The project, its status, and its next two tests
A working thesis on American housing production, published as research. Pre-evidence-collection: no gate cleared, no experiment yet run.
Eleven objections are stated on this site at full strength. By the test that matters — if this objection is true, does the thesis fail regardless of everything else? — six are both dispositive and unresolved: no demonstrated cost advantage; firm size as the real mechanism; whether codes compile at all; a reviewer paid by the party it reviews; precedents that were political rather than technological; and a jurisdiction that can simply buy the automation itself.
So what is published here is the diagnosis — the record of how American housing production is regulated, what four prior attempts hit, and a corrections log of what this project got wrong — and not a case for building the thing. Two experiments would change that: measuring what share of binding code provisions are objective rather than discretionary (E-039 — two weeks, no capital), and a like-for-like price per square foot at matched finish and region (E-098). Neither is blocked on anything but doing it. Until they are done, treat everything here as a diagnosis with its uncertainties published, and read the corrections log before the argument.
The same instinct produced the model autopsy — sixteen structures considered and where each one broke, kept because what was rejected is part of the record. Every assessment in it is a judgement rather than a measurement, and it says so at the top.
01What this is
This site presents a diagnosis — American housing regulation as ~20,000 uncompiled codebases, and housing itself as an industry of projects rather than products — together with everything currently known for and against it.
The argument is confident where the published record supports confidence, and explicit where it does not. Every factual claim carries a row in the evidence register with a source, a grade and a review date. Thirty-two of those rows are gaps: things nobody has measured, recorded as rows precisely so they cannot be quietly forgotten. The claims the whole argument leans on hardest are among the least evidenced — that is the shape of the opportunity, and it is stated rather than obscured.
It is published as research because the claims are meant to be citable, checkable and falsifiable — by a legislator, a journalist, or an economist, whether or not they agree with a word of it.
That standard only means something if the errors are visible too. Where something published here has been found wrong, weaker than stated, or in need of repair, it is recorded in the corrections log and left there after it is fixed — the record of the error is the point, not its removal.
August 2026 — the argument has changed, and the change is published. The claim that used to be the centre of this site, that one number closed the market for compiled building codes, was refuted by this project's own research and has been removed rather than softened. Four of its beats were deleted, one register row was regraded Contested, and the count of things nobody has measured rose from eleven to fifteen, and to sixteen since. Three pages carry the consequences: the research log, which is every position held and given up since this started; what changed and why; and what was already known — which states plainly which parts of this argument established researchers published first, and how much of it that is.
02The standard this project holds itself to
Filed in the thesis before any evidence was collected. Quoted here without abridgement.
Falsification statement — stated before evidence collection, Aug 2026This thesis fails if any one of these turns out to be true.
- a Binding provisions prove predominantly discretionary, and therefore non-compilable. Measured by Gate 0, rewritten Aug 2026 · register rows E-039 and E-107 · unmeasured anywhere
- b Encoding cost per jurisdiction does not fall below the threshold implied by the addressable market. Register row E-051 · tested Aug 2026 — the thesis lost this clause · what changed
- c Builders predominantly identify utility capacity or fire district sign-off, rather than approval-side gates, as the critical path on recent projects. Measured by Gate 1 · ten to fifteen interviews · never run
Materially weakened — not failed — if the firm-size mechanism in Red Team 06 dominates, in which case type certification addresses a problem adjacent to the real one.
Clauses (a) and (c) are cheap to test and neither has been tested. Clause (b) has been, and it failed in a way the statement did not anticipate: encoding cost was never the binding constraint, so the threshold it names was the wrong test. The claim built on it is withdrawn — see what changed and why. That is the current state of this project, stated plainly.
Source: HAP-02 Rev 3.5, Sheet 12 — read it in the paper.
03What happens next
The next steps are not fundraising and not product. They are the three cheapest tests that could kill the thesis — run first, published regardless of which way they come out. They are listed in the order they should be run, which is cheapest first, not most important first.
Two frames, and only one of them is conditional. This is research that publishes, unconditionally: the register, the corrections log and the research log stand whether or not anything is ever built, and every row is graded on its evidence rather than on whether it is useful to a company. It is also aimed at a buildable company — and that half is conditional on the gates below, which have not been run. Where the two conflict, the evidence wins and the record shows it: in August 2026 this project refuted its own central claim and published the refutation rather than quietly editing the pages.
Rewritten in August 2026. Gate 0 used to ask whether a municipal ruleset was compilable at all — a question designed to price a claim this project has since withdrawn. The compilability question survives as register row E-039 and is now measured inside this gate as a cross-tab.
