Why Failure Is Required For Success

The title of this essay is a claim that cannot fail. Failure is required for success: say it back and nothing in it can be checked. No observation would show it false, no result would make it wrong. By the standard I have applied against every other frame I have put in front of anyone, that sentence did no work. So this essay begins by refusing its own title, and then tries to earn it.

Because most failure teaches nothing. It just happens. A person tries a thing, the thing does not come off, and what they carry away is the feeling of having tried. Nothing was measured, nothing was written down, and the same mistake is waiting in the same place for the next attempt. Failure in that form is required for nothing. It is simply the cost of being alive in a world that does not owe anyone a result.

The failure that carries is the one with a size

Mercury's orbit was wrong. Not wrong the way a guess is wrong, but wrong by a number. Newton's theory accounted for almost all of the observed drift in the point where the planet comes closest to the sun, and left about forty-three arcseconds per century unexplained. That figure sat there through the second half of the nineteenth century, small enough to be shrugged at and specific enough to be embarrassing. When general relativity arrived, one of its first achievements was to produce that number rather than to explain it away. The mismatch did its work because somebody had written down how big it was.

That is the distinction worth holding onto. A theory that fails by a named amount tells you where the next one has to start. A theory that fails vaguely tells you nothing, because it does not tell you what has to be accounted for.

Wrong theories that led somewhere

The ancient system of circles within circles, which placed the earth at the centre and made the planets travel in loops, was wrong about what the planets were doing and still gave positions accurate enough to be worth correcting. Phlogiston, the supposed substance released when things burned, was wrong about burning and still organized the experiments that led a chemist to weigh the vessels instead of only describing the flame. Both wrong theories led onward. The condition was not their wrongness but the record: someone had written down where prediction diverged from what was seen, and someone later had that divergence to work against.

Where nobody recorded the size of the gap, the wrong theory does not lead anywhere. It gets rediscovered by the next person and run again.

The frames that cannot fail

Now the other end of the range. Everything is energy. Everything is consciousness. These are not wrong; they are worse than wrong, because they cannot be wrong in any particular way. A frame that cannot be wrong in a specific place cannot be right in one either. It produces no mismatch to write down, so it hands the next person nothing. This is the sense in which failure is not merely useful but load-bearing. Without the possibility of a measured failure, a frame is not a weak claim about the world. It is a way of speaking that leaves the world exactly where it found it.

Where a borrowed shape stops carrying

Most of what I have written borrows a shape from somewhere: geometry taken from fluid motion, a spiral taken from a shell, a spectrum taken from a set of mathematical zeros. A borrowed shape does not succeed or fail as a whole. It behaves like a tangent line. A tangent touches a curve at one point and shares the curve's direction exactly at that point; everything past that point is the line's own business. At the contact the borrowing is exact, the mathematics true there. Diverging afterward is not a defect in the borrowing. It is what a tangent is.

So the honest form of a suspicion is not that a borrowed shape will not hold. It is this: here is the point where it stops carrying, and here is what I would expect instead. That form is the only one another person could pick up without me. It is also the form that makes a wrong borrowing worth doing, because it leaves a residue with a location.

Not every failure has a number

My own work is uneven here, and the unevenness is worth stating. Some of it carries a magnitude now and some of it does not. A claim that matter is vortices in a fluid wave system can be asked whether it bends light by the amount a given mass would bend it, and whether the quantity doing that work is secretly something else that was never supposed to appear in the calculation. That is a knife, and it cuts the claim or it does not. Other claims in the same body of writing are structural: that relationships come before things, that a membrane described as folded in two layers explains what a flat description cannot. Those would fail by showing a case where the relation does not hold, or where the folding fails to do the describing it was brought in to do. That is a real failure, but it arrives in shape rather than size, and it needs a different record form. Both kinds have to be written down. The second kind is the one most easily left unrecorded, because nothing forces a number on it.

The burden arrives after you are gone

Here is the part that makes the writing matter more than the theory. A theorem about what measurements on separated particles can and cannot agree about was proposed in 1964, and its author did not settle it. Decisive laboratory tests came in 1982; the tests that closed the remaining escape routes came in 2015, decades after he died. This is normal. The burden of proof usually arrives after the theorist is safely not around to defend it or enjoy it.

For someone inside an institution, the institution carries the program forward: the journal, the students, the funding line. For someone outside one, the only vehicle is the written record. Which means the archive is not a side project kept alongside the theory. It is the substitute for the apparatus that is not there. And a record holding only the claims, without the size of each mismatch, is a record the next person cannot use.

What this asks

So the question the title was reaching for is not whether failure is required. It is whether you are willing to find out the size of yours. Not the feeling of having been wrong, which most people can produce on request and which settles nothing, but the amount, the place, and what you would expect instead. That is a much smaller and much less flattering thing to write down. It is also the only version of failure that has ever carried anyone forward.

What would one of your own convictions have to do to count as having failed by a named amount? If you cannot say, the conviction is not strong. It is merely comfortable.

Status and exposure

This is a claim about method, not a result. It says that recorded mismatch is what makes one theory useful to the next, and that unrecorded failure does not carry.

Falsifier: it fails if advances routinely proceed from failure nobody measured. If the historical record turns out to show unrecorded failure producing better theories as often as recorded failure does, then the size of the mismatch was never the load-bearing part, and this essay has been arguing with a distinction that does not exist.