The Collation

The corpus rests on one theorem and a small bedrock of classical results. The theorem is Lawvere's, and the corpus states it correctly in all four places it appears — the front matter, the keystone, the apparatus, and the broadside. Re-derived from the bare definitions of a cartesian closed category, the proof is exactly the five lines the corpus gives, and it is valid.

The two directions check out as stated. Backwards — the contrapositive — a fixed-point-free twist forbids the naming: with B = {0,1} and α = ¬, the missed map is the diagonal set {a : a ∉ φ(a)}, and that is Cantor exactly. Forwards — at φ = id on a reflexive object — the construction collapses to g = λx.α(x x), whose fixed point g g = α(g g) is precisely Curry's Y α. The corpus's tightest claim — that Cantor's escapee and the engine of recursion are one string of symbols read at two settings — is therefore not a figure of speech. It is an identity, and it holds.

The reductions of the diagonal genus were checked one at a time, against the four-tuples the Atlas exhibits. Each names a category, a domain of things, a domain of answers, a self-naming, and a twist, and reads off what the construction forces. All fourteen core instantiations are faithful; the six the corpus leans on most — Cantor, Russell, Gödel, Tarski, the halting problem, and the Y-combinator — are genuine instances in the strict sense the apparatus requires, each becoming the construction once finitely many proven bridge-assumptions (arithmetization, representability, the diagonal lemma, the s-m-n theorem) are named.

The two limit-volumes turn on two further classical theorems, and both are stated and used correctly. The Ascent's engine is that the reflection operator is fixed-point-free: among the local reflection sentences Prov(⌜φ⌝)→φ is the instance Prov(⌜⊥⌝)→⊥, which is Con(T); were T to prove its own reflection it would prove its consistency, and by Gödel's second theorem be inconsistent — so R(T) ⊋ T strictly, with no fixed point. The Descent's engine is Löwenheim–Skolem: any first-order theory with an infinite model has models of every infinite cardinality, so none is categorical. The supporting apparatus — the diagonal lemma and Löb, the provability logic GL with Solovay's completeness, the order type ℕ + ℤ·ℚ of countable nonstandard models, Henkin's dilemma, the independence of CH — is throughout correct and carefully qualified.

Finally, the corpus makes several claims that are simply counts, and a collection staking itself on checkability invites the reader to count. They were counted, against the page's own data. Every figure is exact.

The lattice is the corpus's boldest move and its best one — a public ledger that puts each of its eighteen claims on the record as exactly what it is. The audit took it tier by tier and tried, at each, to find a claim wearing a stronger label than it deserved. The search came back nearly empty, which is the relevant result.

The theorems — Lawvere, Cantor, Gödel, Turing, Kleene/Y, the strictness of reflection, Löwenheim–Skolem, Knaster–Tarski — are classical and correctly stated, and they sit at the base where established claims belong. The reductions — that the construction is one, and that enrichment recovers the forbidden fixed point — are genuine and verified instance by instance, not asserted; they rest on the theorems beneath them and claim no more than the bridge-assumptions allow. The framings — the three genera, the five walls, the power face, "no roof," "no floor," "one purchase," "four faces" — are labeled as the author's synthesis and are never smuggled into the theorem tier. And the analogy — echoes in physics, mind, and society — is placed at the rim, flagged as loose, with the one inference that masquerades as a proof, Lucas–Penrose, marked rejected.

What earns the corpus its standing is not that it draws the line but that it is hardest on itself exactly where the temptation to cheat is greatest. The apparatus does not merely defend the unification; it volunteers the unification's limits. Scholia downgrades the Atlas's own organizing slogan — "three genera as one" — from a metatheorem to a heuristic, conceding that the order-theoretic and topological theorems share only a schema with the diagonal and admit no reduction in either direction. The Atlas isolates Yoneda as a false friend that shares the word "diagonal" and none of the phenomenon, and wires it to nothing. The same discipline names renormalization fixed points and strange attractors as vocabulary collisions, not kin. This is the behaviour of a work that means its own standard.

