Correct Is Not Consequence

Validity belongs to the work; consequence begins when someone acts differently because the work reached them.

Elias Kunnas

A finished output sits in one of three states. Valid: sound on its own terms — true, well-formed, faithful, or technically correct. Recoverable: findable, legible, and matchable enough that someone could still use it later. Enacted: it reached someone who did something differently because of it. Only enacted work has changed an actor’s course, and the characteristic self-deception of the capable producer is reading valid or recoverable as enacted — counting a correct, unread artifact as contribution already made. What you owe past valid depends on what the work was for: an intervention is unfinished until it couples; a deposit is finished when it is recoverable; an instrument is finished when someone can use it.

I. The drawer

There is a folder on the capable person’s drive: the analysis no committee answered, the tool no team installed, the proof no field needed yet, the memo that routed to an inbox with no owner. The work in it is finished, and much of it is correct. The producer opens the folder, sees the accumulated competence, and tells himself the contribution is banked: the value is in the world now, waiting, and the only thing missing is a recognition he has chosen not to chase.

The folder is real. The correctness is real. And on the account that tracks what the world has actually had done to it, the folder has moved nothing.

The error points inward. Good work does go unseen, and that points outward at the routers and gatekeepers, which is largely true: civilization suppresses unaffiliated, early, cross-domain signals as a matter of structure (Optionality Has No Router). But the producer commits a second error against himself. He records correctness as consequence. The two are different properties of different things, and the gap between them is where capable people strand most of what they make.

II. Three states

A finished output can be in three states, and one does not imply the next.

Valid is the work against reality: is the claim true, the proof sound, the tool functional, the form faithful? Validity needs no reader. A correct theorem in a lost notebook is fully valid and has done nothing.

Recoverable means the artifact can still be found, read, matched to a problem, and routed by someone who did not make it. Recoverability is held open by specific conditions — retrievability, legibility, relevance, receptor-compatibility (whether any actor, institution, or vocabulary can yet use it), routability — and any of them can fail while the work stays perfectly valid.

Enacted: an output is enacted when it reaches an actor who then acts differently. Enactment is not a property of the artifact. It is a relation across artifact, channel, receiver, and changed action, and it is the only one of the three that registers as work the world has actually had done to it.

Validity is between the work and reality. Recoverability is between the work and a possible future. Enactment is between the work and someone who acted.

The self-deception is the collapse: seeing a valid, recoverable artifact and reading it as enacted. It is seductive because the first two states are the ones a producer can confirm alone at his desk, and the third is the one he cannot.

III. Recoverability decays

Recoverable correct work can keep for decades; dissipated correct work cannot. The hard case is Gregor Mendel. His 1866 paper on inheritance was correct and almost entirely ignored for a generation; around 1900 three botanists rediscovered it and it founded a field. The drawer, it seems, is not a dead end — so does the whole frame collapse?

It sharpens. Account the paper by date. From 1866 to 1900 it had no field-level enactment: it was cited only a handful of times, no program of inheritance theory formed around it, and it set almost no one on a different course. What it had across those thirty-four years was recoverability: it stayed in print, in a notation later biologists could read, indexed in enough libraries that three people independently found it when their own work made them need it. The 1900 enactment came not from the paper waking but from later researchers, live research questions, and a vocabulary catching up to a deposit that had stayed legible.

Mendel shows that recoverable correct work keeps, and recoverability is a holding that decays. Dormant work can also dissipate: the manuscript that goes illegible, the result superseded before anyone reads it, the file whose format no longer opens, the finding whose vocabulary drifts out of range, the memo nobody can locate. Stored possibility and dissipating stock sit on one gradient, and an artifact slides toward the dead end whenever one of its liveness conditions reaches zero.

The cases we cite are the survivorship tail. Mendel and Semmelweis are the rare deposits that stayed recoverable long enough to be enacted; the far larger class — correct work that lost legibility or relevance before any actor arrived — leaves nothing to cite, which is exactly why correctness looks self-preserving. Ignaz Semmelweis was right in 1847 and produced real local enactment: deaths fell in his own clinic. He won no profession-wide adoption in his lifetime, because the receptor his finding needed — germ theory — did not yet exist. Merit did not bind the case. The missing receptor did.

