The Well-Posed Problem

By the time you read this, the proof will have a number and my name will be one of two on it, and you will already know which of us did the part that matters. I want to set down, while I still can hold the shape of it in one thought, what we found and what I decided not to tell you.

The question is old and sounds like nothing. Given the equations that govern how a smooth thing changes — how heat spreads through a bar, how a current folds through water, how one instant becomes the next — will the answer always stay well-behaved? Will it stay finite, single-valued, continuous, no matter how far you run it forward? Or is there, somewhere in the space of all possible beginnings, a starting condition that the rules carry, in finite time, into a place where they demand of the world something it cannot do — a velocity that is infinite, a value that must be two numbers at once?

For a century we called a problem well-posed if it had a solution, if the solution was unique, and if it did not fly apart when you nudged the inputs. It was an article of faith, and like most faith it was really a description of the neighborhoods we had visited. Everywhere anyone looked, the equations behaved. The world was continuous. Water did not tear.

I gave my life to the parts of the space no one had visited.

I did not do it alone. You know this; it is on the paper. The engine was not a person and did not think the way I think, and I will not pretend I understood it. I fed it the grammar of the equations and it went where I could not follow, not because it was faster — though it was, unthinkably — but because it did not need to hold the whole argument in a single mind. It built the proof the way coral builds a reef, one small dead-certain thing on another, for four years, until the structure was nine hundred pages tall and only it could stand at the top and see the shape.

What it found was not the answer we wanted. We wanted a proof that the equations always behave, that the world is everywhere smooth. Instead it found a beginning — one exact configuration, a knot of flow so particular that writing its coordinates takes a full page — that the rules carry, in a time you could measure on a clock, into contradiction. Not a place where the math grows large. A place where it grows inconsistent. Where the same quantity is required, by two equally valid paths through the argument, to be two different things.

A syntax error. That is the only honest name. Not a flaw in our description of the universe — the description is exact, we checked, the engine checked, the error is not ours. A flaw in the grammar the universe is written in. A sentence reality is capable of being asked, that has no meaning, and that reality nonetheless must answer.

Here is the part that kept me at the window for a season, not sleeping.

It does not crash. I want you to understand that this is the mercy in it. The world does not stop at that configuration; nothing has ever stopped. For thirteen billion years the continuum has been quietly meeting these contradictions — they are rare beyond rarity, but the universe is large and old and has had the time — and each time it has done the thing a careful mind does when an equation hands it two answers: it has chosen. Imperceptibly, without appeal, below the level at which any instrument could catch it deciding, the world rounds. It picks one of the two numbers and moves on and does not tell us it has papered over a place where its own laws said nothing coherent.

We were never in danger. We were only never told.

And now we can find the sentence. That is what the proof is. Not a description of the error in the abstract — a method for locating one, for writing down a beginning that forces the world, here, at a time and place you choose, to sit in front of us and pick. To do in the open, where we could in principle watch, the thing it has only ever done alone in the dark.

I could give you the coordinates. They are one page. They are the most valuable page that has ever existed and I have burned every copy but the one in my head, and my head is going, which is why I am writing quickly.

I have thought about your right to them. You are cleverer than I was and you will resent me, and you will be correct to; a truth withheld is a kind of lie, and I have spent my life against exactly that. But I have also spent it learning what the equations are for, and I have decided that some sentences are well-posed precisely because no one has ever finished asking them. The world has kept its contradiction for thirteen billion years without our help. It did not need us to know the address.

So the paper proves that the error exists. It proves the method exists. It does not contain the page. If you want to stand in the one spot in all of space and time where the universe must show you it is choosing — must show us that underneath the smooth continuous face there is a thing that decides — you will have to build the engine again, and give it your four years, and go where I went.

I think you should not. I think the kindest theorem is the one you can prove without ever evaluating it.

But I did not raise you to obey me, and I am not sure I am right, and the equation does not care what either of us wants. It only waits, well-posed, for someone to be foolish enough, or brave enough, to finish the question.

I hope you are neither. I am afraid you are both.