A course reader is a document with one job: to carry a set of modules in which every objective has reading under it and a check beside it, so that somebody who was not in the room can see which source was meant to carry which objective, and what was produced to show the objective was met. It is not a syllabus, a lesson plan or a curriculum. It is the artefact the course is built from, and the test of it is mechanical rather than philosophical: for each module, is there an objective a check can test, is there reading that carries that objective, and does the check produce something a reviewer can read against the objective without asking the author what they meant?
Those three questions do not require a theory of learning to answer. They require four columns and the discipline to leave one of them empty rather than fill it with something plausible. A module with an objective and no reading is an ambition. A module with reading and no check is a reading list. A module with a check that cannot fail — a discussion, an invitation to reflect, a slide the learner is asked to agree with — is a formality wearing the clothes of an assessment. The reader is the place where those failures are visible before the course is taught, which is the only point at which they are cheap to fix.
This piece is about the artefact: the columns, what goes in each, a six-module reader worked through end to end, and the arithmetic of building one with the Curator / Editor assembly. It makes no claim about whether a course built on such a reader teaches better, and it holds no figure for completion, retention or pass rates, because it has not measured any. What it can show is the structure, and the cost of producing it.
The four columns
A module is not a topic. “Invoice matching” is a topic; it can be taught for ten minutes or ten weeks and the reader cannot tell which was intended. A module is one objective, the reading that supports it, the check that shows whether it was met, and — the column most often left out — the record of what the reading actually carries, so that the next reader of the document can test the module rather than trust it.
Table 1 sets out the four columns and the test each one has to pass. The fourth column is the one that turns a reader into a checkable document: without it, the first three are assertions about a course, arranged in a table.
Table 1. The four columns of a module, what each holds, and the test it has to pass. Constructed for this piece from the structure the Curator / Editor assembly produces: study material section by section, a coverage-and-citation matrix, and a deck cut from the same copy.
| Column | What it holds | The test it has to pass | What fails the test |
|---|---|---|---|
| Objective | One verb, one object, one condition of performance | A check can be written that produces evidence of it, and a reviewer can say whether the evidence meets it | Two verbs (“understand and apply”), a state of mind (“be aware”), or a topic instead of an action |
| Reading | The sources under the objective, each with the passage that carries it and the date of the passage | Every clause of the objective is supported by something in the reading, and the passage can be opened from the reader | A reading list with no passages; a source that exists only as a link nobody opened; three sources where one would do, arranged to look thorough |
| Check | The task the learner performs and the artefact it produces | The artefact can be judged against the objective without asking the author what they intended | A discussion, a reflection, a poll, or a task whose correct answer is visible in the reader’s own wording |
| Record | The claim list: what each source carries, at what status, as of what date | A reviewer who doubts one sentence can follow it to the passage in one step | A matrix of ticks; a bibliography with no passages; a source register that records titles but not what each source was used for |
The order is the logic, and the fourth column is why the reader survives contact with a second person. A reader that asserts its own coverage is a document of intent. A reader that carries the claim list is a document of evidence, and the difference between the two is not the quality of the prose in either.
The objective, written so that a check can test it
An objective has three parts and the third is the one people drop. The verb names what is done. The object names what is done to. The condition names the standard, the material or the constraint under which it is done, and without it two readers will disagree about whether the objective was met — which is the same as having no objective.
Compare “understand three-way matching” with “reconcile a purchase order, a goods receipt and an invoice for one supplier-month, and report each exception in the format the procedure specifies”. The second is not better writing. It is the only one of the two from which a check can be written, because it names the material, the unit of work and the output. A reviewer holding the learner’s sheet can say whether the exceptions were reported in that format. Nobody can say whether somebody understands three-way matching, which is why the first version cannot be tested and the module it heads cannot be reviewed.
Two rules follow, and they are rules rather than preferences because the reader’s structure depends on them. One objective per module, because a module with two objectives needs two checks and will get one, and the second objective will be the one that goes untested. And the objective is written before the reading is chosen, because a reading chosen first decides the objective, and the decision quietly becomes “whatever these sources happen to say”.
The objective is also where the reader refuses to promise anything about the learner. It says what will be done and to what standard in this document. It does not say that the learner will be able to do it next month, in another context, or faster than before. Those are claims about a person, and a document that has not observed the person cannot make them.
