Build it yourself
Every other page in this collection is read. This one is operated. Mount a setting and a twist, and the lathe turns the one construction — g(x) = α(φ(x)(x)) — and shows you what it cuts: a wall, where the twist has no fixed point; an engine, where it does; or nothing, where it does not bite. The famous results are not described here but built, by hand, from their parts. And as you sweep the dials you will find the fact the whole collection turns on — that the wall and the engine are the same machine, and that the same twist on the same object cannot both have and lack a fixed point.
mount · twist · turn
A theorem read is a theorem taken on trust; a theorem built is a theorem known. The lathe below is a working model of the one construction this collection is about, wired so that you supply its two inputs and watch it run to its conclusion. Nothing in it is hidden: the result you reach is the result the parts you chose were always going to force.
You choose two things. First the setting — the universe in which the naming map φ lives: sets and their subsets, sentences and their proofs, programs and their halting, or the λ-calculus, where naming is application and every term is also a function. Then the twist — the operation α the construction applies once it has fed a thing its own name. That is all the machine needs. From those two choices, everything else is forced.
Watch the outcome, and watch why. The lathe reports one of three cuts — a wall (a limitative theorem: the construction demands an object that cannot exist), an engine (recursion: it hands you a fixed point to use), or silence (it does not bite) — and shows the two facts that decide which: whether the setting is reflexive, and whether the twist has a fixed point. Turn the dials until you see the pattern. It is the whole argument of the collection, in a single mechanism.
Choose a setting and a twist; the construction runs to its conclusion. Then sweep the dials and watch the pattern appear.
Two bits decide every cut — and the fixed-point equation must be read in its own setting.
Sweep the whole instrument and a law appears, fixing every cut from two facts alone: whether the setting is reflexive, and whether the twist has a fixed point. There are only three outcomes, and which one you get is never a matter of choice once those two bits are set.
The contrapositive bites: φ cannot name every map, and a limitative theorem falls out — Cantor, Gödel, halting.
The contrapositive is satisfied and says nothing; with no reflexive object, there is no engine either. Nothing is forced.
φ = id names every map, so every twist, fed back, must have a fixed point: recursion. Y α = α(Y α).
The first two cells are the ordinary universes — sets, sentences, programs — where a fixed-point-free twist forces a wall and a fixed-point-bearing one forces nothing. The third is the reflexive universe of the λ-calculus, where φ names every map, and so every twist, fed back, must have a fixed point: the engine. Here, a λ-term encoding negation has a fixed-point term. But Y NOT need not reduce to either Boolean value. A term-level fixed point does not make the untyped calculus contradictory: Boolean values and arbitrary terms are different settings.
The Ascent and The Descent found their walls in the ordinary universes, where ¬ has no fixed point. The Engine found its power in the reflexive one, where everything does. The difference is the object and the maps it admits. The same twist cannot both possess and lack a fixed point on that object.
A wall, an engine, or nothing — and only ever the cut the two bits allow. The construction does not care which universe it is turning in; it does the same thing everywhere, and its interpretation depends on the object and the maps that setting admits. You have not read that here. You have built it. One construction; its setting matters.