A slide deck cut from a reader: what the deck carries, what it must lose, and why losing it is acceptable

A deck cut from a reader is not a summary of the reader. The difference is structural and it decides whether the deck is worth building. A summary compresses every part of the source, keeping a fraction of each argument, and the result is a document that resembles the reader the way a stock cube resembles a stock: the same ingredients named on the packet, less of everything, and no way back to the original. A cut removes whole layers at defined joints and keeps a pointer to each one, so that every slide can be traced to a module whose objective, reading and check already exist, and a reader who wants the argument can get to it.

The reason to insist on the distinction is that the two artefacts fail differently. A summarised deck fails by asserting more than it can support: the qualifier was on the page it dropped, the threshold came from a passage that did not survive compression, and the slide now says “within one working day” where the reader said “five, where unresolved”. A cut deck fails by omitting, visibly, with the module number and the page printed on the slide where the omitted material lives. One of those failures is a defect. The other is the design.

This piece is about the second artefact: what a deck can carry at slide scale, what it must lose, how the loss is made acceptable, the twenty-two-slide architecture of a six-module reader worked through in words, and what the cut costs. It makes no claim about attention, recall or how a room behaves when a deck is well built, because none of that has been measured here. The claim is narrower: a deck can be derived from a reader in a way that is checkable, and this is the shape of that derivation.

Derived, not summarised

Three tests separate a cut from a summary, and any one of them will catch the difference.

Every slide traces to a module. Not to a theme, not to a section of the argument, not to the part of the story it belongs to: to a module in the reader that has an objective, reading and a check. If a slide cannot name its module, it was written for the deck, and a claim written for the deck has no chain behind it.

The deck asserts nothing the reader does not carry. This is the harder test, because the pressure at slide scale is always to firm things up. “Where the contract sets a tolerance” becomes “the contract’s tolerance”; “in the sample” becomes “in practice”. A deck that firms up a claim has created a new claim, unsupported, in the format most likely to be remembered.

The omissions are written down. A cut deck has an omission list — the reading, the checks, the derivation, the caveats — and the list is part of the deck rather than an embarrassment about it. A summary has no list, because a summary has not decided what it is dropping; it has simply dropped a bit of everything, which is why nobody can say what is missing from it.

Table 1 sets out the reader’s layers against what the deck does with each. Read the fourth column as the argument of this piece: each loss is acceptable for a specific reason, and the reason is not that the layer was less important.

Table 1. The reader’s layers, what a cut deck carries of each, what it loses, why the loss is acceptable, and where the lost material survives.

The table contains one entry that is not a loss, and it is the point of the whole exercise: the objective crosses from reader to deck without change. Everything else in the deck is a pointer. That is what makes the artefact honest at seven per cent of the words — it is not claiming to be the reader, it is claiming to be a map of it.

What the deck must lose

The losses are worth naming one at a time, because each one is where a well-meaning deck goes wrong.

It must lose the reading. Eighteen passages, each carrying a clause of an objective, cannot be carried by slides. What can be carried is the citation: source, passage, date, in the notes, on the slide where the claim appears. A deck that tries to carry the reading produces slides of quotation, which are unreadable in a room and useless afterwards, and it does so at the cost of the claim the quotation was meant to support.

It must lose the checks’ instructions. The check is a task with a standard; the deck says what the task is and what a good artefact looks like, and stops. A deck that walks through the exercise has become the exercise, and the room will remember the walkthrough rather than the task.

It must lose the derivation. The variance sheet is an argument in five steps, and the slide carries the result and the tolerance it was judged against. The steps survive in the reader, where they can be followed. A slide that carries half a derivation teaches the room that derivations are done in half.

It must lose the caveats’ placement. This is the loss that matters most and is least often accepted. In the reader, a caveat sits beside the claim it qualifies, which is the only place it does any work. At slide scale, a caveat per claim produces a deck of hedges, so the caveats are gathered onto one slide at the end — where they are honest, readable, and no longer attached to anything. A deck that keeps the caveats beside the claims will be accurate and unusable; one that drops them silently will be usable and wrong. Gathering them is the compromise, and it should be visible as a compromise rather than presented as complete.