IV. Intent decides the duty

None of this says the drawer should be empty. It says the producer must know what each output was for, because the intent sets what counts as finishing it.

An intervention is made to change a present actor or decision: a policy model, a bug report, a pitch, a mechanism analysis aimed at a live bill. It is finished not when it is valid but when it couples, and the labor of coupling — making it legible to the receiver, routing it, framing the stakes, surviving the receiver’s filters — is part of the work, not a debasement after it. An intervention left at valid is unfinished, however correct.

A deposit is made to keep faith with reality for a future that may not arrive: an archive, a record of a dead path, a result with no current use, a work of art. It is finished by fidelity and recoverability. Chasing present receptors does not complete a deposit; it deforms it, bending the work toward whoever happens to be watching now, which is the one thing a deposit must not do.

An instrument is made to be used, repeatedly, by someone doing other work: a tool, a dataset, a map, a standard. It is finished when someone can pick it up at the point of need. Its failure is neither “unread” nor “deformed” but “unusable where the work happens.”

Most outputs are mixed, and intent drifts. Classify by the accounting claim you are making now. If you are counting present-world change, you are claiming an intervention and you owe coupling evidence; if you are preserving future-readable truth, you are claiming a deposit and you are judged by fidelity and recoverability. An artifact can be deposited today and converted into an intervention later: the accounting changes with the claim, not with the hope.

The specific error is the intervention made and then accounted as a deposit: you built something to move an actor, it moved no one, and you file it under “kept for the future” so the books balance. Sometimes that is honest. Usually it is the drawer, relabeled.

Capacity can wait inside a living producer; a finished artifact can wait only as long as it stays recoverable. This is the boundary most easily confused with the error. When the ports are genuinely blocked, preserving your own capacity — staying sharp, coherent, able to re-sense and re-couple when conditions change — is a complete and rational move (The Spore Strategy). A spore can act when the season turns; a finished artifact cannot wake, sense, or re-route itself. Treating your output as if it were your capacity, assuming the paper will bloom the way you would, is the same misread in its most flattering disguise.

V. The leak

At civilizational scale, valid interventions leak into drawers when their makers stop at correct. The producers most prone to it are the most capable — the people whose work is valid often enough that “it’s correct” reliably feels like “it’s done.” They make interventions, read the resulting recoverability as enactment, and under-invest in the one operation that would have turned the work into consequence. Then they read the silence as bad luck, or as the world’s failure rather than the work’s incompleteness, and they make the next correct, uncoupled thing. A scarce civilizational stock, the capacity to produce work that is actually right, is lost to drawers.

The vacuum is structural, which is what makes the misread easy to sustain: institutions leave a correct signal unanswered until some legitimate actor is obliged to answer it, and producing the signal does not produce that obligation (The Legitimacy Gate). The producer is right that the system is stacked against him. He is wrong that this releases him from the coupling, because for an intervention the coupling is the work.

Two states sit next to enactment and get mistaken for it.

Maker-development — the way making the thing sharpened your judgment and seeded your next work — is real value, but it is banked in you, not delivered to the world, and it pays out only if your improved capacity later couples to something.

Credit — recognition, status, citation — is also real and also separate. Technical enactment can run far ahead of it: LSTM-based models ran inside widely used speech-recognition and translation systems while most users never knew Hochreiter and Schmidhuber’s names. Credit can run far ahead of enactment: the celebrated paper nobody builds on. Neither development nor credit is consequence.

VI. Discipline and guards

The discipline is a five-question pass over any finished output:

For interventions, budget coupling as part of the work from the start, and record enactment only when you can name the actor and the changed action. Recognition is optional and often never comes; enactment is not, because without it the intervention did not occur.

