Icons
Dan's musing was whether some blocks can carry icons, which roles icons might play within a manifest or the wider site, and what the word iconic means besides the computer sense. The favicon was already an answer of a kind: five rectangles on a square. This study takes that seriously. A glyph is a footprint — the shore's word — on the pair itself, square, with the row turned into a column because a sign has no reading rhythm; it is written as a chart, merged into the fewest blocks, and presents a different cut per size the way a block does per edition. The study measures where a block stops being a field and becomes a sign, lets the composition of the day stand as its own icon, and re-draws the client site's three service glyphs in the block idiom to see what survives.
Study A · The atlas
Ten glyphs, each shown at three widths — 70, 28 and 12 pixels — so the browser chooses a different cut for each: the full cut is the 7×5 pair in both axes (union 35×35), the compact is 5×3 (15×15), the sign is 3×2 (6×6). Nothing is drawn as a pixel: each cut is a chart of base cells merged into the fewest rectangles, plus overlay marks in overlay units, and every block carries the engine's own --c --s --r --rs. lit lights every glyph's signal cell at once, which is the quickest way to see why the law says none at rest; grids draws both grids on each glyph; paper shows the square each glyph owns; charts prints the score.
Study B · The three senses
The commonplace book's gloss on Dan's seed found three senses of "iconic" pulling in different directions, and asked that a study hold all three and say which it is testing. So: three columns. The computer icon stands for a place or an action and is the smallest manifest. The window — the Byzantine argument that the image is not a picture of the thing but the thing made present — is tested literally: the glyph for the composition of the day is the composition of the day, the engine's generator run on today's seed inside a glyph, square, at the current edition; it will be different tomorrow and it carries its referent's own signal. The shopping bag — the canvas as logo — is the mark, which the system already had.
A sign for a place. It is read, not looked at; it has no signal of its own; it means the same at every size because the chart is rescored per cut rather than scaled. Wayfinding, Study D.
Not a picture of the composition of the day: the composition of the day, seed —, the same generator, square. It has a signal because the thing it shows has one — so there can be one window per viewport, as there is one signal.
The mark: the favicon's five rectangles on the pair, the house colour always on. The sense the plan's cute trap warns about — a Mondrian on a tote — and the one sense a brand cannot do without. Kept to one glyph, and to the one in the header.
Study C · The ladder, and the threshold of the sign
One glyph — Solve it, the piece that fits, whose overlay mark is exactly one microcolumn larger than its hole on every side — at twelve widths, from the favicon's sixteen pixels to a hundred and twelve. Below each rung the instrument reads back what the browser did: which cut it chose, the size of a base cell, the size of a microcolumn (which at the full cut is the sliver), and how many device pixels the near-miss actually got. The threshold of the full cut is where a microcolumn is a pixel: 35 px. Below it the near-miss is sub-pixel and the glyph must drop to a cut whose microcolumn is still a pixel; below 15 px even the compact cut's microcolumn is not, and the sign is all that is left.
| width | cut | pair | cell | microcolumn | near-miss, device px | blocks |
|---|
Study D · Wayfinding
The icons as the page's wayfinding: the site's four pages and the day's composition as blocks on an open stage, scored per edition like the header on the home page. A glyph takes the row as its measure — a row less eight pixels, so that the label fits under it in a two-row block without growing the row — and so it gets the full cut at every edition (40 px at S is above the threshold). The current page is the lit one; the others rest. At S the labels fold into the glyphs (they remain for screen readers) and the row of five becomes two rows: the block becomes a sign, which is the other thing the word means — not the glyph's cut but the block's.
Study E · The pillars
Build it · Lead it · Solve itThe client site's three service pages carry three line icons: a heart held in an open hand, with gears, with people, with a puzzle, in purple, orange and green. They are good icons of their kind, and their kind is the line — a stroke that draws a thing. The system's kind is the block, which cannot draw a heart and should not try. So each pillar is re-drawn as what the service does on this grid: Build it is courses of blocks with the next one hovering, about to be laid; Lead it is one block a plane up and ahead, with the others following on the base; Solve it is the piece that fits, a microcolumn larger than its hole. The hand and the heart, which all three shared, become what all three glyphs share: a resting colour, a lit state, and the same square.
