An Insolvent Class Problem

A diagnostic lexicon for structural governance failures, institutional capture, and policy dynamics

The Insolvent Class Problem — A Diagnostic Lexicon (v3)

Part One: The Problem Nobody Can Fix

The punchline first

Some problems can only be fixed when nobody benefits from them being broken. Most governance problems aren't like that. Most governance problems make someone money, win someone votes, or justify someone's department — specifically because the problem exists. Housing isn't broken despite decades of policy. It's broken because decades of policy have arranged themselves around the brokenness, and the brokenness now pays rent to everyone in the room.

This lexicon names that structure. It identifies a class of governance failures where the thing that makes the problem unfixable is the same thing that makes it politically valuable. Not because anyone planned it. Not because anyone is evil. Because the incentive structure doesn't permit success.

The tragic punchline: systems will only allow a problem to be fixed once it's no longer fixable. The political will to act arrives after the capacity to act has been consumed.

What kind of problems are we talking about?

You know the feeling. A problem everyone agrees exists. Solutions everyone agrees would work. Years of commissions, reports, expert panels, and bipartisan concern. And nothing changes. Or worse — things change in ways that make the problem look different while keeping it structurally identical.

Housing affordability. Birth rate decline. Climate adaptation. Welfare dependency. Infrastructure decay. AI governance. These look like separate issues in separate domains. They're not. They share a structural property that makes them members of the same class — different problems with the same shape.

The shared property: there is no politically survivable path where dominant actors can simultaneously preserve their rents and close the liability ledger.

Unpack that:

Rents — the power, money, votes, and institutional authority that the current players derive from the system as it is. Not just financial. A minister gets rent from announcing housing policy. An opposition party gets rent from attacking it. Media gets rent from the argument. Consultants, lobbyists, developers, advocacy groups — everyone eats. The brokenness is load-bearing.
Liability ledger — the accumulated cost of the unresolved problem. Deferred maintenance, future obligations, displaced consequences. This ledger is real and it compounds. It can be moved around — shifted to future taxpayers, exported to other countries, hidden in demographic projections — but it cannot be reduced without someone losing their rent.
Politically survivable — any actor who genuinely closes the ledger destroys the rents of every other actor at the table. That actor does not survive the next election cycle, board meeting, or budget round. So nobody does it.

The system is structurally insolvent. It owes more than it can pay off without destroying the income streams of the people who'd have to authorise the payment.

What happens inside the trap

Once a problem enters this class, a specific set of dynamics takes over. These aren't metaphors. They're recognisable patterns — and once you learn to see them, they're depressingly reliable.

The system converges on maximum survivable dysfunction. Not maximum dysfunction — that would cause collapse, and collapse is expensive for everyone. The system finds the sweet spot: broken enough that every political actor can extract rent from the brokenness, not so broken that the whole thing falls over. Working institutions get degraded toward this point. Already-broken institutions get propped up just enough to stay above it. Think of every public service you've interacted with that's just barely functional enough to not cause a crisis, while being permanently, irritatingly inadequate. That's not an accident. That's the equilibrium.

Solutions get eaten. When someone does implement a genuine fix, the political system doesn't ignore it — it recaptures it. A working institution becomes invisible (nobody campaigns on things that work). Invisible institutions lose political defenders. Undefended institutions become targets for rent extraction. The fix gets defunded, restructured, or loaded with new mandates until it's broken enough to be politically useful again. Success → invisibility → vulnerability → recapture → dysfunction. The cycle has a name in the lexicon: recapture. It's why good institutions die.

Problems change shape but not structure. Apply policy pressure to smoking? The liability rotates into vaping — different product, identical structure (addictive inhaled consumer product, regulatory lag, youth uptake, corporate capture). Ban one financial instrument? The risk migrates to the next one. Tighten one border? The flow redirects. The problem transforms under pressure while preserving its essential architecture. Same molecule, new spatial arrangement. We call this conformer rotation.

The bill gets moved, not paid. Can't fix housing domestically? Import migrants to paper over the demographic consequences. Can't fund pensions now? Borrow against future taxpayers. Can't absorb the liability locally? Export it to another country's balance sheet. Every apparent solution is a transfer, not a resolution. The bill changes form but it never disappears. This has a termination date: you run out of places to put it.

The three ways out

The lexicon identifies exactly three structural exits. Two are genuine. One is a deferral that eventually runs out.

The Montreal Condition — the democratic exit. Requires two conditions simultaneously: (a) the solution space must be collapsible into a binary choice — on or off, with no gradient for the political system to exploit, and (b) incumbent rent must be alignable with the switch — the actors who could block the change must profit from allowing it. Neither condition alone is sufficient. Binary framing without rent alignment still gets blocked by incumbents who lose from the switch. Rent alignment without binary framing still gets gamed on the gradient — partial measures create new surfaces for political exploitation. The Montreal Protocol on ozone-depleting substances is the only identified case. CFCs were banned — not regulated, not traded, not taxed — and DuPont had pre-positioned to profit from substitutes, collapsing industry resistance. The politics drained out because there was nothing left to argue about and nothing left to lose from compliance. This dual requirement makes the Montreal exit far narrower than it first appears, which matters for practitioners assessing whether it's available for a given problem.

The Singapore Condition — the authoritarian exit. Eliminate the political competition that sustains the rent gradient. If the politics become monopolar, the deadlock breaks — not because the problem is solved well, but because the negotiation cost drops to zero. One actor decides, and the decision sticks. This isn't better. It's cheaper. And it comes in flavours, distinguishable by where the monopolar actor's rent sits:

Terminal — the monopolar actor's rent is tied to system performance. Lee Kuan Yew's legitimacy depended on outcomes: housing, corruption, infrastructure, ethnic tension — all resolved because failure threatened the actor's own position. This is structurally a form of Harmed Engineer alignment: the monopolar actor is harmed by the dysfunction they have authority to fix. Terminal exits are rare because this incentive alignment is rarely present and has no known reliable method of external engineering.
Oscillatory — the monopolar actor's rent is extractive. The old rent structure is replaced with a new one serving the actor themselves. Argentina is the canonical case: each monopolar intervention creates fresh rents rather than closing existing ones. The deadlock breaks temporarily and reconstitutes around the new beneficiary.
Imposed — external restructuring by a more powerful actor. Outcomes depend on the external actor's exit strategy and whether new rent structures are seeded during restructuring. Marshall Plan vs. structural adjustment programmes: one built institutional capacity, the other created dependency traps.

