they could be (and were) incompetent, even destructive, and face no consequences. Meanwhile, competent people who challenged decisions got systematically pushed out.

This, unfortunately, isn’t rare. I’ve advised enough people through similar situations, and have seen these firsthand, to recognize the pattern instantly. When leadership treats laws as discretionary, when discrimination is a management style, when HR operates as intelligence rather than support, you’re not dealing with fixable dysfunction; you’re dealing with a system that’s working exactly as designed, just not for the purpose you’d incorrectly inferred.

Why Traditional Reporting Fails Structurally

My colleague tried every internal channel. Direct conversations with the CEO. Documentation to HR. Formal complaints about retaliation and legal violations. None of it mattered. The people responsible for addressing misconduct were either committing it themselves or protecting those who were.

This is the structural problem with reporting mechanisms. They’re designed on the assumption that misconduct is aberrant, that leadership acts in good faith, that the problem is isolated bad actors. When leadership is the bad actor (when the CEO is violating law, when HR reports to executives who are complicit, when the board provides no oversight) internal reporting accomplishes nothing except identifying you as someone who needs to be managed out.

The 2023 Global Business Ethics Survey confirms this. 72% of people who observed misconduct reported it. But 46% of those who reported faced retaliation.2 Nearly half. It’s seductive to think that this is an implementation failure, but that’s the expected outcome when you report problems in organizations where leadership is the problem.

The mechanisms exist to protect the organization from legal liability, not employees from harm. When someone reports misconduct, the organization now has information about a potential lawsuit. HR’s job becomes managing that risk, which usually means building a case to terminate the reporter for “performance issues” or “cultural fit” before they can file anything.

I watched this happen with my colleague in real time. Once he raised concerns, everything changed. Suddenly his work wasn’t meeting expectations. He stopped being invited to budget meetings scheduled randomly for Saturdays and Sundays. He wasn’t demonstrating leadership. He wasn’t aligned with company culture. These were manufactured justifications for what the organization was going to do anyway. They were establishing a casus belli for termination.

What Actually Works

When my colleague came to me, the first thing I told him: consult an employment attorney immediately. Not when things get worse, but now. Many offer free consultations. You need to understand your legal protections, what qualifies as protected activity, how to document effectively. Everything else depends on understanding your legal position.

He’d already been documenting, but not strategically. I advised him to keep detailed records on personal devices only, never company systems. Make them factual and objective. And in New York, where he was located, a single-party consent state, record every conversation with HR and leadership. Not to prevent retaliation, but to have evidence when it happens.

This proved essential. HR tried repeatedly to get him to say things they could spin as insubordination or poor performance. The recordings showed what they were actually doing, building a case for constructive discharge. Without that evidence, it would have been his word against theirs, and organizations always believe leadership over employees.

But documentation doesn’t prevent harm. It just gives you leverage later. The retaliation still happens. You still get targeted. The misconduct still continues. You just have proof for legal action if you choose to pursue it.

I also told him to continue performing professionally. This feels absurd when you’re dealing with systematic misconduct, but it’s essential. Toxic organizations will use any pretext to claim you’re failing in your role. Don’t give them ammunition. Maintain performance so if they escalate, it creates more evidence of retaliation rather than legitimate termination.

And most critically: protect yourself financially and start planning your exit. Build an emergency fund. Update your network. Talk to recruiters. You’re leaving one way or another; either you choose when, or they force it. Better to control the timing.

The Cost

My colleague stayed six months after recognizing the pattern was systematic. Six months trying to document while the organization retaliated. Six months of HR building their case. Six months of manufactured performance concerns designed to justify forcing him out.

The financial cost was significant: legal consultations, lost salary, the gap between jobs. The professional cost was harder to measure, as resume gaps, difficult interview explanations, relationships damaged with people who chose silence over risk.

But the psychological cost was what he struggled with most. Twenty-plus years building a model of how professional environments work. Experience with dysfunction, politics, difficult people. Then encountering something operating under completely different rules, where laws are optional, employees are disposable, reporting makes you the problem, and every framework he’d built becomes useless.

That realization changes how you think about work. Makes you suspicious of organizational promises, skeptical of policies, aware that power matters more than ethics. That’s not damage you recover from quickly.

He eventually left. The bad actors are still there, still operating the same way. That’s usually how it ends. The bad actors stay in power. The people who reported problems move on. The organization continues.

What The Statistics Actually Mean

The Global Business Ethics Survey says 65% of employees observe misconduct, 46% of reporters face retaliation, and only 13% work in strongly ethical cultures.3 These aren’t measuring edge cases. They’re describing normal organizational reality.

But they understate the problem because they measure incidents, not systems. They don’t distinguish between organizations where misconduct is aberrant (and fixable through better enforcement) and organizations where misconduct is structural (and only fixable through leadership replacement).

In the first type, the statistics suggest solutions: better training, stronger protections, increased accountability. In the second type, those solutions are useless. The systems that should provide protection have been subverted to enable harm instead.

My colleague learned this at great cost. The organization had all the right policies: employee handbook, compliance hotlines, whistleblower protections, anti-retaliation policies. None of it mattered because the people who controlled those mechanisms were the ones benefiting from the misconduct. You’re asking them to investigate and sanction themselves. They won’t.

The challenge for employees is figuring out which type of organization they’re in before investing energy in internal remedies that won’t work. Internal reporting in a systemically toxic environment doesn’t fix problems, it just identifies you as a problem to be solved.

What Leaders Should Actually Do

For leaders genuinely interested in building ethical cultures, not just policies that provide legal cover, the implications are clear.

Recognize that reporting mechanisms only work when employees trust they won’t face retaliation. That trust can’t be mandated by policy. It has to be earned through consistent response: when someone reports a problem, does it get addressed or does that person become a target?

Understand that ethical culture can’t be built from the middle. If executives are the problem, no amount of training or policy enforcement at lower levels matters. Board oversight has to be real, not performative. Independent investigation has to mean actually independent, with consequences that apply to everyone including leadership.

Most critically: accept that some organizations can’t be reformed from within. When misconduct is intrinsic, when it serves leadership’s interests, when the systems designed to constrain it have been captured, the only solution is external intervention or leadership overhaul. Internal reform efforts just identify reformers for removal.

The Real Problem

The systems we’ve built to address workplace misconduct operate on the assumption that organizations want to prevent misconduct and just need better tools. That assumption is often wrong.

Many organizations want to minimize legal liability from misconduct, which is different from preventing it. They want plausible deniability, documentation that they had policies in place, mechanisms to identify potential lawsuits early so they can be settled quietly.

This creates systems that look like accountability but function as risk management. Employees report problems thinking they’ll be addressed. Instead, they’ve identified themselves as legal risks to be managed out. The organization protects itself, not from misconduct, but from the legal exposure that reporting misconduct creates.

My colleague experienced this directly. The reporting mechanisms worked exactly as designed, they identified someone who might create legal liability and triggered processes to eliminate that risk. The fact that he was reporting genuine misconduct, that the CEO was actually breaking laws, that people were being discriminated against, none of that mattered. What mattered was managing the legal risk he represented.

Until we’re honest about what these systems actually do versus what they claim to do, the statistics won’t improve. We’ll keep building mechanisms that protect organizations from employees while pretending they protect employees from harm. And people who encounter truly toxic environments will keep learning, expensively and painfully, that the systems they thought would help them are designed to do something else entirely.

That’s the reality beyond the handbook. Everything else is just policy documents that exist to limit liability, not to create ethical workplaces.

Paul Graham’s recent piece on “founder mode” has people nodding along like he discovered something new. He didn’t. What he described, founders who stay deeply engaged with their companies rather than delegating everything to direct reports, has always existed. The reason it’s getting attention now is because the alternative, what Graham calls “manager mode,” has produced so many mediocre outcomes that people are looking for permission to do something different.

I’ve run the full spectrum. Founder of my own company for four years. Engineering manager at Etsy learning participative leadership. VP at Shutterstock managing 105+ people across continents. CTO at multiple companies, one catastrophically toxic. I’ve seen founder mode work brilliantly. I’ve seen it become a weapon for incompetent narcissists to justify their inability to delegate.