The audit registers one standing reservation and one consequence, both already half-acknowledged inside the corpus. The reservation: the "three genera, one schema" picture is a genuine organizing insight but, as Scholia itself concedes, a heuristic — "richness + regularity ⟹ a fixed point" is a family resemblance across the diagonal, order, and topological theorems, not a single metatheorem from which all three descend. The corpus says this in the apparatus; it could say it more loudly in the Atlas, whose section-one prose ("three genera, one schema") reads more confidently than the apparatus permits. The consequence is the broader one this Collation exists to draw: a corpus this disciplined about the line between proof and proposal in mathematics turns out to have left one proposal about itself uninspected — how many things it is. That is the next movement.

Set the eleven copies side by side — nine HTML pages, two print plates — and ask each the same question: how many volumes is Quodlibet? The answers run from three to six. They are not random. Read in the order the pages were written, they are a stratigraphy: each witness was frozen at the size the collection happened to be on the day it was made, and the cross-references inside it were never brought forward. The collection grew, and its account of itself grew in layers that never quite reconciled.

The layers resolve the apparent contradictions completely. There is no disagreement about what the collection is — only a record of what it was, preserved at four depths.

It would be easy to file this under typography and move on. It is more interesting than that, and the corpus itself supplies the reason. The Atlas ends on a line it means as its deepest: an atlas of self-reference is the one map compelled to contain itself — the diagonal that draws the map sits on the map's own inner ring, and the act of mapping is one more fixed point of the thing being mapped. That is exactly the move this Collation found the corpus failing to complete. The collection charts self-reference with total command, and then, at the one place its chart must turn back on itself and count its own faces, it loses the thread — preserving, across its witnesses, four different answers to how many things it is.

This is the diagonal turned on the diagonal, and it is precisely the error the subject predicts. A formal system rich enough to name itself cannot, from inside, settle every fact about itself; the reflexive facts are the ones it cannot close. A corpus rich enough to chart self-reference proves no less prone: the fact it leaves uninspected is the reflexive one — not any theorem, but its own extent. The mathematics, which points outward, is flawless. The bookkeeping, which points inward, is where the seams show. A work on self-reference miscounting itself is not an embarrassment to be hidden; it is the work's own thesis, demonstrated one last time, on the only body left to demonstrate it on.

So the correction is offered in the corpus's own spirit, and it is small. The established structure stands; the earlier layers are brought forward to meet it; the broadside is reissued at six. What follows is the exact list — every locus, its frozen reading, and the reading that reconciles it — so that the map may, at last, contain itself, correctly, and count.

class A — the binding.  Two defects betray how the edition was assembled, and the graver one breaks it outright. The front door, quodlibet.html, is a true table of contents — eight links that open the volumes and the back matter in place, as its own caption promises. But every link points to a bare filename (the-diagonal.html, scholia.html), while the files as distributed carry a quodlibet- prefix; in that state not one link resolves, and the collected edition does not connect at all. The lighter defect is cosmetic but pervasive: every page but two carries the same browser title, “The Ascent — why the diagonal cannot be escaped,” regardless of which volume it is — each cloned from The Ascent with its <title> never changed (only The Diagonal and The Atlas were given their own; The Ascent's is right only because it is the source). Both are corrected in this edition — the files named to match the links, and each page given its true title.

class B — the count of the corpus.  The six fixes below carry the earlier layers forward to the established six-volume structure. They are the collation's substance: each is a place where a witness froze at a smaller corpus. The resolution throughout follows the folio and the contents — six volumes, with The Engine as the power face, making five faces bound by the keystone.

class C — minor imprecisions.  None is load-bearing; each is a loosening of a date, a figure, or a phrase that a careful reader will catch and that the corpus's own standard would rather not leave standing.