Rwanda proves the category includes its most extreme cases. Singapore proves it's not reducible to them. The diagnostic question is not whether the exit is authoritarian but where the monopolar actor's rent sits — and that determines whether the resolution holds.

Parasitic rotation — the false exit. Transfer your insolvency to someone else's balance sheet. Consume another system's surplus to service your own dysfunction. Migration as fertility policy. QE as solvency policy. Geographic or temporal displacement of the bill. This works until the supply of non-insolvent systems runs out. It's a painkiller, not a cure, and the dose keeps increasing.

Why this matters now

Compound insolvency — where multiple insolvent-class problems are mutually load-bearing — is a pre-authoritarian condition. Not because authoritarianism is good, not because democracy is weak, but because compound insolvency closes the Montreal exit on every node simultaneously while the parasitic rotation options run dry. The structural arithmetic narrows toward monopolar resolution whether anyone wants it or not.

New Zealand is a useful case because the compound structure is legible: housing, capital flight, birth rate, and fiscal sustainability are four faces of one problem, each feeding the others. Fix any one in isolation and the others compensate. The simultaneous intervention required across all nodes is itself politically unsurvivable.

The lexicon also describes how the governance debate itself — the argument about what to do — can prevent governance. Moral and identity framings dominate technical framings because they generate more political energy. The conversation about AI alignment consumes the governance capacity that could be spent on AI infrastructure governance, which is the boring part that would actually help. The debate is the disease pretending to be the cure.

None of this is a counsel of despair. The lexicon identifies intervention windows (before systems lock in), design principles for solutions that survive (make them boring), and actors capable of building genuine fixes (people whose rent is negatively correlated with the dysfunction). But it is honest about the structural constraints. Optimism that ignores the trap isn't optimism — it's the trap working as designed.

What follows is the vocabulary: the named patterns, diagnostic tools, and worked examples that practitioners can use to identify and describe what they're seeing. If Part One gave you the shape, Part Two gives you the words.

Part Two: The Vocabulary

What this is — and what it isn't

This is a diagnostic lexicon. It provides a set of named patterns for governance failures that share a common structure. Each term names something practitioners already recognise but currently lack efficient language for.

It is not a predictive theory in the point-prediction sense. It does not claim to forecast specific outcomes, specific timelines, or the arrival of exogenous shocks. It describes the now and the before — what's happening and what happened — with enough precision that practitioners can see the structure, communicate about it, and identify intervention opportunities earlier.

But the framework does make structural gradient claims. Given a set of structural features (rent/liability deadlock, collateralisation, recapture dynamics), it identifies the direction and steepness of incentive surfaces: the slope runs toward MSD, recapture, and conformer rotation, and the slope steepens as collateralisation deepens. Reform capacity tends to degrade over time, and the rate of degradation steepens as collateralisation deepens — but local reversals are possible (the framework's own T6 mechanism, exogenous shocks, crisis windows, individual leadership). Those reversals work against the gradient and tend not to persist absent sustained structural change to the incentive landscape itself.

The framework's epistemic claims have three tiers. Gradient claims (which the framework makes): given structural conditions, the incentive surface slopes in a specific direction, and practitioners can read that slope. Trajectory claims (which it does not make): no claim that systems will follow the gradient on any specific path, or that reversals are impossible. Point predictions (which it definitely does not make): no timelines, no specific outcomes, no specific triggering events. It has no opinion about black swans, technological breakthroughs, or individual leadership; it describes what the incentive gradient will do conditional on those events. Any predictive application of this vocabulary should be framed as scenario envelopes — ranges of structurally likely behaviour — not trajectory or point predictions about specific systems or timelines.

It makes no claim about intent, morality, or competence. No actor in any system described here is accused of malice or stupidity. The claim is structural: rational actors, operating within rational systems, producing outcomes that no individual chose and no institution can correct — not because people are failing, but because the structure doesn't permit success.

The Core Patterns

Five organising observations. These are the recurring dynamics that appear across every member of the class. The vocabulary items in the next section are specific expressions of these patterns on specific substrates.

Rent-Maximising Convergence

In Insolvent Class Problems, political systems converge toward configurations that maximise sustainable political rent extraction. Configurations that reduce extractable rent are unstable — the system actively pushes back against them.

This is the central observation. Domain-independent. Makes no reference to quality, purpose, or any specific institution. It describes what political systems do in the presence of insolvent-class problems the way gravity describes what objects do in the presence of mass.

A good solution isn't just attacked — it's structurally unfavourable. The system treats it like a ball on top of a hill.

Rent Surface Area (RSA)

The amount of available surface for political rent extraction from a given institution, policy, or problem. RSA is what determines which configurations the system converges toward.

When multiple configurations are available — multiple framings, multiple policies, multiple targets — political attention flows toward the one with the highest RSA. This produces a recognisable hierarchy: moral and identity framings dominate technical framings because they generate more extractable surface. The governance debate consumes governance capacity.

Political attention is finite. It goes where the rent is. Boring policies survive because they're not worth attacking. Exciting policies get eaten.

Maximum Survivable Dysfunction (MSD)

The equilibrium point. The level of brokenness where rent extraction is maximised without causing system collapse.

Two useful definitions of "working":

Working-for-purpose — does the institution achieve its stated goal?
Working-for-system — does the institution sit at a stable equilibrium?

An institution can be working-for-system because it is broken-for-purpose. The brokenness is load-bearing.