The nuance is non-trivial and significant. Graham’s essay treats “founder mode” as a monolithic concept. It’s not. It’s a set of tools that work in some contexts and fail catastrophically in others. The question isn’t whether to operate in founder mode. It’s when founder mode serves the organization versus when it serves the founder’s ego.

What Founder Mode Actually Is

Graham describes Steve Jobs holding annual retreats with Apple’s 100 most important people, regardless of org chart position. Skip-level meetings becoming normal rather than exceptional. Founders maintaining direct engagement across the organization instead of operating only through VPs.

This works when the founder has genuine insight into the business and the judgment to know where their engagement adds value. Jobs knew product and design at a level his VPs didn’t. His skip-levels weren’t theater, they were how he maintained direct connection to the work that mattered most. That’s founder mode as a devastatingly effective feature.

But I’ve also seen founder mode as a bug. A CEO who insisted on being in every product decision, every engineering standup, every customer call, not because he had unique insight or deep curiosity but because he fundamentally didn’t trust anyone else’s judgment. Including people who were objectively more competent than he was in their domains. This was control expressed as bumbling interference.

The difference isn’t subtle. One creates alignment and speeds up decision-making. The other creates bottlenecks and organizational paralysis. One empowers people by giving them direct access to leadership context. The other disempowers them by making everything contingent on a single person’s approval.

When I Was In Founder Mode

Running mohchi from 2011 to 2015, I was necessarily in founder mode. Six people, limited runway, moving fast. Of course I was in every decision. There was no organizational structure to delegate through. The question wasn’t whether to stay engaged, it was whether I was engaged in the right things.

Some of that engagement was productive. I understood the business model inside and out. I knew our technical constraints. I could make product decisions quickly because I had full context. That’s founder mode working as designed.

Some of it was destructive. I made unilateral decisions about team processes without involving my co-founder. I changed direction frequently because I was learning in real-time and didn’t have the discipline to communicate why. Looking back, I was operating in “founder mode” the way a lot of founders do, conflating speed and brute force with effectiveness, conviction with correctness. The startup didn’t fail because of this, but it certainly didn’t help.

What I didn’t realize at the time was that I was bringing the command-and-control patterns from banking into a startup context and calling it “founder mode.” The real problem wasn’t the mode, it was that I hadn’t learned participative leadership yet. That came at Etsy, where I discovered that involving people in decisions didn’t slow things down, it, perhaps unintuitively at the time, made them better. But that learning happened because I was forced to operate differently. If I’d stayed in founder mode at mohchi, I would have likely kept making the same mistakes (though my co-founder would, and did, put a stop to my worst offenses—thanks, 🙏🏻).

When Founder Mode Becomes Pathological

Graham acknowledges the risk: “Founders who are unable to delegate even things they should will use founder mode as the excuse.” This is understated. I’ve seen founder mode become a cover for pathological behavior.

The CEO who can’t distinguish between domains where his intuition is valuable and domains where he’s just wrong, but insists on being involved in everything because he’s the founder. The one who treats every VP as a threat to his authority rather than a competent leader he hired to own a function. The one who holds skip-level meetings not to gather information but to undermine his directs by going around them.

It’s incompetence with a fashionable narrative, regardless of attempts to spin it as something else. But the language of “founder mode” gives it legitimacy. “I’m staying engaged across the organization” sounds better than “I don’t trust anyone and I’m a bottleneck.” The pathology gets rationalized as hands-on leadership.

I watched this destroy a company recently. Founder CEO, convinced of his own brilliance, in every decision at every level, in every discipline. Product, engineering, data, sales, marketing, operations, customer success, brand, finance, HR, legal—he knew better than everyone; a “true” (self-witnessed) polymath. The domains he admittedly excelled in were self-branding (“I want one of the company’s top goals this year to be about *my* personal brand.”) and giveaways of luxury items for low-yield customer acquisition. He was profoundly incompetent in most domains but was utterly incapable of seeing it. His “founder mode” meant constant direction changes, arbitrary decisions, and systematic undermining of anyone competent enough to threaten his self-image. That’s nothing but dysfunction.

The organization couldn’t function because everything required his approval, but his approval was contingent on whim rather than strategy. People couldn’t make decisions autonomously because the founder might reverse them without explanation. The “skip-level” meetings he insisted on weren’t about gathering information, they were about demonstrating his authority and identifying people to blame when things went wrong. Documenting mercurial decisions in order to structurally impose accountability when inevitable contradictions were made without notice also held no purchase.

That’s the dark, bizarro, version of founder mode. And the problem is, from the inside, the incompetent founder thinks he’s doing exactly what Steve Jobs did. He’s engaged. He’s involved. He’s maintaining the culture. What he can’t see, what the Dunning-Kruger effect prevents him from seeing, is that Jobs had judgment and he doesn’t.

What Actually Matters

The useful distinction isn’t between “founder mode” and “manager mode.” It’s between competent leadership and incompetent leadership. Between leaders who know when their engagement adds value and leaders who can’t tell the difference.

Competent leaders in founder mode are selective. They stay deeply engaged in the domains where they have unique insight or where alignment is critical. They delegate everything else, actually delegate it, not just nominally. They use skip-levels to gather information and build trust, not to undermine their VPs. They maintain high standards but trust their leaders to meet them.

Incompetent leaders in founder mode are indiscriminate. They’re in everything because they can’t distinguish between what matters and what doesn’t. They skip-level not to gather information but to maintain control. They confuse presence with value. They think delegation is weakness rather than recognizing it as the only way to scale.

The same behaviors (skip-levels, direct engagement, staying close to the work) produce completely different outcomes depending on the leader’s competence and intent. Which means “founder mode” isn’t a methodology you can adopt. It’s a description of how some effective founders operate, and an excuse for how some ineffective founders justify their inability to scale.

What I Learned At Scale

At Shutterstock, managing 105+ people across multiple continents, I couldn’t be in founder mode even if I’d wanted to. The organization was too large, too distributed, too complex. My job was to create systems that worked without my constant involvement. Clear frameworks, consistent processes, strong leaders who could operate autonomously, yet always be assured of my air cover.

But I still did skip-levels. Regular 1-1s with people three levels down. Not to micromanage, but to gather signal about whether the systems were working, whether my directs were effective, whether people felt heard. This wasn’t founder mode. It was maintaining connection to the organization without creating myself as a bottleneck.

The difference is intent and constraint. I wasn’t trying to be involved in every decision. I didn’t need to; I trusted my leaders. I was trying to understand whether the organization’s architecture was sound. When I found problems, I worked through my directs to fix them, not around them. That’s the fundamental distinction. Founder mode as Graham describes it risks becoming a bypass mechanism, going around your VPs when you should be either coaching them or replacing them.

At Candid, as CTO, I was back in something closer to founder mode, smaller team, earlier stage, more direct engagement with the work. But by then I’d learned the difference between productive engagement and controlling behavior. I stayed close to product engineering and manufacturing systems because those were domains where my judgment mattered. I delegated platform engineering and security completely because my Director was better at them than I was.

The question isn’t whether to operate in founder mode. It’s whether you have the judgment to know where your engagement creates value versus where it creates friction. And whether you have the humility to recognize when someone else’s judgment is better than yours.

What This Means

Graham’s essay is less a discovery than a permission slip. Founders who’ve been told they need to “scale themselves” by delegating everything now have language to justify staying engaged. That’s fine if they’re competent. It’s dangerous if they’re not.

The real insight isn’t that founder mode exists. It’s that there’s no universal management methodology that works at every stage, in every context, for every leader, for every personality. The pathology is thinking there is, whether that’s “delegate everything” or “stay engaged in everything.”

What matters is accurate diagnosis of where you add value and where you don’t. Where your intuition is reliable and where it’s not. Where your engagement accelerates the team and where it constrains them. Most leaders are bad at this self-assessment. The best ones are constantly questioning it.

Founder mode isn’t a solution. It’s a description of how some effective leaders operate in some contexts. The danger is treating it as a prescription; another management framework to adopt wholesale rather than a set of tools to deploy selectively based on judgment and context.

I’ve been on both sides. Founder who stayed too engaged in things I shouldn’t have. Leader at scale who had to build systems that worked without me. The lesson from both: the mode doesn’t matter. The judgment does.

Over twenty years building and fixing organizations, and I thought I understood the failure modes. Technical incompetence. Political dysfunction. The occasional ethical lapse. Then I encountered something I’d never seen before: a system where the toxicity wasn’t a bug. It was the architecture.

