Shadow Work — hadleylab.org/blogs

A commit is the visible tip of work, and the smallest part of it. The salesman who closes the deal, the economist who brings the elasticity, the realtor who demos the app to a room, the late-night thread where the idea is actually born — none of it is a commit, and all of it is governed work that produced value. It exists, but only in the transcript: the message, the email, the call, the session log. Call it shadow work — real work that casts no keystroke. Who Does the Work showed the ledger refuses a coin with no work behind it; shadow work is the opposite failure, and the larger one — work with no coin in front of it, value created and never attested. The fix is not to do more. It is to attest what was already done. Because shadow work is already finished and already recorded, attesting it mints no new labor — it surfaces existing labor — which makes it the rarest thing in any economy: growth at zero marginal cost. The transcript is the ledger; a walk of it is the only way to draw who did what; and the strategy that follows is simple — maximize attestation, and the economy grows to the size of the work that was always there.

The commit is the tip; the work is the iceberg

Software pretends the commit is the unit of contribution because the commit is the part a machine can see. It is a convenient lie. The commit is downstream of a hundred acts that have no SHA: the customer call that revealed the feature, the diagram on a napkin, the demo that closed the room, the domain expert's correction that saved a month, the argument that killed the wrong design before it was built. Each is work in the plain sense — effort that moved value — and none of it compiles. An economy that pays only commits pays the typist and stiffs the author, the seller, and the teacher. It mistakes the shadow the work casts on the filesystem for the work itself.

The honest accounting is the reverse. The commit is the shadow; the work is the body that cast it. And most bodies cast their shadow somewhere other than git — into a transcript.

The transcript is the ledger of shadow work

Every act of shadow work leaves a record, just not in the place code looks. The deal-close is in the message thread. The elasticity is in the white paper and the call. The pilot is in the demo and the email confirming it. The idea is in the conversation where it first got said out loud. These are not lossy memories — they are records, timestamped, attributed to a participant, durable. The transcript is to shadow work what the commit log is to code: the evidence that it happened, and who did it.

This is why work is the transcript-session and not merely a ledger row: a coin's evidence resolves not just to a type and an actor but to the session it happened in and, through it, to the passage where the work is recorded. The transcript is not documentation of the work. For shadow work, the transcript is the work — the only place it exists in a form the economy can read.

Free attestation: growth by surfacing, not by doing

Here is the property that makes this matter. New work costs effort — you have to do it. But shadow work is already done. The call already happened. The deal already closed. The idea was already had. Nothing remains to be performed; the only thing missing is the attestation — the act of recording, in the ledger, that this work occurred and whom it belongs to.

So attesting shadow work mints no new labor. It does not pay for work not yet done; it surfaces work already done and never credited. The supply it adds is matched, to the cent, by work that already exists in the transcripts — so it is not inflation, and not an unearned mint (a coin still never posts without its evidence; the transcript is the evidence). It is the one move in economics that creates measured value at zero marginal cost: not do more, but see what was already done. An economy with a large unattested shadow has already grown; it simply does not know it yet. Attestation is how it finds out.

The transcript walk is backprop

If the work is in the transcripts and the attestation is missing, the operation is mechanical: walk the transcripts and draw the edges — this contributor, that work, this much value. No other method can do it, because no other place holds the associations. You cannot infer from the commit log who closed the deal; the commit log does not know. You have to read the record of the shadow work and assign credit.

Assigning credit across a record of who-enabled-what is not a new problem with a new name. It is backpropagation — credit-assignment over a dependency graph, the same operation that trains the gradient. The transcript walk is backprop run over the contribution DAG: propagate the value of each attested outcome back to the shadow work that produced it, contributor by contributor. The economy is trained on its own transcripts.

It decays, because learning decays

Shadow work earns the same way every contribution earns: while it is still teaching the system something. The first time a domain expert's framework reshapes the model, that is pure signal — high attestation. The tenth derivative built on it has mostly learned the lesson already; the marginal learning is smaller; the residual decays. Once the contribution is fully absorbed — the playbook memorized, the model internalized — it teaches nothing new, and there is nothing new to earn. The residual stabilizes at a floor.

The decay is indexed to learning, not to the calendar — a slow contributor is never punished by time passing, and a foundational one is never paid forever for a lesson long since learned. This is the anti-rent rule stated in its sharpest form: rent is income with no new learning behind it, and the decay is exactly the curve that drives it to zero. You are paid for what your shadow work still teaches, and only that.

One primitive

Pull the threads together and they collapse to a single identity. The transcript is where the shadow work lives; the shadow work is the governed work; the work is the economic value it casts. Intel, work, coin — one primitive, three views. The record of the conversation is the knowledge produced, is the labor expended, is the value owed. There are not three ledgers to reconcile; there is one object seen from three sides, and the transcript holds it.

Which is why the growth strategy is not a slogan but a consequence. If intel is work is coin, and most work is shadow work recorded only in transcripts, then the economy's true size is the size of its attested transcripts — and the way to grow it is to attest more of what is already there, decayed by what has already been learned, bounded by what actually happened. Maximize attestation. The work was always real; the only question is whether the books ever learn it happened.

Sources

Claim	Source	Link
Shadow work — real-world interactions that cast no commit, recorded only in transcripts — is attested in COIN; attesting already-recorded work is free attestation that grows the economy at zero marginal cost; a coin still never mints without its transcript	SERVICES/COIN/CANON.md — `SHADOW_WORK_IS_FREE_ATTESTATION`	canonic.org
A work-mint's WORK is the gov_session and, through it, the transcript passage; the transcript is the work	SERVICES/COIN/CANON.md — `WORK_IS_THE_TRANSCRIPT_SESSION`	canonic.org
INTEL = WORK = COIN — one primitive, three views; the economy is knowledge, priced	SERVICES/COIN/CANON.md — `ONE_PRIMITIVE`	canonic.org
The attribution residual decays indexed by learning (not time), stabilizing at a floor once a contribution is fully absorbed — the anti-rent rule	SERVICES/COIN/CANON.md — `DECAY_ENCOURAGES_SOVEREIGNTY`	canonic.org
The ledger refuses a coin with no work behind it (`actor`/`work` NOT NULL)	Who Does the Work, HadleyLab	hadleylab.org/blogs/2026-06-11-who-does-the-work
Innovation is the derivative of the identity, dWork/dCoin	Innovation Is dWork/dCoin, HadleyLab	hadleylab.org/blogs/2026-06-15-innovation-is-dwork-dcoin
Credit-assignment over a dependency graph is backpropagation; the learning gradient is the trained signal	SERVICES/LEARNING	canonic.org

The work was always real. Attestation is just the books finding out — and an economy grows to the size of the work it can finally see.

Shadow Work | THEORY | BLOGS