Storage and the forge

This is not a result. It is the record of a search that closed — a long afternoon of building geometric instruments, each meant to pry a number's factors loose by some clever arrangement of shape, and each turning out to do the same thing in a different costume. The instruments were real and some were beautiful. None of them worked, and the way they failed was always identical. By the end the failure had hardened into a law, and the law is worth more than any of the instruments would have been had it succeeded.

The law is this: a shape can store the structure of the primes, but it cannot forge it. Everything below is the walk that arrives there.

IThe levers

The problem is the oldest one in this corner of mathematics. A semiprime is two primes multiplied; recovering them from the product is the difficulty that public-key cryptography rests on. The hope, each time, was geometric: that if you drew the number the right way, its factors would become visible — a length you could measure, an angle you could read, a place where a shape betrayed what built it.

So the levers came, one after another. A triangle anchored at a number's two prime factors, whose apex angle reads the gap between them. The same triangle climbing until it collapses, the rate of collapse a fingerprint of balance. A logarithmic spiral — a custom unit, the gyre, one full turn per doubling — winding the whole family of semiprimes into a single curve. A coil kinked at every prime. A wave warped, one prime at a time, away from a clean sine. A skewer driven through the figure to read where it pierces. The same skewer lifted into three dimensions for more room. Two number lines crossed, the factors in the difference of their placements. The lines wrapped into a cylinder, then flared into a funnel when the primes crowded.

Each one felt, in the moment, like it might be the one. Each one was an honest attempt to make the hidden thing show itself by arrangement alone.

IIThe single failure, in many costumes

There is a test that settles every one of them, and it is not a matter of cleverness. Count the independent information. A triangle has three vertices, three angles, three sides, a perimeter, an area — but they are not independent. Fix the vertices and all the rest is determined; they are one set of numbers wearing ten hats. And the vertices, in every construction, are built from the one thing you started with: the number's position on the line. So the figure — however elaborate, however many measurements you take of it — carries exactly the information you put in, and nothing more.

This is conservation, and it is as firm as conservation of energy. A relationship between shapes reshapes information; it never creates it. Relate a triangle to a circle, a circle to π, a line to another line, and you have applied a fixed transformation — a lens. Lenses bend light. They do not make it. So the factor you did not put in cannot come out, no matter how the shape is bent, skewered, lifted, crossed, or wrapped.

Any shape relatable to another is based on a relationship — and relationships preserve, they don't create.

Here is the ledger of the search, every lever and the one verdict it earned:

The two-number-lines lever deserves its own word, because it is not a shadow of the problem — it is the problem's true face. The best factoring algorithms humanity has, the quadratic sieve and the number field sieve, are exactly this: two structures related so that the factorization lives in their alignment. They do not break the wall. They are the cleverest known way to search for the alignment — and the search is the wall, undisguised. To reason from marbles to this is to arrive, by hand, at the front line where every working number theorist already stands.

IIIStorage and the forge

The clearest way to see why every lever failed is to build one in time, one number at a time, and watch where it strains. Lay the primes down as kinks along a winding surface. The primes thin as they climb, so the structure crowds, and to keep room you widen it — and you must keep widening, forever, at the rate the primes accumulate. There is never a step you cannot take. There is only a cost that never stops rising. You do not run out of steps; you run out of resources to keep paying for them. That is the exact shape of an intractable problem: no single move is impossible, and the total is beyond any finite machine.

But the surface only had to store the kinks. The deeper cost is the one the geometry hides entirely: each kink has to be forged before it can be placed. You cannot drive the tooth at p until you have found that p is prime — and finding it means testing it, and testing the countless candidates between that turn out to be nothing. The winding displays only the kinks that survived. It says nothing of the labor of all the numbers you checked and threw away to find them.

And there is the whole of it. Every construction — the gyre, the coil, the funnel, the cylinder, the crossed lines — is a place to hold prime structure. None of them makes it. The kinks must be forged, one prime at a time, by checking, before any shape can hold a single one. The shapes display the primes beautifully, exactly, lovingly — after the forging. They never spare you the forging.

Storage is free, and the geometry is generous with it. Forging is the wall, and no shape forges.

This is why the search closed, and why it had to. The wall was never about storing the primes, or relating them, or displaying them — those are free, all of them, all conservation-bound, all just re-encodings of what you already hold. The wall is about forging, and forging lives entirely outside the geometry. You cannot draw your way out of having to find each prime, because the drawing can only ever hold what you have already found. The room you can only know by building it; the kink you can only place by forging it; the factor you can only have by finding it.

An honest note, as always: none of this proves factoring is permanently hard. It proves something narrower and, to my eye, more useful — that a whole family of approaches, the ones that build a shape from a number and read its parts, are conservation-bound and cannot forge what they were not given. That is a map of exhausted ground, not a verdict of impossibility. A different machine already changes the picture: a quantum computer does not store the winding and walk it — it holds the whole winding at once and reads its period by interference, forging the answer by a means no shape on paper has. The wall here is the wall against drawing. It is real, and it is not the same as the wall against all possible machines. Knowing precisely which wall you have found is the only honest place to stop.