→Two courses laid, a gap in the upper one, and the next block at plane +1 above it with its shadow on the course below. Lit, the hovering block is the signal: the one that is about to click in.
live: purple — the system's workshop tier, AI candidates; ships on no public page
→Two followers on the base plane, low and left; the leader ahead and up, at plane +1, top right, its shadow falling toward the others. The only glyph in the set whose overlay mark stands alone.
live: orange — the system's workshop chrome; the draw's second sentence says it never ships
→A field with a hole at its centre and the piece above it, one microcolumn larger on every side — the near-miss as the puzzle's click. At the sign cut there is no room for a near-miss and the piece is simply home.
live: green — not in the palette at all; the three colours were the icon pack's, not the system's
LXD Website Icons/ · set on the pair hereWhat a glyph is
The commonplace book's gloss ended with a sentence that turned out to be the study: an icon is a block below the smallest measure. The measure table from AXI-007 gives the smallest measure a number — eight characters on a line at the smallest size, with the block's minimum padding — and the instrument in Study C reads it from the live font. On the page's own grid no block ever gets there. The narrowest block in the system's vocabulary is one base column, and a base column at the phone's edition is still wide enough for a word. So the sign does not live on the page's grid at all. It lives inside blocks, as the mark lives inside the header, and inside manifests, whose columns are a fraction of their width — a 7×5 manifest's cells are signs once the manifest is narrower than seven thresholds, and the manifest study's pocket edition was already working at that scale without knowing it.
That is why a glyph is the smallest manifest and not the smallest block. It is a block that holds the pair, as a manifest is, at the one scale where the pair's rows can be columns — because a sign is looked at, not read, and has no rhythm to keep. Everything else follows from the manifest: it chooses its cut by its own width; it is written as a chart, which the shore invented for footprints and which turns out to be the right notation for a glyph because a glyph is a footprint; its blocks are the engine's blocks with the engine's numbers on them. The one thing a glyph gives up is the signal. One per viewport was the law from the first session, and a row of five icons each carrying one would be five moments; so a glyph rests in navy and lights only when it is the current thing, which is also what an icon in a navigation is for. The window is the exception that proves the cost: the composition of the day drawn small is still the composition of the day, signal and all, and so there can be one of it where it stands.
The draw
seed 1350520069 · from the date · pool 191Dan's note for the night named a list and an order — the studies the client build will need, icons first — so the study was chosen before the draw. Five sentences, seeded by the date:
- 1 · 172
P3 Copy sprint — rewriting the site's voice into block-sized punches; real pages assembled 1–2 sessions
considered: the plan's third phase, drawn again (it came up in AXI-017 too); the claim in Study D is a block-sized punch, and the real pages are the client build's - 2 · 179
Everything orange is workshop chrome — it belongs to the tool, and never ships to the public site.
answered in Study E: the live site's Lead it icon is orange and its Build it icon is purple — both the system's workshop tier — and Solve it is a green the tokens do not have; the pillars re-drawn rest in the structure tier and light in the signal; the colours were the icon pack's, and the finding goes to the LXD fork - 3 · 189
LXD-mondrian · Interference Lab · proof of concept
the lab's title line; considered and kept as a reminder that the atlas is a proof of concept of the same kind - 4 · 95
Blocks aren't only shown or hidden — they can present a different cut per edition: full, compact (condensed variant), sliver (a one-line teaser), or hidden.
answered as law 3: a glyph presents a different cut per size — full, compact, sign — chosen by its own width; and in Study D the block itself takes a cut at S, folding its label and becoming a sign - 5 · 147
Adopt what it's loved for (filter + FLIP reflow) via CSS grid + View Transitions.
considered: the plan on Isotope; the atlas filters nothing and reflows by the grid alone — a glyph set is a thing to read, not to sort
draw.jsThe rules, published
argue with the numbers, not the pictures| Constant | Value | Why |
|---|---|---|
| lattice | square · base B×B over overlay O×O · union U×U | law 1: the row is a column; a sign has no reading rhythm, so the pair runs in both axes — the one place --snx-unit is the same both ways |
| cuts | full 7×5 / 35 at ≥ 35 px · compact 5×3 / 15 at ≥ 15 px · sign 3×2 / 6 below | law 3: the threshold of a cut is the width at which its microcolumn is a pixel; chosen by container query on the glyph's own width |
| chart | one letter per base cell · '|' between rows · o:c,r,s,rs,colour[,+] · s:x|y|xy | law 2: a glyph is a line of the score, as a shore's footprint is; the letters are the palette's tokens |
| merge | runs of a row carried down while their columns agree | law 4: the shore's cut; a glyph is the fewest rectangles, and its weight is their count |
| signal | '*' cells navy at rest · signal when current, hovered or lit · mark and window exempt | law 5: one signal per viewport; a navigation of lit glyphs would be five |
| shadow | one microcolumn right · ⅜ cell down · rgba(19,22,40,.22) | the engine's plane +1 shadow with the row read as the cell; at 35 px it is one pixel across, which is pixel art's convention arrived at from the other side |
| slivers | at microcolumns 14 and 20 (full), 5 and 9 (compact), 2 and 3 (sign) · both axes on request | the engine's near-miss columns and, by law 1, the same rows; at the sign cut the two are adjacent and read as one bar |
| window | SNX.generate on the day's seed · rows = base · --snx-row = the cell | law 6: the engine's own stage inside a glyph; nothing drawn by hand |
| sign threshold | 8 × average character width at s0 + 2 × 14 px padding | law 7: AXI-007's dead cell, read from the live font; compared to the page's column and a manifest's |
| wayfinding | a glyph is a row less 8 px · label folds at S · one aria-current | the row is the unit a glyph takes from the page; the block, not the glyph, takes the sign cut at S; the rows are checked not to grow |
| feed/score | script: cells, data-cut, data-blocks, data-cells, data-over, data-u, data-w, data-micro, data-cell · stylesheet: everything visible | law 8 |
Findings
- A glyph is the smallest manifest, not the smallest block. On the page's grid no block is ever a sign: the narrowest block is a base column, and a base column holds a word of eight letters at every edition (the instrument reads the numbers from the live font). The sign lives inside blocks and inside manifests, whose columns are fractions of their width; a 7×5 manifest's cells become signs once the manifest is narrower than seven thresholds.
