The pages
The working papers themselves. What stays in the repository instead is set out below the cards.
The apparatus, published on purpose: what changed and when, every error we caught in our own work, how the pages are built for screen and for paper, and what the measuring tools still cannot see.
These are the receipts: a dated record of what we got wrong and who caught it, the design system, the print rules, and the reasoning behind the calls that shaped the site.
They are public for a plain reason. A site that keeps telling you to check the source cannot then ask you to take its own work on trust.
What is a page here, what stays in the repository, and why this whole section sits outside the reading order, are below the cards.
The working papers themselves. What stays in the repository instead is set out below the cards.
Publishing our own errors is the method, not an embarrassment.
These are living documents, and a living document should show its work.
Every piece as it arrived, every substantial revision, and every fact-check — including the errors we found in our own work and how we fixed them.
The public half of the fact-check ledger, backfilled from the collection's full commit history and kept up from there. It records three kinds of change and no others: a piece added, a piece substantially revised, and a fact-check or attribution audit — typo and styling passes do not earn an entry. The corrections are the point of the page rather than something buried in it: a paraphrase we had dressed as a Sagan quote, a line credited to Baldwin that is Dr. King's, a forest-wide fungal network asserted where the literature is contested, a coinage returned to its author. Each names what we got wrong, what is right, and who caught it. Newest date first, every entry tagged New, Revised, Fact-check or Site.
Accessibility-first, printable, and credited.
The visual language of this collection is itself an argument about access.
One typeface chosen for legibility, five accents held at near-uniform lightness, and body text that never fades — with its numbers and its receipts.
Six sections: Typeface, Palette, Palette lineage, Principles, Print and Shared assets. Atkinson Hyperlegible, chosen because it was drawn for readers the standard face failed. Five accents held within sixteen points of each other in lightness, so no one collection outshouts another. Body text at 18.84:1, because full opacity is the house rule and a faded grey is a hierarchy decision made at the reader's expense. The lineage section is the honest one: the palette began as a Solarized variant and no Solarized value remains, so Solarized is cited for the method and never for the colours. Every artifact here is a single self-contained HTML file with no dependency beyond a web font.
Every number on the page, and why it is that number.
Every number here was arrived at by printing something and finding out it was wrong.
The working behind the print rack — the spec, the inks, the gates, and the faults that no gate can see.
The working behind the print rack, where every number was arrived at by printing something and finding out it was wrong: 190 × 259 mm as the intersection of A4 and US Letter, 9 mm margins rather than 10 because 0.4 mm of slack tipped a sheet onto a third page, ink on white in both media because browsers omit background graphics, borders instead of backgrounds, and five house hues darkened until they clear AA on white. Plus the three gates that measure it — and a standing list of the faults none of them can see.
A resonant frame, not a proof — and it says so.
Reference notes behind Love You Down To Your Star Stuff.
Backing material for anyone who asks why difference gets to be the starting point — a summary of Helen Edgar's essay “Why Difference Comes First.”
A summary of Helen Edgar's essay Why Difference Comes First, held here for anyone who asks why difference gets to be the starting point rather than a deviation from a norm: the cosmic microwave background needing tiny perturbations to make any structure at all, Deleuze on difference before identity, relational quantum mechanics, and ethodiversity. Kept as backing material rather than promoted to a zine, precisely because it is a resonance across four fields and not an argument any one of them makes.
The method is named here, and almost nowhere in the zines.
A reading of our own shelf, not a measurement — and a reading can be argued with.
ARLES mapped spread by spread onto the pieces that climb it: the one built on the ladder on purpose, three that arrive at it anyway, and two that sit on a rung and say so.
Attention, Relational, Lived Experience, Environment, Systems — the Stimpunks Design Method, upward a design order and downward a diagnosis. This page states why the acronym stays off the spreads (a named framework asks for assent before the argument does, and the fact stops doing the work the moment it becomes an example), then maps it: No. 59 runs the full ladder one rung per spread, each rung paid for by an animal on the spread before it, and its companion No. 58 runs the same rungs from a reading made before it was built while naming ARLES nowhere at all — not on a spread, not in its colophon, not in the ledger; No. 17 runs four rungs in order and has no Attention spread, which is a real gap; No. 15 descends to the floor for three spreads and then climbs; No. 42 reads downward the whole way. Against those, No. 31 occupies two rungs and says so, and No. 16 runs them out of order because its claim is about the Environment rung itself. Ends by publishing the exception rather than tidying it: four pieces already name ARLES on the page, and one draws it.
