From caf382d7bbec9dfb317a3b35c6a0744975edab63 Mon Sep 17 00:00:00 2001 From: Talon Date: Sat, 29 Aug 2026 20:14:58 -0400 Subject: [PATCH] Build carousel art studio foundation Actor: human --- .claude/skills/carousel-art-studio/SKILL.md | 169 ++++++++++++++++ .../carousel-art-studio/references/atelier.md | 59 ++++++ .../references/composition-and-camera.md | 86 +++++++++ .../references/computational-systems.md | 68 +++++++ .../references/light-material-depth.md | 87 +++++++++ .../references/linkedin-native.md | 72 +++++++ .../references/review-gates.md | 67 +++++++ .../references/styleframe-brief.md | 108 +++++++++++ .../references/visual-thesis.md | 78 ++++++++ examples/art-atelier/README.md | 29 +++ examples/art-atelier/slides/slide-01.html | 142 ++++++++++++++ examples/art-atelier/slides/slide-02.html | 122 ++++++++++++ examples/art-atelier/slides/slide-03.html | 129 +++++++++++++ examples/art-atelier/slides/slide-04.html | 145 ++++++++++++++ examples/art-atelier/slides/slide-05.html | 131 +++++++++++++ examples/art-atelier/slides/slide-06.html | 143 ++++++++++++++ knowledge/carousel/REFERENCE_ATLAS.md | 120 ++++++++++++ knowledge/carousel/VISUAL_CEILING_AUDIT.md | 182 ++++++++++++++++++ 18 files changed, 1937 insertions(+) create mode 100644 .claude/skills/carousel-art-studio/SKILL.md create mode 100644 .claude/skills/carousel-art-studio/references/atelier.md create mode 100644 .claude/skills/carousel-art-studio/references/composition-and-camera.md create mode 100644 .claude/skills/carousel-art-studio/references/computational-systems.md create mode 100644 .claude/skills/carousel-art-studio/references/light-material-depth.md create mode 100644 .claude/skills/carousel-art-studio/references/linkedin-native.md create mode 100644 .claude/skills/carousel-art-studio/references/review-gates.md create mode 100644 .claude/skills/carousel-art-studio/references/styleframe-brief.md create mode 100644 .claude/skills/carousel-art-studio/references/visual-thesis.md create mode 100644 examples/art-atelier/README.md create mode 100644 examples/art-atelier/slides/slide-01.html create mode 100644 examples/art-atelier/slides/slide-02.html create mode 100644 examples/art-atelier/slides/slide-03.html create mode 100644 examples/art-atelier/slides/slide-04.html create mode 100644 examples/art-atelier/slides/slide-05.html create mode 100644 examples/art-atelier/slides/slide-06.html create mode 100644 knowledge/carousel/REFERENCE_ATLAS.md create mode 100644 knowledge/carousel/VISUAL_CEILING_AUDIT.md diff --git a/.claude/skills/carousel-art-studio/SKILL.md b/.claude/skills/carousel-art-studio/SKILL.md new file mode 100644 index 000000000..f15b255b0 --- /dev/null +++ b/.claude/skills/carousel-art-studio/SKILL.md @@ -0,0 +1,169 @@ +--- +name: carousel-art-studio +description: Art-direct and construct show-stopping, evidence-grounded computational artwork for Texas AI Docket LinkedIn carousels. Use when researching visual references, inventing a carousel treatment, making or comparing styleframes, planning shots, coding bespoke slide art, or reviewing whether a deck reaches the publication's 10x visual ceiling. Produces visual theses, rendered alternatives and production decisions; use carousel-engine separately for rendering and machine QA. +--- + +# Carousel art studio + +Make an image worth stopping for before making a deck worth approving. This skill is the +creative studio upstream of `carousel-engine`: it decides what the art says, finds a singular +visual world, proves that world in pixels and develops the shots. The engine renders and checks +the result; it does not provide the art direction. + +The standard is `knowledge/carousel/VISUAL_CEILING_AUDIT.md`. The source shelf is +`knowledge/carousel/REFERENCE_ATLAS.md`. The publication rules in +`knowledge/carousel/DESIGN_DOCTRINE.md` and the accuracy/technical contracts in +`carousel-engine` still apply. + +## Read by task, not by habit + +Load only the references needed for the present decision: + +- concept or treatment: `references/visual-thesis.md` and + `references/composition-and-camera.md`; +- styleframe competition: `references/styleframe-brief.md` and + `references/linkedin-native.md`; +- dimensional scene: `references/light-material-depth.md`; +- generative, data-bound or simulated form: `references/computational-systems.md`; +- full-deck sequencing or pixel review: `references/linkedin-native.md` and + `references/review-gates.md`; +- coding an advanced construction: `references/atelier.md`, then inspect only the relevant + source under `examples/art-atelier/`. + +Do not read an atelier frame as a template. Read it to understand how camera, value, light, +material, typography and code resolve into one shot. + +## Non-negotiable output + +Do not select a treatment from prose. A selected direction requires: + +1. a one-sentence visual thesis tied to the verified claim; +2. a cited reference cut with study/transform/do-not-copy notes; +3. at least three materially divergent candidate directions; +4. rendered pixels for each candidate — cover plus one interior turn when selecting a deck + language, not a cover alone; +5. full-size and 432-pixel comparisons; +6. a written selection decision and discarded-direction reasons; +7. a shot card for every production slide. + +If the task does not permit rendering, label the result **unproven treatment**. Never imply +that prose has passed the visual gate. + +## Studio sequence + +### 1. Lock the visual truth + +Extract one claim the image can honestly carry. Separate: + +- verified fact, quantity, place and time; +- measured or modeled data; +- bounded metaphor; +- unknown or unavailable information. + +Geometry is a sentence. Size, position, direction, connection, material and absence all make +claims even without labels. Do not invent a magnitude, map location, causal link, crowd, +machine detail or record fragment for drama. Read `references/visual-thesis.md`. + +### 2. Cut references with intent + +Use three to six references from the atlas and primary story-domain sources. Research a craft +property, not a vibe. For each reference write what to study, what evidence it serves, how it +will be transformed and what must not be copied. At least one source must come from the actual +material world of the story and at least two must break from the current AI-design defaults. + +Do not collect a wall of similar images. That narrows invention while pretending to expand it. + +### 3. Write the visual thesis + +The thesis names a visible event, not an adjective stack: + +> The record becomes **[specific world/object/system]**, seen from **[decisive vantage]**, where +> **[verified tension/change]** is expressed by **[spatial/material behavior]** and illuminated +> by **[motivated light]**. + +Then name the cover's memory sentence: the concrete description a reader could repeat tomorrow. +If it is “text over a nice background,” restart. + +### 4. Diverge in construction, not decoration + +Make at least three directions that differ in all of these: + +- image architecture: object, environment, section, field, machine, evidence excavation, etc.; +- vantage and lens: plan, macro, low wide, orthographic, cutaway, impossible evidence camera; +- material and light behavior; +- type's spatial role; +- motion or generative rule; +- feed-size silhouette. + +A palette swap, font swap or alternate arrangement of the same plate is one direction. +Explicitly challenge the cream/serif archival plate, near-black/bright-accent object plate and +hairline broadsheet layout. They may win only for a claim-specific reason. + +### 5. Prove the hard parts in pixels + +For each candidate, build a graybox before finish: + +1. three-value composition and crop; +2. subject silhouette and camera; +3. depth planes and type reserve; +4. key/fill/practical light and contact; +5. material response; +6. evidence layer and annotations; +7. finish: atmosphere, grade, grain and optical detail. + +Render at 1080×1350 and inspect at 432 pixels. If the graybox is weak, do not texture it. +Use `references/styleframe-brief.md` for the artifact and `carousel-engine` for rendering. + +### 6. Choose against the ceiling + +Score candidates with `references/review-gates.md`, beginning with feed stop, story specificity +and memorability. Accuracy and legibility are vetoes. Select the direction with the strongest +image system, not the one that was easiest to code or most eloquently described. Preserve the +losers and the decision so tomorrow does not rediscover them. + +### 7. Direct the deck as a sequence + +Give every slide a different visual job inside one world. Alternate pressure and release: +scale, density, camera distance, value, tempo and evidence mode. A quiet frame can be +exceptional; it still needs a decisive image, proportion or absence. Read +`references/linkedin-native.md`. + +Write a shot card per slide before production. If three cards differ only in copy, the deck is +still a template. + +### 8. Build by passes + +Keep construction passes explicit so criticism can target the real problem: + +- `G0 truth` — encoded geometry is supported; +- `G1 read` — graybox and silhouette work at feed size; +- `G2 space` — camera, overlap, occlusion and scale read; +- `G3 light` — value hierarchy and motivated illumination reveal form; +- `G4 material` — surfaces respond differently and specifically; +- `G5 evidence` — labels, data and source matter integrate honestly; +- `G6 type` — display and body type participate without becoming furniture; +- `G7 finish` — atmosphere, grade and microdetail unify the resolved shot. + +Do not jump from G0 to G7. Finish amplifies what is already there, including weakness. + +### 9. Review the image and the system + +Use the full render, the 432-pixel thumbnail and the contact sheet. Review the frame with type +briefly hidden to expose compositional dependence. Review the deck in grayscale to expose value +collapse. Then run the machine QA. A green harness means the artwork is safe to judge, not that +it is great. + +## Studio prohibitions + +- No invented facts disguised as visual specificity. +- No adjective-only art direction: premium, cinematic, tactile and bold are outcomes. +- No named technique standing in for a shot. +- No full deck before divergent styleframes. +- No generic background whose only job is filling space behind type. +- No panel score used as permission for repeatable sameness. +- No copyrighted style or living artist requested “in the style of.” Study transferable craft + properties from original sources and transform them. +- No external runtime dependency or network asset in slide HTML. + +The goal is not maximal visual noise. It is maximal authorship: the most inevitable, surprising +and beautifully resolved image this particular public record can truthfully support. diff --git a/.claude/skills/carousel-art-studio/references/atelier.md b/.claude/skills/carousel-art-studio/references/atelier.md new file mode 100644 index 000000000..b1cdce958 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/atelier.md @@ -0,0 +1,59 @@ +# Code atelier — reading finished constructions + +`examples/art-atelier/` is a set of executable visual studies made with the same offline assets +as production. Each frame isolates a larger construction — shot, camera, light, material, +evidence and type — rather than advertising a primitive. + +## How to use it + +1. Choose the construction whose **problem** resembles the story's, not the one whose look you + like most. +2. Render it through `carousel-engine`; do not judge source code names or a browser at arbitrary + zoom. +3. Inspect the full render and 432-pixel thumbnail. +4. Read how macro composition is separated from micro detail. +5. Extract a mechanism: depth ordering, highlight construction, type reserve, data-to-form + mapping, evidence integration. +6. Rebuild the mechanism with the day's own nouns, camera, palette, material and behavior. + +Never duplicate an atelier frame and replace its text. That would turn a craft reference into +the template this skill exists to prevent. + +## Construction index + +| Frame | Problem it demonstrates | Inspect for | +|---|---|---| +| 01 — evidence monolith | Make a record feel physical and consequential | Extreme crop, lit edge, embedded evidence, type as counterforce | +| 02 — volumetric cutaway | Reveal concealed layers/process without a dashboard | Impossible evidence camera, depth separation, motivated internal light | +| 03 — topographic signal | Turn a verified gradient or threshold into a material field | Macro density transition, contour survival, one literal anchor | +| 04 — data sculpture | Give a small verified series scale and spatial presence | Bound geometry, camera, lit ground, quantitative legend discipline | +| 05 — document archaeology | Use real record matter without making a paper collage | Layered depth, raking light, selective focus, exact fragment provenance | +| 06 — typographic machine | Let language participate in physical behavior | Type as structure, moving parts, legibility protected by construction | + +These frames are technique-agnostic at the concept level. Production may implement the same +construction with Canvas, SVG, relief, SDF or Three.js depending on the shot budget. + +## What to copy and what not to copy + +Copy: + +- explicit value hierarchy; +- camera and depth planning; +- motivated light; +- material differentiation; +- a macro read plus micro reward; +- deterministic mechanics; +- protected vector type; +- comments explaining visual decisions. + +Do not copy: + +- palette; +- type lockup; +- object silhouette; +- exact shader/noise parameters; +- frame layout; +- story metaphor. + +The atelier is successful when tomorrow's work looks nothing like it but is built with the same +degree of intention. diff --git a/.claude/skills/carousel-art-studio/references/composition-and-camera.md b/.claude/skills/carousel-art-studio/references/composition-and-camera.md new file mode 100644 index 000000000..4b156b9d0 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/composition-and-camera.md @@ -0,0 +1,86 @@ +# Composition and camera + +## Compose the thumbnail first + +At 432 pixels wide, detail collapses before structure. Begin with three flat values and ask: + +- Is the main silhouette recognizable with type hidden? +- Is one region dominant, one subordinate and one quiet? +- Does the eye enter at a deliberate point? +- Is the crop active, or is the subject politely centered? +- Does negative space create pressure, scale or anticipation? + +Make six postage-stamp grayboxes before polishing one. A successful graybox should still carry +the visual thesis in grayscale. + +## Name the camera + +Every dimensional frame needs a camera note: + +```text +Vantage: +Lens / projection: +Camera height and tilt: +Target: +Horizon: +Near / middle / far planes: +Crop and intended offscreen continuation: +Why this camera tells the claim: +``` + +Useful choices have meaning: + +- low wide: monumentality, exposure, instability, distance; +- long lens: compression, congestion, stacked relation; +- macro: material evidence, hidden mechanism, fragility; +- orthographic: authority, comparison, system legibility; +- plan: territory, allocation, pattern; +- cutaway: concealed structure or process; +- tilted/oblique: tension, transition, competing planes. + +Do not pick a dramatic lens because it looks dramatic. Pick it because perspective changes the +reader's understanding. + +## Build depth before texture + +Use at least three legible spatial layers where depth matters: + +1. foreground occluder or scale cue; +2. subject plane; +3. ground/environment; +4. atmosphere or distant field when useful. + +Separate layers by overlap, scale, focus, value, haze and motion direction. A stack of same- +value translucent rectangles is layering, not depth. + +## Type has a spatial role + +Choose one, deliberately: + +- **monument:** type is a massive form the image negotiates; +- **aperture:** type or its reserve reveals/conceals the world; +- **instrument:** type belongs to a measurement or machine layer; +- **caption:** quiet type observes an image that already works; +- **counterforce:** type pushes against the subject mass; +- **evidence:** exact record language is materially embedded. + +Avoid placing a self-contained text box into leftover space. Type and art can remain separate +for legibility while still sharing proportion, axis, scale or rhythm. + +## Deck rhythm + +Map the sequence before coding: + +| Beat | Typical job | Spatial move | +|---|---|---| +| Stop | Singular hook | Extreme silhouette/crop, shortest copy | +| Orient | Establish world | Wider camera, literal anchor | +| Reveal | Expose mechanism | Cutaway, macro or layer separation | +| Measure | Make evidence felt | Data-bound form, comparison or scale shift | +| Turn | Change understanding | Reversal of value, vantage or behavior | +| Consequence | Human/institutional meaning | Specific object, place or absence | +| Resolve | Useful takeaway | Clearer space, decisive final image | + +Not every deck needs seven beats and a nine-slide deck may split them. It does need contrast. +Alternate full/empty, near/far, dense/quiet, orthographic/perspectival and dark/light without +abandoning the same world. diff --git a/.claude/skills/carousel-art-studio/references/computational-systems.md b/.claude/skills/carousel-art-studio/references/computational-systems.md new file mode 100644 index 000000000..52afa5652 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/computational-systems.md @@ -0,0 +1,68 @@ +# Computational systems — behavior with meaning + +## Pick a behavior, not an effect + +Describe the system as a sentence: + +> **[entities]** move/grow/partition/accumulate according to **[rule]**; **[verified input]** +> controls **[visible property]**; the viewer should perceive **[claim]**. + +Examples of meaningful mappings: + +- delay controls queue length, not arbitrary hue; +- uncertainty controls blur or boundary instability, not fake precision; +- load controls density or deformation when the scale is documented; +- jurisdiction controls real polygon membership, not invented cells; +- time controls position or erosion along one declared direction. + +If there is no verified input, label the system illustrative and make no legend that suggests +measurement. + +## System families + +| Behavior | Good for | Common lie/failure | +|---|---|---| +| Advection / flow | movement, transmission, drift | random streamlines imply routes or volume | +| Accumulation | queue, backlog, compounding | area looks quantitative without a scale | +| Field / contour | threshold, gradient, influence | decorative noise masquerades as measured terrain | +| Packing / collision | capacity, congestion, constraint | circle size implies magnitude accidentally | +| Growth / branching | propagation, dependency, buildout | invented topology reads as a real network | +| Partition | allocation, territory, ownership | Voronoi cells replace real jurisdictions | +| Erosion / fracture | loss, stress, exposure | damage metaphor overstates an administrative change | +| Oscillation / wave | cycles, instability, signal | amplitude or frequency appears data-bound when it is not | + +## Authored generative work + +Randomness is not art direction. Define: + +- invariant structure — what every seed must preserve; +- bounded variation — which properties may move and within what range; +- composition constraints — quiet zone, dominant mass, collision rules; +- selection policy — whether a seed is fixed before or after styleframe review; +- failure detection — empty field, even density, hair-like lines, moiré, bad crop. + +Use deterministic seeds. Never sample seeds until one “looks cool” without recording the +selection, because that hides art direction inside roulette. + +## Data, model and metaphor labels + +Classify every visual layer: + +- **DATA** — directly bound to a cited value; +- **MODEL** — produced by a documented transformation or simulation; +- **MAP** — real geographic geometry/projection; +- **ILLUSTRATION** — explanatory but not measurable; +- **ATMOSPHERE** — non-semantic finish. + +The dossier should say which is which. Data may drive illustration, but the reader must not be +invited to recover a number the mapping cannot support. + +## Thumbnail stability + +Generative detail dies first at 432 pixels. Design two scales: + +- macro structure: one silhouette, density transition or directional field; +- micro reward: marks, particles, hachure, interference or grain. + +If removing the micro layer erases the claim, the system is under-composed. Avoid line spacing +and repetition frequencies that moiré after downsampling; render the actual thumbnail. diff --git a/.claude/skills/carousel-art-studio/references/light-material-depth.md b/.claude/skills/carousel-art-studio/references/light-material-depth.md new file mode 100644 index 000000000..68c192e01 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/light-material-depth.md @@ -0,0 +1,87 @@ +# Light, material and depth + +## Light is part of the claim + +Write the light plot before shading: + +```text +Key: source, direction, size, color, story-world reason +Fill: source, ratio to key, what must remain readable +Practical/emissive: physical source and narrative job +Ground: value and material under the subject +Contact: where weight is proven +Atmosphere: what catches light and why it exists +``` + +One strong key with controlled fill usually reads better than many undirected lights. A rim +light is meaningful when it separates a subject from the environment or belongs to a real +source; otherwise it becomes a “cinematic” sticker. + +## Value pass + +Before color, identify: + +- darkest anchor; +- lightest event; +- subject value group; +- type reserve; +- separation between adjacent planes. + +The lightest and darkest regions need different