I should have recognized it earlier. The signals were there. I rationalized them because I believed, wrongly, that good process could constrain bad actors. That frameworks could contain malice. That you could debug a system that was designed to be corrupt. You can’t. When the kernel is compromised, everything running on top of it is compromised too.

This is what I learned the hard way.

The Fundamental Mistake

For most of my career, I’ve thought about organizational problems as systems failures. Build better structures, implement better processes, create better incentives. This works when leadership is fundamentally well-intentioned but inexperienced or misguided. The problem is a mismatch between intent and execution. Fix the execution, solve the problem.

But some organizations don’t have a mismatch between intent and execution. The execution perfectly reflects the intent. The toxicity isn’t a failure of the system, it’s the successful operation of a system designed to extract value for leadership at everyone else’s expense. Employees aren’t stakeholders to be developed. They’re resources to be depleted. This is a different kind of problem entirely, and all your frameworks become useless when you encounter it.

I used to think that was hyperbole. It’s not. It’s diagnostic reality.

What a Corrupted System Actually Looks Like

Some patterns took time to recognize because they operated at a level I’d never had to think about before. In healthy organizations, even normatively dysfunctional ones, there’s a baseline assumption that laws are constraints, that employees have rights, that governance structures exist for a reason. Remove those assumptions and you’re operating in an alternate dimension.

The inner circle problem was obvious early but I misread its significance. Certain people (founders, childhood friends, whoever had been there longest) operated under different rules. They could be, and were, gloriously incompetent. They could be actively destructive to the business. They repeatedly violated policies that would get anyone else terminated. None of it mattered. They were untouchable.

This isn’t just unfair. It’s how you know the organization’s stated values are fiction. When competent people who challenge decisions or advocate for employees get systematically pushed out while incompetent loyalists remain protected, you’re seeing the real incentive structure. Loyalty to leadership trumps everything: competence, ethics, performance, the actual success of the business. Once you understand this, every other dysfunction makes sense. It’s not a broken system trying to function. It’s a functional system serving interests you didn’t realize were primary.

Accountability flowing in only one direction took longer to see as a pattern. Decisions would be made, often over explicit objections, and when they predictably failed, the people who’d objected would be blamed for the failure. Not in a casual “well someone has to take responsibility” way. Systematically. With consequences. Meanwhile, the people who’d made the decisions faced nothing.

This inverts the normal problem-solving process. In healthy organizations, failures generate information. You do post-mortems, identify root causes, fix systems. Here, failures generated scapegoats. The goal wasn’t learning, it was protection. Leadership couldn’t be wrong, so someone else had to be responsible and made an example of. This makes honest feedback impossible. The moment you identify a problem accurately, you’ve marked yourself as the “true” problem.

The mercurial decision-making was something else entirely. I’ve worked with indecisive leaders. I’ve worked with leaders who changed their minds as they learned. This was neither. Direction would shift arbitrarily. Commitments made one week would be abandoned the next without explanation or acknowledgment. Goalposts would move mid-flight, not because priorities changed, but because moving the goalposts was itself the point. It kept everyone off-balance, scrambling, unable to build anything stable.

This is a control mechanism disguised, or misinterpreted as disorganization. When nobody can predict what leadership wants or whether today’s mandate will still be valid tomorrow, everyone becomes dependent on leadership’s real-time interpretation of reality. You can’t operate autonomously because the ground is quicksand. Some leaders do this deliberately. Others are just so incompetent that the effect is the same. Either way, teams can’t function.

Then there’s HR. In every organization I’d worked at before, HR had at least tried, insofar as incentives allowed, to mediate between employee welfare and business interests. The competence varied, but the intent was recognizable. Here, HR was intelligence. When they started having conversations like “I know this isn’t what you signed up for, so it’s okay if you want to leave,” they weren’t expressing sympathy. They were delivering a message: comply or be pushed out. Anything you told them would be used to build a case, not to address your concerns.

This transforms HR from a neutral function into an active threat. You can’t report problems through official channels because the channels exist to identify and eliminate people who report problems. Once you recognize this, you understand why toxic organizations stay toxic. The mechanisms that should constrain bad behavior have been captured to enable it.

Work-life boundaries became a compliance test. Meetings scheduled during pre-approved vacation. Surveys sent on weekends with next-day deadlines. Expectations of immediate availability regardless of time or circumstance. Some leaders do this because they’re workaholics who don’t know how to separate work from life. That’s bad enough. Worse is when leaders claim they respect boundaries but systematically punish anyone who actually uses them. Reduced responsibilities, blocked advancement, manufactured pretexts for termination. The second version is more insidious because it’s duplicitous. At least openly demanding leaders are honest about expectations.

But here’s what broke my model entirely: the relationship to law itself. I’ve worked with leaders who bent rules, who operated in gray areas, who pushed boundaries. All within the law. This was different. Regulations simply didn’t apply. When you tell someone that something violates federal law and their response is to argue about whether the law is actually relevant to their specific situation (it was), not as a legal strategy but as a genuine belief, you’re dealing with someone operating under assumptions you didn’t know were possible in professional contexts.

This creates liability for everyone around them. They make decisions that expose the company and its employees to legal risk, and they do it with confidence because they genuinely don’t believe the rules apply. I wouldn’t believe this behavior existed in the professional world if I hadn’t watched it happen. The certainty with which some leaders dismiss legal constraints would be impressive if it weren’t so dangerous.

Governance became performance art. Situations misrepresented to boards and investors. Company funds and personal expenses deliberately blurred. Decisions made that benefited leadership at the company’s expense, presented as sound business judgment. The fact that investors sometimes allow this doesn’t make it okay, it just means you’re operating in an environment where oversight has failed at multiple levels and has, in effect if not intent, become complicit. Everyone who should be providing checks has abdicated responsibility for profits and/or expedience.

But here’s what I’d never encountered before, not in over twenty years: genuine, breathtaking incompetence at the leadership level, paired with an equally breathtaking, absolute conviction of brilliance. The Dunning-Kruger1 effect is usually a joke. Leaders who couldn’t understand basic business concepts but were utterly certain they were geniuses. Who made obviously irrational decisions and punished anyone who questioned them. Who genuinely lacked the cognitive capacity to recognize their own incompetence.

This is different from “ordinary” arrogance. Arrogant leaders can at least be reasoned with if you present evidence clearly enough. This was cognitive incapacity. You can’t reason with someone who lacks the mental framework to even understand why they’re wrong. Every conversation becomes circular. Every attempt to provide feedback gets reinterpreted through a lens that confirms their brilliance and your obvious disloyalty. It’s like trying to debug code written in a language that doesn’t have logic operators.

I’d worked for bad leaders before. Leaders who were political, inexperienced, occasionally unethical. But they’d all operated within a familiar spectrum of organizational dysfunction and, without exception, within the law. There were still shared assumptions about how businesses work, how people should be treated, what laws mean. This was outside that range entirely. The rules I thought governed professional behavior simply didn’t apply. Laws were optional. Employees were disposable. The business existed to serve leadership’s personal interests, and everything else (product, customers, employees) was instrumental to that purpose.

What I Did Wrong

Due diligence. I did some. Not enough. I researched the company, talked to the team, requested documents. What I got should have been disqualifying. I rationalized it. The opportunity seemed good on paper. My team seemed strong. I thought I could influence the situation from the inside. In hindsight, that was hubris.

The documents were revealing if you knew how to read them. I knew how to read them. I just didn’t want to believe what they were telling me. When you see financial structures that exist primarily to enrich founders regardless of business performance, when you see governance that provides no real oversight, when you see a track record that doesn’t match the narrative, those aren’t yellow flags. Those are air raid sirens.

I talked to people who would report to me. They were smart, aligned with my principles, frustrated with current leadership. I thought: good team, bad CEO, I can be a buffer. What I didn’t understand was that you can’t buffer people from a system that’s toxic by design. You can maybe delay the damage, shield the team from the worst of it. You can’t prevent the rot from intruding. Eventually the corruption reaches everyone, including you.

I had instincts that something was wrong. Multiple conversations where things didn’t add up. Red flags in the interview process. Defensiveness about reasonable questions. I talked myself out of them because I believed, stupidly, that my experience and frameworks would be sufficient. That I’d built enough organizations, fixed enough problems, that I could handle this. I was more wrong than I could possibly have imagined.

