The Layer Walk

A protocol for using The Stack on a real failure

Elias Kunnas

The Layer Walk is the short operating manual for The Stack: how to take a concrete failing system, walk it through the twelve layers, find which layer is binding, locate the unowned channel through which reality is entering, and write a repair specification in four parts — owner, authority, capacity, feedback. It is not the same artifact as a Mechanism Analysis (the pre-enactment legislative document) or the production discipline behind one. It is the general-purpose diagnostic procedure for any failing institution, policy, or program.


I. What this page is

The framework's operational front matter now has five pages. The Stack is the corpus map — the twelve failure layers and the essays at each. Capital Stocks is the operational reference for the Ledger layer. The Mechanism Analysis defines the pre-enactment legislative artifact. How Mechanism Analyses Are Made defines the production discipline behind that artifact. This page is the fifth: the short procedure for diagnosing a failing system using the Stack and producing a repair specification.

The output is a repair specification, not a recommendation memo. The protocol's eight steps are: bound the case; state the failure; walk the layers; identify the binding layer; locate the unowned channel; write the repair specification; apply the viability test; tag confidence. The order matters; the artifact is only as strong as its weakest step.


II. Inputs — the evidence bundle

Before the layer walk, assemble the evidence bundle. The bundle is the first artifact the protocol produces and the first thing a reviewer should ask to see.

ItemWhat it contains
System boundaryWhat is inside the system being analysed; what is outside.
Time horizonOver what period the failure is observable and over what period the repair should produce visible movement.
ActorsThe named decision-makers, beneficiaries, cost-bearers, and intermediaries with material influence.
Observable failureThe concrete failure-as-symptom, stated in one or two sentences as a measurable gap between stated purpose and produced outcome.
Capital stocks affectedWhich stocks from the Capital Stocks reference are being built, depleted, or held flat.
Counter-explanationsThe strongest alternative diagnoses already in circulation, named explicitly. The protocol's diagnosis has to outperform them.
SourcesEmpirical evidence with provenance. Unsourced claims are flagged as such.
Success criterionWhat it would look like for the repair to have worked, stated in advance.

III. The layer walk

For each of the twelve layers in The Stack, ask the diagnostic question for the case in the evidence bundle and write a short answer. Thin or vacuous answers are evidence — they indicate the failure is not binding at that layer.

#LayerDiagnostic questionFailure signal
0TelosWhat is this system actually selecting for?Stated purpose and revealed purpose diverge under stress
1MechanismHow does it produce its outcome?Intent treated as cause; legal text assumed to produce behaviour
2ResponseWhat does the next actor find cheapest?Gaming, exit, substitution, withdrawal dominate compliance
3CarrierWhich harms have a procedural voice?One side has standing; the other absorbs cost silently
4MeasurementWhat signal is treated as reality?Metric optimised; underlying reality drifts
5LedgerWhich stocks are drawn down?Hidden depletion masked by visible flow
6CompilationDoes the decision frame exist?Pieces scattered; assembled public-usable frame missing
7Computation / adoptionDoes the system compute with the frame?Frame refused, captured, sloganised, or moralised away
8DecisionWho is forced to choose?Diagnosis exists; no one is required to act
9ExecutionCan the receiver implement?Repair exceeds the institution's bandwidth
10FeedbackDoes reality update the rule?Evaluation observes but does not correct
11ReproductionDoes the generator-chain continue?Output survives; the generator dies

IV. Finding the binding layer

Most real failures involve more than one layer. The binding layer is the one whose repair would unblock the others. Three patterns distinguish it from a merely contributing layer:

Joint-binding diagnoses (two layers binding together — most often Carrier–Decision or Compilation–Computation) are common and should be named explicitly, not implied.

Triage heuristic. In this corpus's applied cases, the binding failure often sits at Carrier (3), Decision (8), or a Carrier–Decision interaction — the architecture is wrong about who can challenge it or about who is forced to decide. Walking all twelve layers is the rigorous protocol; checking Carrier and Decision first is the rapid-triage version when time is short.


V. Locating the unowned channel

At the binding layer, the diagnosis is incomplete until it names the specific channel through which reality is entering the system without being received. The master heuristic is interpretively elastic — almost any failure can be loosely described as an unowned channel — so the template below operationalises it. Fill in every slot; if a slot cannot be filled, the diagnosis is not yet specific enough to support a repair specification.

SlotWhat it specifies
Reality signalThe concrete real-world signal the system should be receiving but is not.
Current routeThe path the signal currently travels — through which actors and intermediaries.
ReceiverWho in the current architecture is structurally positioned to receive the signal.
Who can ignore itThe actors who can decline to act on the signal without consequence.
Who bears the costThe class — often silent, future, statistical, or diffuse — whose interest is being absorbed by the failure.
Who should own itThe actor or institution that should take responsibility under the proposed repair.
Required authorityThe legal, statutory, or constitutional authority that would make the ownership real (advisory ownership is not ownership).
Feedback that bindsThe consequence — reputational, procedural, budgetary, statutory, constitutional — that ensures the owner actually corrects in response to outcomes.

VI. The repair specification

The repair specification has four parts. Each is required; an incomplete specification produces an incomplete intervention.