MSD requires two conditions to be present: competing actors (more than one player extracting rent) and a continuous gradient (a spectrum of possible policy positions to exploit). The two known exits each work by removing one of these: the Montreal Condition removes the gradient (binary solution, nothing to argue about) while simultaneously aligning incumbent rent with the switch. The Singapore Condition removes the competition (one actor decides).

The Transition Boundary

The point where a system crosses from dysfunctional but failure-permissible to dysfunctional and failure-impermissible.

Before the boundary: the system may be broken, but it can die and be replaced. Failure is a potential reset.

After the boundary: the system cannot die (too much depends on it) and cannot heal (the political system won't allow it). The dysfunction becomes permanent.

Firms are generally not insolvent-class problems because bankruptcy exists — assets and liabilities can be reallocated, shareholders can be wiped, failure is designed in. Governance systems, critical infrastructure, welfare states, and systemically important institutions lack this reset mechanism. Their failure is politically or structurally non-viable. That's what makes the deadlock permanent.

The transition between these regimes is the highest-leverage intervention window. Everything is cheaper and easier before a system crosses this boundary.

The Temporal Trap

The incentive to enact genuine reform emerges only after the system's capacity to reform has materially degraded.

The will to act and the capacity to act move in opposite directions. The window doesn't close gradually — it accelerates. Each year of inaction makes the next year's intervention harder and more politically expensive.

This is the lexicon's most important observation, and its tragic punchline. It must remain on the front page of any presentation.

The Vocabulary

Named patterns that appear across the class. Each one names something specific and recognisable. They stand on their own utility — you don't need to accept a theoretical structure to use them. If you recognise the pattern, the word is doing its job.

Conformer Rotation

The problem transforms under policy pressure while preserving its essential structure. Same molecule, new spatial arrangement.

Apply pressure to smoking → the liability rotates into vaping. Ban a financial instrument → the risk migrates. Tighten a border → the flow redirects. The surface changes. The architecture doesn't.

This is why policy "victories" often feel hollow. The problem isn't solved — it's rotated to a surface the policy can't reach. And the new surface is, by construction, outside the scope of whatever intervention triggered the rotation.

Parasitic Rotation

Transfer your insolvency to someone else's balance sheet. Consume another system's surplus to service your own dysfunction.

Migration as fertility policy — consuming other societies' reproductive surplus. QE — displacing insolvency from the present onto the future. Geographic, demographic, or temporal displacement of the bill.

Distinguished from conformer rotation: conformer rotation transforms the problem, parasitic rotation transfers the cost. Both preserve the underlying liability.

Has a termination date: when the supply of non-insolvent systems runs out. Birth rate decline is the canonical example — it's global, convergent, and driven by universal forces. When exporting societies hit the same walls, the rotation collapses.

Temporal Parasitic Rotation

The governance debate oscillates between two framings of the same problem. Neither framing produces governance output. The oscillation itself consumes the intervention window.

The mechanism: one framing is boring but technically productive (infrastructure standards, audit requirements, technical regulation). The other is engaging but politically unresolvable (moral arguments, identity positions, values debates). Attention flows to the high-energy framing because it generates more political rent. That framing becomes toxic or exhausted. Attention retreats to the boring framing. The cycle restarts with less time remaining.

Each cycle is locally rational — every actor follows the incentive. The aggregate is a governance window consumed by oscillation.

The AI alignment debate is a worked example: the infrastructure conversation (API standards, model evaluation, compute governance) keeps getting crowded out by the moral conversation ("whose values?", existential risk rhetoric, culture-war engagement). The boring work that would actually help keeps losing to the exciting argument that won't.

Recapture

The mechanism by which technically stable solutions are re-ingested by the political layer. Success → invisibility → no political defenders → rent extraction → destabilisation → visibility → more rent from arguing about fault.

Working institutions are invisible. Invisible institutions have no political constituency. Institutions without constituencies become targets. The surrounding system converts a working solution into a source of rent-extractable instability.

Example: ACC (New Zealand's no-fault accident compensation). A genuine structural solution that the political layer cannot leave stable because instability is rent-generating. The right tightens eligibility. The left expands coverage beyond funding. Both extract political capital. Neither is incentivised to leave it alone.

Compound Insolvency

When multiple insolvent-class problems are mutually load-bearing. Each member's dysfunction is a structural dependency of the others. Partial intervention on any single member is self-defeating — the system self-heals by compensating through the other nodes. The simultaneous intervention required across all nodes is itself politically unsurvivable. Strictly harder than any individual member.

New Zealand's housing, capital gains tax absence, migration settings, and birth rate decline are four faces of one problem:

Fix housing → requires CGT → collapses retirement vehicle → circular.
Implement CGT → deflates property → requires alternative wealth accumulation → requires higher wages → requires tighter labour → requires less immigration → accelerates demographic decline → circular.
Fix birth rates → reduces migration dependency → reduces housing demand → deflates property → triggers CGT problem → circular.
Reduce immigration → eases housing but accelerates demographic decline, shrinks tax base, widens fiscal gap → circular.

Every intervention on one member destabilises the rent structure of at least one other. NZ isn't trapped by ignorance. It's trapped by arithmetic.

Collateralisation

A system becomes collateralised when the cost of allowing it to fail exceeds what powerful actors can tolerate — not because people care about it, but because failure produces cascading harm that forces continuation.

Six conditions contribute:

T1 — Formal commitment. Legislation, mandates, explicit guarantees, repeated public commitments that make reversal politically fatal.

T2 — Bailout precedent. Once rescue becomes expected, actors price it in. Behaviour adapts. The system becomes functionally too-important-to-fail. The emergency response is the lock-in event.

T3 — Critical dependency. Other systems depend on it. Failure cascades. Supply chains, payment systems, health capacity, welfare delivery.

T4 — Identity fusion. The system becomes symbolically fused to national identity or regime legitimacy. Failure becomes humiliation, not mere loss. Amplifies other triggers but is not sufficient alone.