Writing it is what found the thing it had to disclose.
Almost nothing, and this page says exactly which almost.
No cookies, no accounts, nothing written to your device, and a search that never leaves your browser — then the small remainder, named rather than rounded to zero.
The sentence “most pages contact nobody” could not be written, and could not be fixed by rewording: until the day this page shipped, every page here fetched its typefaces from Google, handing over a reader's address before a word was rendered — on a site with no analytics script and no cookies. So the fonts were self-hosted first and the page written second. What is left is disclosed in the detail a reader needs to decide whether to care: our host counts pageviews server-side and treats distinct IP addresses as distinct visitors, quoted from its own documentation; and 292 embedded players on 13 pages still reach Google when those pages load. Including the part we got wrong the same morning. We moved 66 embeds to youtube-nocookie.com and called it a privacy win — then read Google's own description, which promises only that a view will not personalise what you are shown, and never that nothing is collected. The domain says nocookie; the documentation says non-personalised. Also: why there is no consent banner, why we decline the word anonymous, and why the Stimpunks policy for stimpunks.org — comments, logins, HubSpot — describes almost none of this site.
Most projects keep this material in a drawer. Ours is in the open for one reason, and it is a consistency argument rather than a virtue: a site that spends this much time telling you to check the primary source cannot then ask you to take its own work on trust.
Too Good to Check, over in Foundations, names six ways a fact goes wrong here. That essay is only worth anything if there is visible evidence of the checking actually happening — a dated record of what we got wrong, what is right, and who caught it. Otherwise it is a page about rigour, which is the cheapest kind of page there is.
So this collection is the receipts. It is also, frankly, the part most likely to be useful to someone building something of their own: the design system and the print rules are the accumulated results of getting it wrong first, and they are CC BY-SA like everything else.
There is a gradient here rather than a wall, and it is worth naming since this page is where a reader would wonder:
/changelog, because it is the one people link to.FACTCHECK.md — the full per-piece ledger, keyed by number, with every status and open item. Linked from the changelog for anyone who wants the long version.DECISIONS.md says on its own face that it is not published — it's for us. It tracks choices made and choices still open, in a register that would be misleading as a public statement.Nothing here is hidden; some of it is simply not a page. The line is about what has been written for a reader versus what is still shop talk, not about what we are willing to show.
One working paper has moved out. Glimmer Wire — the weekly graded scan behind Glimmers — was the sixth page here until 4 September 2026, when it became a collection of its own so that each edition could have its own page. It is still apparatus, and it is still off the reading chain; it simply grows every week, which is not a shape a single card can hold. This page kept the five that do not, and gained a sixth on 9 September 2026.
Nearly every page in the collection is linked into a single prev/next reading order — a hundred and forty-three of them, running from Start Here through to the last broadside. All six pages above sit outside it, and so does this one — and it is no longer the only collection page that does: Glimmer Wire is off the chain too, for the same reason.
That is deliberate and it follows from what these are. A changelog is something you consult when you want to know whether a claim moved; a design system is something you open when you are building. Threading them into a reading sequence would put a maintenance document between two zines and imply an order nobody wants. About and Cosmic Connections are off the chain for the same reason.
So they are reached the way reference material should be: from the front page, from the nav, from search, and from the pages that cite them. Being off the path is not the same as being buried.
Not a disclaimer page. Nothing here is hedging the rest of the site. The changelog exists because the claims are meant to be checkable, which is the opposite of a caveat.
Not a completeness claim. The changelog records the errors we found. It cannot record the ones still sitting in a page nobody has re-read — and the ledger keeps an open-items list precisely because that set is never empty.
Not process theatre. These pages exist because they get used: the print rules came out of forty-four pages that printed blank, and the design system out of a contrast audit that found real failures. A document that recorded no failures would not be worth the file.
Not the interesting part. If you have limited attention today, spend it on a zine. This collection will still be here, and it is written for the person who has already read something and wants to know how it was made or whether it has changed.