The belief that logic and good process can constrain bad actors is a form of professional naivety that gets rebutted painfully. Some people can’t be constrained by process because they don’t acknowledge that process exists or applies to them. Some systems can’t be fixed because the dysfunction is intentional.

What You Should Do Differently

Research leadership not just thoroughly but correctly. Don’t just look at LinkedIn and company materials. Look for their actual track record. If they claim extensive experience but have no verifiable accomplishments, that’s data. If former employees uniformly describe the culture as toxic, believe them. If current employees are carefully evasive when you ask direct questions, that’s signal too.

Request documents and actually read them. For senior roles, cap tables, board decks, financial information matter. Not just to verify the business is sound, but to verify the governance structure isn’t designed for extraction. If they refuse to provide them or give you sanitized versions, don’t join. Companies hiding their governance structures are hiding them for a reason.

Talk to multiple people at different levels. Not just people who report to you, but people across the organization. Watch for fear, excessive deference, careful language that suggests people can’t speak freely. Ask about turnover. Ask why people left. Ask how decisions get made. Listen for what’s not being said as much as what is.

But here’s the hard truth: even with perfect due diligence, you can miss things. Toxic leaders are often skilled at impression management. The real dysfunction only becomes visible once you’re inside and see how decisions actually get made, how conflicts actually get resolved, what happens to people who challenge leadership.

Which means you need to know what to do when you realize you’ve made a mistake. Document everything. In single-party consent states, record conversations with HR and leadership. Keep contemporaneous logs of problematic incidents. Don’t conflate this with paranoia. This is protection. When things escalate legally, and they might, documentation is your defense.

Be cautious with HR to the point of paranoia. In toxic environments, HR is a proxy for leadership. Consult an employment attorney before making protected complaints. Understand your legal protections before you trigger them. I learned this early enough to protect myself. Not everyone does.

Build support networks outside the company immediately. Don’t wait until things get bad. You need people you can talk to honestly who aren’t embedded in the system. Start looking for your next role the moment you realize the environment is toxic. Don’t wait for it to get better. When toxicity starts at the top, it doesn’t get better. It metastasizes.

If you’re a manager, protect your team as much as possible. Create a buffer. But recognize the limits. You can’t fix systemic dysfunction from the middle. I tried. The system is stronger than you are. At some point, protecting yourself means leaving. I stayed longer than I should have. That was a mistake. The longer you stay, the more it costs financially, professionally, and psychologically.

What This Actually Cost

Over twenty years in tech. Goldman Sachs, Deutsche Bank, startups, Etsy, Shutterstock, Candid. I thought I understood organizational pathology. I’d seen incompetence, politics, occasional ethical failures. But all within the normal range of organizational dysfunction underscored by basic competence. Everyone was still operating under the same basic assumptions about how businesses work, how people should be treated, what laws mean.

When you encounter a system where those assumptions don’t apply (where laws are optional, employees are pawns, governance is theater) all your frameworks become useless. You can’t process-design your way out of purposeful malice. You can’t create guardrails that constrain someone who doesn’t acknowledge that guardrails apply to them. The belief that you can is professional naivety.

The cost isn’t just the time. It’s the damage to your model of how organizations work. To your faith in human nature. I built a career on the assumption that good systems constrain bad behavior, that frameworks create fairness, that you can architect your way to health. That’s mostly true. But it breaks down completely when leadership is fundamentally corrupt. When the system isn’t failing, it’s succeeding at objectives you didn’t realize were primary.

Some systems can’t be fixed. Some leaders can’t be reasoned with. Some organizations exist to serve interests that are fundamentally at odds with employee welfare or business success. Recognizing this early and leaving isn’t failure. It’s understanding pathological organizational patterns. The only winning move is not to play.

That’s the lesson. Everything else is details.

When I joined Shutterstock in 2017, I discovered that engineers with years of proven, high-impact work (people who’d shipped critical features, who understood the systems intimately) had to run the full external hiring gauntlet just to transfer to another team. Full-day interview loop. Whiteboard algorithms. The whole theater.

This wasn’t some edge case. It was policy. And it was insane.

More than that, though, it was a declaration of institutional distrust. Your track record, your context, your demonstrated value? Worthless the moment you wanted to move from Payments to Editor, or from Contributor Experience to Enterprise. The organization was treating its own people like strangers. Worse than strangers. Strangers at least get the benefit of the doubt. Internal candidates got skepticism.

This is depressingly common. It’s organizational self-harm disguised as rigor.

What This Actually Reveals

Requiring a full re-interview of someone you already employ is a diagnostic. It tells you several things about the organization, none of them good:

• Teams don’t trust each other’s judgment. Team A doesn’t believe Team B’s assessment of an engineer is valid. Maybe Team B has low standards. Maybe they promote people for the wrong reasons. Maybe they just assess differently. Whatever the reason, there’s no federated trust. You’re not one engineering organization, you’re a collection of suspicious fiefdoms with internal borders.

• There’s no calibrated standard for “good.” Without an organization-wide leveling framework and a consistent interview process, every team’s assessment is proprietary. Non-transferable. Each team is running its own private credentialing system, and nobody else accepts the credentials.

• The organization can’t adapt. When moving talent internally is harder than hiring externally, you’ve created a brittle, static structure. Talent gets trapped in one area while another starves. Business priorities shift, but people can’t flow to where they’re needed because the friction is too high.

• But here’s the core pathology: This practice is institutional contempt. It tells every employee that their contribution, their context, the trust they’ve built, none of it matters. You’ve spent three years shipping critical projects? Great. Now do a whiteboard algorithm so we can verify you know how to reverse a linked list. This isn’t a tax on morale. It’s an insult to professional identity. And it’s why your best people leave for companies that actually value their track record.

Imagine an engineer who has successfully shipped multiple critical projects over three years. Forcing them through a full-day interview, including basic algorithmic challenges, isn’t just a waste of time, it’s an act of institutional contempt.

What We Did: Engineering for a Borderless Organization

Fixing this required deliberate organizational re-engineering. The goal: create a single, trusted currency for talent assessment. Make every team’s judgment portable across the organization.

We built a standardized technical interview process and a robust leveling framework; transparent, consistently applied, calibrated across all teams. This wasn’t about bureaucracy. It was about writing down the rules so the game was fair.

The structural pieces:

• A consistent set of interview domains (system design, domain expertise, behavioral competencies) led by functional representatives. Every interview, regardless of team, assessed against the same core competencies.

• A codified career ladder with explicit behavioral expectations and impact definitions for each level. The difference between Senior and Staff wasn’t opinion anymore, it was documented, ratified standard.

• Mandatory pre- and post-interview huddles. Brief, structured meetings where the panel aligned on what they were looking for and calibrated their assessments before making a decision. This forced consistency.

• Comprehensive documentation for everyone: interviewers, candidates, hiring managers. Not red tape. Clarity.

For internal transfers: The loop focused on what was actually new—team-specific context, domain knowledge, mutual fit. It respected the engineer’s existing record. If you’d already proven you could code at a Senior level, we didn’t make you prove it again on a whiteboard.

Some team leads resisted. Predictable. They’d been operating autonomously, and this felt like a loss of control. We overcame it by involving them directly in the design and by showing them, with data, that a calibrated process would give them better hiring outcomes. Not just better for the organization, but better for their teams.

What This Actually Is: Lessons for the System Architect

Leadership is systems architecture. You’re engineering the conditions under which people can do their best work. That requires diagnosing pathologies and fixing them at the root.

Some takeaways:

• Involvement neutralizes resistance. Engage stakeholders in designing the new system. Co-creation is more effective than mandate. People will accept constraints they helped create.

• Structure enables agility, not prevents it. A standardized system isn’t a cage if you build in appropriate flexibility. Allow for team-specific needs, but never compromise the core principles: fairness, calibration, portability.

• Systems are products. They need iteration, feedback loops, continuous improvement. The first version won’t be perfect. That’s fine. Ship it, learn, improve.

• Narrative matters as much as mechanics. The best-designed system fails if people don’t understand why it exists. Explain the benefits, address the fears, show the positive outcomes. You’re not just implementing process, you’re changing culture.