One guard, because the diagnostic is dangerous turned outward. Absence of enactment is evidence of non-coupling, not of non-merit: a valid, uncoupled artifact is a routing failure, not a quality verdict. Use it on your own drawer first. It does not license downgrading someone else’s work as “just inventory” without knowing what it was for, which receiver it sought, what routing and preservation paths existed, and what would count as its later activation. Turned into a weapon, it arms the same gatekeepers who make the vacuum.

Validity remains real whether or not the world receives it. Truth is not downstream of an audience, and a correct deposit is a genuine deposit into the world’s recoverable intelligence. Keep the meaning of the work. The honest account refuses one move only: marking an intervention enacted on the strength of its being correct.

Correctness earns the making. Only consequence is the contribution an intervention was built to make — and for anything built to move the world, the distance between correct and enacted is not someone else’s job. It is the part of the work you have not finished.

Sources and Notes

The corpus dependencies. This essay names a producer-side self-accounting error; the system-side machinery it depends on is developed elsewhere. Optionality Has No Router is why the vacuum exists: the four routers (money, institutions, social graphs, platforms) suppress unaffiliated, early, cross-domain signals. The Legitimacy Gate is why a correct signal does not bind until a legitimate actor must answer it (“the signal problem is being solved; the coupling problem is the work”). Mechanism Realism is the parent ontology: replace the quality claim with the mechanism that converts it into changed outcomes. The Spore Strategy owns the live exception — preserved producer capacity, which (unlike a finished artifact) can re-sense and re-couple. Powerless Intelligence is the same gap in the AI era: cognition abundant, response duty scarce. There Is No Altruism supplies the two-level discipline used in the close (keep the internal meaning; account the mechanism honestly).

Ignored-then-recovered specimens. Mendel presented his inheritance work in 1865 and published in 1866; it gained field-level traction only with the 1900 rediscovery by Hugo de Vries, Carl Correns, and Erich von Tschermak (historians dispute how fully Tschermak grasped the laws). The point here is narrow: the paper’s recoverability — in print, legible, library-indexed — is what let later researchers enact it. Ignaz Semmelweis introduced chlorinated-lime handwashing in Vienna in 1847 and reduced maternal mortality in his own clinic, but did not win profession-wide adoption in his lifetime; uptake followed only once germ theory supplied the missing receptor. Both are read as the survivorship tail: the larger class of correct work that lost recoverability before any actor arrived leaves no trace, which is why correctness looks self-preserving.

Delayed recognition (the academic name for “recoverable, not yet enacted”). A. F. J. van Raan coined “Sleeping Beauties” for papers unrecognised for years then suddenly cited; Ke, Ferrara, Radicchi, and Flammini, “Defining and identifying Sleeping Beauties in science,” PNAS 112:24 (2015), give the quantitative treatment. These describe enactment delayed by a live, recoverable deposit, not by the artifact acting on its own.

Adoption and execution. Everett Rogers, Diffusion of Innovations (1962), names the uptake pathway (channels, time, social system, adopter categories, critical mass) that correctness alone does not traverse. Derek Sivers’ “ideas are a multiplier of execution” is the compressed startup-canon version of correctness-is-insufficient; it collapses coupling, distribution, and realised work into one bucket, where this essay separates them.

The receiver-side bottleneck. Herbert Simon’s observation that “a wealth of information creates a poverty of attention” (1971) names the scarce substrate that an intervention must win to be enacted.

Credit is a separate state. Robert Merton’s Matthew effect (1968) describes how recognition concentrates on the already-eminent, and Stigler’s law of eponymy how it drifts from the original discoverer — which is why enactment and credit must be accounted apart. Long Short-Term Memory: Hochreiter & Schmidhuber, Neural Computation 9:8 (1997); LSTM-based models later ran widely in speech-recognition and translation systems while remaining little-known to users outside the field, a clean case of enactment running ahead of public credit.

The silent graveyard. Nassim Taleb’s “silent evidence” (The Black Swan, 2007) and the file-drawer / publication-bias literature describe the uncoupled, unseen mass that makes survivorship reasoning about “correct work eventually wins” unreliable.

Related: