The lattice
Kill the head and keep the mind: the first screen is a ruled lattice of every speaker and every session of the programme, set in one face on one flat teal, and the mind is attention: asked who speaks on Tuesday, it gathers Tuesday's people into one bright diagonal that falls into its own answer, while everything it does not mean waits at a quarter of the ink.
The story
The energy industry's word for its own body is "the grid". This site takes it literally and typographically at once: the programme is laid out as a ledger of cells, like a share listing or a switchboard, and the mind is not a figure on a backdrop but a behaviour in that material. On arrival the visitor sees a masthead: "Free Your Mind" set extra-wide across the full width, framed by four facts at its four corners (ETS27 and Tickets above, Austin and May 3–6 2027 on the dateline below), with the fork printed on the dateline between them. Beneath it runs the lattice of the whole programme, 194 words on cells, running off the bottom edge. One answer is already composed. The words print row by row; then Tuesday's people slide along their rows and down their columns into a slanted column that steps from under the date, top right, to the heading of "01 Who's speaking Tuesday" in the lower left: the day, then each session's time with its people, in time order, morning at the top and evening at the bottom. The answer itself is a merged cell framed in the grid: a 34 px heading, the mind's say-line, four of the day's sessions, a plain-words source, and "02" waiting on a rule with three questions sitting on it. When nobody is pointing, the lens rests on the heading and every few seconds walks down the diagonal, re-reading its answer; nothing it does not mean gains ink. Scroll and the same programme is read in order: the theme as a compressed black statement, the lineup ledger, the four days, the price ledger, the place, and a state line in the footer. A rail in the label column names the five sections and keeps the current one lit, the same selection logic carried down the page.
Typography
One face, Science Gothic (Google Fonts; axes wdth 50–200, wght 100–900, CTRS 0–85, slnt −10–0), and the width axis says the role. There is no second face: the mind's voice is a change of axis, the way selection is a change of ink.
- The lockup at the display size (fitted edge to edge, 90 px at 1440 wide), wdth 200, resting in focus at wght 220 / CTRS 85, with thick stems and hairline bars. It lets go to the hairline (wght 100 / CTRS 0) over the first 640 px of scroll, as the visitor leaves it.
- Labels (corners, dateline, fork, rail, section names, buttons): 11 px caps, wdth 125, wght 500, tracking 0.08 em, wght 850 / CTRS 38 for the emphasised word.
- Readouts (times, dates, prices' context, notes, the source line, the footer state line): 11 px, wdth 70, tabular figures. This replaces the monospace; the condensed width is the readout voice.
- Lattice words: 11 px, wdth 100, dim at wght 330. Lit words rest at wdth 104 / wght 620 / CTRS 16 and sharpen to wdth 115 / wght 720 / CTRS 45 under the lens. Times use the same five levels on the condensed width (wdth 70 to 84). Contrast stays capped at 45 below 17 px.
- The answer's heading: 34 px, wdth 140, wght 450, CTRS 85, sentence case. This is the page's middle voice, the second thing the eye reads.
- The mind's say-line: 17 px on a 22 px leading, slnt −10, wght 280, CTRS 40, set at a fixed 560 px measure. It is the lattice's own face, leaning.
- Prose (theme, tickets, place): 17 px at 1.5 leading, wdth 96, wght 320, CTRS 20; emphasis is the slant.
- Lists below the fold (lineup and agenda): 13 px, so an executive reading at 7 am does not lean in.
- The theme statement at the display size, wdth 50 / wght 900. "Austin" at wdth 200 in the lockup's resting axes, fitted to the full body width (166 px at 1440), coming into focus as it enters the viewport.
Sizes: 11, 13, 17, 34 and the display size. Apostrophes and quotes are typographic throughout. Science Gothic has no arrow glyph, so the arrows on Tickets and Buy are drawn hairline SVG in the text colour.
Colour world
- Ground: Barton Springs teal
#1F7A6Eat 100 %, every screen. - Ink: cream
#F2E9D8. Words the mind does not mean sit at 28 %, so selection is a change of ink, never a second colour; below the fold, unselected rows rest at 66–74 % so they stay readable. - Cell rules: cream at 13 %; the answer's frame, section rules and the composer rule at 50 %.
- Portraits keep their own colour at 40 px; a speaker without a portrait gets an empty place, never a frame or a monogram.
- No second hue anywhere.
Layout grammar
At 1440 × 900 the masthead takes the top 173 px; the lattice runs from under the dateline to beyond the bottom edge, 12 columns of 116 px and 43 rows of 17 px. Each word takes one to three cells, measured at its lit resting width with room for the squeeze when the lens is far away. Cells carry the ground, so the rules never run through a word. The answer is a merged cell of 7 columns by 19 rows, framed at 50 %, with one empty column to its right and one empty row above it, and nine rows of lattice beneath it so the composer sits at about y 770 and a consent strip covers only texture. The gathered answer is a slanted column five cells wide that falls from the top-right to the answer's heading. Everyone else is spread evenly over the remaining cells in programme order. The answer's rows never bulge; the rest of the grid swells around the lens along rows and columns only.
