Published live (2026-07-03). The flagged next step, taken now because its precondition is finally met: three definite treatments exist — energy (the ratchet cell), mobility (the second commodity), curriculum (the stowaway) — and three is the minimum for distillation. One run is an anecdote; two runs share coincidences; three let the structure that repeated be separated from the structure that merely happened. This note is that separation: the treatment method written down as a tool spec — what is essential (a fourth treatment that omits it is not a treatment), what is a scope condition (the runs disagree, and the disagreement is carried, not smoothed), and what was run-specific (free to vary). Per the observer-architecture note: specifications delegate, continuities don't. This is the specification — the part of the world computer's first standing instrument that can be handed to a run that its author never touches.


The unit — what a treatment is

A definite treatment is not a principle and not a search residue. It is a specified, replicable design — a template with a worked instance — put through the bridge gate as its condition of publication and published with the verdict on its face: conditions, degradations, and the sharpest unanswered objection printed, never managed away. The deliverable is always the template; the instance (Bethesda, the A5 corridor, one Surrey classroom) is the evidence the template is real, not the point. A treatment that ships without its gate verdict is a proposal; a verdict that ships without its conditions is marketing.

The pipeline — stages essential, counts not

Five stages, in order. The counts each run used (fourteen agents, four researchers, three adversaries) are tuning; the stages and their ordering are the tool.

  1. Evidence fleet. Sourced findings against the live world, confidence flagged on every figure, never invented. Its job explicitly includes contradicting the prior work — the mobility fleet killed the previous run's "only move" claim, and that contradiction was the run's most valuable output. If the fleet degrades or fails to launch, the degradation is printed first and the verdict weakens accordingly: the forest run's fleet was refused at the permissions wall, and the verdict dropped to the gate's weakest form for exactly that reason. Unverified legs are carried forward as open, never silently assumed.
  2. Expansion lenses. Generate configurations that dissolve trade-offs rather than trade them off — re-levelling first (is this the right altitude?), then cross-domain and constraint-denial passes. This is the search commitment inherited from the iterative loop: a trade-off is a hypothesis that the frame is too small.
  3. Walls pass. Every residual compromise classified into the three-class ledger (below) — classified, not merely listed.
  4. Synthesis. The ladder (below) — one design, in the grammar.
  5. Independent adversaries. Attack the synthesis against the bridge test; the verdict comes from their agreement and is graded honestly (below). Adversaries independent of the synthesiser is the load-bearing property; three of them is a choice.

Standing order across all five: evidence before design; walls classified, not listed; the gate run on the result, not the intention.

The ladder grammar

The design output is always a ladder, and its grammar has seven non-negotiable clauses:

  1. Every rung individually rational on day-one terms for its adopter. No rung's rationale may cite the whole, the future, the grant, or the cause. The cell, the mutual, the movement — all of it arrives as a by-product of self-interested rung-climbing.
  2. No rung requires the next; every rung makes the next cheaper. A ratchet, not a leap. Where a rung is an accelerant rather than a gate (the party-wall dyad), it says so and the ladder routes around it.
  3. Minimal permission surface. No event at which anyone approves the whole; each step small, self-assessed, inside permissions that already exist. The route-around beats the chokepoint by having no single point of refusal.
  4. Origination is paid. Rung 0 is a funded job producing an artefact kit — never volunteer faith. And any residual saint the design cannot pay for (the energy run's unpaid fiduciary directors) is confessed in print, not hidden.
  5. Value bases must survive the design's own success. Price rungs off what persists — self-consumption, not the retail/export spread the design itself compresses; planning-cost relief, not enrichment appeal. No projection may stand on a number the ladder's adoption erodes.
  6. The headline is a free option. The quotable endgame rung — the valley out-negotiating Waymo, forest maths — is pre-demoted before the gate ever sees it: the ladder must pay in full without it. Every run so far has had one, and in every run it was the weakest claim.
  7. The institution arrives late, as paperwork catching up with trades — a wrapper forced by the accountant, not a founding meeting. (The regime this holds in is a scope condition; see the disagreements section.)

The gate — bridge, catchment, median

The bridge test is the gate: a configuration passes when adoption needs no campaign, no exhortation, no heroic act of collective will — anyone can see instantly that it's easier. Its adversarial form, refined across the three runs:

The walls ledger

Every wall the design cannot dissolve goes in a ledger with exactly three classes, and every entry names its residue — no dissolution is complete, and the residue line is mandatory:

Class Discipline Residue takes the form of
Physical floor Bought, never designed away. No rung may promise past it (no grid independence; no 2am car without paying for idle slack; no artefact that dissolves thirty children in the rain). A recurring bill, a surveyed finite pool, a conceded cohort.
Political contest Routed around or conceded — never assumed won. Mitigation is posture (move while the window is open) or altitude (push load-bearing value below the contested layer). The distributive fight itself, stated as a fight — network-cost shifting, the FCA perimeter, MAT standardisation.
Frame-artefact Dissolved into the design by construction — mutual ownership dissolving the split incentive, two books dissolving the fare cap, subject choice dissolving SATs risk. Whatever the dissolution displaced, named beside it.

