Lintel
Lintel · Model Autopsy Working document · HAP-10 · Rev 1.0 · August 2026

Sixteen models.
None passed all three.

Every structure tested against the same three requirements. Each one satisfied two and broke on the third. This records where each broke, and — more usefully — which component layers came out of the wreckage already solved.

Models tested16
Passed all three0
Layers solved5 of 11
Open, all testable6

Every assessment on this page is a judgement, not a measurement.

Each PASS, PART and FAIL is one working session’s reading of one model, made against no measurement of any kind. Nothing here was tested. Nothing here carries a row in the evidence register, and that is deliberate rather than an oversight: a register ID would imply a graded claim behind a stamp, and there is no such claim.

What the page is for is the salvage lines — what each failed model left behind that a later one can reuse. Those are the substance. The stamps are the filing system.

HAP-10 · Rev 1.0Dated August 2026 Working documentAppend, do not overwrite
The three requirements

What every model was tested against

A model passes only if it satisfies all three at once. Two out of three is the trap — it produces something that works, but not the thing being looked for.

TEST A

Delivers

Does a household actually pay meaningfully less for housing because this exists? Not indirectly, not eventually.

TEST B

Scales

Does it work nationally without being bound to one locality, one negotiation at a time, or one dollar of equity per unit?

TEST C

Funds

Can it raise capital on terms someone will actually accept, and return it on a timeline that capital tolerates?

The record

What was tried, and where it broke

Ordered roughly as explored. The salvage line is the part worth keeping — the component that survived even when the model didn't.

Show
01

Conformance authority

Sell a compiled determination — signed by a professional, insured — that a jurisdiction accepts in place of first-pass review.

Broke on

Assent is granted one jurisdiction at a time and cannot be bought. Insurance is priced against correlated software error with no loss history.

Salvage

The placement compiler itself. Still needed, just not as the product.

A DELIVERSFAIL
B SCALESFAIL
C FUNDSFAIL
02

Advisory screening

Compiled envelope queries sold to builders and homeowners. What Symbium, Canibuild and CivCheck already do.

Broke on

Advisory information without legal weight moves a fraction of the timeline. Already commoditised by better-distributed incumbents inside city hall.

Salvage

Proof the compilation works at commercial scale. Someone else paid to demonstrate it.

A DELIVERSFAIL
B SCALESPASS
C FUNDSPART
03

Zoning indemnity

Don't determine the answer — insure it. Sell through the existing title endorsement channel rather than to jurisdictions.

Broke on

Title insurance has low losses because title is a records search. Zoning is interpretation, and the claims concentrate in the ambiguous tail where the machine is weakest.

Salvage

One national counterparty replaces twenty thousand local ones. The insight survives even if this product doesn't.

A DELIVERSPART
B SCALESPASS
C FUNDSPART
04

Accessory dwelling wedge

The homeowner is the developer. Backyard unit on land they already own, utilities already in the street.

Broke on

One or two units per lot. Fifty homeowners is fifty sales, not one project. And unless the jurisdiction has opted into separate sale, it produces rentals rather than owners.

Salvage

The only configuration with zero land cost, zero entitlement and zero horizontal infrastructure. Still the best unit economics found.

A DELIVERSPART
B SCALESFAIL
C FUNDSPART
05

Community pad infill

Acquire manufactured housing communities, fill vacant sites, capture the lift in net operating income on refinance.

Broke on

Residents stay on chattel loans at roughly double the rate, on land they don't own. The returns are real and they come from the residents. This is a trade, not a solution.

Salvage

Community lending programs finance the homes at near-zero sponsor capital. Home financing is not the hard part.

A DELIVERSFAIL
B SCALESFAIL
C FUNDSPASS
06

Rental portfolio backyards

One institutional owner of scattered single-family rentals, one credit decision, fifty-plus backyards.

Broke on

Produces rental stock rather than ownership. And the July 2026 restrictions on large institutional purchases of single-family homes narrow the counterparty set considerably.

Salvage

Batch scale is available if you find a counterparty who already owns the land.

A DELIVERSPART
B SCALESPART
C FUNDSPART
07

Community land trust

A trust holds the land permanently; the household buys the improvements under a long ground lease with a resale formula.

Broke on

Once the lease is 99 years with capped ground rent, the appreciation is real but decades away and illiquid. Almost no capital accepts that, which is why these entities are nonprofits.

Salvage

Both agencies purchase mortgages on these leaseholds, explicitly including manufactured homes. Permanent affordability with conventional financing already works.