T5 — Mass constituency. Large populations become dependents or beneficiaries. Creates political veto mass. Pensions, housing wealth, welfare entitlements, consumer technology adoption.

T6 — Ledger legibility. A discrete moment when the liability becomes widely visible and attached to accountable decision-makers. Audit revelations, crisis reports, investigative journalism. Converts fog into recognised insolvency.

These interact. T5 can actively defend T3: when a large user base is entangled with a systemically embedded technology, any attempt to impose technical safeguards on the infrastructure is experienced as a personal loss by the constituency. The users resist the regulation for you. AI is building this compound right now — enterprise embedding (T3) deepens consumer dependency (T5), and the constituency organically defends the embedding against intervention.

De-Collateralisation (DCC)

A practical intervention lever. Not a fix for the dysfunction — a way of changing the failure landscape so that the system is no longer hostage to its own collateral damage.

DCC works by making failure tolerable. If a system can fail without bringing everything else down with it, it can be permitted to fail — and that permission breaks the deadlock.

Four design moves:

Partition failure. Ringfencing, circuit breakers, structural separation. One part can die without killing the whole.
Cap external harm. Insurance, compensation schemes, guaranteed minimums. Failure hurts, but not catastrophically.
Assign losses credibly. Clear liability chains. Known consequences for known actors. No ambiguity about who pays.
Precommit to letting a component die. Living wills, mandated wind-down plans, automatic triggers.

Post-2008 banking resolution (ringfencing, bail-in, living wills) is DCC implemented in practice before it was named. The Depositor Compensation Scheme in NZ may be functioning as DCC for the banking system — it's what allowed the RBNZ to accept higher risk in its 2025 capital review.

The Boringness Constraint

DCC interventions survive recapture only when their RSA is below the recapture threshold. The cost of attacking a boring institution exceeds the yield from attacking it.

Boring infrastructure evades recapture not by being defended, but by being uneconomical to attack. Nobody marches over loan-to-value ratios. Nobody campaigns on deposit taker capital requirements.

Survivability test for any intervention:

Can you describe this to a non-specialist without them getting angry?
Does this appear on any party's policy platform?
Would a newspaper put this on the front page?
Does this have a constituency that would march for it?

If the answer to any is yes, the design is not boring enough. The test is not "is this good policy?" The test is "is this boring policy?" Those are different questions, and only the second one predicts survival.

The boringness constraint is not a one-time gate. It is a continuous condition. A single identity-coding event — one policy announcement, one ministerial speech, one governance decision — can permanently poison an intervention by associating it with a tribal position. Even if the identity payload is subsequently removed, the memory of the framing persists.

Case contrast — Three Waters vs RBNZ prudential tools (NZ): Same country, same decade, same structural goal ( partition risk from a collateralised system). RBNZ tools (LVR, DTI restrictions): boring, low RSA, no identity payload. Still standing. Three Waters: sound DCC design, but the governance wrapper became identity-coded through co-governance provisions. RSA spiked. Recapture was immediate. Dead on arrival.

The Harmed Engineer

The only actors capable of building genuine fixes are those whose rent is negatively correlated with the dysfunction. If everyone benefits from the problem, nobody is structurally positioned to fix it.

For most insolvent-class problems, no Harmed Engineer exists. The actors with authority to build solutions are the actors whose rent is positively correlated with the brokenness. This is not coincidence — the system converges toward configurations where the people who could fix it are the people who benefit from not doing so.

The ECB in the euro doom loop is an exception — its institutional rent (credibility, authority, independence) is destroyed by sovereign default. Aviation safety is an exception — a single crash destroys massive rent simultaneously for every actor with authority. These exceptions exist because of specific institutional architectures that accidentally created actors with the right incentive alignment.

The design principle: If solutions are to work, Harmed Engineers must be engineered into existence before they are needed. Create institutional actors whose rent depends on system resilience rather than system persistence.

Central bank independence is the canonical example. Central banks are Harmed Engineers for monetary instability — their institutional rent is destroyed by currency failure.

The structural rarity of Harmed Engineers, design principles for engineering them into existence — including loss function templates, the conflicted actor diagnostic, mandate/prestige separation, and manufactured T6 — are addressed in the companion document: Harmed Engineer — Design Addendum.

The Double Bind

Post-transition-boundary systems experience a double bind:

The system cannot be allowed to die — too much depends on it.

The system cannot be allowed to heal — healing deletes rents across powerful actors.

This is how MSD feels from the outside. It's not a separate mechanism — it's the observable consequence of a system that's too broken to work and too embedded to fail.

Membership

A governance failure belongs to the Insolvent Class when four conditions are present:

N1. Rent/liability deadlock — no politically survivable path for dominant actors to simultaneously preserve rents and close the liability ledger.

N2. Disallowed failure — the system has crossed the transition boundary. System-level failure is externally unacceptable due to collateralisation.

N3. Recapture dynamics — reform attempts are absorbed, inverted, or neutralised to preserve existing rents.

N4. Temporal trap — the incentive to enact genuine reform emerges only after the system's capacity to reform has materially degraded.

Systems where failure is unrecognised rather than disallowed are candidate members — entering the class undetected because the causal chain between the structural failure and visible harm is long enough that liability accumulates before anyone notices. They become identified members upon collateralisation. This covers cases like AI alignment and AI capital flight: they are entering the class but have not yet locked in.

Exit Eligibility

Not all exits are available for all problems.

Problem type	Montreal	Singapore	DCC
Binary technical	✓	✓	✓
Continuous technical	✗	✓	✓
Moral / identity	✗	✓ (requires imposed framework)	✓
Universal rent (everyone benefits from brokenness)	✗	Practically blocked	Limited

For moral-class problems, Montreal is a mirage. For universal-rent problems, the exit landscape is maximally constrained: perpetual dysfunction, externally-imposed monopolar resolution, or catastrophic reset.

