The library — every public artifact
Fixed, citable copies of everything this site presents. Each document states its own revision, its own grades, and its own uncleared gates. Nothing here is final; the register says exactly what is and is not established.
The argument in five movements: the gap, the 1976 precedent, the quarantine that followed it, what the other attempts prove, and an ending that concedes what it has not proved. This is the current statement of the case — read it before the thesis. Available as a scroll narrative at two lengths, a keyboard deck, and a fixed PDF.
The paper the argument grew out of. Diagnosis, pathologies, prior art, four cities, the threshold, donor discipline, the seven-layer architecture, the loop, cycle time, graded evidence, adversarial review, the experiments, references. Revisions 3.2 and 3.3 correct the Red Team 04 response, re-render the PDF in the document’s own typefaces, signpost the Sheet 02/08 overlap, and bring the Tokyo claim into line with its own grade — all four recorded in the corrections log.
This paper and the narrative had become two documents arguing the same case at different lengths — the same duplication this project exists to remove from housing, reproduced in its own filing. HAP-08 is now the single source of record, and the web adaptation of this paper was retired in August 2026.
Every argument it carried is on this site, in the story and the instrument. Before the page was deleted, each of its sections was checked against HAP-08 line by line; one sentence was missing and was moved across. The PDF stays exactly as published, because a citation has to keep resolving to what it cited. Cite this; read those.
It closed before the August 2026 correction, and it still carries claims that have since been withdrawn — the encoding-cost climax, the thirty-two-to-seven comparison, the twenty-five months, the 504-day EIR figure and the Auckland band, twenty-seven passages in all. It is not edited. What changed is recorded in the errata, which is what a citation of this paper should travel with.
The errata cannot reach a copy already downloaded. A PDF carries no link home, so every copy of this paper in circulation is uncorrected and will stay that way. The errata is published as its own PDF for that reason — file it with the paper. The paper is not re-rendered with a cover sheet bound in, because a citation that resolves to different bytes than it cited fails in a worse way than an uncorrected one.
The synthesis. Three binding constraints — politics, information, capital — the stage rail on which each binds, the four orderings that have actually been run and failed, the defended ordering, and the five named results that would rewrite it. Now the first half of the solution page — this argument and the instrument it implies were one case split across two pages, and were merged in August 2026. The PDF remains the citable form.
The argument compressed to a single sheet: the 32-month strip, what has been tried, why it all fails the same way, the sequenced answer, and — boxed in red — what is honestly unproven. Made to be printed and handed over.
Sixteen structures considered for solving housing affordability, each tested against the same three requirements — delivers, scales, funds — and each one breaking on a different one. The salvage line under each records what survived the failure and can be reused. Published for the same reason the revisions page is: recording what was considered and rejected is the method, not an exception to it.
Every assessment in it is a judgement, not a measurement. Each stamp is one working session's reading against no measurement, and nothing in the document carries a register row — deliberately, because a row would imply a graded claim behind a stamp and there is no such claim. It is dated and versioned so later sessions append rather than overwrite. Two of its sections, the component inventory and the template, are marked as Phase 2 candidates and are not part of the argument on this site.
The critical path as thirty-two squares — twenty of information, four of site, eight of building, and the compiled row underneath.
No longer offered for download. The compiled row is the thirty-two-to-seven comparison, withdrawn in August 2026 because four of its seven months are factory build — it set site-built today against factory-built tomorrow. HAP-02 has the same problem and is still offered, because a paper is cited in a footnote where an errata sheet can travel beside it. A poster is made to be shared on its own, as an image, with no route back here: whoever meets it meets the withdrawn claim and nothing else, and no errata can reach them. So this one stops being handed out.
The files stay at the addresses they have always had, because anything already linking to them has to keep resolving, and what they still say is recorded in the errata. The title survives the correction — eight months of building is sound, and rounds from the figure that was the right column all along (C-016).
The spine. Two hundred and thirty-one claims, each with value, source, grade, where it is used, and a review date. Thirty-two rows are recorded gaps nobody has measured. Opens in any spreadsheet; rendered sortable and filterable on the evidence page.
Two studies on how long housing takes to deliver: 183 country-by-segment records across 38 countries, and a 255-row workbook covering every state and the District of Columbia. The finding that survived both is an absence — no country in the compilation publishes approval, waiting and construction as one regular series, and no state-level American construction duration exists at all. Both studies are published in full at research, and the graded claims drawn from them are argued on what the world publishes.
They are the working material the studies were built from, carrying another project’s source tiers and confidence ratings. No row in either is graded by this register — only the figures quoted on the page are, and each of those is a row. Both sit here rather than in the register because importing them would put two different instruments in one column under one name, and because 52 of the 183 records carry no figure at all.
Six defects were corrected before publication, found by an independent review that checked the studies against primary source files rather than against the sessions that wrote them. One of the six is the reason the rest are worth reading about: a prior pass took a figure the source actually prints, replaced it with one back-solved from the source’s own rounded phrase, and described the substitution as a verification. Every change is recorded against the unedited originals in the knowledge base.
The row count flatters the evidence. Only 60 of the 183 records carry a construction figure, and 26 of the 38 countries have one at all. Both numbers are printed wherever the scope is stated.
Everything published here that was later found wrong, weaker than stated, or in need of repair — with the corrected form alongside it. Entries stay after they are fixed. All fifty-three are resolved and all fifty-three stay published — including the one where this project miscounted its own objections in four places.
These are all the public artifacts. Working documents stay internal until their claims are in the register and graded.