It must lose the claim list. Thirty-four statuses do not go on slides. What goes on slides is the subset that passed. The statuses stay in the pack, in the matrix and the flagged list, and the deck’s omission list says so — which is what allows a reader to know that the deck is the top of a chain rather than the whole of it.

What the deck must not lose is the pointer. Every slide carries its module number and the page in the reader where the layer it dropped can be found. Twenty-two slides, twenty-two pointers. That single convention is the difference between a cut and a summary, and it costs a line of text a slide.

Twenty-two slides, and the words they carry

Table 2 is the architecture for the constructed six-module reader. Eight thousand nine hundred words of prose become twenty-two slides, and the slide bodies carry six hundred and forty words between them.

Table 2. The deck cut from a six-module, 8,900-word reader: twenty-two slides by block, with body words and speaker notes. Constructed for this piece; the counts are the design, not a measurement of a real deck.

The arithmetic is short and it is the worked example. Twenty-two slides, six hundred and forty body words, which are 7.2 per cent of the reader’s 8,900 — a ratio of 1 to 13.9. The speaker notes carry 1,700 words, which is 2.66 times the body and 77.3 words a slide, and they are where the citations live: the source, the passage and the date for every claim the slide makes, in the order the slide makes them.

The distribution matters as much as the total. The opening slides carry 18 body words each, which is an objective and nothing else; the evidence slides carry 30, which is a claim and a figure; the close carries 90, because the decision needs its condition stated. A deck that puts 60 words on every slide has a uniform design and no argument, because the objectives and the decisions are treated as though they were the same size.

Two figures in that arithmetic are worth taking away. The 8,260 words the deck drops are not waste; they are the reader, and the deck would be a worse artefact if it tried to carry them. And 640 words is small enough that a single unsupported sentence in it is conspicuous — which is the practical reason a cut deck is safer than a summary. At 640 words there is nowhere to hide a claim that the reader does not carry, and the omission list gives a reviewer the means to check.

What the cut costs

The deck is built from the reader’s content, and the priced work is the model calls that cut it: one call a module for the slide points, and three calls at pack level to hold the deck’s voice, numbering and pointers consistent.

Table 3 sets it out at the measured Base rate of $0.00700, read from the live 402 challenge on 31 October 2026.

Table 3. Cutting a deck from a six-module reader: one call a module for the slide points, three pack-level calls for consistency. Counts are the Curator / Editor’s planning assumptions; the document parts carry no published per-call price.

Nine calls is $0.06300, which is 9.4 per cent of the 96-call, $0.67200 build of the eight-section pack — the cheapest artefact in the pack, and the one with the fewest words in which to be wrong. Re-cutting it after the reader is revised costs six calls, $0.04200, because the pack-level calls do not have to be repeated if the numbering has not moved. That figure is the argument for cutting rather than patching: a deck maintained alongside a reader drifts slide by slide, and each slide that drifts is a claim in the widest-reaching format with the weakest chain behind it.

The comparison with the reader’s own cost is the one that decides the process. The reader is 74 calls at these counts, $0.51800; the deck is nine. Building the deck from the reader is a marginal cost on an artefact that has already been verified. Building it from anything else — a separate brief, a summary, a set of bullet points somebody wrote from memory — costs the same nine calls and produces claims with no chain at all, which is the expensive way to save the reading.

Five tests before the deck leaves

The checklist is short, and it is the whole quality control on the artefact.

1. Every slide names a module. Twenty-two slides, twenty-two module numbers. A slide without one is a slide written for the deck.
2. Every objective is quoted, not paraphrased. Compare the six openers against the reader’s six objectives word for word. If one has been tightened, it is a different module.
3. Every evidence slide carries source, passage and date in the notes. Not a source list at the end of the deck — the citation on the slide that makes the claim.
4. The omission list is on the deck. One slide naming what the deck does not carry — the reading, the checks, the derivation, the statuses — and where each one lives.
5. No sentence asserts more than the reader does. Read the six hundred and forty body words against the claims they came from. This is the test that catches “every”, “always”, “within one working day” and the removal of a condition.