For some members — particularly universal-rent problems where all exits are structurally blocked or practically unavailable — the honest diagnostic endpoint is that the system is stuck. No exit is accessible within the current structural conditions. Acknowledging this is not a failure of the framework. It is the framework doing its job. A vocabulary that can only diagnose problems with solutions is less useful than one that can also name the cases where solutions are absent. In those cases, the lexicon's value is diagnostic clarity: helping practitioners stop wasting intervention resources on structurally closed pathways and redirect effort toward the problems that still have open windows.

When is this vocabulary useful?

When a practitioner — policy maker, journalist, analyst, engineer — encounters a governance failure and needs to:

Name what they're seeing. "This is recapture" is faster and more precise than explaining the full dynamic every time. "This intervention will fail the boringness constraint" is actionable in a meeting.
Recognise the shape. Different problems in different domains sharing the same structure. If you've seen one compound insolvency, you can recognise the next one faster.
Identify the intervention window. Before the transition boundary, everything is cheaper. After it, the options narrow. The vocabulary helps locate where a system sits.
Design solutions that survive. The boringness constraint, the Harmed Engineer principle, and DCC design are practical tools for anyone building institutions or policy.
Explain to others why things aren't working. "The debate is consuming the governance window" (temporal parasitic rotation) is a diagnosis that points toward action — stop oscillating between framings and do the boring technical work.

What this lexicon does NOT do

Make trajectory or point predictions. The framework identifies structural gradients — see What this is for the full three-tier distinction between gradient claims (which it makes), trajectory claims (which it doesn't), and point predictions (which it definitely doesn't).
Assign blame. Every actor described in this lexicon is behaving rationally within their incentive structure. The claim is about structural constraints, not individual failures.
Prescribe solutions. It identifies intervention windows and design principles, but the specific policy content is domain work. The lexicon tells you where to intervene and how to design for survival. It does not tell you what to build.
Claim completeness. The vocabulary covers patterns that have been observed across multiple domains. There may be dynamics the lexicon does not yet name. Terms earn their place by appearing in at least three independent cases.

Identified Members

Welfare

The case that produced MSD. Technically broken, politically stable. Both left and right need it broken: the right needs visible failure for austerity narratives; the left needs the programme's existence for compassion narratives. A system that eliminated poverty traps would reduce political rent on both sides. The dysfunction is the equilibrium.

The rent structure is symmetric and self-reinforcing. Caseworkers' institutional rent depends on caseload volume. Advocacy organisations' funding depends on the problem's persistence. Policy consultants are retained to study the problem, not to end it. Every reform attempt is recaptured: work-testing programmes become compliance theatre that increases administrative cost without reducing poverty; benefit increases are absorbed by landlords through rent adjustment. No Harmed Engineer exists — no institutional actor's rent is destroyed by welfare failure. The closest candidates (Treasury, who bear fiscal cost) lack operational authority, and their fiscal concern is easily reframed as austerity by actors whose rent depends on the narrative.

NZ Compound Insolvency (Housing + CGT + Migration + Birth Rate)

Four members that are mutually load-bearing. Every intervention on one destabilises the rent structure of at least one other. Partial intervention is self-defeating. The simultaneous intervention required is itself politically unsurvivable. Migration does triple duty: demographic duct tape, housing demand engine, and scapegoat absorbing blame the absence of CGT actually causes.

AI Alignment

No rent-preserving reason for incumbents to accept meaningful constraint. Harm not yet manifested. The governance window closes before the incentive opens. The alignment debate is a temporal parasitic rotation — the moral argument consumes the bandwidth the infrastructure work needs.

Infrastructure at Scale

Invisible outcomes, no campaign payoff, admitting need implies admitting decades of neglect. Every dollar spent on an underground pipe is a dollar a politician cannot spend on a ribbon-cutting. Maintenance budgets are extractable rent.

The rent gradient runs against maintenance at every level. Politicians extract rent from visible new builds, not from invisible pipe replacements. Engineering consultancies extract rent from crisis remediation at premium rates, not from routine upkeep. Insurers extract rent from risk pricing against degraded infrastructure. The recapture mechanism is specific: when maintenance budgets are allocated, they are redirected to visible projects through "deferred maintenance" — a euphemism for parasitic rotation of future capacity into present political rent. The temporal trap is acute because infrastructure degradation is non-linear: a pipe at 80% life costs nothing to maintain; at 95% it costs moderately; at 100% it fails catastrophically and the remediation cost is orders of magnitude higher. No Harmed Engineer exists in the political system. The closest structural analogue is the insurance sector, which is harmed by infrastructure failure — but insurers' response is to reprice or withdraw, not to fix.

Climate Adaptation

Has perfect attribution, scientific consensus, and known interventions — still stuck. Because adaptation is pure cost. Seawalls don't generate winners. Proves that the trap isn't about knowledge.

The rent structure is uniquely hostile. Mitigation generates diffuse, unattributable benefits decades in the future — no politician can claim credit for a disaster that didn't happen. Fossil fuel incumbents extract enormous rent from the status quo; their resistance is not irrational, it is structurally rational. The recapture mechanism operates through "green transition" branding: genuinely structural interventions (carbon tax, binding emissions caps) are replaced by non-binding pledges and offsetting markets that create new rent surfaces without reducing emissions. The conformer rotates from denial to delay to greenwash, each rotation consuming the political bandwidth that binding action requires. The temporal parasitic rotation is global: wealthy nations export adaptation costs to developing nations through climate debt, consuming their surplus capacity. No Harmed Engineer exists at the scale required — reinsurers are harmed by climate risk but lack political authority, and their structural response (withdrawal from markets) accelerates the problem rather than fixing it.

Insurance

The core business model is the insolvency. The customer pays for coverage; the company profits by not providing it. ACC demonstrates the recapture mechanism: a genuine structural exit pulled back into the class by political rent extraction.

Birth Rate Decline

Potentially the purest member. The temporal trap is biological — the childbearing window expires, creating a harder lock than any other member. No country has reversed it despite billions in pro-natalist policy. The system "solution" is parasitic rotation via migration — and that rotation has a termination date.

Two-Party Policy Convergence

The incumbents are the two parties. Rent preservation is the trap. Two parties competing within the same incentive structure converge on the same policy equilibrium — different narratives over the same attractor.

NZ Rugby

A small-market sporting body running a globally elite product it can no longer afford. Record revenue and consecutive annual losses simultaneously. The All Blacks brand is the system's only asset and the mechanism by which every other layer extracts rent. Provincial unions control constitutional voting power and depend on cross-subsidised funding from All Blacks revenue. They cannot vote to reduce their own funding or governance power — and they have demonstrated this, rejecting an independent board proposal (the Pilkington Report) and voting in a counter-proposal that preserved provincial representation.

The Silver Lake investment ($262 million for a minority commercial stake) demonstrates the recapture mechanism through traceable dollar flows. Intervention capital designed to professionalise commercial operations was absorbed by the existing rent structure: direct distributions to provincial unions, guaranteed interest payments to Silver Lake, and projected revenue initiatives that have not materialised. The deal also failed the boringness constraint — "selling the silver fern to American investors" was identity-coded from announcement. The talent drain runs as reverse parasitic rotation: NZ develops players through years of provincial and Super Rugby investment, then exports them at peak value to wealthier French, English, and Japanese leagues. The system exports finished product and imports nothing. The temporal trap is legible in broadcast cycles — each renewal at lower relative value cascades into smaller player budgets, more departures, and a weaker product. No Harmed Engineer exists. The coaching position — the closest candidate — is a sacrificial component the system replaces to maintain MSD, not an actor structurally positioned to reform the architecture.

Managed Unemployment (NAIRU) — borderline

Not clearly a full member, but structurally interesting as the only case where MSD is formalised as explicit policy. Orthodox economics prescribes a managed unemployment rate — the dysfunction is the stated objective, not a hidden consequence. Other members deny the dysfunction is dysfunction. NAIRU defends it as optimal design, and builds institutional machinery to maintain it. Whether this constitutes full ICP membership (rent/liability deadlock, collateralisation, recapture) or is better understood as MSD without the surrounding trap structure is an open question. It sits here as a marker, not a settled diagnosis.

Defeated Counter-Examples

Montreal Protocol / CFCs

Proposed as: Counter-example (governance solved it pre-crisis). Defeated: Both Montreal conditions were present. The solution was binary (ban, not regulate — no gradient to exploit) and incumbent rent was aligned with the switch (DuPont had pre-positioned to profit from substitutes, collapsing industry resistance once compliance preserved incumbent rents). Not a counter-example. The only identified Montreal exit — and an illustration of why the exit is so narrow: it required both conditions simultaneously.

Smoking / Public Health

Proposed as: Counter-example (policy eventually shifted). Defeated: The industry rotated the conformer to vaping. Same structure, new surface. The structure was not broken. It pivoted faster than governance could follow.

Seatbelts

Proposed as: Counter-example (industry resisted, then rapid adoption). Defeated: Class exclusion. The causal chain is zero — no seatbelt, through windshield. Not an insolvent-class problem.

Testing the vocabulary

The lexicon earns its place if:

Each term names something real. Practitioners recognise the pattern when they see the word. If a term doesn't trigger recognition, it's not doing work.
Each term names something that existing vocabulary doesn't. "Recapture" must do more than "political interference." "Compound insolvency" must do more than "complex problem." If standard language covers it, the new term is unnecessary. The overfitting test: the lexicon can describe it, but does the lexicon explain it better than existing vocabulary? If not, you're using expensive terminology for a cheap problem.
The vocabulary generates useful hypotheses. A practitioner who knows the boringness constraint will ask different questions about policy design. Someone who recognises temporal parasitic rotation will see the AI governance debate differently. The vocabulary should change what people notice.
Terms can be wrong. If a term is applied to a case and the expected pattern doesn't appear, the term is either misapplied or the pattern doesn't hold. A vocabulary that can't be wrong isn't a vocabulary — it's a slogan.
The class boundary holds. If something is identified as an insolvent-class problem and then exits through a mechanism the lexicon doesn't describe, the boundary needs revision.
Applications are falsifiable. Every application of the vocabulary to a specific system must name at least one observable that would falsify the diagnosis within a stated timeframe. "If we observe Y within N years, this ICP diagnosis or exit classification is probably mis-specified." A vocabulary that cannot be wrong about any specific case is unfalsifiable by construction, and unfalsifiable frameworks decay into pattern-matching. The discipline of naming what would prove you wrong is what keeps the vocabulary honest.

This is a living document. Terms earn their place by appearing in multiple independent cases. Terms that don't do work get retired. The vocabulary serves practitioners, not the other way around.

Other works by this author: creative.andyclarke.xyz

Licensed under CC BY-NC-SA 4.0.

Harmed Engineer — Design Addendum

Why this addendum exists

The Harmed Engineer was introduced in the lexicon as a diagnostic observation: the only actors capable of building genuine fixes are those whose rent is negatively correlated with the dysfunction. The lexicon identified existing examples (ECB, aviation safety) and noted that for most insolvent-class problems, no Harmed Engineer exists.

This addendum does two things. First, it explains why Harmed Engineers are vanishingly rare — and why that rarity is structural, not accidental. Then it extends the concept from diagnosis toward design: if Harmed Engineers almost never occur naturally, can they be engineered into existence? The answer is conditionally yes — with specific constraints, only before the system locks in, and with failure modes that the addendum's own design principles cannot fully close.

Why Rarity Is Structural

Harmed Engineers are not merely uncommon. They are actively destroyed by the systems they would need to fix. This is not an accident — it is a direct consequence of rent-maximising convergence.

