Design record · Sketch 08 · Not part of the site
A way to say where a piece of work stands without either overclaiming it or hiding it. The ladder is W0–W6 from the Foundation Note §7; this is a proposal for how a stage is set on a page.
It was drawn as an issue stamp — the block on a drawing sheet reading preliminary or approved for construction, which governs what the drawing may be used for. Sketch 07 took the conditions off, because this record had invented them, so the analogy is now an argument for a form the page no longer shows. It is left standing because it was the reasoning, and because a stamp with a condition on it may still be right — with conditions someone had the standing to name.
Deliberately not a progress bar. A bar says completion is inevitable and only a matter of time; §7 says graduation is evidence-bearing and reversible, and that a project may stay a valuable internal pattern without ever becoming a public product. Marks on a scale can move backwards. A filling bar cannot.
Sketch 02. Three faults in sketch 01, all found by looking at it rather than by reasoning about it. The current-stage mark was a pale oxide ring and read lighter than the solid rungs behind it, so the one thing the stamp exists to say was the faintest thing on it. The rung labels were SVG text that scaled with the container and went illegible at 390px — you could see a mark was current but not which. And the record page left its right two-fifths empty at desktop width. The ladder is now an HTML grid with labels in rem, the current mark is solid oxide and the heaviest on the line, and the specimen notes sit in the margin.
Sketch 03. The margin claim above was only half true. Sketch 02 gave the specimens a right-hand column but left this header a single column inside the same sheet, so the top of the page went on running its right third empty — the fault the header was describing as fixed. The header now takes the same two columns as a specimen, and the illustrative-values caveat sits in the margin where a caveat belongs. The kicker also overflowed its line at 390px and dropped site onto a line of its own; the tracking comes down at narrow widths instead.
Sketch 04. The claim above holds only above 1088px, and it names no width. The margin column arrives at 68rem but the sheet was 62rem at every width, so between the two the sheet stood at full width with nothing in its right third and the double rule ran on past the last block — worst at 1080px, eight pixels short of the breakpoint that fixes it. The sheet is now the measure itself until the margin column exists to fill the rest, so the rules end where the content ends. Three sketches have now found the same fault in a width the previous one did not look at.
Sketch 05. Two things found by looking at 04. The compact variant was the one section with a note that never reached the margin — it was written like a specimen but not built like one, so its right third stayed empty while every specimen above it filled. It is a specimen now. And the evidence links wore the browser's own focus ring, a hard black box tight against the text and the only thing on the sheet drawn by something other than this page; the ring is oxide and offset now. The open questions below stay in one column on purpose — a sheet may end in its text column, and saying so is the point, since it was an unstated claim that cost sketches 03 and 04.
Sketch 06. The first two open questions are answered, so this sketch is drawn from decisions rather than from faults. “May not be said” is not mandatory, which means the stamp has to hold without it — specimen 4 is that case, at the one stage where there is genuinely nothing yet to withhold. The issue condition is derived from the stage, which means the mapping has to exist rather than being implied by three demonstrated stages out of seven; it is written out above, proposed from what each stage already claims. Both questions keep their place below with the answers attached, because a decision read without the question it answers is half a decision. 390px is now the narrow floor.
Sketch 07. Everything above is a fault found by looking at the page. This one was found by being asked where the words came from. “May be said”, “may not be said”, “next gate”, “evidence”, and the three issue conditions appear nowhere in the Foundation Note or anywhere else in the repository: this record invented them in sketch 01 and then spent five sketches setting them in the same mono, at the same weight, inside the same ruled box as the stage names, which are sourced. A reader had no way to tell the two apart — and the record's whole subject is a component not claiming more than its evidence supports. They are gone, with the rows, the specimen built on them and the mapping derived for them. What is left on a stamp is a subject, a stage line quoted from §7, and the mark. If the mark cannot carry the reading on its own, that is worth knowing before more wording is built on top of it.
Sketch 08. Two things, and the first is a fault of the same family as 04. This record had been showing a stamp the site stopped rendering: the component gained §7's sentence for the current rung, and a header naming the scheme, when it became a component — the record did not follow. Specimens 1–3 were therefore judgements of a form no reader sees, which is the fault of judging fiction that sketch 07 spent 208 deleted lines on, arriving by a different road. They are at parity now.
The second is the standing question, and the answer proposed is that the mark needs nothing invented — it needs the one sentence of §7 the stamp was leaving off. Graduation is evidence-bearing and reversible. A project may remain a valuable internal workshop pattern without becoming a public product. Every per-rung sentence in §7 is already on the stamp; this is the sentence about the ladder itself, and it is the precise denial of reading W5 as nearly finished. The stamp has obeyed it since sketch 01 in its form — marks that can move backwards, never a filling bar — but only the code comments said why, so a reader got the form without the reason. It takes a foot band answering the head band: the head says which ladder is being read, the foot says how a ladder is to be read, and both are the scheme speaking rather than the subject.
The band was first drawn unattributed, on the reasoning that the head already names the scheme and every other string in the box is §7's, so a reader would take the foot as §7's too. Proteus settled it the other way: say §7 explicitly. The reasoning was sound and still an inference asked of the reader, and this record has now twice been caught leaving a reader to work out where a string came from — once by inventing words and setting them beside sourced ones, once by arguing a form was self-evident. The citation sits right-aligned under the sentence in the label vocabulary, so it reads as a reference under a note rather than as a heading over what follows. It is gunmetal rather than oxide because oxide already carries the scheme head and the current mark, and the quietest line in the box should not compete with the loudest.
