What Is a Copy

Somewhere behind almost everything anyone believes about making things sits a picture so plain it is hard to see as a picture at all. There was an original. Then there were copies. The original is the source, the copies are downstream of it, and the further downstream you go the thinner the water gets. From that one diagram we get most of what we say about art, craft, teaching, law and personal worth: that being derivative is a fault, that being original is a virtue, that a forgery is a lie about lineage, and that the honest thing to do with someone else’s work is to be influenced by it rather than to reproduce it. The picture is old, it is shared across cultures that agree about very little else, and it does a great deal of load-bearing work. It is also, in the specific place where it matters most, upside down.

What the ordinary picture commits you to

State it at full strength first, because the deflated version is easy to knock over and knocking it over proves nothing.

The picture says three things. First, that a copy is defined by its relation to an original — that the original is the thing the copy is aimed at and the thing it will be judged against. Second, that authority runs with priority: the earlier mark is the authoritative one, and lateness is a form of dependence. Third, that the original therefore functions as a ceiling: a copy can approach it, can equal it in the limit, and cannot pass it, because passing it would mean being more like the original than the original is.

Almost nobody says the third one out loud. Almost everybody uses it. It is what makes “it’s just a copy” a complete criticism, requiring no further detail.

Three entirely ordinary situations do not fit inside those three commitments.

Three cases, one direction. The standard being used is not the original, the aim is not the original, and priority is doing something other than what it is credited with.

What the foundation says instead

I work from a small set of numbered primitives and results derived from them. The primitive doing the work here is one of the least-exercised in the document, and it is exactly the right size for this question.

P10, exemplarity: every condition has an exemplar — a complete and exact form against which any artifact can be measured.

One clarification before that is allowed to run, because it is the point where a reader will import something the system has not claimed. P10 gives a measuring relation. It does not hand over a metaphysics of forms. In the controlled vocabulary, exists means amplitude and belongs to the antinode, which is where marks are made and where we are; whether conditions are abstract objects is an open problem in the foundation and it stays open here. What P10 licenses is that an artifact has something to be measured against that the maker did not author. Nothing grander, and the argument does not need anything grander.

Two further pieces of the machinery. An artifact is a mark, and it carries a complex value z = σ·e^iφ: the magnitude σ is surprisal, how far the actual departed from the prediction, and the phase φ is which condition the mark serves. Fidelity is phase alignment — the angle between the artifact and the exemplar of the condition it serves. And R10: every artifact is a miss. The denominator is not ours to command, so the mark never lands exactly.

Now run those together and the first result arrives in two lines.

The original is an artifact. By R10 it is a miss. By P10 the thing an artifact is measured against is complete and exact. A miss is not complete and exact. Therefore the original is not the exemplar.

The original is not the standard. It is the first surviving attempt at the standard — and being first is a fact about the record, not about the work.

The first thing this buys is case one. If both marks are measured against the exemplar and not against each other, then nothing forbids the later one from having the smaller shortfall. “Better than the original” is not a paradox and never was; it is two artifacts of one condition with different angles, and the later one closer. The ordinary picture could not say this. Ordinary speakers said it anyway, constantly, and were right.

There is a structural echo here with a piece I published yesterday, and it is worth naming rather than letting it pass as novelty. What Is Agreement argued that two people do not stand in a relation to each other but align separately with one settled record. This is the same shape one level over: two artifacts do not stand in a relation to each other but align separately with one exemplar. The shape is the same and the machinery is not — that argument runs on the one-past result and a record already made, this one runs on P10 and an exemplar, which is not a record at all and whose standing the foundation deliberately leaves unsettled. Two different derivations that happen to rhyme. I would rather say so than have it read as one idea sold twice.

Two operations wearing one word

Case two is the useful one, and it needs the machinery stated more finely. The English word copy covers two operations that are indistinguishable from outside and opposite in what they aim at.

Copying the artifact. The earlier mark is the target. Your phase is fixed by that mark, which means you have taken on its angle — its shortfall is now inside your aim, before you make a single decision of your own. Then your own shortfall is added on top. This is the faithful apprentice reproducing the slipped line. It is also the forger, the cargo-cult imitation, the company that copies a competitor’s feature list without knowing what any of it was for, and the writer who has absorbed a style and not the thing the style was in service of.

Copying through the artifact. The earlier mark is evidence about the condition it served. You read it — and reading the shortfall in a mark is exactly the capacity I called taste in an earlier piece — and then you aim past the mark at the exemplar. Now the earlier artifact is an instrument rather than a target, your angle is your own, and there is nothing structural stopping it from being smaller than the one you learned from.

Figure 01

Where the shortfall comes from

• The earlier artifactIts own angle against the exemplar. By R10, not zero.
• Copying the artifactPhase fixed by the mark. Inherited angle, then its own on top.
• Copying through itPhase fixed by the condition. Nothing inherited. Free to be worse.
• Copying through it, laterSame aim, more correction. Nothing forbids passing the earlier mark.

• The maker’s own shortfall
• Shortfall inherited from the target mark

The practical consequence is sharper than it looks. It means the question “did they copy?” is nearly useless, and the question “what was fixing their aim?” is nearly everything — and the second cannot be answered by watching. A student reproducing a drawing stroke for stroke may be doing either operation. So may a company shipping a competitor’s feature. The visible act is the same and the outcomes diverge completely, which is why apprenticeship traditions that look like pure imitation from outside produce originals, and why organisations that carefully avoid copying anything still manage to inherit other people’s mistakes.