As organizations scale, they have to evolve from implicit, trust-based networks to explicit, principle-based frameworks. You can’t run a 300-person engineering org on handshakes and vibes. But the goal isn’t just efficiency. It’s building a place where trust is the default, fairness is operational, and people feel valued enough to build careers instead of occupying jobs.

Dismantling internal borders isn’t about making transfers easier. It’s about replacing petty tribalism with actual organizational citizenship. The work is architecting systems that treat people like the valuable assets they are, not like interchangeable resources that need constant re-verification.

When you get it right, people stay. They invest. They build. Because they’re working for an organization that actually believes in them.

For a decade, I carried around a leadership operating system that was fundamentally corrupt. I contracted it in investment banking (Goldman, Bank of America, Deutsche Bank) where the model was brutally simple: decisions flow down, compliance flows up, fear is the primary management tool. This wasn’t some rogue manager’s pathology. It was institutional design.

Then I left, ran my own company for four years, and somehow managed to bring all those broken patterns with me. When I joined Etsy in 2016, I thought I’d shed the banking mindset. Turns out I’d just taken off the suit.

The Infection: Banking’s Management Pathology

Goldman, 2002. Post dot-com bust, investment banking was still the destination for a specific kind of ambition, the kind that accepted brutal hours and authoritarian culture as the price of admission. In Likert’s framework, this was textbook “System 1: Exploitative Authoritative.”1 Its tenets were control, fear, and top-down decision-making.

Public humiliation was standard operating procedure. Engineers got dressed down by their managers in the open (outside offices, in hallways) for mistakes both major and trivial. The performance was the point. Everyone absorbed the lesson. Fear works. It enforces compliance. Employee well-being was an HR checkbox, not operational reality. Professional development meant mandatory compliance training on SOX and KYC regulations. The concept of a one-on-one as a coaching space didn’t exist. You met with your manager twice a year: annual review, comp discussion. That was the relationship.

By Deutsche Bank, I’d encountered managers who tried to soften this, people who were genuinely more humane. When I moved into management myself, I tried not to be an asshole. But I was still running the same OS. Management, as I understood it: make decisions, communicate them, hold people accountable. The idea that management might be a craft, a practice with learnable skills that required deliberate development, wasn’t part of my model. I thought I was being a good manager because I wasn’t actively cruel. The bar, little did I know, was underground.

Four Years in the Wild

mohchi, 2011-2015. Running a startup should have been the reset. Small team, no corporate bureaucracy, the freedom to build whatever culture I wanted. And in some ways it was. I learned to ship fast, iterate, fail forward. But here’s the problem with mental models: they’re invisible until something forces you to see them.

A founder-CEO with six people operates with enormous authority. You make unilateral decisions because the company will die if you don’t move fast. You skip process because there’s no time. This feels different from banking. Jeans instead of suits, collaboration instead of hierarchy. But the underlying pattern is identical: concentrated authority, top-down decision-making, benevolent dictatorship. The startup gave me founder chops and product sense. It didn’t deprogram the prior decade.

The Inoculation: A Hard Collision with a Healthier Culture

It is more difficult to move from an authoritarian style to a democratic style than vice versa.

Etsy was different. Radically different. One-on-ones weren’t performative check-ins, they were the connective tissue of the organization. Decisions weren’t announced, they were made participatively. Caring about employee development wasn’t mission-statement theater, it was baseline operational expectation. I had no reference for this. It was like watching people have a conversation in a language I didn’t know existed.

A few weeks in, I did what felt natural. Designed a new team process, sent it out via email. Top-down, efficient, done.

Five minutes later: “Can we talk?”

My manager took me to one of Etsy’s many dedicated 1-1 rooms, a thing that had never existed in my professional universe, and asked a simple question: “Did you run this by the team before you sent it?”

He knew I hadn’t. He understood where I’d come from. Though I suspect he thought the problem was banking, not realizing I’d spent four years as a founder and still hadn’t learned this. The feedback was humane and surgical: that approach doesn’t work here.

First reaction: defensive. A flicker of “are you serious?” Then: embarrassment.

I’d recently read Carol Dweck’s Mindset,2 and I’d been trying to internalize growth mindset thinking, the idea that you’re not captive to your history, that patterns can be unlearned. If I hadn’t been actively working on that, I don’t think I could have set my ego aside. But I did. And what I heard was that involving people in decisions that affect them isn’t a nice-to-have, it’s foundational. The skill didn’t exist in my toolkit. The category of thought wasn’t there.

Four years running a startup with total freedom to do things differently, and I’d replicated the exact same authoritarian patterns I’d learned in banking. Just without the org chart to formalize them.

In hindsight, this conversation was one of the most pivotal moments in my career.

What Actually Changed: The Compounding Returns of Trust

The more I practiced participative leadership (at Etsy, Shutterstock (105 people), Candid (scaling through hypergrowth), RepeatMD) the more unambiguous the results became. A model built on trust and genuine collaboration isn’t just more humane. It’s dramatically more effective. When you meaningfully involve people in decisions, when you actually care about their development as more than a retention strategy, you unlock engagement and ownership that command-and-control can never achieve.

This sounds obvious. It’s not. Or it wasn’t to me. Deeply embedded assumptions don’t announce themselves, they just run in the background, shaping every interaction. Seven years in banking, four years as a founder replicating those patterns, and it took one difficult conversation to begin seeing the operating system I’d been running.

The deprogramming was slow. Old habits surfaced constantly, defaulting to unilateral decisions under pressure, forgetting to loop people in, treating collaboration as a time sink rather than an investment. But the shift from “managing through people” to “leading with people” compounds. The teams I’ve led since have been stronger, more resilient, more capable of solving complex problems. Not because I got smarter, but because I stopped treating leadership as the exercise of authority and started treating it as the cultivation of conditions for others to thrive.

Banking taught me rigor. Running my own company taught me to ship. But Etsy taught me something neither of those experiences could: that the best outcomes are achieved with people, not through them. You can’t learn that from a book. You need someone to point out that the way you’re doing it is broken, and you need to be humble enough to listen when they do.

A battlefield promotion (or field promotion) is an advancement in military rank that occurs while deployed in combat. A standard field promotion is advancement from current rank to the next higher rank; a "jump-step" promotion is advancement from current rank to a rank above the next highest.¹

Every experienced leader has seen this: you hear about a promotion that makes no sense. Not questionable. Nonsensical. Someone with three years of experience and mediocre IC performance suddenly managing a team. No process, no announcement, just a quiet org chart update.

This is the battlefield promotion. It happens when a key person leaves, panic sets in, and leadership makes the expedient choice instead of the right one. It’s organizational malpractice disguised as pragmatism.

Consider this scenario. You’ve probably seen some version of it:

Jane: Hey John, did you hear about Jack's promotion?
John: Huh? What do you mean? It's not promo season.
Jane: Well, after Jenny resigned last week, they promoted Jack into her role as the team's software development manager.
John: Wait a sec. Jack's a level 2 engineer with 3, maybe 4, years of work experience! I thought the manager track was level 4. Not to mention that he wasn't a particularly strong engineer. How's that make any sense?
Jane: I don't know... there wasn't an announcement. Guess you have to be in the right place and time to get promoted here... 🤷🏼‍
John: 🤬 My shields are down.

We instinctively recoil at this because it violates basic organizational fairness. One instance might be a fluke. When it becomes a pattern, it’s a symptom of deep dysfunction. As Donella Meadows observed, “Purposes are deduced from behavior, not from rhetoric or stated goals.”2 Repeated battlefield promotions reveal an organization’s true purpose: stability at any cost, including the cost of integrity.

Why Leaders Do This

The battlefield promotion is weak leadership, not clever problem-solving. It’s driven by predictable pressures:

The vacuum fallacy. A key person leaves, creating a void that triggers panic. Instead of running a proper search, leadership rushes to fill the role; any warm body will do. The goal isn’t finding the best leader. It’s having someone, anyone, in place to restore the appearance of order.

Path of least resistance. A proper hiring process is work. Define the role, source candidates, run rigorous interviews, make a difficult decision. Promoting the person already there is easy. It’s an expedient shortcut that trades the team’s future for the manager’s present comfort.

No succession planning. Healthy organizations constantly identify and develop future leaders. Battlefield promotions signal this work hasn’t been done. There’s no bench strength, forcing reactive, desperate decisions.

Loyalty over competence. In toxic environments, battlefield promotions reward relationships rather than capability. The “Jack” in our scenario is often someone’s protégé or a “founder bro” whose primary qualification is proximity to power. This transforms a mistake into deliberate cultural corruption.

The Ripple Effects

The immediate damage is obvious: an unqualified person in a role they’re not ready for. The secondary and tertiary effects are worse. A single battlefield promotion is a virus that spreads through the organization.

Trust evaporates. The message to everyone else: your performance doesn’t matter. What matters is being in the right place at the right time or knowing the right people. Why strive for excellence when rewards are distributed arbitrarily?

The career ladder becomes fiction. All those carefully defined levels and competencies? Theater. High performers get cynical and disengaged. The politically savvy learn that the real game isn’t impact, it’s alignment and visibility.

Your best people leave. The ones who value fairness and have options elsewhere; they’re gone. This creates a downward spiral. The talent pool concentrates with people who either can’t leave or are willing to play politics.

Incompetence gets normalized. The promoted individual, now operating beyond their capabilities, will likely struggle. To protect themselves, they surround themselves with loyalists or weaker performers. Their presence lowers the bar for what leadership looks like. This is the Peter Principle in action, but accelerated.

The Antidote

Preventing battlefield promotions requires systemic discipline; choosing principle over convenience. Not a single fix, but a reinforcing system.

Transparent, calibrated promotion processes. Promotions should happen on a predictable cadence, judged by a calibrated committee against clearly articulated criteria. Any off-cycle promotion must face the same scrutiny and require a higher burden of proof. Make the bar explicit: what specific evidence demonstrates this person is ready for this level?

Robust succession planning. Every leader should be able to answer “Who are your potential successors?” This should be a standard question in leadership reviews. If the answer is “I don’t know” or “Nobody,” that’s a problem that needs immediate attention.

Interim leadership as default. When a leader departs, appoint a qualified interim, often a senior IC or a leader from an adjacent team. This stabilizes the team without making a rash long-term commitment. It buys time to run a proper search for a permanent replacement.

Leadership courage. Ultimately, the antidote is courage. The courage to withstand the discomfort of a leadership vacuum. The courage to have a difficult conversation with a loyal but unqualified employee about why they’re not ready. The courage to prioritize long-term team health over short-term org chart aesthetics.

What It Reveals

A battlefield promotion is never just one bad decision. It’s a referendum on organizational character. It reveals whether a company is committed to its stated principles or if those principles are negotiable when convenience demands it.

I’ve seen organizations recover from battlefield promotions, usually by quietly course-correcting within 6-12 months, often through a “restructure” that undoes the damage. But the trust lost in that period doesn’t come back easily. People remember. They remember who got promoted and why. They remember that fairness was sacrificed for expediency.

Healthy organizations play the long game. They understand that the integrity of their promotion process is foundational to their culture. They refuse to sacrifice it, even when a role sits empty for months. Because they know that the cost of getting it wrong (in lost trust, departed talent, and normalized dysfunction) far exceeds the discomfort of an extended search.

Every leader knows what a bad management hire does to an organization. It’s not immediate, and that’s what makes it dangerous. The damage spreads slowly: morale craters, trust evaporates, your best people start taking calls from recruiters. A great manager creates conditions for their team to thrive. A bad one injects toxicity and triggers attrition among exactly the people you can’t afford to lose.

The old adage is true: people don’t leave companies, they leave managers. Given the stakes, the standard hiring process for this role isn’t just inadequate, it’s organizational malpractice.

Most companies default to a process that’s basically what they use for individual contributors, with a few tweaks. Screen for technical depth, architectural judgment, “culture fit.” When it comes to assessing leadership itself, they rely on behavioral interviews. This systematically fails to detect the most dangerous failure modes in leadership. You end up with a critical blind spot at the most leveraged point in your organization.

The Problem: Behavioral Interviews Are Theater

The traditional behavioral interview assumes that past behavior predicts future performance. Ask a candidate to “describe a time when you managed a poor performer,” and you should get signal on their competence. For any seasoned manager, this isn’t a diagnostic at all, it’s a performance. The questions are predictable. The “correct” answers are documented online. The whole exercise tests a candidate’s ability to perform authenticity, not their capacity to actually lead.

Consider the “poor performer” question. A savvy candidate knows the script:

• The Diagnosis: Frame it with empathy. Seek to understand the root cause: personal issues, unclear expectations, skill gap.

• The Action: Describe a structured process. Clear feedback, a PIP framed as supportive, regular check-ins.

• The Outcome: Whether the employee improved or was let go, take accountability, emphasize fairness and learning.

This sounds great. It’s also completely unverifiable. The interview provides zero signal on how they actually behaved. Did they belittle the employee? Set impossible goals? Ignore the problem for months until HR forced their hand? The behavioral interview can’t uncover this. It rewards polished storytellers and penalizes honest leaders whose real experiences are messy.

What you actually need is insight into the candidate’s core leadership philosophy and their judgment under pressure. Standard behavioral questions don’t get you there.

A Better Approach: Simulation Over Storytelling

To get real signal, replace theatricality with simulation. Observe the candidate’s thinking in scenarios that mirror the complex reality of the job. This requires a multi-part process that goes beyond canned questions.

The Leadership Philosophy Deep Dive

Instead of asking for stories, probe for models and first principles. Best conducted by a senior engineering leader, this moves from “what did you do” to “how do you think.”

Weak Question: “Tell me about a time you resolved a conflict.”

Strong Question: “Walk me through your mental model for resolving conflict between two high-performing engineers who disagree on a technical approach. What principles do you rely on? When do you intervene versus let the team work it out? How do you ensure the outcome strengthens rather than weakens trust?”

Other examples:

• “What’s your framework for delegation? How do you balance autonomy with accountability?”

• “Describe your system for managing performance across a team. How do you define it? Measure it? Handle calibrations?”

• “What is psychological safety to you, and what specific, observable actions do you take in the first 30 days to create it?”

This forces candidates to articulate their underlying philosophy. You’re not looking for a single right answer; you want evidence of a thoughtful, coherent approach.

The Socio-Technical System Design

This is the leadership equivalent of an architectural design interview. Present a realistic, messy team scenario and ask them to solve it.

The Prompt: “You’re inheriting a 6-person team responsible for a critical service. The previous manager just left. You’ve heard that morale is low, deadlines are slipping, and there’s tension between two senior engineers. Your new director wants a 30-60-90 day plan within your first week. Walk me through your process.”

This reveals:

• Diagnostic Skills: Do they jump to solutions, or start by gathering data? Who do they talk to first—the team, stakeholders, their boss? What questions do they ask?

• Prioritization: What do they tackle first? Rebuilding trust? Clarifying technical direction? A quick win?

• Communication Style: How do they plan to communicate with the team, their peers, leadership?

• Systems Thinking: Do they see this as just a personnel issue, or investigate systemic causes—unrealistic roadmap, tech debt, broken processes?

This simulates the core of the job: diagnosing and solving complex, human-centric problems.

The Upward Feedback Loop (The “Meet the Team”)

This is the most critical part of the process and the most often mismanaged. Too often it’s an unstructured “vibe check.” To make it effective, it must be a structured interview where the team is the interviewer.

Don’t ask the team “did you like them?” Task them with assessing specific competencies:

• “Does this person listen more than they talk?”

• “When you described a team challenge, did they respond with curiosity or judgment?”

• “Did they create an environment in this 45-minute meeting where you felt safe to be candid?”

Provide a structured feedback form. The team’s collective signal on a candidate’s ability to foster safety, listen, and demonstrate empathy is far more reliable than any answer the candidate gives to senior leadership. This is also where you catch the people who perform well upward but poorly downward, a failure mode behavioral interviews consistently miss.

The Outcome: Hiring for Judgment, Not Performance

This process requires more effort. It’s harder to design and execute than running through a behavioral script. But the return is significant.

You stop optimizing for people who are good at interviewing. You start identifying people who are good at leading. You hire for judgment and character, the things that actually matter when someone is managing humans through ambiguity.

The leaders you bring in will have more than compelling stories. They’ll have principles and demonstrated ability to navigate complexity. These are the managers who build resilient, high-trust teams that can weather challenges. They become healthy nodes in your organizational graph, strengthening the entire system.

