Where It Lands

Return to the proof and notice a strangeness it passed over in silence. The construction is unconditional: given the mirror, it forces the point s = φ(a)(a) and proves α(s) = s for every twist whatsoever. Yet the simplest twist of all — negation on the two truth-values — plainly has no fixed point: nothing equals its own opposite. Both facts cannot stand unless something quietly gives. What gives is the unspoken premise that s must be one of the answers in B. It need not be. The forced fixed point is never absent; it is only relocated — pushed out of the values and into whatever the surrounding world will let it be. Limitation is the name we give that relocation when the world is too rigid to seat it at all.

So the dial's single bit — does α rest somewhere in B? — resolves, on inspection, into four destinations for the one forced point, sorted by whether the mirror genuinely exists, whether B is roomy enough to host the value, and whether the twist is a real, total negation.

The sorting deserves to be made exact, because the four are not loose impressions but the leaves of a short decision. Run the construction and ask, in order. Does a genuine mirror exist? Where its existence is the open question — is there a surjection, a decider? — a fixed-point-free twist answers it: no, and that no is the whole result. This is Absent. Take the halting decider: let A be programs, B = {halts, loops}, φ(e) the function program e computes, and α the swap. The diagonal g(e) = ¬(φ(e)(e)) — loop if e halts on its own code, halt if it loops — becomes programmable the instant a decider exists, and running it on its own code c forces g(c) = ¬g(c). So no decider exists; the forced point cannot be located, because the mirror that would name it cannot be built. Cantor is the same sentence with B = Ω: no set lists its own subsets.

If instead the mirror is stipulated — naïve comprehension grants membership as a total predicate, naïve truth grants the T-schema — then the identical fixed-point-free twist refutes not the mirror but the system: the forced point is a sentence equal to its own negation, and the theory is inconsistent. This is Paradox — and it is worth seeing that Paradox and Absent are one diagonal pointed two ways: aimed at an open mirror it reads no such map, aimed at a stipulated one it reads no such system. The choice of which is forced by nothing in the construction; it is forced by whether the world had already promised the mirror. If the mirror is genuine and the twist does rest, there is no obstruction at all — Recursion. And in the one case left — mirror stipulated, twist fixed-point-free, but the predicate weaker than a total negation, obeying only Löb's laws and not the full T-schema — the contradiction cannot complete, and the forced point lands as a sentence the system can neither prove nor refute. This is Incompleteness: the paradox that could not detonate.

The decision is clean; the instances at its leaves are worth naming precisely, because each is one tuple (A, B, φ, α) in the single construction. Set in a row, they show how little changes as one walks from a theorem of recursion to a theorem of limitation — only whether the twist rests, and whether the world had already promised the mirror.

Absent is not three coincidences but a class. By Rice’s theorem — a corollary of the recursion fate — every nontrivial property of a program’s behaviour is undecidable: a decider for such a property would, through the recursion theorem, let one build a program that does the opposite of what the property predicts of it — a fixed-point-free twist on the would-be deciders, forbidding its own. Halting is only the most famous line in an endless ledger of no such map.

The two fates that made B a space of truth-values need a mirror inside syntax — a way for a formal system to name its own sentences and substitute them into themselves. That mirror is a theorem, not an assumption, and it is the exact point where Gödel’s argument touches Lawvere’s.

The proof is the diagonal. The operation that substitutes a formula’s own code into itself, diag(⌜θ⌝) = ⌜θ(⌜θ⌝)⌝, is primitive recursive, so it is represented by some formula. Given ψ, let θ(x) := ψ(diag(x)) and put G := θ(⌜θ⌝). Then G = ψ(diag(⌜θ⌝)) = ψ(⌜G⌝): the sentence asserts ψ of its own code. Read in the category of formulas this is Lawvere exactly — diag is the mirror φ, the map carrying a thing to itself-applied-to-itself; ψ is the twist α; G is the forced point s = φ(a)(a) with a = ⌜θ⌝. Lawvere noted the identity in 1969. It is not an analogy; it is the same fixed point in another cartesian closed category.

Feed that mirror the twist unprovability. With Prov(x) := ∃y Proof(y, x), the lemma yields G with T ⊢ G ↔ ¬Prov(⌜G⌝) — a sentence that says I am not provable. If T is consistent it cannot prove G: a proof would yield Prov(⌜G⌝) by necessitation and ¬Prov(⌜G⌝) from G, at once. If T is ω-consistent — Rosser sharpens this to mere consistency — it cannot prove ¬G either. So G is independent: true in the standard model, unprovable in T. The forced point exists, a perfectly good sentence; what it has lost is a truth-value the system can reach.

That missing step is itself a theorem. Löb (1955): T ⊢ Prov(⌜S⌝) → S holds only when T ⊢ S already. The reflection a system would need to finish the Liar is available to it for nothing but its own theorems — never for G, whose unprovability is the failure of Prov(⌜G⌝) → G. The modal logic GL, with □ for provability and the lone axiom □(□A → A) → □A, axiomatizes precisely the schemata every such system must obey; the Gödel sentence is the fixed point of ¬□, and de Jongh and Sambin proved those fixed points always exist and are unique. The gap between truth and provability is not vague. It is the exact width of Löb’s theorem.

The same seam, seen from the other fate, is Tarski’s theorem — and it is why Gödel had to reach for provability in the first place.

The proof is a single turn of the diagonal: were there such a True, the lemma applied to ¬True(x) would give L ↔ ¬True(⌜L⌝) ↔ ¬L, and T would be inconsistent. So truth is the Absent fate for a predicate — no such map — the very same diagonal as the Liar, read against an open mirror instead of a stipulated one. And now the picture closes. Truth is undefinable because it carries reflection and would explode; provability is definable because it lacks reflection and merely incompletes. Gödel’s move was to substitute the definable, weaker mirror — provability — for the undefinable, stronger one — truth — so that the diagonal’s contradiction, denied the reflection it needs to detonate, deposits a true-but-unprovable sentence instead. Incompleteness is the controlled demolition of the Liar: the same charge, the same fuse, defused by exactly the reflection the system cannot prove of itself.

Read across the four and one axis organizes them. They are sorted by how the world answers a single question — where may s live? — which has just two moving parts: whether the mirror genuinely exists, and whether the answer-space B is roomy enough to seat the value the twist demands. Both hold and the point is recursion; the room runs short and the point is forced into paradox, into undecidability, or clean out of existence. The fates are therefore a ledger of demand — an accounting of what the forced point becomes when nothing in the values will seat it.

A ledger of demand implies a matching ledger of supply, because the same construction, run on a kinder world, hands the point back whole. By what right does any fixed point exist? That is the question the next movement takes up — and the two ledgers, laid side by side, turn out to be one, balanced to the entry.