ICP systems converge toward configurations where no actor has incentives negatively correlated with the dysfunction. A Harmed Engineer is, by definition, an actor whose rent is destroyed by the very dysfunction that everyone else profits from. That actor is structurally unstable within the system. The system doesn't ignore them — it eliminates them through the same mechanisms it uses to maintain MSD:

Budget extraction. The Harmed Engineer's resources become a target for the standard N3 recapture cycle. A working institution with a boring mandate is invisible — and invisible institutions lose political defenders, making their budgets available for reallocation.
Identity contamination. A single identity-coding event — one political association, one culture-war adjacency — permanently spikes RSA and fails the boringness constraint. The institution becomes interesting. Interesting institutions get eaten.
Metric redefinition. The loss function is decoupled by redefining what the controlled metric measures. If leadership survival depends on audit compliance, redefine compliance. If it depends on projection accuracy, change the projection methodology to something more politically convenient. The institution persists in name while the mechanism that made it a Harmed Engineer is quietly removed.

The result is that systems optimise Harmed Engineers out of existence under normal political pressure. The 99% case for regulators is that their budget expands with crisis severity, not inversely. More dysfunction means more staff, more authority, more political relevance. That's a positive rent correlation — the opposite of what a Harmed Engineer requires.

The canonical survivors — aviation safety, central bank independence — are rare, but not purely accidental. In both cases, the institutional architecture that created a genuine Harmed Engineer emerged from specific historical crises ( catastrophic crashes; hyperinflation and sovereign default) where acute salience happened to coincide with statutory independence being granted before peacetime politics could prevent it. The sequence was:

Crisis creates acute salience (T6 — ledger legibility forced by catastrophe).
A single actor receives statutory independence during the crisis window.
Loss function couples tightly to a metric the actor controls.
Upstream power cannot easily redefine the metric or redirect the budget — because the charter was written during the crisis, when the political cost of interference was temporarily too high.
No identity payload contaminates the institution's RSA during the consolidation period.

All five steps must occur in sequence. Central bank independence has been deliberately designed and exported across jurisdictions since the 1990s — the RBNZ's 1989 Reserve Bank Act is an example of intentional Harmed Engineer construction, not accident. But even deliberate designs required a crisis-origin template to copy, and every instance faces continuous recapture pressure once peacetime politics resume. Step 5 almost never survives indefinitely — central bank independence is under constant political assault in every jurisdiction that has it. Aviation safety survives partly because the loss function is so binary and so public (a crash is an undeniable, televised event) that metric redefinition is structurally unavailable.

These remain rare outcomes. They exist because of specific institutional conditions — some accidental, some designed — that are difficult to reproduce. But the fact that some were engineered deliberately is relevant: it means the design problem is hard, not impossible.

Diagnostic implication: absence is confirmation. When examining a candidate ICP member, the absence of a visible Harmed Engineer is not a gap in the analysis — it is evidence that the system has already optimised them out. Systems without Harmed Engineers aren't missing them by oversight. They've been converged away. If a Harmed Engineer were structurally viable, the system wouldn't be at MSD in the first place.

This means the design principles that follow are pre-crisis engineering specifications. A Harmed Engineer becomes progressively harder to create as collateralisation deepens. Post-N2 — once the system has fully crossed the transition boundary — every actor with authority to build the institution is already extracting positive rent from the dysfunction. The intervention window for Harmed Engineer design tracks the same gradient the temporal trap describes: it narrows before the political will to use it arrives, and the narrowing accelerates.

The practical consequence is uncomfortable: if you want Harmed Engineers to exist when you need them, you must build them before anyone believes they're necessary. That is the design challenge this addendum addresses — incompletely, and with open failure modes.

The Loss Function

A Harmed Engineer is defined by a loss function: a mechanically legible relationship between dysfunction and institutional pain. If the dysfunction metric rises, the actor's core institutional position degrades — not eventually, not reputationally, not in some diffuse political sense, but through a specified, auditable mechanism.

The loss function must be directly actionable within the actor's authority surface. This is the constraint that separates a Harmed Engineer from a concerned bystander. An actor who suffers when things go wrong but can't do anything about it isn't a Harmed Engineer — they're a victim. An actor who has tools but doesn't suffer when things go wrong isn' t a Harmed Engineer — they're a rentier.

Template:

Actor X's [leadership tenure / institutional prestige / charter continuity] degrades by mechanism M when metric Z ( which X directly controls through its operational mandate) worsens past threshold T.

Canonical examples:

Actor	Metric they control	Mechanism of pain	Why it works
Central bank (ECB)	Price stability via rate decisions	Independence destroyed by sustained inflation	Tools and loss function are coupled to the same metric
Aviation authority (NTSB)	Safety standards via enforcement	Public trust and institutional legitimacy destroyed by crashes	Investigation mandate is statutory; prestige is earned by outcomes

The common structure: the actor has entrenched tools (they can act regardless of political weather) and exposed prestige (they personally suffer when the metric they control moves the wrong way). The tools don't depend on the political cycle. The pain does depend on the outcome.

Open problem: who guards the metric? The addendum identifies metric redefinition as one of the three mechanisms that destroy Harmed Engineers, then proceeds to propose institutions whose survival depends on metrics. This is a genuine vulnerability. A captured board can degrade its own methodology to improve its apparent performance. "Independent audit" is the standard answer, but it defers the question rather than closing it — who audits the auditor is the same infinite regress that alignment discourse faces. The canonical survivors avoid this partly through binary, publicly visible metrics (a plane crash cannot be redefined away; sustained inflation is experienced by every voter). For metrics that are more abstract, technical, or methodology-dependent, the redefinition attack surface remains open. Any Harmed Engineer design must be evaluated against this vulnerability, and honest assessment will often conclude that it cannot be fully closed.

Mandate vs. Prestige: The Separation That Matters

A correctly designed Harmed Engineer has two structurally independent components:

Mandate and tools — legally entrenched, budget-protected, non-discretionary. The institution can act regardless of who is in government, regardless of political mood, regardless of whether anyone is paying attention. The mandate doesn't scale with the problem — it persists. This is what prevents the institution from becoming incapable precisely when it's needed most.

Prestige and leadership survival — inversely coupled to the dysfunction metric. The people running the institution are personally and professionally harmed when the metric they control worsens. This is what prevents the institution from becoming comfortable with failure.