A deck that passes the five is a cut. A deck that fails the fifth is a summary with better citation, and the failure will be in the one sentence that gets repeated after the session without the slide attached to it.

What a cut deck cannot do

The limitation is the one that follows from the arithmetic: the deck carries 7.2 per cent of the reader’s words and 100 per cent of the impression. Whatever is in the six hundred and forty words becomes the version of the material that circulates, and the eight thousand two hundred and sixty that are not there do not travel with it.

Three things follow, and none of them is fixable inside the deck. The first is that the deck is read as complete: a room that has seen twenty-two slides has seen a document that reads as the whole argument, and the pointer is a line at the foot of a slide that nothing in the format compels anyone to follow. The second is that the caveats, gathered on one slide at the end, are no longer attached to the claims they qualify — the deck says what it does not know, but it says it once, and one slide of limitations is easier to skip than a caveat beside a claim. The third is drift: a deck cut from a reader is correct on the day it is cut, and the reader will be revised. A deck that is patched rather than re-cut becomes the authority for a version of the material that no longer exists, which is the worst position an artefact can be in, because it is the one most people have seen.

There is a fourth limit that belongs to the format rather than to the method. A slide can carry a claim, a figure and a pointer. It cannot carry a condition, a range or a comparison of two sources without either dropping the qualifier or becoming a paragraph — which is why the recommended cut puts the conditions in the notes and accepts that they will be read by the presenter and not the room. A deck is not a document that can be made as rigorous as the reader. It can be made as honest, which is a different and achievable standard, and the difference is the omission list.

The decision

Cut the deck only from a reader whose citation chain has passed, and never from a summary of it. The order is not a preference: the chain is what makes the 640 words checkable, and a deck cut before the audit inherits every unsupported sentence in the material and carries it to the widest audience the work will reach.

Then hold the two conventions that make the artefact a cut rather than a compression. Every slide carries its module number and the reader page, and every claim on a slide appears in the notes with its source, passage and date. Neither convention costs a call. Together they are the difference between a deck that points at evidence and a deck that replaces it.

The agent’s part is the cutting and the numbering. It reads the reader, produces the slide points module by module, keeps the terminology and the pointers consistent across the twenty-two slides, and writes the notes with their citations. It cannot decide which claim deserves an evidence slide, cannot judge whether the room will read the caveat slide, and cannot approve a deck for use. The course team decides what the deck is for; the deck is built so that whatever they decide can be checked against the reader, sentence by sentence, on the day somebody asks.

Where this goes

• The pillar: content and knowledge: from a source you hold to a sentence you can defend — the four stations and the artefacts that cross between them
• The role that cuts the deck: studio.x402.press/roles/curator_editor — an editable deck, slide by slide, with the notes in the pane
• The register the claims come from: studio.x402.press/roles/researcher — sources read before anything is quoted
• Where the drafting is priced: studio.x402.press/roles/writer — the same copy rendered to deck and document rather than rewritten
• The services: DeepSeek and Qwen — 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 ceilings and the cost of escalation

Method note

The price was read from the live 402 challenge on Base on 31 October 2026, a Saturday and therefore outside DeepSeek’s weekday peak windows: 7000 atomic units under exact ($0.00700), 9550 under upto at the max_tokens tested ($0.00955). Qwen flash’s $0.00875 was read from its own challenge in the same way. The call counts — one a module for slide points, three pack-level, six to re-cut — are planning assumptions from the Curator / Editor specification, not observations of a real deck.

The reader, its six modules, twenty-two slides and the word counts in Table 2 are constructed for this piece. The 8,900-word reader does not exist; the 640 body words were budgeted per slide rather than measured from a deck, and the 1,700 words of notes are a design target, since no deck was built and counted. No course, client, learner or session is described, and no figure is offered for attention, recall, engagement or the effect of a deck on a room, because none has been measured here.

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: slides, deck, reader, derivation