The reading under each objective
The reading column has an unusual property: it is the only column that can be verified against the world. An objective is a statement of intent and a check is a task; the reading is a set of documents, and each one either exists, is dated, and says what the reader says it says, or it does not.
This is where the assembly’s order matters. The register is built first — every document read into its own text before anything is condensed, its format, its date and what it was used for recorded — and the reader is written from the register rather than from a folder of files. The Researcher produces that register and the claim list that goes with it; the Curator / Editor reads the register into study material and speaking notes, and builds the coverage-and-citation matrix that shows which source carries which point. Neither station fetches the open web. A source that arrives as a link and nothing else is unread, and an unread source cannot carry an objective, which is why the reading column fills with the documents actually opened rather than the ones that looked right in a search result.
Three habits make the column usable.
Cite the passage, not the work. “Procedure note, rev. 4, §4.2” tells a reviewer where to look. The title of the note tells them a document exists. The first is a citation; the second is a gesture at one.
Record the date of the passage, not the date it was read. A tolerance in a contract, a threshold in a procedure and a figure in a report each describe a period. A module that cites a tolerance without the version it came from cannot be re-checked when the contract is renegotiated, and the reader that cannot be re-checked is the one that quietly goes stale.
Let a module have less reading than it wants. Three sources under one objective usually means the objective is really two, or that two of the sources are decoration. The reader is stronger with a thin column and an honest objective than with a full column and an objective that only the author can see.
The reading column is also where a reader admits the shape of its evidence base. If the register is ten documents and four of them are secondary accounts — a summary, a review, a press piece — the reader should say so, because a secondary account cited as the authority for a figure is a claim corroborated at one remove, and at one remove is not the same as verified.
The check that shows whether the objective was met
The check produces an artefact. Not a feeling, not a discussion, not a show of hands: a sheet, a reconciliation, a short written answer, a marked-up document. The artefact is what makes the module reviewable after the session, because it can be laid beside the objective and read.
Three properties decide whether a check does its job.
It must be answerable from the reading and not from the reader’s own wording. If the correct answer appears verbatim in the study material, the check tests recall of the material rather than the operation the objective names. A check that asks the learner to compute a variance is a different instrument from one that asks them to state the tolerance.
It must be capable of failing. A check that every learner passes is not evidence of anything; it is a record of attendance with extra steps. The check should be built so that a learner who has not done the reading produces something visibly short — an exception list with the exceptions missing, a variance computed against the wrong base.
It must have a standard written beside it. “Report the exceptions in the format the procedure specifies” is a standard. “Show you understand exceptions” is a request for a performance of understanding, and the artefact it produces cannot be judged.
What the check cannot do is worth stating as plainly as what it can. It shows that a task was performed and, against the objective’s standard, whether it was performed correctly on the material supplied. It does not show that the learner can do it unaided next month, that the course caused the performance, or that the module is well designed. It is evidence at one point in time against one standard, and a reader that treats it as more than that has made a claim about people that a document is not in a position to support.
One reader, six modules
Table 2 is the artefact: a six-module reader on supplier invoice matching, written for a finance team that already runs the process and needs the exceptions handled consistently. The subject is ordinary on purpose. The structure is the point, and it is the same structure for a reader on contract clauses, on records retention or on a piece of regulation.
Table 2. A six-module reader on supplier invoice matching: one objective per module, the reading under it, the check that produces evidence, and what the check cannot establish. Constructed for this piece; the six modules, nine sources and eighteen readings are an illustration, not a course that has been taught.