Below, each section has a label column (columns 1–2) and a body (columns 3–12). The label column is a sticky rail of the five sections with the current one lit and its note beneath. The statement wraps at the display size. The lineup is a two-column ledger of 77 rows: stamp, name, title wrapping to a second line (never truncated), and a readout of days. The four days are four columns at 13 px, with Tuesday sharpened and the other days at 66 %. Tickets are four ledger rows, then the note beside a readout ledger of what is not published yet. The place is "Austin" across the body width, the prose beside a readout ledger (coordinates, time zone, venue, hotel block, the teal, contact). The footer is the state line.
On the phone (390) the lockup turns and runs up the left edge in its resting axes. The four facts sit in two lines at the top. The lattice keeps three real columns: each Tuesday session starts its own row (time, then people), and dim cells close every row so no column stands empty. The answer spans the columns with its four sessions, and "02" is a 44 px input. The three questions stack one per line, and the fork takes two lines, breaking after "answers".
Motion and life
- Entry: the words print row by row, 2.4 ms apart, done in 1.3 s, with Tuesday's people printed where the programme order puts them. At 1.5 s they slide along their rows, then down their columns, into the diagonal, earliest session first, 10 ms apart; settled by about 3.5 s. Words move only along rows and columns, and they pass under the answer, never over it.
- Attention: a field with exactly two attractors, the pointer and the answer's heading. Columns within about 130 px of the lens widen by up to 60 % and rows within 110 px by 35 %, normalised so the ledger keeps its size. Words within 200 px sharpen across five levels.
- Honest attention: with no pointer moving, only lit words gain ink, so bright always means the answer. Under a moving pointer every word sharpens, a reading glass.
- Idle reading: 5.2 s after load, and whenever the pointer has been still for four seconds, the lens hops every 3.6 s to the next lit word in reading order, walking down the diagonal and back to the heading.
- Scroll: the lockup lets go of its focus over the first 640 px; "Austin" comes into focus as it enters. Both are CSS scroll-driven animations.
- The caret blinks on the composer rule. Loops pause when the first screen leaves the viewport or the tab is hidden. Reduced motion starts settled, with no gather and no hops.
The fork and the mind
The fork is one sentence on the dateline, two verbs that weigh differently: "TYPE and the lattice answers" in wght 850 against "SCROLL and read it in order" in wght 200. There are no cards and no nav bar. The standard route is the page read downward, with the rail to jump. The interactive route starts with any keystroke anywhere: "02" takes focus and every character re-selects the lattice live, in place, by substring or by day word ("wed" lights Wednesday where it lies while the Tuesday diagonal dims), so the visitor watches the attention move. Enter composes a numbered section in the page's own grid, "Answer 02", with the question as a 34 px heading, a slanted say-line composed from the snapshot (days, prices, place, theme or a text match), the components beneath and a plain-words source; the three questions on the rule do the same with one click. The omnipresent assistant on the standard route is the same selection carried down the page: the lineup ledger and the agenda sharpen what is lit, and clearing the question returns Tuesday everywhere. The state is exposed once, in the footer: "Snapshot 13 Oct 2026 · 77 speakers · 50 sessions · sample data".
What to build next
- Replace the rule-based composer with the real mind: a model that returns a selection (speaker and session ids) plus a 45-word say-line, rendered through the same lit and dim mechanism, the same gather and the same numbered sections.
- Gather on every answer, not only the first: when a typed question settles, slide its lit set into the diagonal the way Tuesday's is, and let the previous answer's cells slide back to programme order.
- Short viewports: below about 800 px of height the diagonal has too few rows and becomes a slanted block; give the answer a compact form there (heading, say-line, composer) so the gather keeps its slope.
- Portraits on hover in the lattice (a stamp beside the lit name), and a tap on a name opening its ledger row.
- One live readout in the rail (ERCOT frequency or wind), so the lattice is never the same twice.
- Partners, press and the FAQ as further ledger sections; a moderated "asked on this page" line rotating in the composer placeholder.
- A real accessibility pass (the 194 cells need an aria-friendly summary), a Windows hinting check of Science Gothic's lowercase at 11 px and of the slanted say-line, and ETS27 data in place of the ETS26 stand-ins as it is announced.
Risks
- It can drift back into a word cloud or a dashboard. The cure is the grid discipline: every word on a cell, movement only along rows and columns, no lines between anything, two attractors and no more.
- The diagonal is the poster. It needs rows: at 1440 × 900 it is 5 cells wide over 10 rows; on short screens it fattens into a block, and if a selection grows past about 70 cell-widths it cannot hold a slope at all, so large selections stay lit in place.
- Contrast: cream on this teal is 4.3:1 at full ink, and the words at 28 % are texture at about 1.6:1, by design. The lit set and the answer carry everything a visitor needs; the pointer is a reading glass for the rest.
- One face carries every voice. It depends on Science Gothic's width, slant and contrast axes rendering consistently; check older Safari and Windows before building.
- Portraits are hotlinked from the Whova CDN and the ETS site for the mockup only. Long session titles are shortened at the colon in the lattice; full titles live in the agenda and in the answer's components.
- The 2026-10-14 version of this concept is kept unchanged in
v1/.