PartWhat it answers
OwnerWho owns the unowned channel under the repair? Named actor or institution, not a function.
AuthorityWhat legal, statutory, or constitutional authority lets the owner act? Advisory authority is rarely sufficient.
CapacityWhat resources, skills, and operational infrastructure let the owner execute? If the receiver lacks capacity, the repair's first deliverable is capacity, not action.
FeedbackWhat binding consequence ensures correction? Specify the level (reputational / procedural / budgetary / statutory / constitutional) and why that level is sufficient.

Before recommending the repair, apply the viability test: does the receiving institution have the capacity to implement the change without producing worse distortion than the original failure? If no, the capacity-build step is the repair's first deliverable, not an afterthought.

Finally, tag each claim with its confidence type: mechanism logic (deductive from causal architecture — highest confidence); structural inference (architecture as currently observed — depends on evidence bundle); behavioural prediction (how actors respond — lower confidence because behaviour depends on factors outside the mechanism); speculative cascade (second- and third-order effects — useful for scenario planning, not specification). Magnitude claims should never be presented with mechanism-logic confidence unless anchored to a quantified source.

The success criterion from the evidence bundle becomes the movement test: define in advance what observable change in outcomes would count as the repair having worked, and the time horizon over which it should be observable.


VII. Worked miniature — housing-supply collapse

Compressed for length. The intent is to illustrate the walk, not to validate the protocol.

Evidence bundle. System: state-and-local housing-supply governance in California, 2000–present. Observable failure: permits well below population-implied need; gap compounding over two decades. Capital stocks affected: demographic, social, fiscal, human (out-migration). Counter-explanations to outperform: developer rationality alone, NIMBY moralism, construction labour shortage.

Strong layer signals. Carrier (3): incumbent homeowners have full procedural standing; future residents have none. Ledger (5): local councils book wins on the visible ledger; demographic and fiscal depletion unbooked. Decision (8): state delegated implementation with no override; councils evaded by zoning required housing onto unbuildable parcels.

Binding layer. Joint Carrier (3) + Decision (8). The carrier asymmetry (homeowners have standing; future residents do not) makes the decision-layer evasion structurally available. Either repair alone under-performs.

Unowned channel. The decision-relevant interest of would-be residents enters the local-council system through no procedurally-recognised channel.

Repair specification. Owner: state housing agency. Authority: statutory provision stripping or weakening local zoning veto on demonstrated non-compliance — California's Builder's Remedy is the live form this repair has taken. Capacity: agency staffing and legal infrastructure to certify non-compliance and defend it in court — earlier programs (RHNA without enforcement) lacked exactly this. Feedback: statutory level; non-compliance produces concrete legal exposure. The statutory level appears to be the current live test of sufficient binding force; lower binding force had already proved insufficient over prior planning cycles.

What the protocol does not solve by itself. The local carrier asymmetry is backed by political-power asymmetry: homeowners vote in local elections; future residents, displaced workers, renters, and would-be household-formers have weaker or no local voice. A local process repair cannot overcome that by clever design alone. The repair requires a higher-tier owner — state, statutory, or constitutional — that can give the silent class procedural standing or override the captured local channel. Mechanism design does not substitute for political power; it specifies where political power must be reallocated, and what owner, authority, capacity, and feedback would make that reallocation real.


The argument in three sentences. The Layer Walk turns a failing system into a repair specification by walking the Stack, finding the binding layer, locating the unowned channel through the eight-slot template, and writing the four-part repair: owner, authority, capacity, feedback. Each step has an artifact; an incomplete artifact at any step produces an incomplete diagnosis or an incomplete repair. The page is short by design — it tells you how to run the procedure, not what the layers contain (Stack), what stocks to count (Capital Stocks), what a legislative mechanism analysis looks like (Mechanism Analysis), or how the production discipline works (How Mechanism Analyses Are Made).


Sources and notes

Status. The Layer Walk is an operational procedure, not yet a discipline (no shared epistemic community, no codified empirical methodology, no third-party replication). The page presents the procedure honestly — what it claims and what it acknowledges as unfinished — without claiming the field-formation work that would have to happen for it to be cited as a discipline.

Adjacent traditions. The procedure overlaps substantially with realist evaluation (Pawson and Tilley: context + mechanism = outcome), public choice theory (Buchanan, Tullock, Olson) on the Carrier layer, institutional economics (North, Williamson, Ostrom) on the underlying treatment of rules and incentives, the safety-engineering Swiss-cheese model (Reason, Leveson) on layered failure, and implementation science on the Execution and Feedback layers. The Layer Walk does not claim to have invented these treatments; the contribution is the specific layer-walk + four-part-repair organisation.

Methodological caveat on the housing miniature. The §VII miniature is a demonstration of use, not a validation. The diagnosis is constructed after the fact and against a case where the eventual repair had multiple advocates. The procedure's status as a useful frame depends on its evidence base growing through additional applications, particularly to cases the analyst had not previously studied.

Housing miniature. The California housing example draws on California housing-element law, RHNA compliance debates, and Builder's Remedy enforcement under the Housing Accountability Act. Useful public references include California HCD materials on housing elements and RHNA, ABAG / UC Davis primers on Builder's Remedy, and public reporting on California's continuing housing underproduction and permitting shortfall.

Known gaps. The protocol is strongest at Mechanism, Response, Carrier, Ledger, Compilation, and Reproduction. It is weaker at Execution and Feedback authority, and at the freshly-separated Measurement and Computation/adoption layers. Future essays will fill those gaps.

Related references: