Every candidate on the roster was attacked the same way: name the assumption it quietly imports, test the assumption against the constraints record and canon arithmetic, and let the verdict fall where it falls. The pass produced three kinds of result — assumptions that inverted (the constraint runs the other way in a habitat), assumptions that broke (deal-breakers), and assumptions that held with conditions (the technique survives, revised). It also produced the bundle’s meta-finding: adversarial analysis here did as much repricing as killing. Several techniques came out stronger than their advocates claimed, because their terrestrial objections were terrestrial facts.

Reference figures used throughout, all derived from shipped canon: a diet is ~10 MJ/day ≈ 100 W per person, ~100 MW for a Coriopolis-scale population of 853,000; the shell of a Coriopolis-scale habitat can reject ~400 W/m² over ~5,000 km² ≈ 2 TW, of which the sky at ~200 W/m² average already spends about half.

The inversions#

A1 — “Land is the scarce input”#

Source: all of terrestrial agricultural history; rides in on S1, S4, S15. Test: the nitrogen column. Open main-level ground carries ~3,500 kg of N₂ per square meter; a 3 m deck carries ~2.8 kg. Meanwhile feeding a person from open ground takes only ~60–120 m² (sky at ~200 W/m² average, ~1% conversion to edible energy ≈ 2 W/m² against a ~100 W diet) — 50–100 km² for the whole city, one or two percent of the main level. Verdict: inverted twice. Land is not scarce in area (a few percent feeds everyone) but is absurdly dear in nitrogen — and the same arithmetic makes decked volume nearly free. The scarce input is neither land nor light nor energy; it is open sky, which food production specifically does not need. Consequence: production moves into decks; open ground survives only where the ground was wanted anyway (S2).

A2 — “Vertical farming fails on energy costs”#

Source: the 2020s terrestrial verdict on S3, imported wholesale. Test: lamp-lit staples at photosynthesis’s few-percent wall-plug-to-edible efficiency cost ~4 kW per person, ~3.3 GW for the city — a third of a percent of the thermal budget remaining after the sky, powered from collectors the Works produce at negligible marginal cost. Verdict: inverted. The objection that killed vertical farming on Earth was an electricity price, and there isn’t one. Deck farming is not the expensive option here; it is the workhorse. The genuine constraint on S3 is containment discipline (C6), not power.

A3 — “Sunlight from the sky is free, so farm under it”#

Source: planetary instinct; rides in on S1, S4. Test: C1 + C2. The sky’s photons are indeed marginally free to a crop standing under them — the light was already bought, for residents. But the crop gets no control: spectrum, day length, and season are amenity settings, and no operator will run the sky for wheat. Verdict: held, with the price relocated. Under-sky growing is thermally free and agronomically hostage. That combination selects exactly the crops whose value survives it — perennials, vines, paddies, grazing — i.e., S2, and rules out S1’s row-crop staples, which want the one thing the sky will never give them: light scheduled for yield.

A4 — “A field in a habitat behaves like a field on Earth”#

Source: S1 again. Test: the gradient-invitation corollary plus month-scale dynamics (C5, C6). A monoculture is an unexploited gradient held open on purpose; on Earth, winter, distance, and deep ecology slow the race for it; a habitat has none of those brakes, and its excursions run to completion in months. An unwalled field is a standing invitation the farmer cannot withdraw. Verdict: broke. Open production must either go behind walls (S3, S6, S11–S13) or dissolve into the ecology as patchy polyculture with the biome’s own predators and a sky-scheduled winter as sanitation (S2). Open-field monoculture inside a biome is dead. This is the analysis’s cleanest kill, and it came entirely from shipped canon.

A5 — “Storage is expensive in space”#

Source: spacecraft intuition, where every kilogram is fought for. Test: a year of staple calories is ~230 kg of dry store per person (3.65 GJ at ~16 MJ/kg); for the whole city ~200,000 t, which at bulk density fits under a tenth of a square kilometer of deck — deck volume being the thing C3 makes nearly free. Verdict: inverted. Multi-year reserves are trivially affordable, so doctrine can demand them (R7), and famine reclassifies from physical risk to governance failure. This single inversion does more for the setting’s food security than any production technique.

A6 — “Closed loop means self-sufficient”#

Source: the CELSS literature; rides in on every “the habitat feeds itself” framing. Test: trace the fixed flows for a habitat that imports food. Diet moves ~4 kg of fixed nitrogen and ~0.5 kg of phosphorus per person per year — ~3,500 t N and ~440 t P annually at city scale — and every imported meal carries its share inward with no return leg unless one is built. The sink for the excess is the habitat’s waters, whose buffering capacity is canon’s scarcest reservoir (C5). Verdict: held, with a new requirement. Importing food without exporting recovered nutrients is a slow eutrophication program. The loops close, but at cluster scale, and the return manifest is as real as the outbound one. This became R4’s second clause and a standing fact on the production page.

The deal-breakers#

D1 — Agricultural gyrealms (S15)#

A habitat’s defining cost is its nitrogen-filled open sky (C8). Bulk food production is the one activity the analysis shows specifically does not need open sky (A1–A3). A farm habitat therefore spends the Solar System’s only priced input on the one customer that cannot pay for it: the proposal is self-refuting at the level of its own premise. Survives only as a cultural pattern — habitats whose residents choose large landscape-agriculture identities — which is S2 at scale, not S15.

D2 — Bulk food from Earth (S17)#

Lift is ~34 MJ/kg (C9); dry staples contain ~16 MJ/kg. Bulk food up the well costs roughly twice its own caloric content in lift energy and competes for the same launch capacity as the nitrogen the habitats actually need. Dead on one line of arithmetic. Luxury imports (S18) are untouched — provenance has carried price across every transport barrier in history.