This isn’t just better hiring. It’s a statement about what you actually value as an organization; not polish, but substance. Not theater, but reality.

“Engineering Culture” shows up everywhere in tech—job descriptions, strategic mandates, conference talks. To outsiders, it probably reads as pretentious jargon. For those who’ve built teams at scale, it’s something else entirely: the invisible physics that holds an organization together or tears it apart.

Culture is the operating system. Not written in Python or Go, but in the accumulated weight of accepted behaviors, shared assumptions, and power dynamics. It’s the kernel that schedules tasks, manages resources, and either enables or corrupts every process running on top of it. This is the gravitational force binding together the sprawl of interests, responsibilities, roles, functions, priorities; all the machinery of an engineering organization.

We tend to dismiss culture as “soft,” things like perks and mission statements. This is a fundamental misunderstanding. Culture is the most critical piece of your technology stack. A buggy kernel guarantees system-wide failure. You can’t hire or spend your way out of it.

The Default Pathogen: Command-and-Control

For most of organizational history, the default has been command-and-control. Military hierarchy applied to civilian contexts. Top-down, authoritarian, treating employees as fungible resources expected to execute orders without questioning them. Its core functions: control, compliance, tight management of information flow.

There are contexts where this works. Organizations facing existential threats where swift, unquestioning action is paramount. But in knowledge work, creative and technical contexts, it’s a pathogen. It actively destroys the psychological safety and autonomy required to solve complex problems. Fear replaces curiosity. Compliance replaces collaboration.

When you treat people as cogs, they stop bringing their full faculties to the work. Questions don’t get asked. Risks don’t get named. A corrosive silence settles over the organization. You get adherence, but you strangle innovation. It’s an OS designed for a factory floor, misapplied to work that requires judgment and creativity. Highly effective teams cannot be commanded into existence, they have to be cultivated.

If not command-and-control, then what? Not chaos. A more sophisticated architecture.

Building Better Systems: Culture as Deliberate Design

High-performing engineering culture doesn’t happen by accident. It requires deliberate organizational design, replacing brittle hierarchy with a resilient framework built on clarity and distributed decision-making. This means engineering the foundational layers with intention. Moving from abstract principles to concrete structures. The codification and implementation of an organization’s roles, processes, and formal relationships.

Structural Integrity: Roles, Levels, and Reporting

Scrappy startups can get away with ambiguous responsibilities when everyone’s doing a bit of everything. Once you grow past a certain size, usually around 50 people, that ambiguity becomes a fatal flaw. It unleashes anxiety and the politics of arbitrary decision-making.

A robust career framework is more than bureaucracy. It’s an act of institutional respect. It transforms the abstraction of “career growth” into something concrete and navigable. Without it, you’re telling people their future is subject to whim. Implementing this structure is often painful. You’ll lose some early employees who valued the unstructured environment and the breadth of their technical jurisdiction. That’s the cost. But the alternative, organizational drift and attrition at scale, is worse.

Equally fundamental: sensible, unambiguous reporting lines. In the absence of clear accountability, power vacuums form. They get filled by politics and shadow hierarchies. This is organizational cancer.

The Paradox of Process: Liberation Through Structure

Most engineers resist “process.” Usually because they’ve experienced bad process, cumbersome procedures divorced from value, serving only bureaucracy. But good process is liberating. It automates trust and reduces cognitive overhead, letting teams focus on complex problem-solving instead of reinventing operational workflows.

Nowhere is this more critical than recruiting. An uncalibrated, ad-hoc hiring process is a primary vector for cultural decay. Individual teams often insist their needs are unique and demand bespoke processes. This is dangerous. It injects systemic weaknesses:

• Internal perceptions of inequality. Different teams operating by different rules breed resentment.

• Objective miscalibration of skills. You poison the well for internal mobility, which is vital for a dynamic organization.

• Expression of bias. Both conscious and unconscious. “Culture fit” becomes a synonym for “looks and thinks like me,” corrupting hiring decisions and undermining meritocracy.

A strong recruiting process, built on a well-defined leveling framework, is a systemic corrective. A defense against our worst instincts. It forces us to be as objective as we claim to be. By standardizing on expected behaviors and impact for each role, it allows teams to make minimally-biased, defensible hiring decisions.

Instantiating Values: From Abstract to Observable

Beyond formal structures, healthy culture depends on less obvious interpersonal protocols; expectations around behavior, approaches to work, relationships between teams and stakeholders. A mission statement is useless until it shows up in daily practice.

This is where artifacts like leadership principles or communication charters become critical. I’m biased toward organizations that enable collaboration based on trust and mindfulness. Etsy’s Charter of Mindful Communication wasn’t just a document, it was a protocol for interaction designed to build psychological safety. Amazon’s Leadership Principles aren’t posters on a wall, they’re heuristics used in decision-making, performance reviews, and daily work. They give everyone a shared language for execution.

These artifacts are the user-space libraries of the cultural OS. They translate abstract values into concrete, observable behaviors.

The Outcome: A Self-Healing System

Culture is complex. Multifaceted. An absolute requirement for building and sustaining strong engineering organizations. You won’t get it right the first time. That’s fine. The key is treating culture and its supporting processes as living systems that must evolve as the business, product, and teams change.

By building on a stable foundation (roles, processes, principles) you create an environment where trust becomes the default state. Collaboration friction drops dramatically. Alignment happens not through top-down enforcement but through shared context and internalized principles. The organization develops an immune system capable of identifying and rejecting the pathogens of bias, politics, and bureaucracy.

That’s the ultimate function of well-designed culture: creating an environment where extraordinary people can do their best work, at scale, with minimal systemic friction.

Introduction

Organizational design is the deliberate act of encoding a company’s values into its structure. Before we can diagnose the pathologies that undermine an organization (e.g. the subtle viruses of ambiguity, favoritism, and inconsistency) we must first understand the anatomy of a healthy system. This post, originally from 2018, serves as an archival look at the raw mechanics of building a scalable engineering culture at Shutterstock. It is a case study in creating the foundational bones: the clear pathways and defined expectations that form the first, best defense against systemic dysfunction.

Clear definition of levels and active enablement of individual career development are cornerstones of healthy, high-performing, highly-engaged engineering organizations. At any given point in time, it is important for engineers and managers of engineers to comprehend where they stand, the possible paths before them, and the specific behaviors and impact expected of them at each level from both performance and career progression perspectives.

Although tech companies generally share many traits, what expectations are deemed important at a given company are highly dependent on their own unique engineering cultures. As such, there are no drop-in, one-size-fits-all career leveling/development frameworks that can be leveraged between companies without (potentially significant) loss of fidelity to the characteristics and culture that companies choose to define themselves.

As such, this post describes how we’ve implemented a structured framework that is aligned to the unique traits and culture of Shutterstock but does not claim to define a generalized solution for all engineering organizations.

Prior (Ground)work

One of the most notable, relatively recent enlightenments to engineering tracks is the bifurcation of management and individual contributor careers within software engineering as parallel, and not sequential, tracks. Before this shift, when an engineer reached a certain level of seniority and was ready for promotion, upward mobility usually necessitated a move into management, or to otherwise see their careers stall. Given that many engineers are not interested in or suited for the far different demands of people management (vs. individual contributor (“IC”)), this change was a momentous (though with the benefit of hindsight, painfully obvious) advance. Fast forward to present day—though everyone begins their careers as individual contributors, after a certain level, a fork in the path appears where an engineer, should they be suited for the role and identified as such by leadership, may embark upon the management track or remain in the IC path which continues before them.

Today, most large tech companies now define IC and management (M) tracks as parallel.

Individual Contributor (IC) Track

When we look at the traditional IC track, progress is generally linear along a single dimension. For example, in an organization with 7 career steps for engineers (e.g. L1…L7), an individual will progress incrementally from L1 through L5 over the course of their career at the company, with L7 being the highest level (both L6 and L7 are reserved for “exceptionally exceptional” individuals and are not considered part of “normal” career progression for the vast majority of ICs). In many places, as one progresses in their career, this not only means that they’re getting progressively better at the specific competencies and expectations of them as a software engineer, but also means, especially after the Senior Engineer level (usually L3), that they may need to acquire/demonstrate skills that may not be in their interests or wheelhouse.