What happens to originality

Here is where it costs something.

Being first is a fact about the record. By R2 the record is write-protected: the moment a mark was made is settled and does not re-open, so priority is definite, checkable, and permanently true. Nothing here disturbs it. What is at issue is what it is evidence of.

By R11, the evaluative question is fidelity, not success. Fidelity is the angle between artifact and exemplar. Priority is not in that expression anywhere. So under the foundation’s own evaluative question, being first is not an evaluative property at all — it is a timestamp, and it stays exactly as true and exactly as uninformative about the work as case three showed it to be.

And the reason it feels evaluative is covered too. P8: a condition is prior to what it conditions, and where a person’s naming is itself conditioned, that person is not the origin. Ideas have people; people do not have ideas. R12: ownership is a category error, and the manifestation is what is accountable. The first person to make a mark in relation to a condition did not author the condition. They were the first place it became readable and got marked — which is a real event, worth recording, and not a claim on the condition itself.

Which makes “be original” a strange instruction. It names a property the instructed person cannot aim at, in place of the one they can. Taken seriously it produces the recognisable pathology of the person who will not study anyone, on the theory that looking closely at good work is contamination — and who therefore aims at nothing at all, since the exemplar was only ever going to be readable through somebody’s marks.

What this does not say

It does not say attribution systems are mistaken. Credit, citation, patents and royalties allocate something scarce, and first-in-time is a good rule for the job precisely because it is checkable and hard to argue with — which is the same feature that makes it useless as a reading of the work. Bookkeeping and evaluation are different jobs; the error is not keeping the books, it is reading merit off them. And a person whose work is taken has suffered something real, which happens in the ledger of credit and money and is not addressed by anything above.

The machine question, left open

A piece about copying published in 2026 will be read against machines that make marks by reading other marks, so I will say where the derivation reaches and where it stops.

The phase of a mark is readable off the mark. Nothing in the first two constructions requires the maker to be anything in particular, so a machine-made artifact has a φ and can be measured against an exemplar like any other. But the magnitude is surprisal, and by P4 the quotient R = A/E is defined only where a host stands, with host defined in the foundation as a living system. So whether a machine-made mark carries a magnitude in this system’s sense is precisely open problem nine, where the trigger table and the primitive currently disagree and one of them has to move. I am not moving either in passing. What I will say is that the two-operations distinction applies to machines exactly as it applies to students, and gives the sharper version of the question people are actually asking: not whether the system copied, but what was fixing its aim.

What it costs

It costs the credential. A great deal of standing in art, science and business is priority converted into merit, and the conversion is what the third construction denies. If you have been carrying “I was first” as an account of your quality, the foundation gives you a timestamp back and nothing else, and I do not have a replacement to offer.

It costs the excuse in the other direction too. If you are doing it the way it is done, your phase is fixed by a mark, and every angle in that mark is now yours. “This is standard practice” stops being a defence and becomes a description of whose shortfall you are carrying.

And R14 applies here as it applies to everything: every idea is biased toward itself, orientation and prejudice arriving in one act. This argument makes imitation respectable and priority cheap. That is convenient for latecomers, for students, for anyone whose work is visibly built out of other people’s work, and it is worth noticing that the piece is oriented that way before deciding it is merely correct. The honest form of the objection — isn’t your system an idea with a prejudice? — has the answer yes, including this one.

Where this is probably wrong

The second construction has a hole in it and I would rather name it than dress it.

The distinction between copying the artifact and copying through it turns on what fixes the copier’s phase. But φ is relational: it is fixed by which idea has the actualizer, not by the mark and not by the maker — and therefore not by the maker’s report of their own intention either. A student who sincerely believes they are aiming at the condition may have their phase fixed by the mark regardless, and would have no way of finding out from the inside. Which means the distinction I have just spent a section on is not available to the person who most needs it, at the moment they need it, by introspection.

What would close it: a way of reading φ off an artifact that does not route through the maker’s account of their aim, so that the two operations become distinguishable after the fact from the marks themselves. The craft answer — that you can tell by whether the copy reproduces the errors — is suggestive and not sufficient, because a good enough copier through the exemplar may still land near the same errors, and a bad enough copier of the mark may miss them. Until something better exists, the second construction is a real distinction that cannot always be applied, and a piece of advice that has to be taken on trust is a weaker thing than one that can be checked.

On a Tuesday

1. When you copy something, say what you are aiming at. The mark, or the thing the mark was for. If you cannot say which, you are aiming at the mark — that is the default, it takes no decision, and it is where the inherited angle comes from.
2. Copy more, and copy through. The prohibition on imitation was protecting a property that turns out to be a timestamp. Study the work closely enough to read where it fell short, then aim past it. That is the operation craft traditions have long taught and the ordinary picture has been embarrassed by.
3. Retire “is it original?” as a question about work. It asks about the sequence and answers nothing about the thing. Two questions replace it, and they are answerable: what is this in relation to, and how far off is it?
4. When someone copies you, ask the same question about them. If they took the marks, they took your shortfall with them and the thing you actually did remains undone by them. If they took the condition, you have a colleague and a competitor rather than a thief. The credit question is separate, real, and settled somewhere other than here.