The falsification statement above was filed before evidence collection, and one of its clauses has already been tested and lost. This is the same instrument one level down: what will be concluded from each Gate 0 outcome, written before the gate runs, so the conclusion cannot be chosen after the result is known.
Two of those three outcomes end the company path. That is what a kill criterion is for, and stating it in advance costs nothing except the option to reinterpret the result later.
Not funding for a conformance authority. There is no entity, no pilot and no cleared gate, and a proposal in that shape is not fundable however good the analysis is — the programs that exist fund organizations that are already executing, and two of the three cannot legally accept an application from an individual at all.
The ask is narrower and it is the first line of Gate 0: nobody has measured where a home an American can actually buy is legally allowed to go. That figure is register row E-107 and it is empty. Not whether a district permits mobile homes — the National Zoning Atlas already codes that across 102,000 districts — but whether a real, purchasable model clears the width, floor-area, roof-pitch, materials, foundation and setback standards attached to the permission, on a real parcel. One county, three jurisdictions, no capital, publishable whichever way it comes out. Alongside it, and from the same fieldwork, the objective-versus-discretionary split that E-039 has recorded as unmeasured since this project started.
Nobody owns that number. It belongs in the literature rather than in a pitch deck, and whoever publishes it first becomes the person cited on it — which is a better credential than anything a pre-evidence venture proposal can offer, and the thing that would make the larger argument fundable later. That is the order: measure, publish, then build.
Recorded because they are dated and easy to miss, not because any has been pursued. The Ivory Prize takes free self-nominations across three focus areas. The posted cycle closes 30 October 2026 at noon MST — the terms say September to December, and this site said so too, which would have missed it by two months (C-041). Whether an individual rather than an organization may enter is contested: the terms say individuals over eighteen, the prize page says organizations. That is dispositive here and has not been resolved. HUD's Advancing Building Technology program runs $500,000–$1.5M awards across roughly fifteen agreements, and requires an entity — individuals and sole proprietorships are ineligible. The Housing Affordability Breakthrough Challenge, funded by Wells Fargo and administered by Enterprise, is the largest per award at $2M, five of them per cycle; its 2026 round closed in May, and it asks for existing results and a scaling strategy, so it is a future-cycle target rather than a near-term one. A bipartisan congressional housing caucus exists, with founding members including Arizona's sixth district, which is the state this project's live natural experiment sits in.
04The name, and why the site looks the way it does
The name is doing two jobs at once, and both of them are the argument.
A lintel is the beam over an opening that carries whatever sits above it. It is the least visible structural member on a building and the one everything over the doorway depends on. That is the position this work describes: a conformance authority is the span across the gap between permission granted and homes built, and financing, production and sale all bear on it from above. Arizona is the standing illustration — a doorway cut in 2026, with nothing yet across the top.
Lint is the program that checks source code before it ships. It runs ahead of the compiler and catches the defects that would otherwise surface in production. A determination is exactly that, applied to a building: the rules evaluated before land is committed rather than discovered at review afterwards. The narrative names the system, where both halves of the word are drawn.
The wordmark is set in the site's own type, and the mark is the beam form — a lintel in profile — in vermilion, the one accent the identity permits. The system underneath it is deliberate: a near-black charcoal field; soft-white plates that carry the diagrams and documents, because this project's diagrams are its only imagery; and a second color, amber, that the brand is never allowed to touch, because it belongs to the word Untested. Vermilion marks structure and amber marks doubt, and no component is permitted both — so the design cannot flatter a claim the register will not support. The identity enforces the honesty system rather than decorating it.
This site publishes research, and it names the instrument that research proposes. Those are two different things and the reader is entitled to know they come from the same hand: the author intends to build the conformance authority described here, and would benefit if it were adopted. That is disclosed rather than separated, because a concealed interest discovered later would damage this work far more than a declared one does now.
The protection against a motivated argument is not a different letterhead. It is that every claim carries a graded row with its source, that the claims the thesis most depends on are marked Untested on the front page, that the falsification statement above was filed before any evidence was collected, and that errors stay in the corrections log after they are fixed. Judge the argument on those, and discount it for the interest declared here.
See the full identity system → — the mark's construction, the color spec with measured contrast, the type scale, and the one rule that governs all of it: vermilion marks structure, amber marks Untested, and they never appear in the same component.
Stated for the record. Lintel is a working name. A search of construction and proptech turned up no obvious conflict, but no trademark or domain clearance has been run — which is not the same as clearance.