Reference notes. Source retained; historical claims may need rechecking.
Format: vision paper
First entry in the formats library (9 - System/ui-ux/formats/). A format
is a reusable document archetype - a fixed anatomy plus a rule for when to
reach for it - distinct from the house kit (documentation-kit.css, the
shared tokens and blocks every format is built from) and from a one-off page.
Reference build: template.html in this folder. Sample content ("A vision
for didactic government documents") is house material, not a real
publication - copy the file, replace the words, keep the shape.
What it is
A vision paper is one long, scrollable document that argues a single
thesis, then walks it through a fixed number of numbered sections, each
opening with a short summary before the detail. It reads start to finish,
like an essay, not like a dashboard or a reference manual. The reader is
expected to stay on the page for minutes, not seconds.
It generalises the long-form vision/manifesto page: a public body or team
sets out a position (why we believe X, what we will do about it) and wants
it read as an argument, with evidence, not skimmed as a spec.
When to reach for it
| Need |
Format |
| Argue one thesis end to end, meant to be read start to finish |
Vision paper (this format) |
| One idea, one screen, no scroll needed |
One-pager (documentation-design.md grammar, no dedicated template yet) |
| Reference material - every block type on one page, browsed not read |
Compendium (documentation-compendium.html) |
| Teaching page with a sidebar for jumping between many topics |
Documentation page (documentation-kit.css §3 ORIENTATION, .on-this-page rail) |
| A screen or interaction to prove |
Clickable mockup, not a document at all |
Rule of thumb: if the page's job is to be argued and read, use the vision
paper. If its job is to be looked up, use the compendium or a
documentation page with the on-this-page rail. The vision paper deliberately
refuses that rail - see below.
Anatomy
- Thin header - sticky, one line: wordmark left, two plain text nav
links right (
Contents, Authors). No logo art, no menu.
- Hero - kicker label, document title, one-line dek, a plain
Start reading → text link. Not a marketing headline, not a button.
- Lead / abstract strip - date, authors and format in a narrow left
column; the abstract paragraph beside it. One hairline rule above and
below, no card, no shadow.
- Jump to section - one
<details> control, centred, closed by
default. Opens a plain numbered list of the section titles. This is
deliberately not the kit's .on-this-page sidebar: a vision paper is
read in order, not browsed, so it gets one small control instead of a
permanent rail competing with the text for width.
- Numbered sections - five is the house default (fewer reads thin,
more asks too much of one sitting). Each section is a
kit's details.doc-section, opens with a TL;DR summary box
(.callout, no colour modifier - grey, never tinted), then subsections.
- Pull-quotes - sparing, one or two for the whole paper, never one per
section. A short attributed sentence, not a slogan.
- Inline citations - a normal accent-coloured link inside the sentence
that needs it. No footnote numbers, no endnote list.
- Authors / contributors block - plain list of roles near the end,
before the footer.
- Footer - plain, light, one hairline top border. No colour fill.
House adaptations (do not build this format any other way)
- Light theme only. No dark hero, no dark footer. The public long-form
vision-paper page this generalises uses a saturated colour band across
its footer; the house version drops that entirely - footer is plain
paper background with a hairline top border, same as every other house
page.
- Hero stays small. Title, one-line dek, one text link. No pill
buttons, no stacked CTAs, no marketing energy. The hero's job is to name
the document, not to sell it.
- Grayscale-first, one accent. The TL;DR box is the kit's default
.callout (grey, --ink-soft border) - never .info/.warning/
.success. Table headers are untinted (kit default). The single accent
(--accent) appears only on links, the current nav state and the pull
quote's rule - never as a section-colour system.
- Reading column, not full width. Unlike house working documents/boards
(which run full container width per
feedback_html_output_rules.md rule
16), a vision paper is a continuous read: the whole body is capped at
--measure-prose (68ch) inside a wider fixed header, matching the
≤65–72ch prose rule. This is the format's own justified exception to the
boards' full-width default, not an oversight.
- Italic Georgia as the signature voice, used narrowly. The hero title
and each section's number+title are
--font-body (Georgia) italic -
the format's one recognisable voice. Subsection headings (h3) are kept
in the plain sans (--font-ui), deliberately not the signature serif, so
only the top of each section gets the "full pop" (Kennedy: one thing per
page gets all the emphasis, everything else leans one direction only).
No external font is loaded - Georgia is a system font, one self-contained
file.
- No invented numbers. The stat row in the sample shows the pattern
directly: one real, cited figure (20%, Nielsen Norman Group, linked) next
to a second figure deliberately left as
- because no equivalent number
exists yet. That is the house rule from feedback_html_output_rules.md
#4 in action, not just stated.
A decision recorded, per rule 7's own exception clause
House rule 7 (feedback_html_output_rules.md) requires every section and
subsection to be a collapsible <details>. This format collapses only the
top-level numbered sections (open by default, so the page still reads as
one continuous scroll) and leaves subsections as plain flowing prose,
not individually collapsible. Folding every subsection of a five-section
argument would fragment the one property that defines this archetype - a
thesis walked through start to finish - into a picklist, which is the
compendium's job, not the vision paper's. This is the documented exception
the rule itself allows ("exceptions only where a format's own decision
record says so").
Build checklist for a new vision paper
- Copy
template.html to the new document's folder.
- Link
documentation-kit.css at the correct relative path (two levels up
from formats/vision-paper/ if left in place; adjust if moved).
- Replace: wordmark, title, dek, kicker, date, authors, abstract, all five
section titles/content, quote(s), citations, stat row, table, footer
line. Keep the block count roughly the same - that discipline is the
point of a format.
- Do not add a second accent colour, a sidebar rail, or a hero button.
- Run
python3 "9 - System/ui-ux/token-audit.py" <file> - zero
violations before shipping. Add any new bespoke measure (a ch cap, a
bespoke font-size) as a named token in a local :root block inside the
page's own <style>, never as a bare literal.
- Verify no horizontal scroll at 375 / 768 / 1280px.
- Open in the browser (
open <file>.html) before calling it done.
Provenance
Format anatomy generalised from a public agentic-government long-form
vision-paper layout (structure only - header/hero/lead-strip/TOC/numbered
sections/pull-quotes/inline citations/authors block/footer). No text,
brand, colour or asset from that source was copied; the palette, type,
spacing tokens and every block come from the house kit, and the sample
content is original, written for this workspace.