D3 — Surface farms exporting upward (S16)#

A lunar-surface kilogram still costs ~3 MJ of well to reach orbit before any capital, against ~0.5 MJ/kg for a free-flyer that starts and finishes there — and the surface farm needs the same enclosed growing volume a free-flyer needs, plus the regolith works, minus the free solar pointing. Strictly dominated for habitat supply. Survives as local supply for surface populations, which was never this bundle’s question.

D4 — Diluent-free crop atmospheres (the aggressive version of S14’s pressure tailoring)#

Canon is explicit that flammability tracks oxygen mole fraction (C10). A low-pressure pure-O₂ greenhouse is a dry-biomass fire waiting for an ignition source. Revised, not killed: hypobaric growing atmospheres keep a nitrogen (or lean) diluent and hold the O₂ fraction down; the free-flyer’s nitrogen bill stays negligible anyway (see A7).

D5 — Algal staples (S10) and chemotrophic exotics (S19)#

Algae’s terrestrial advantages — land and water efficiency — are non-binding constraints here (A1), its lamp photons cost exactly what crop photons cost, and it loses the palatability comparison to everything above it on the roster. Demoted to feed and specialty lipids. The exotics fail on yield by orders of magnitude and are dismissed without ceremony.

D6 — The terrestrial livestock ladder (S7)#

Three independent attacks, any one sufficient at scale: conversion (ruminants burn 7–10 dry calories per edible one, multiplying the entire upstream chain); epidemiology (a shared airshed with month-scale dynamics cannot host large epidemic reservoirs — this caps flocks and herds by ordinance, not economics); culture (industrial slaughter in a sealed world where nothing stays out of sight). Survives in fragments, each for a specific reason: poultry contained and capped; ruminants in the thousands as landscape instruments with dairy as by-product; insects as residue-fed feed converters; pigs and feedlots priced out entirely, their products ceded to S12. The fragments are load-bearing; the ladder is gone.

Assumptions that held#

• A7 — Free-flyer nitrogen economics (S14). Checked rather than assumed: a resident’s food footprint in low-pressure growing layers holds tens of kilograms of gas; the same resident’s habitat share is ~27,000 tonnes. Nearly six orders of magnitude. Free-flyers are as nitrogen-cheap as anything pressurized can be. Held.
• A8 — Robotic tending (S14, S3). Canon already makes habitat construction fully automated; greenhouse horticulture is not the frontier case. Held.
• A9 — Radiation outside (S14). Chronic dose is tolerable for crops; solar particle events are met with shuttered bays and shielded seed stores rather than hull shielding; germplasm redundancy (A5’s cheap storage) absorbs the residual. Held with standard mitigations.
• A10 — SCP acceptability (S11). History’s verdict is consistent: microbial food fails as a dish and succeeds as an ingredient. The system design assumes only the ingredient role — flours, feed, reserve stock — and the requirement (R6) is carried by other tiers. Held as scoped. Purine processing and B₁₂ are solved problems and stayed solved.
• A11 — Containment-by-chemistry (S11). The reversal that pays for the whole pass: the gradient-invitation corollary, which killed S1 in A4, protects S11 — an escaped hydrogenotroph finds no hydrogen gradient in an open biome and starves. Substrate choice is intrinsic containment. Canon’s most dangerous principle, run in reverse, is a safety mechanism.

Revisions the pass forced#

1. S1 deleted; S2 promoted to the sole legitimate form of open-ground production, run as ecology (polyculture, patchiness, biome predators, sky-winter sanitation), priced as amenity by-product.
2. S3 promoted to workhorse, with the constraint restated: sealed decks with independent air handling — the wall, not the watt, is the scarce discipline.
3. S5 deleted; S6 standard. Open waters stay artisanal; production fish live in closed loops.
4. S7 shattered into its surviving fragments (poultry / ruminants-as-landscape / insects-as-feed), with S12 absorbing the demand the fragments release.
5. S11 + S13 designated the survival floor: biome-independent calories and protein, sized so the biome can fail without anyone missing a meal — a design requirement, not an accident.
6. S14 ratified as the bulk tier, hypobaric-but-diluted (D4), distributed, robot-tended — paired mandatorily with A5’s reserves and a domestic deck floor, because the scoop-attack precedent makes concentrated external dependence the food system’s real vulnerability. Efficiency arguments do not outrank the L5 events pages.
7. The trade rule made explicit: importers export recovered nutrients (A6). Written into the production page as ordinary practice.

Standings after the pass#

Reconciliation notes and open questions#

• “Agricultural zones” reconciled. Gyrealm puts agricultural zones on the main level; the analysis makes the main level the least production-suited ground in the habitat. No contradiction once the phrase is read as what shipped: the zones are S2 — landscape agriculture — which is exactly what a visitor-visible main level would carry. The bulk system is enclosed and invisible, which is why the wiki could go this long without describing it.
• The largest technology bet is S13 (commodity cell-free carbohydrate synthesis), extrapolated from 2020s bench demonstrations. It is load-bearing for the “survival floor” doctrine. If the user ever wants it out, S11 plus tuber-heavy decks can carry the floor alone at modest cost to the argument.
• Deliberately left open in-universe: which habitats keep pigs or eat insects directly (cultural variance is canon’s own pattern); whether specific named habitats specialize as exporters; the scale and institutions of the food trade; any named cultivars, dishes, or signature foods — all naming waits on the user per standing rule. Logged as CN-12.