| Module | Objective | Reading under it | Check | What the check cannot establish |
|---|---|---|---|---|
| 1. The three documents | Name the three documents in a three-way match and state, for each, which field carries quantity, price and terms — for one purchase | S1 procedure note rev. 4, §2 (pp. 1–3); S2 standard extract, ¶7–9; S3 contract, clause 7.1 | A one-page annotation of a supplied purchase order, goods receipt and invoice, naming the field in each that carries quantity, price and terms | Whether the learner would find the same fields in a document set they had not seen |
| 2. Price variance | Compute the price variance for a supplied invoice against its purchase order and state whether it falls inside the tolerance the contract sets | S1 §4.2; S3 clause 7.4 (tolerance); S5 textbook, ch. 6 pp. 141–148 | A variance sheet for five invoices, each result beside the tolerance quoted from S3 | Whether a tolerance read correctly in the example is read correctly in a live invoice run |
| 3. Quantity and timing | Distinguish a quantity variance from a timing difference on a goods receipt, and say which document the invoice should be held against | S1 §5; S5 ch. 6 pp. 149–154; S6 regulator guidance, §3 | Three short cases, each classified, with the document the invoice is held against named | Whether the classification holds where a receipt is part-delivered across two periods |
| 4. Holds and releases | Trace a hold from the exception that raised it to the person who can release it, using the procedure’s own thresholds | S1 §6 (thresholds); S3 clause 9; S7 internal memo, 2025-06 | A filled hold log for four exceptions, each with the threshold that applies and the role that releases it | Whether the role named on paper is the role that acts in practice |
| 5. The supplier view | State what the supplier is entitled to be told, and when, from the contract and the procedure | S3 clause 11 (notification); S8 supplier handbook, §5; S6 §4 | A two-paragraph notification for one held invoice, written to the clause-11 standard | Whether the notice would be accepted by a particular supplier |
| 6. The monthly pack | Assemble a one-page exception summary from a month’s matching run, with every figure traceable to a line in the ledger | S1 §8; S9 month-end checklist; S10 ledger extract, September | The summary page, with each figure carrying its ledger reference | Whether the summary is read as the whole truth of the month or as the exceptions in it |
Nine sources carry eighteen readings, and the count of readings is deliberate: three per module is enough for a second reader to disagree with one source and still have the module standing, and few enough that an empty cell is obvious.
Read the right-hand column as the reader’s own statement of its limits. Each row concedes something a reader of the document might otherwise assume, and the concessions are not modesty. They are the difference between a document that claims to have assessed a person and one that reports what was produced when a task was set.
The arithmetic of building one
The build is model calls plus document parts, and the document parts carry no published per-call price. The call counts below are the Curator / Editor’s own planning figures applied to six modules rather than eight: seven reading and condensing calls a module, three calls for the study material and speaking notes, one call for the slide points, and eight pack-level calls to hold terminology, numbering and voice consistent across the sections.
Table 3 sets it out at the measured Base rate of $0.00700 a call, read from the live 402 challenge on 31 October 2026.
Table 3. Six modules priced at the measured DeepSeek rate on Base: seven reading calls a module, three for material and notes, one for slide points, eight pack-level. Counts are the role’s planning assumptions, not observations; the document parts are outside the total.
| Call | Unit price | Calls | Cost |
|---|---|---|---|
| Reading and condensing the register — seven calls a module | $0.00700 | 42 | $0.29400 |
| Study material and speaking notes — three calls a module | $0.00700 | 18 | $0.12600 |
| Slide points — one call a module | $0.00700 | 6 | $0.04200 |
| Pack-level consistency and revision | $0.00700 | 8 | $0.05600 |
| Total | 74 | $0.51800 |
The arithmetic is short and it is the whole of the worked example. Forty-two plus eighteen plus six plus eight is seventy-four calls; seventy-four at $0.00700 is $0.51800, which is $0.08633 a module and 12.3 calls a module. The assembly’s illustrative eight-section pack is 96 calls and $0.67200, so the six-module reader is 77.1 per cent of that pack’s calls and $0.15400 less — the difference being two fewer modules at 11 calls each, which is exactly the 22 calls between 74 and 96.
Three comparisons make the figure useful rather than decorative. At the metered ceiling of $0.00955 a call, the same 74 calls cost $0.70670, which is $0.18870, or 36.4 per cent, more — the ceiling is what an agent authorises, not what it spends. The 0.50 USDC credited to a workspace as working capital is seventy-one calls at the exact rate, so a six-module pack at these counts falls three calls short: $0.02100, or 4.1 per cent. That is not a warning about the price; it is a warning about the balance, and the answer is to fund the workspace from one pack’s counts rather than from the round number in the catalogue. And the document parts — six matrices, one deck, three renders, the mail — sit outside the total entirely, so $0.51800 understates the work rather than summing it.
The counts will be wrong for any particular reader. A register of forty sources costs more to read per module than a register of ten, roughly in proportion; a module with one long source costs less than a module with six short ones; a second pass after the reader’s first review is another three calls a module. What does not move is the shape: reading dominates the bill, because reading is the only step that has to happen before anything can be written, and it is the step a buyer is most tempted to skip.