A DELIVERSPASS
B SCALESFAIL
C FUNDSFAIL
08

Land bank conveyance

Public entities convey serviced, platted, utility-connected lots for nominal cost — sometimes with per-unit subsidy attached — to whoever will build affordable housing on them.

Broke on

Every land bank is a separate negotiation with its own design standards, which frequently exclude factory-built homes outright. Lots are scattered and often too narrow to set a multi-section home.

Salvage

Land acquisition financing is not required. The hardest thing in real estate to fund is being given away — to entities that can build.

A DELIVERSPASS
B SCALESFAIL
C FUNDSPART
09

Ground lease, uncapped

Occupant trades land appreciation for cheaper monthly cost; owner realises it on eventual sale. Different time horizons, same asset.

Broke on

Three contract terms, not the structure. Short lease means rent rises and reclaims the discount. No owned land means the home stays personal property. No exit means a captive tenant.

Salvage

The same structure works with a 99-year renewable lease, an agency-approved form, and formula-capped rent. That's model 07.

A DELIVERSFAIL
B SCALESPART
C FUNDSPASS
10

Shared appreciation

Keep the land in the sale; sell the appreciation separately. Household keeps the deed and the mortgage; an investor takes a share of future value.

Broke on

The investor's return requires the home to get more expensive. That is structurally opposed to the household needing it to stay affordable.

Salvage

Proof you can separate appreciation from occupancy without destroying real-property status. Applied to purchase rather than cash-out, it's unexplored.

A DELIVERSPART
B SCALESPASS
C FUNDSPASS
11

Remote master plan

Buy cheap land far from the centre, build at scale, pair a nonprofit developer with a for-profit affiliate.

Broke on

Cheap land is cheap because it has no infrastructure — you pay to build the network instead of tapping one. Households repay the saving in commuting. The entity pairing risks private inurement.

Salvage

Nothing. This is the model to remember as a warning.

A DELIVERSFAIL
B SCALESFAIL
C FUNDSFAIL
12

Buyer registry

Aggregate verified demand — advertising to deposits — before committing to land or production.

Broke on

Not a standalone business; an input to one. And housing advertising cannot be targeted by parcel, ZIP or demographic, so acquisition is broader and blunter than assumed.

Salvage

An order book no appropriation can cancel — the thing every prior industrialisation attempt lacked. Cost per verified buyer is the single most load-bearing unmeasured number.

A DELIVERSPART
B SCALESPASS
C FUNDSPART
13

Small-dollar origination

A national lender for factory-built homes: convert to real property, route each borrower to the product they qualify for, originate at a cost that makes a $75,000 loan viable.

Unresolved

Fifty-state licensing, supervision, and fair-lending obligations in a channel with documented racial disparities. Demand economics unmeasured.

Salvage

The affordability delta and the revenue event are the same transaction — roughly four points of rate, permanently, on the same house. Nothing else tested has that property.

A DELIVERSPASS
B SCALESPASS
C FUNDSPASS
14

Matching layer

Produce matched, pre-verified, pre-financed transactions — a buyer, a placeable parcel, a certified type, a lender. Local partners build.

Status

Current candidate. Requires all three gates to pass where earlier models needed one thing at a time. Supply is the hard side, not demand.

Salvage

The asset stays local; the transaction goes national. It is the only model where all three requirements are simultaneously plausible.

A DELIVERSUNTESTED
B SCALESUNTESTED
C FUNDSUNTESTED
Read this stamp before that line Model 14 is a working hypothesis and nothing more. It is stamped UNTESTED on all three tests and stays that way until something measures it. Note also what its salvage line does: it argues for the model rather than recording a component recovered from a failure, which no other salvage line on this page does. That is left exactly as written, because rewriting a working note to look more modest afterwards is a worse dishonesty than printing it with the tension marked.
15

Vertical demonstration

Build ten to twenty units yourself in one jurisdiction where the law already permits it, and prove the cycle exists.

Broke on

Local and capital-bound by construction. It is not a scalable company and was never going to be.

Salvage

The only path that produces a demonstration rather than an argument. Every prior failed attempt had conviction; none had a measured result.

A DELIVERSPASS
B SCALESFAIL
C FUNDSPART
16

Measurement and model law

Publish what nobody has measured — placement gaps, certification tracking, the origination decomposition — and supply the enabling language states need.

Broke on

It is not a company. It delivers through others, on a policy timeline, with no equity value.