For example, in many traditional career tracks, expectations for L4+ (i.e. Staff Engineer+) come with the implication of broader focus on more cross-cutting concerns across the engineering organization, as opposed to diving deeper into the product engineering stack for their teams; this almost always means less coding coupled with i) taking on broader, organizational initiatives (e.g. architecture reviews, process design, general “thought leadership”), ii) mentoring, iii) cross-team collaboration/communication, iv) public speaking and, again (and importantly), v) less coding.

Not everyone is suited for or interested in leading organizational initiatives and, as anachronistic as it was in forcing engineers to become managers in the past, it is equally disjointed to take the engineers who are best at what they do (i.e. coding) to make them do less of that, just because of arbitrary career definitions which fail to account for more nuanced differentiation in interests between individuals. At smaller companies (i.e. startups), you typically have no choice but to be a generalist and work on everything that comes your way, but at larger companies, specialization becomes an imperative to achieve broad technical efficiencies and, as far as Shutterstock goes, existing career models did not satisfy the specific needs of the organization.

Management (M) Track

The entrypoint to the engineering management track typically comes after the Senior Engineer level (i.e. L3), as a promotion up and out of the IC track into leadership, which implies that an entry-level manager is on par with L4 engineers in terms of experience/leveling. The nature and roles of Engineering Managers vary widely across the industry, however, the fundamental, defining characteristic of the role is that of being responsible for managing people. The extent of their day-to-day technical involvement depends on the specific needs of companies, however, the explicit mandate of leadership positions generally broadens further along the management track. As with the IC track, we believe there’s a clear difference between types of technical leaders that we’ll get into below.

Building for Purpose: Overall Structure

As referenced in Shutterstock CEO, Jon Oringer’s tech blog relaunch post, the company’s priorities in the early stages were about building product and moving quickly to support the company’s growth, resulting in tacit, undocumented understandings of career development. Like virtually all startups, thinking about how to structure the organization to support career development for what started off as a handful of engineers isn’t a priority until that cohort has “suddenly” grown to several hundred in number. With the benefit of hindsight, it becomes clear that a lack of clearly-defined titles, levels and expectations is detrimental to long-term retention of talent and therefore poses a challenge to scale the organization past a certain point. These are the non-trivial challenges which this framework is intended to counteract.

What We’ve Done

Shutterstock launched a new career track framework across the Engineering organization in early February 2018. In our case, we have a need for both deep, product-level and broad, cross-team/organizational specialists. In order to support growth in what we see as clearly differentiated specialties, we created two distinct and parallel paths within the IC and management tracks, corresponding focus at the team or organizational levels. This is best illustrated with a diagram:

IC Track

After IC3 (Senior SE), the career track splits into two lanes, corresponding to team or organizational scope. We define Staff Engineers to be deep product-level experts who are tied to specific, product teams and who report into the (Sr.) SDMs for those areas. On the other hand, we define Principal Engineers to have broader, organizational focus on cross-cutting initiatives that generally impact more than one team if not the entire org. Principal’s, therefore, report into organizational-level leadership at Shutterstock (i.e. Directors and above).

This does not mean that an IC in either lane is 100% focused in one area at the expense of the other; it’s more about the proportional allocation of focus (e.g. 70% organizational/30% team for Principal SEs and the opposite for Staff SEs) which may shift over time. Being a Staff Engineer at Shutterstock, however, doesn’t restrict their involvement in broader initiatives outside their immediate teams. By the same token, a Principal Engineer is not myopically focused on organizational initiatives alone; they still contribute to discrete team initiatives, but that is not the overarching focus of their attention.

This structure does not preclude the ability to switch between Staff and Principal paths once an IC has chosen a lane.

M Track

We’ve adopted a similar bifurcation with the leadership track, where we’ve identified team- and organizational-scopes for career development therein. Entry-level engineering managers (M1—a level which is equivalent to overall experience as IC4) and more experienced engineering managers (M2) are classified as team-scope leaders. We expect our leaders to have attained a senior level of technical expertise before taking on the challenges of leading engineers.

Though we expect our Engineering Managers to be technical and available as technical advisors/coaches, we don’t generally expect them to write code or make day-to-day technical decisions for their teams; their responsibilities are primarily concerned with maintaining, cultivating and growing healthy engineering teams along the dimensions of individual skills growth, project execution, job satisfaction, and career development. At the M3 (Director) level and above, the mandate of leadership responsibilities shifts towards a more organizational scope (i.e. across more than one team) and typically involves managing managers. This lane goes to M5 (VP).

Breakpoints

We’ve also introduced the concept of “breakpoints” in our career framework (this is not unique to Shutterstock). These are levels at, and after, which a given role is terminal. In other words, once reaching a breakpoint, an individual can conceivably remain there for the remainder of their career. However, up to a breakpoint, all individuals are expected to continue growing professionally through their respective tracks. This concept supports individuals that, for any number of reasons (e.g. family obligations), are not willing/able to make the contributions necessary to advance to the levels above a given breakpoint, yet are fully-productive and valued employees.

Putting Meat on the Bones

Visual diagrams of frameworks are nice and tidy, but without well-defined supporting documentation and processes, they are toothless and wholly ineffective. How we define the behavioral characteristics and expectations of personal impact at every level along both tracks is largely informed by what we value as an engineering organization, but, more importantly, is informed by the clear, aspirational goals we want to set for our employees. In other words, “What sort of organization are we now and what sort of organization do we aspire to be?" This is the core essence which we codify in a career track framework.

Baselines

Most experienced engineers and engineering leaders have strong, internalized views on what it means to be an IC1 vs an IC3 for example. Though these views may intersect to some degree across individuals, they are personal opinions and not universal, empirical truths; they are based on relative, not absolute, measures that may or may not be objectively important from a given organization’s perspective: relative experience, relative skill, relative seniority, relative communication skills, etc. Since an individual’s perceived value of a given measure is pegged to internalized benchmarks derived from their own unique experiences, it is important to define externalized anchors that create objective alignment and shared understanding across individuals, emphasizing the specific behaviors and impact that are important to succeed within the context of the organization.

We use the concepts of “behaviors” and “impact” to define leveling which allows us to scale the framework across the diverse engineering organization (product engineering, services, data science, devops, infra, etc.). These are “stable” characteristics which we believe measure what’s important for individuals within the organization. We deliberately avoid articulating specific technologies as requirements for any level. Instead, technical skills, if applicable for a given individual, are framed as behavioral expressions or magnitudes of impact. These are detailed for every level and each track. Though we don’t drill into the specific behaviors and impact for levels and tracks in this post, we may go into some level of detail in a future post.

Supporting Processes

Like many large companies, at the company level, Shutterstock sets and measures against individual goals/objectives annually. However, career development and human processes are too fluid to hew to such an infrequent cadence of reviews. As such, in parallel to annual objectives, each manager works with their direct reports to create career development plans that are reviewed throughout the year to track progress and (re)set expectations as needed, understand challenges, clarify progression, etc. The career development plan sits apart from and does not directly impact the annual review; it may help inform and measure achievement of annual objectives, but there is a loose coupling between these disparate, but related, processes. Moreover, the contents of such career development plans are confidential between each manager and their direct reports.

Conclusion

The key objectives of why we implemented a comprehensive career track framework are to:

• create a shared, objective understanding of what we expect from each level;

• provide a framework within which an individual can understand their own performance;

• set realistic expectations on career development and timelines; and

• calibrate levels and expectations across the organization and across the industry at large

We view what we’ve created thus far the start and not the end. The organization will change, as will the people, and the framework must adapt to new realities and changing expectations; like organizations themselves, these frameworks must evolve as necessary to remain relevant over time.

Reflecting on this framework years later, its true value becomes even clearer. While the specific levels and tracks detailed here have undoubtedly evolved, the underlying principle is timeless: a well-defined system is the first line of defense against the organizational pathologies of ambiguity and favoritism. A career ladder is more than a map for employees; it is a systemic antidote. By codifying expectations, you inoculate the culture against the ad-hoc, relationship-based decisions that erode institutional trust and undermine merit. The mechanics described here are the tools; the goal is a culture where clarity itself is the operational standard.

[This post has been republished on the Shutterstock Tech Blog].