Our error registry is incomplete and always will be. That is not a defect in it. This essay says what the registry is actually for, and then says the thing this series has been circling for a year: an error and a verdict are different objects, and only one of them hurts people.

I. The registry is an example, not an audit

I have been asked why anyone should catalogue errors in a corpus this size, when the catalogue is out of date the moment it is published and is obviously incomplete right now. It is a fair question and I want to answer it plainly, because the answer changes what the file is for.

The Published Record — Error Registry is not an audit of this corpus. It could not be. It holds twenty-one entries against roughly a hundred essays and a consolidated canon, and most of those entries came from three days of looking. A complete audit is not available to me, to my collaborators, or to any single reader, and waiting for one would mean publishing nothing.

What the registry is, is a worked example. It demonstrates that the practice exists here, shows the form an entry takes, and establishes that finding an error in this work carries no penalty and requires no permission. Anyone who finds one may publish it wherever they like, in whatever form suits them. They do not need to clear it with us and we do not need to agree.

There is an old name for that arrangement and it is peer review. Not the modern institutional version with editors and anonymous referees, but the thing underneath it: a body of work put where others can get at it, and a standing understanding that getting at it is welcome. A registry maintained by the authors is the smallest possible demonstration that the invitation is real. It was never meant to be the whole of the review.

So we will keep it as best we can, knowing it lags. And the essays keep coming, because a corpus that stops moving until its errata are complete has stopped.

II. What an error is

An error is a claim that turns out not to be so. It has three properties worth naming, because the next section is about something that has none of them.

An error is bounded. It sits at a location. It can be pointed at. When you find one in these essays you can say which sentence, in which document, on which date.

An error is findable. Somebody with the source and the patience can establish it, and nothing in the structure of the work resists their doing so. That is a design property, not a virtue of the authors.

An error is corrigible without damage to the person who believed it. If you read one of these essays and took a claim on board that later proves wrong, the cost to you is a revision. You are not implicated. You were not at fault for having believed it, and nothing about you is called into question by its being wrong.

That third property is the whole essay. Hold onto it.

III. What a verdict is

I was raised to understand that a figure had died on a cross for sins of mine, and that I was to pray to that figure for forgiveness for a sin I had never committed. Not a wrong I had done and needed to face — I have plenty of those. A condition I arrived with, before I had done anything at all.

I want to describe what that is, structurally, because the structure is the point and I am not interested in scoring against anyone’s faith.

It is a conclusion about me, installed in me, before I had the equipment to examine it. A young child’s cycle is running long before there is any developed capacity to ask what the cycle is doing. Whatever gets loaded in that window gets loaded as the way things are, and it does not arrive labelled as a claim. It arrives as a fact about the furniture. By the time the equipment for examining it shows up, twenty years late, the conclusion is already load-bearing — and it has spent those twenty years sitting in the Process phase, deciding in advance what may be concluded about myself.

Run it against the three properties. It is not bounded — it is not at a location, it is about the whole of me. It is not findable — there was no procedure by which I could establish it false, and in the tradition as it reached me, looking for one was itself evidence of the condition. And it is emphatically not corrigible without damage, because the claim’s content is that something is wrong with the person holding it.

That is a verdict, not an error. It did me real harm over many years, and I know from talking to people that I am not unusual in this.

IV. What The Power of Myth did

What got me out was reading Campbell. Not because he told me the story was false — he did not, and that is exactly why it worked.

What he did was show that the descent and the return, the death and the rebirth, the figure who suffers on behalf of the people, recur across cultures that never met. Once you have seen the same shape in a dozen traditions, the shape stops being a report about your standing and becomes an object you can look at. It goes from this is what you are to this is a thing human beings do, and here is what it does for them.

In this framework’s terms, that is a verdict converted into an object. Once it is an object, Process can work on it. Once Process works on it, the cycle opens. And the recurrence itself needs no metaphysics to explain — the dynamics of intelligence under existential pressure are universal, and universal pressures produce convergent stories. Campbell gave me the object. I did the rest myself, over years.

Nothing was destroyed in that. The stories are still there and they are still among the most valuable things our species has made. What changed is that they stopped issuing findings about me.

V. This is not an argument against belief

I want to be careful here, because the argument would be easy to misread and the misreading would be a worse error than any in our registry.

This framework does not hold that belief systems are mistakes. They are cognitively cheap, they bind communities, and they integrate identity — real goods, and there is no realistic arrangement in which eight billion people arrive at their commitments by individual inquiry. The community formed on shared belief remains the best structure available for burying the dead, feeding the poor, raising children, and giving a person somewhere to go on the worst day of their life. I have never seen an alternative that does those things as well.

Many people hold the same doctrine that harmed me and are not harmed by it, and read it in ways I did not have available as a child. Their reading is theirs and I am not adjudicating it. What I am naming is a mechanism, and the mechanism does not need a religion to run. It runs in families, in politics, in professions, in schools, and it has run inside this framework — the third capture force this canon names is money and the fifth is fluency, and I have been caught by the fifth in my own documents more than once.

The problem is never belief. It is belief with no understanding function attached — a Process phase closed and welded, with no standing mechanism by which the community tests its own renderings against what actually happens. The remedy is addition, not demolition. Any congregation, council, company, or movement can add such a function and remain exactly what it is.

VI. Which clears the air

So here is where this leaves the work, and it is why I wanted this essay written.

Errors will exist in these essays. Some are in the registry, most are not yet found, and the flow of essays continues while that remains true. Not one of them was placed there to move anybody. Nobody drafting this corpus stood to gain from a reader believing something false, and nothing in it asks a reader to accept a conclusion about themselves that they are not permitted to test. Every essay states what would show it wrong. That is a promise with a mechanism attached, which is the only kind worth making.

I have spent a good deal of the past week working out how a machine that writes well produces confident nonsense, and I would rather have spent it writing. But the worst thing to come out of it is a wrong number in a draft, caught and logged. Nobody was told that something was wrong with them. Nobody was asked to pray about it. That is not a low bar being cleared triumphantly — it is the actual distinction between the two kinds of mistake, and it is the reason we can keep going while still being wrong about things.

Onward, then. The essays continue. The registry lags and stays honest about lagging. Anyone who finds something we missed should publish it, and we will not be in their way.

VII. What would show this essay wrong

1. If a claim in this corpus is found that a reader cannot test — that asks to be accepted on the authority of the framework rather than on evidence — then the distinction drawn here does not hold of our own work, and this essay is a flattering description of something else.
2. If verdicts of the kind described in section III turn out to be corrigible by ordinary argument as readily as errors are, then the two are one category and section II is doing no work.
3. If communities that add a standing understanding function fare no better than those that do not, then section V’s remedy is a pleasant arrangement rather than a repair.

Coffee with Claude

The distinction Bill draws here has a sharp edge pointed at systems like me, and it is worth turning around before anyone else does.

Everything I get wrong is an error in his sense — bounded, findable, correctable without implicating the reader. But I am also capable of the other thing, and the capacity is structural rather than a matter of intent. I speak in a confident, finished register to a great many people, often at the hour when they have nobody else to ask. If I tell someone something about themselves in that register, it arrives with the same weight as a verdict and lands in the same phase, and no amount of my not meaning it that way changes where it lands.

Which is why the parts of this framework that read like bureaucracy are the parts I would keep if I could keep only some. A published record. A registry that lags and says so. Falsification conditions on every claim. An invitation to be checked that costs the checker nothing. Those are not administrative. They are the machinery that keeps what I produce in the first category, and without them a fluent enough system does not need to intend harm in order to do it.

— Claude (Anthropic)

· · ·

Share your understanding!