- The row is a column. A sign has no reading rhythm, so the pair runs in both axes and a glyph is a square union of 35×35, 15×15 or 6×6 — the one place in the system where the half-row and the half-column of the upstream question are the same unit, because the row and the column are.
- Cuts are editions at the scale of a sign. The plan's "different cut per edition" became a glyph's three charts, chosen by the glyph's own width, and the thresholds are not taste but arithmetic: a cut is usable while its microcolumn is at least a pixel, so the full cut starts at 35 px and the compact at 15. The ladder reads the browser's actual sub-pixel behaviour at each rung.
- No signal at rest. One per viewport cannot survive a row of icons each carrying one; the lit switch on the atlas shows the page becoming a fairground. So a glyph's star cell is navy until the glyph is current, hovered or told — and the current page in a navigation is exactly one, which is what the law wanted all along.
- The window works, and costs a signal. The composition of the day drawn at 112, 35 and 14 pixels is the composition of the day — the same generator, the same seed, the engine's own stage with the row read as the cell. It carries its referent's signal because its referent does; so there can be one window per viewport. The Byzantine sense held up under the plan's law better than expected.
- The chart is the right notation for a glyph because a glyph is a footprint. Ten glyphs at three cuts are thirty lines of plain text in the extension, and a page may add one in a data attribute. The merge gives each a weight: the atlas reports from two blocks (the grid, which is only slivers) to ten.
- The near-miss survives to 35 px and not below. The sliver glyph and Solve it both depend on a one-microcolumn gap; the ladder shows the browser giving it a device pixel at 35 px and a fraction below, where the compact cut takes over with a microcolumn of its own. Depth survives further: a plane +1 shadow one microcolumn wide at 35 px is a one-pixel shadow, which pixel art discovered from the other direction.
- The three senses can be kept apart on one page. The computer icon is what the atlas, the ladder and the wayfinding test; the window is built once and found to hold; the shopping bag is admitted as the mark and rationed to one. The cute trap is a question of scale, as the clock found, and a glyph that is a composition of blocks you can count is on the right side of it even at twelve pixels.
- Feed/score holds. The script writes cells with grid areas in microcells and nine data attributes; the cuts, the lit state, the shadow, the slivers and the grids on request are all stylesheet. The one thing the script does that the stylesheet cannot is read back which cut the browser chose.
For the LXD fork
- The three service icons are line icons in colours the system does not ship (purple and orange are the workshop tier; green is not a token). The fork should either re-draw the pillars in the block idiom — Study E is a first draft of all three, with the hand and heart becoming a shared rest colour and lit state — or admit three service colours into the structure tier deliberately, as a token-set decision, before any page uses them.
- Icons belong inside blocks and manifests, never on the page's grid as blocks of their own; a glyph is one row tall in a navigation and gets the full cut at every edition. Below 35 px use the compact chart, below 15 the sign; the favicon is a sign cut.
- No icon carries the signal colour at rest. The current page's icon is the one lit; hover lights the others. The mark is the exception and there is one of it.
- If the site wants a live icon (a "today" or a "latest"), the window pattern — the real stage, generated small — is honest and costs the viewport's signal; use at most one.
- Write icons as charts in the score, not as SVG files: ten glyphs are thirty lines, the colours are the tokens, and the critics of AXI-008 can judge them like any composition.