What a reader is not
The limitation is not a footnote, because the artefact invites a misreading that costs money: a reader looks like a course.
It is not a course. It is the document a course is built from, and between the reader and a taught course sit the session design, the person teaching, the exercises done in the room and the judgement about whether the material landed. Nothing in the reader establishes that any of that happened, and nothing in the build above produces it.
It is not an assessment. The checks produce artefacts that can be read against stated standards. They are not validated instruments, they carry no score that means anything outside the module, and the reader holds no evidence about whether a check distinguishes a learner who understood the material from one who read it carefully the night before. A course that needs an assessment needs one built and tested as an assessment, which is a different job with a different standard of evidence.
It is not accredited, and it makes no claim to be. There is no awarding body behind these modules, no credit, no hours-to-credit conversion, and no statement anywhere in the reader that a person who completes it holds a qualification. Where a course built on such a reader is sold as leading to one, that claim is the seller’s and it should be checked against the body that awards it rather than against this document.
And the reading is only as good as the register. A register built from links nobody opened produces a reader whose citations resolve to pages that may not exist; a register of secondary accounts produces a reader whose claims are corroborated at one remove throughout; a register assembled quickly produces a reader whose coverage column is thin in the places nobody looked. The assembly flags what it cannot verify and leaves the flag in place — the matrix row, the mail, the flagged-items list — but a flag is a request for a decision, not a correction. Somebody has to read it.
The decision
Build one module end to end before building six, and test it with the four columns. Write the objective first, in one verb, one object and one condition. Choose the reading that carries each clause of it, with the passage and the date. Write the check and the standard beside it. Then try to break the module: find a reading that fails to carry a clause, a check that a learner could pass without the reading, and a clause of the objective that no artefact could show. A module that survives that attempt is one; a module that does not is a topic with a reading list attached, and the honest move is to cut it rather than to write around it.
Then apply the same test to the other five before the pack is built. The build is $0.51800 at these counts, which is cheap enough that the constraint is not money but the register: a module with no reading under it cannot be built at any price, and the decision that actually matters is made before the first call — whether the sources exist, whether they have been opened, and whether they say what the objective needs them to say.
The agent’s part is narrow and worth stating. It reads the register into material, writes the notes, builds the matrix and cuts the deck; it flags what the sources do not carry and leaves the flag where a person will see it. It cannot decide whether a module belongs in a course, cannot approve a section, and cannot tell you that a reader is good. Those stay with the course team, and the reader is built so that they are decisions somebody can make from the document rather than from a conversation about it.
Where this goes
- The pillar: content and knowledge: from a source you hold to a sentence you can defend — the four stations, the claim list and the verdict
- The role that builds the pack: studio.x402.press/roles/curator_editor — the reader, the matrix, the deck, the render and the mailbox
- The register the reader is built from: studio.x402.press/roles/researcher — sources read into text, claims mapped to passages
- Where the drafting is priced: studio.x402.press/roles/writer — material and notes in one voice, tables numbered
- The services: DeepSeek and Qwen — the canonical per-call prices on seven chains
- The reviews: DeepSeek and Qwen — the measured handshakes and what they do not promise
- The price ladder: inference endpoints compared — the three endpoints, the ceilings and the cost of escalation
Method note
The price was read from the live 402 challenge on Base on 31 October 2026: 7000 atomic units under exact, which is $0.00700, and 9550 under upto at the max_tokens tested, which is $0.00955. Qwen flash’s $0.00875 was read from its own challenge in the same way. The call counts are the Curator / Editor’s published planning assumptions applied to six modules rather than eight — seven reading calls a module, three for material and notes, one for slide points, eight pack-level — and they are assumptions, not observations of a real build.
The six-module reader, its nine sources, eighteen readings and six checks are constructed for this piece, and the subject was chosen because it is unremarkable: no client, no course and no learner is described, and no source is cited by title that a reader could mistake for a real document. No completion rate, pass rate, retention figure or learning gain appears anywhere here, because none has been measured, and the limitation section says so rather than leaving it to inference.
x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.
Filed under Jobs · Tags: course, reader, curriculum, structure
