Coda: The Retained Invariant

THE RETAINED INVARIANT

There is no epigraph on this one, and there is no table at the end of it. The apparatus has done what it was for and I am setting it down.

Eleven chapters, and the honest thing to say now is that none of them tells you what to keep.

That was never available. Physics has no opinion about you. What it establishes — and this is not nothing — is that certain things are not optional. Keeping everything was never on the table. There is a floor under the cost of forgetting and a ceiling over the capacity to hold, both of them measured, and every structure that has ever persisted has operated between them. The choice was never whether to lose. It was only ever which loss, and most of the time nobody was choosing.

Because you have a compression policy already, and you did not select it.

It was set by what your attention happens to be caught by, which you did not design. By what your institutions instrument, which you did not build. By what was rewarded early enough to become automatic. And in substantial part, as Chapter One argued, during the years behind your own wall — the period you cannot access, where the parameters were fixed by a process that has left you the output and none of the episodes.

That is not a grim finding. It is a locatable one. A policy you did not choose is still a policy, and unlike a temperament it can be examined by looking at what comes out of it. You cannot inspect the years. You can inspect this week's discards, which are the same policy running in the present tense.

I want to be careful about the next part, because it is where a book like this reaches for consolation and I have refused that twice already.

Whatever survives of a person is a compression performed by other people. The CV, the obituary, the anecdote a niece repeats at a table thirty years on. Each of those has a selection function, and the selection function belongs to whoever is doing the sampling, in whatever channel they happen to have. Chapter Ten's warning applies without modification: a catalogue is partly a map of the instrument, and you do not own the instrument.

You will also be tempted to run the Golden Record policy on yourself. Curate toward the favourable. No war, no famine, no atrocity. It is a defensible choice — the committee made it, and their reasoning was sound — and it produces an artefact from which the thing summarised cannot be reconstructed, biased in a known direction, with nothing in the record to flag that anything was removed.

The pulsar map on the cover was the honest part. Not the letter. The address.

So the question is not what will survive of me.

That is the phoenix question, and Chapter Five refused it for reasons I still hold. The star that makes the oxygen gets nothing. The material drifts for millions of years and becomes part of objects that will never know it existed. There is no version of that story where the benefit returns to the thing that paid, and I could not convert it into comfort then and cannot now.

What is available is smaller and it is happening continuously.

What am I discarding.

That is a live operation. It is running right now, at some rate, according to some policy, and it is the only part of any of this that has any purchase in it. Not the legacy, which is not yours. Not the meaning, which Chapter Five located downstream and external. The discard rate, and what is going out with it.

And one more, which is the only piece of method I would carry out of this book if I could keep just one.

You cannot audit your own compression. I established that in Chapter Ten at my own expense: the correction that raised my forecast is the one I am least able to certify I checked as hard as the others, and no amount of resolve fixes that, because the asymmetry does not require any dishonesty to operate.

What you can do is make it auditable. Lay the thread while you are walking, not afterwards. Write down the assumption that seemed too obvious to write down — the declination cutoff seemed too obvious to write down. Keep the derivation attached to the conclusion, because a conclusion that has come loose from its derivation cannot be corrected, only contradicted, and contradiction without a located error does not accumulate into anything.

That is not a technique for being right. It is a technique for being findable by someone who can tell you where you were wrong, which is a different and more achievable thing, and the only one that has ever worked at any scale in this book.

Two of each kind. Six numbers, not a plasma. A period, not a star. The schema, not the population.

Decide what you are discarding. Lay the thread as you go. Keep the small general thing and let the rest go, because the rest is going regardless, and the only question was ever whether you chose it or inherited it.

`[ILLUSTRATIVE]`

I have flagged every claim in this book according to how much weight it will bear, and that last one will bear less than any of them. I have left it in, and marked it, because a book about what survives a compression should end with its weakest sentence clearly labelled rather than its strongest one dressed up as a conclusion.

The gods do not come back. The pieces are in the grass.

Someone will pick them up, and they will not know whose they were, and they will play.