If you couple budget to dysfunction (more dysfunction = less budget), you get an institution that's correctly incentivised but structurally incapable. If you entrench the mandate but don't couple leadership survival to outcomes, you get a regulator that persists without urgency. You need both pieces and they must be independent.

Hardwire rules:

Leadership auto-replacement if the controlled metric worsens past threshold for a defined period (e.g., two years running).
Prestige scoring via independent audit — leadership compensation, reappointment eligibility, or public standing tied to audited performance on the controlled metric.
Charter revocation trigger if the metric exceeds a catastrophic threshold for a sustained period — the institution's existence is conditional on minimum performance.

The Conflicted Actor Problem

The most common false positive when identifying Harmed Engineers is the conflicted actor — an actor that appears to satisfy the criteria on the harm axis but whose rent structure is mixed rather than cleanly negative.

A conflicted actor is simultaneously harmed by the dysfunction (real costs — operational difficulty, legitimacy erosion, downstream damage) and extracting rent from it (political capital, regulatory authority, campaign material, budget expansion). The harm is real. The rent is also real. And because rent operates on a shorter timescale than harm, the actor's survival function resolves the conflict toward rent.

Diagnostic test:

Does this actor's core institutional position improve when the dysfunction gets worse?

If the answer is "partly yes" — if there is any channel through which worsening dysfunction increases the actor's budget, authority, visibility, or political capital — they are a conflicted actor, not a Harmed Engineer.

National governments are almost always conflicted actors for the problems they regulate. They suffer from housing dysfunction but extract rent from housing policy. They suffer from platform harms but extract rent from regulating platforms. They suffer from climate costs but extract rent from climate debate. The harm is on a generational timescale. The rent is on an electoral timescale. The discount rate does the rest.

The three-question recapture-proofing test for any proposed Harmed Engineer:

Under a plausible political change, could this actor's budget or mandate be redirected to serve other purposes?
Could the metric they control become politically redefined to measure something more convenient?
Could the actor begin benefiting from mild dysfunction — gaining relevance, budget, or authority from the problem's persistence?

If two or more answers are yes, the design is recapture-vulnerable. The actor will converge toward conflicted status under normal political pressure.

Manufactured T6

A Harmed Engineer doesn't have to fix the problem directly. The most survivable design is an institution whose boring operational mandate mechanically produces the political precondition for action by others.

The mechanism is manufactured T6 — forced ledger legibility on a schedule. The institution's job is to make the dysfunction visible, attributable, and locally specific, on a recurring cycle, through audited data products that upstream actors cannot ignore.

This works because fog is load-bearing in most insolvent-class problems. Actors who benefit from dysfunction also benefit from the dysfunction being illegible. Nobody is accountable for a diffuse national problem. But an institution that produces regular, granular, auditable data attaches the liability to specific actors who received the data and chose not to act. The fog lifts. The upstream actor now stands next to a scoreboard.

The incentive chain:

The institution controls data quality, methodology, and timeliness (its authority surface).
Bad data quality triggers leadership consequences (its loss function).
Good data quality produces legible, attributable, locally specific liability (its output).
Upstream actors who receive good data and fail to act become the next generation of Harmed Engineers — their inaction is now documented and auditable.
The institution remains boring plumbing throughout. Nobody campaigns for or against data quality.

The institution doesn't solve the problem. It makes the problem impossible to hide from. That shifts the political economy from "nobody owns this" to "the person who received the data owns this."

The Boringness Paradox

This addendum names Harmed Engineers, explains their structural dynamics, and describes design principles for building them. In doing so, it raises RSA on the concept itself.

The boringness constraint says that surviving institutions must be too dull to attack. But a named concept in a published framework is, by definition, no longer invisible. The moment "Harmed Engineer" becomes a term in policy discourse, any institution identified as one becomes interesting — and interesting institutions get eaten.

The implication is that the best Harmed Engineers would be ones nobody recognises as Harmed Engineers. The design manual is in tension with the design principle. A central bank governor doesn't describe themselves as a "Harmed Engineer" — they describe themselves as doing monetary policy. The NTSB doesn't use the vocabulary of this lexicon. The boringness constraint implies that successful application of these principles would be invisible as application of these principles.

This is not a defect that can be patched by better design. It is a structural limitation of any framework that names the thing it is trying to protect. The vocabulary is useful for diagnosis and for the pre-deployment design phase. Once an institution is operating, the vocabulary should not be attached to it.

Design Principles — Summary

Separate mandate from prestige. Entrenched tools that persist regardless. Exposed leadership that suffers from failure on the controlled metric.
Couple the loss function to the authority surface. The metric that triggers pain must be one the institution directly controls through its operational mandate. Never tie leadership survival to downstream outcomes they can't influence.
Test for conflicted actors. The three-question recapture test. If two or more are yes, the design doesn't survive political weather.
Manufacture T6. If direct intervention is structurally unavailable, build boring plumbing that produces forced ledger legibility on a schedule. Make the dysfunction visible, attributable, and locally specific. Let the data create Harmed Engineers upstream.
Pass the boringness constraint. Every design decision. Every name. Every public-facing surface. If it's interesting enough to campaign on, it's interesting enough to recapture. The institution survives by being too dull to attack.
Acknowledge open failure modes. Metric redefinition cannot be fully closed for non-binary metrics. The boringness paradox means the framework's own vocabulary is a contamination risk. The transition boundary is a gradient, not a gate — design gets harder as collateralisation deepens, without a clean cutoff point. Honest assessment of any specific design will often conclude that not all vulnerabilities can be eliminated.

This addendum extends the Harmed Engineer concept from Part Two of the lexicon. It should be read alongside the core vocabulary — particularly Recapture, the Boringness Constraint, De-Collateralisation, and Collateralisation Triggers ( T1–T6). The design principles here are practical tools with acknowledged limitations. They describe how to build institutions that are more likely to survive — not institutions guaranteed to survive. Whether anyone builds them is a separate question the lexicon does not answer.