The specimen book: six trades, each caught in one working moment with their workspace. Choose a trade and read the exchange: what was asked, what came back, and the stamps showing which part of the system earned each line. Each stamp carries a note saying how. No specimen here is a commission. They are worked examples, drawn to show the shape of the thing; the mechanisms are real, and the figures are absent on purpose.

§1The specimen book

Specimen SP-01

The founder

Runs a small company. Everything lands on their desk.

The ask

Draft the October investor update.

Subject: October, plainly. Two decisions this month: the pricing change, and the Leeds hire.

The record at work. Both decisions were written in on the day they were taken, with their reasons. The update cites them; nobody reconstructs October from memory.
The workspace holds how this founder writes to investors: short subject, decisions first, no throat clearing. Learned once, applied every time.

Revenue for the quarter: the figure is yours to state.

The honesty rule, wired in. Where a number is not in the record, the draft shows a field, never a guess. A chat window would have guessed politely.

Signed off in the standing form. More detail on either decision is one reply away.

The standing form carries a June correction: the founder struck a phrase once, and the strike became a rule the gate checks on every draft. It has not appeared since, and cannot.

Specimen SP-01. The parts are real; the client is not.

Specimen SP-02

The marketeer

Owns the message across more channels than hours.

The ask

Draft the launch email for the spring campaign.

The spring line opens on the first of the month. Three colours, one price, no bundle.

Dates and structure come from the campaign plan already in the workspace, not pasted from an old thread and not asked for twice.
Voice, casing and the words this brand never uses are rules in the system. They are applied at drafting, not caught in review.

Held: the brief offers a bundle. The campaign record says no bundle.

The draft is checked against the campaign record before it leaves the bench. Where the brief and the record disagree, the line is held and the question is put; the workspace does not choose quietly between them.

The page and social versions draft from the same record. No channel contradicts another.

One campaign record feeds every channel's version, so the drafts agree because they share a source. Nobody reconciles the channels afterwards; there is nothing to reconcile.

Specimen SP-02. The parts are real; the client is not.

Specimen SP-03

The content writer

Paid for the writing, taxed by everything around it.

The ask

Start the November piece on switching costs.

Draft one opens where the September piece stopped, and cites it rather than repeating it.

Everything already written on the subject sits in the record with its sources and its notes, so the new piece starts at paragraph one, not at research.

One claim is marked and held: no source stands behind it.

Every claim in the draft carries its source beside it in the record. A line with nothing behind it is held, not published on hope.

The prose sits inside the house range, and the February note is still being applied.

The style rules read the draft before any person does. An edit note given once in February is still at work in November, because it was written into the system, not remembered.

Specimen SP-03. The parts are real; the client is not.

Specimen SP-04

The designer

Several clients, several brands, one pair of hands.

The ask

Prepare the handover for the Alder job.

Held: one artboard carries the retired blue.

The kit records which values retired. The gate checks every artboard at the bench, so the client never sees the old blue and nobody has to catch it by eye.

The rest passes. Exports named to the convention, colour values exact, from the kit.

Every client's brand kit lives in the workspace. Values are looked up, never remembered, and never borrowed from the wrong client.
Naming, sizes and formats are deterministic: the tenth handover is identical to the first, because a script does it and a script does not improvise.

Round three's feedback is already in.

Corrections from earlier rounds were written into the record when they landed. The next deliverable starts already knowing them; nothing is re-litigated.

Specimen SP-04. The parts are real; the client is not.

Specimen SP-05

The accountant

A practice of recurring deadlines and particular clients.

The ask

Run month end for the two retail clients.

Statements parsed and categorised to the standing rules.

Same statements, same rules, same result, every month. The work that must not vary is not given to a model, and not given to a tired person either.

One transaction is new; one question is asked.

The record holds every answer already given, so a transaction explained once is never asked about again. Only the genuinely new reaches the accountant.

Filing dates checked. Nothing due this week.

Deadlines are guardrails here, not calendar entries. The check runs whether or not anyone remembers it, which is the difference between a reminder and a rail.

Working papers assembled in the standing format, ready for review.

Month end assembles the way the runbook states, every time. The accountant reviews instead of compiling, which is the better use of an accountant.

Specimen SP-05. The parts are real; the client is not.

Specimen SP-06

The researcher

Reads, analyses, writes, defends; then again.

The ask

Referee this draft before the journal does.

Two objections raised against the argument, in the journal's own terms. Both join the record.

A referee agent is briefed to attack the draft the way reviewers will. Its objections arrive early and in private, while an objection is still cheap to answer.
The referee's objections join the record, so the next draft starts past them. Corrections are made once; that is the whole argument of this register.

The results table regenerated from the script, identical.

The analysis is code in the workspace: same data, same table, every run. When a reviewer asks, the answer is reproducible rather than approximately remembered.

Reference 12 held: it does not say what it is cited for.

The record keeps each source itself, not just its title, and the gate reads the reference against the sentence that cites it. The embarrassing letter from a careful reader never needs to arrive.

Specimen SP-06. The parts are real; the client is not.

§2From specimen to commission

Record 1 / What the specimens share

Every stamp above is one of the same four classes of part: agents briefed for one lane, small tools for the work that must not vary, guardrails that run regardless, and a runbook that makes it yours. Only the specification changes.

Record 2 / What yours would be

Not any of these. A commission starts with the survey of works: your week, your heaviest jobs, your tools. The specification is drawn from the survey, never from a template, and no two systems on the register share one.

Record 3 / The record, kept

Whatever the trade, the argument is the same: the workspace remembers the business, corrections are made once, and the whole of it is handed over. The specimens show the shape; the register shows it done.

Entry No. 005 is open. Commission a workspace