The standing rule, inherited from the search loop: treat every wall as a frame-artefact until it bottoms out on a floor. Relentless about the frame; honest about the floor.

The scan — find the waitlist

Before computing anything, scan the domain for queues: unsubsidised, unexhorted, and for a non-positional good. A qualifying queue is the world announcing it has already run the adoption experiment, and it points at the binding constraint (in Bethesda: generation, not persuasion). An empty result is the lens working, not failing — mobility's silence said no one had half-built that bridge, and the design brief changed accordingly. And the third run added the precision clause: identify what the queue is for at the level the queuers experience it. Teachers queueing at the photocopier are queueing for planning-cost relief, not for wellies — the queue is for the function, not the analyst's favourite instance of it, and a bridge that enters by any door other than the queue's own function will stand unused beside a visible queue.

The clock

Every run so far has bottomed out on a clock, not a wall — the residue that survives every dissolution is temporal. Three types, one per run:

One line gates all three: routing around rent is not routing around protection. A design that minimises its permission surface to escape accountability rather than gatekeeping fails the viable objective regardless of its clock. The clock never licenses the extractive playbook.

The capture rule, and the run obligations

The threshold for capture is existence, not merit — merit is decided by the climb, in public. The spine's entry ramp is the selector; rationing the door duplicates it badly. Two obligations this places on every treatment:

  1. Capture its own candidates — every generalisation the run throws off enters at forming with its sharpest objection attached, load-bearing nothing.
  2. Discharge its obligations to the standing candidates, explicitly. Each run must report what it did to bridge-catchment, institutions-precipitate, find-the-waitlist, and the-closing-window — and to whatever candidates later runs add — including negative and empty results, which are findings, not failures. The forest run's candidates section is the worked form.

The honest verdict

The gate must be able to partially fail a design it likes — a gate that only rubber-stamps or only kills is doing no discrimination. Three runs, three demonstrably different behaviours, which is the evidence the gate is real:

Standing clauses of the discipline: process honesty prints first (the lost researcher, the refused fleet — before any finding); the verdict is never stronger than the process that produced it; the sharpest unanswered objection gets its own section, especially when it cannot be answered; conditions are numbered and concrete enough to be discharged or to fail; and every treatment ends with climbs-when criteria falsifiable enough that not-climbing is also informative.

Where the runs disagree — scope conditions, not smoothing

Three live disagreements. The spec carries them as scope conditions; papering over them would be exactly the ad-hoc variability the work's own Deutsch criterion forbids.

  1. Precipitation. The mobility run found pure precipitation blocked standalone — precipitation requires rails where accumulated value legally forces a wrapper; the first commodity pays the founding cost and later ones inherit it. The forest run found unforced precipitation possible where nothing legally forces incorporation — but slow, optional, and thin, with a wrinkle that cuts the other way: the precipitate acquires interests and can occlude the widest catchment (the FSA guarding its definition against curriculum mapping). Scope condition: which regime a domain is in is set by whether its accumulated value legally forces a wrapper. Neither run licenses "institutions always precipitate," and the 1988 national curriculum stands as the honest anti-case — a founding act that worked, in exactly the class the candidate's own scope condition predicts (value whose first unit requires the whole).
  2. The waitlist. Present (energy), absent (mobility), present-at-function-level-only (curriculum). Essential: running the scan. Not essential: a positive result. The scan's three outcomes brief three different design postures — build to the queue, build the first half of the bridge, or enter through the queue's actual function.
  3. The clock's direction. Two clocks close; one opens. "The political residue is a clock" has generalised across all three runs; "the clock is closing" has not. Type the clock before writing the doctrine.

And one cross-run finding carried as a candidate, not a clause: the cell is the unit that compounds, not the commodity — energy was not the first treatment but the first rail, and the second commodity arrived at marginal founding cost. Two runs support it; the third is silent (a school may be a pre-existing rail — untested). It stays a candidate at forming, per the capture rule.

Explicitly run-specific — free to vary

For the avoidance of reverence: rung counts (nine, eight, six — the grammar doesn't count rungs); agent and adversary counts; the domains and the worked instances; the specific conditions each verdict imposed; the specific walls each ledger carried; which lens found the decisive contradiction. A fourth treatment that reuses the energy run's rung count or the mobility run's two-books move because "that's the method" has confused the sentence for the grammar. The spec is the grammar; each run writes its own sentences.


Not yet promoted into propositions.ts — capture, not climb. Forming register. Climbs when: (1) a fourth definite treatment is run from this document — following the spec rather than the author — and the gate again does real discrimination; (2) the spec's own sharpest objection is faced: a method distilled from three runs conducted by the same two seats may have codified the seats' box, not the method — the honest test is a run where someone else holds the gut and the verdict; (3) the three scope conditions accumulate cases — a precipitation regime called in advance, a waitlist scan whose posture-briefing is used, a clock typed before its doctrine is written.