Salvage

Grant-fundable now, and it is the only item on this list that is already achievable with what has been assembled. Fifty legislatures are on a twelve-month clock and nobody is tracking them.

A DELIVERSPART
B SCALESPASS
C FUNDSPASS
Component inventory

What the wreckage left behind

Phase 2 candidate — pending, not merged The component inventory below maps onto register rows and would strengthen the argument about what is unmeasured. It has deliberately not been folded into the case on this site, and nothing elsewhere cites it. It is recorded here as thinking, and stays that way until Phase 2.

The models failed. The layers underneath them did not fail uniformly — five are genuinely solved and reusable by any future model. The six that are open are the entire remaining problem, and every one is cheap to test.

Solved

The home

A federally certified factory-built house averaging roughly $115,000, against a median new home near $500,000. Production sits near 100,000 units a year against more than 350,000 in the late 1990s.

Source: HUD Code, 1976. Idle capacity is a feature.
Solved

Construction approval

Federal preemption of local construction codes, with design approval delegated to accredited private agencies since 1976. Type certification, already running.

Source: 24 CFR 3282. This is the aviation model, domestically.
Solved

Land access

Three channels require no acquisition debt: a lot the homeowner already owns, a pad an operator already owns, or a serviced lot conveyed by a land bank for nominal cost.

From models 04, 05 and 08.
Solved

Buyer financing instruments

Both agencies purchase mortgages on trust-held leaseholds including manufactured homes; renovation products finance a factory unit as an accessory dwelling; chassis-free homes became federally definable in July 2026.

Solved on paper, barely used in practice. That gap is the opportunity.
Solved

Tenure

A 99-year renewable ground lease on an agency-approved form with formula-capped rent converts the ground lease from an extraction device into permanent affordability with conventional financing.

Three contract terms separate model 07 from model 09.
Open · Gate A

Placement

What share of residential parcels legally permit a factory-built home, and are the binding provisions objective or discretionary? Never measured anywhere. Design-compatibility standards are the exclusion mechanism, so measuring the discretionary share is measuring the exclusion.

Two weeks, no capital. Publishable either way.
Open · Gate B

Demand economics

What does a verified, deposit-paying buyer cost, and what share of them qualify for a mortgage? If the people who want a cheap factory home cannot get financed for one, that is the whole thesis measured from the demand side.

Under $25,000. Broad targeting only — housing ads cannot be narrowed by parcel or demographic.
Open · Gate C

Origination economics

What does it cost to originate a $75,000 loan on a standardised type, and how far does homogeneity drive that down? Small loans are uneconomic because fixed cost per loan doesn't fall — unless the transactions are near-identical.

A federal pilot for mortgages under $100,000 now exists. Nobody has the evidence to design it.
Open

Valuation

Agency programs for site-built-equivalent factory homes have existed since 2018 and still lack comparable sales in many markets. The rules were written to treat these as houses; they need sales that don't exist.

Only volume creates comps. Whoever concentrates in one market first manufactures them.
Open

The value gap

Development cost exceeds what a moderate-income buyer can pay, and unlike rental housing there is no federal tax credit to fill it. The claim that a $115,000 home closes this gap without subsidy has never been run on a real parcel.

Run the arithmetic on one market before claiming it anywhere.
Open

Placement politics

Fifty states must certify equal treatment for chassis-free homes within one to two years or they cannot be sold there. Nobody is tracking who is on schedule or what the enabling language needs to say.

The window opened in July 2026 and closes next summer.
For the next one

Template

Phase 2 candidate — pending, not merged The template below — the questions any proposed structure has to answer — is not part of the published argument and is not cited anywhere else on this site. Same rule: recorded now, adopted later or not at all.

Before adding a seventeenth model, fill this out. If it can't be completed honestly, it isn't a model yet.

  1. What is sold, and to whom. Name the payer. If the payer isn't the party who benefits, say who bridges the gap.
  2. Which solved layers it reuses. Reusing solved layers is the whole point of this document.
  3. Which open layers it depends on. Each dependency is a way it dies. Count them honestly.
  4. Test A — delivers. Name the household and the dollar amount they save. If you can't, it fails.
  5. Test B — scales. What is the marginal cost of the hundredth unit versus the first? If it's the same, it's local.
  6. Test C — funds. Who writes the cheque, on what terms, and what returns it?
  7. The cheapest experiment that could kill it. If there isn't one under $25,000, the model is too vague to test.
  8. Who already tried it. Several of the models above were being run by someone else the whole time.