Stage values below are illustrative. None is a statement about any component's actual maturity — they exist so the design can be judged at low, middle and high readiness.
The test is whether the mark lets you name something at W0 without either embarrassment or inflation. Nothing here is doing that work but the ladder, which is the point of sketch 07: if the mark cannot carry the reading alone, no amount of wording around it was going to.
Workshop graduation path
Ecphorium
Stage W1 — Workshop experiment
The idea is tried aboard AstralMaris with explicit scope, rollback, and honest limitations.
The stage a component sits at longest, and the one where overclaiming is most tempting because there is real work to point at.
Workshop graduation path
ming-qiao
Stage W3 — Structural assurance
Identity, authority, failure behavior, intervention, recovery, evidence, and adversarial cases are specified and tested in proportion to the component's claims.
A high stage is not a clean bill, and until sketch 07 the stamp said so in words. Now the only thing on it holding W5 short of the end is the unfilled W6 rung. Whether that reads as “not finished” or merely as “nearly finished” is the question the rows were papering over, and it is now visible. This specimen is the control for sketch 08 below, which is the same stamp with one sentence added.
Workshop graduation path
ming-qiao
Stage W5 — Stand-alone project
The component has a neutral configuration model, installation path, public contracts, conformance tests, upgrade and recovery story, threat model, and clean source history suitable for independent use.
The same W5 stamp as above, with one addition: the last sentence of §7, verbatim. It is the only sentence in §7 that is about the ladder rather than about a rung — every per-rung sentence is already on the stamp, and this one was the one being left off.
It is also the exact denial of the reading specimen 3 exposes. “Nearly finished” requires that W6 be where the ladder is going; the sentence says stopping is a legitimate end and the mark can move back. The stamp has been enforcing this since sketch 01 — marks rather than a bar — but only in its form, and only the code comments said so. A reader had the form and not the reason.
Workshop graduation path
ming-qiao
Stage W5 — Stand-alone project
The component has a neutral configuration model, installation path, public contracts, conformance tests, upgrade and recovery story, threat model, and clean source history suitable for independent use.
Graduation is evidence-bearing and reversible. A project may remain a valuable internal workshop pattern without becoming a public product. Foundation Note §7
The check that costs the proposal its case if it fails. At W1 there is nothing to overclaim, so a sentence about not overclaiming has to earn its ink some other way — and it does: at W1 the same words say the experiment is allowed to stop here. One line reading correctly in both directions is why it is one line and not a condition keyed to the stage.
Workshop graduation path
Ecphorium
Stage W1 — Workshop experiment
The idea is tried aboard AstralMaris with explicit scope, rollback, and honest limitations.
Graduation is evidence-bearing and reversible. A project may remain a valuable internal workshop pattern without becoming a public product. Foundation Note §7
Where the full stamp would outweigh what it labels. Same marks, same colour logic, no prose — it says only where the thing sits, and links to the full stamp. It carries no foot band: the band is the ladder's caveat, and a chip not showing a ladder in full has no ladder to caveat.
Not placed anywhere on the site, and not to be placed yet. Three specimens are shown because a row of one demonstrates nothing — the repository has exactly one component carrying a stage. The third open question below holds the condition that would give this form a set to live in.
This is the answer to “how would a reader know?”, and it is the only reason the sketch line is allowed to keep proposing anything. If a string is not below, it is not on a stamp.
Withdrawn The question was about wording this record invented and never had the standing to propose. Both rows are gone; the question goes with them.
Withdrawn Same fault: the three conditions were this record's invention, not the Foundation Note's. Answered derived in sketch 06, which was answering a question that should not have been asked.
Neither yet One item in the whole content layer carries a stage — ming-qiao, at W2. A compact variant exists to make a set comparable at a glance; at a set of one it is not a variant but a smaller copy of the stamp already on that page. On the plate grid it also fails on its own terms: A.1, A.2 and A.3 are plates of the vessel, and §7 grades what could mature into an independent project. Marking all four invents three positions on a ladder those cards are not on. Marking one leaves a seven-rung ladder at W2 beside three unmarked cards, which reads as behind the others when the others are not on the course. The card already states its difference in a sourced word — the Exercised aboard signpost against Plate A.x — and carries it without a ladder. Revisit when two or more items in one set each carry a stage; until then the chip has no comparison to make, and the index it would live on builds itself once that set exists.
Accepted, and shipped Proteus, 2026-08-17. The foot band is in StageStamp.svelte and on /ming-qiao. Both strings moved into src/lib/content/stages.ts beside §7's stage names, so the component quotes them rather than carrying its own copy — the same discipline as the rung sentences, and the reason a paraphrase cannot drift in later. It closes the question this record opened in sketch 01 and answered wrongly for six sketches: what a mark needs beyond the mark was in §7 the whole time.
Proposed, not settled Sketch 08 offers one answer that requires nothing invented: §7's closing sentence, in the box, as a foot band. The two specimens above are it at W5 and at W1. This is a proposal because the question is yours — but it is a narrower proposal than the rows were, since it adds no vocabulary at all. It restores a sentence the stamp was already obeying silently. If the answer is that the mark needs nothing, the case against this is that a stamp carrying its own caveat is a stamp arguing with itself, and the ladder's form said it well enough alone. Superseded by the decision above.
Attribution settled The band names its source on the stamp — Proteus, 2026-08-17. It was drawn unattributed first, on the reasoning that the head band already names the scheme; the decision is that a reader should not have to infer it. The proposal above is still open; only this part of it is closed.