jobs. Check the thumbnail in grayscale. If +subject and ground merge, color will not rescue them after platform compression. + +## Material sentence + +Do not say “make it tactile.” Define each important surface: + +```text +Material: +Base value/hue: +Roughness / gloss: +Edge behavior: +Microstructure and scale: +Thickness / translucency: +Wear or manufacture: +What distinguishes it from the adjacent surface: +``` + +Material comes from response, not texture pasted on top. Metal needs coherent highlights and +edge behavior; paper needs fiber scale, thickness and soft grazing light; stone needs mass, +fracture and broad value variation; glass needs reflection/refraction logic and something to +reveal by transmitting. + +## The depth stack + +Build in this order: + +1. camera and silhouette; +2. occlusion and overlap; +3. lit ground and contact shadow; +4. key-to-fill ratio; +5. local material response; +6. aerial perspective or depth fog; +7. optical finish. + +Contact shadow on a black ground is invisible. Light the ground first, then subtract the +shadow. A floating hero object should float because the claim wants suspension, not because a +contact solution was skipped. + +## Browser-native benches + +- `txthree.js`: PBR, ACES, soft shadows, procedural environment, lighting rigs and hero-object + helpers. Use for one consequential dimensional shot, not every slide. +- `txsdf.js`: ray-marched continuous forms, unions/cuts and impossible sections. +- `txrelief.js`: fast relit heightfields and 2.5D form. +- SVG diffuse/specular lighting plus turbulence/displacement: useful for paper, relief and + surface detail that must remain browser-native. +- `txpost.js`: final grade, bloom, grain, sharpen and dither after structure is resolved. + +Use the lightest bench that produces the shot. Complexity hidden behind a flat camera has no +value. + +## Finish restraint + +Bloom belongs to luminous energy, not every bright edge. Grain binds layers at final size; +visible repeating grain is texture wallpaper. Chromatic aberration and glow are not proof of +depth. Finish should make the materials feel photographed through one optical system, not make +the frame announce software. diff --git a/.claude/skills/carousel-art-studio/references/linkedin-native.md b/.claude/skills/carousel-art-studio/references/linkedin-native.md new file mode 100644 index 000000000..bc10c3ebc --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/linkedin-native.md @@ -0,0 +1,72 @@ +# LinkedIn-native carousel direction + +## The real canvas is 432 pixels wide + +The source is 1080×1350, but the first judgment happens around 432 pixels wide in a moving +phone feed. Inspect both. At feed size: + +- one image event must dominate; +- headline shape should read before every word does; +- body copy must be optional to understanding the visual turn; +- thin rules, fine particles and subtle value differences need intentional survival tests; +- the outer UI-safe band cannot contain the only evidence or focal detail. + +Hide type during one review. If nothing remains, the artwork is a background. Hide art during +another. If the complete slide remains, the art is decoration. + +## One visual sentence per slide + +Write each as a verb: + +- the queue **compresses**; +- the boundary **moves**; +- the filing **opens** to reveal the omitted layer; +- the network **narrows** at one verified point; +- the material **changes state** across a documented threshold. + +“Shows three facts” is not a visual sentence. Choose the relation that advances the swipe. + +## Swipe choreography + +The right edge is an invitation, not mandatory furniture. Create continuation with cropped +motion, interrupted geometry, a question in the image or a changed state the reader wants to +resolve. Do not add arrows to every slide. + +Use the deck as time: + +- reveal hidden layers across frames; +- carry one object through changes of scale or state; +- reverse camera to expose consequence; +- let a system accumulate, then interrupt it; +- repeat one anchor while changing everything it touches. + +The frames should not all peak at once. A quiet frame earns attention through precision, +absence or scale, then makes the next dense frame feel larger. + +## Publication furniture + +Counters, source cues and the Docket mark create continuity. Keep their grid stable enough to +orient without turning the artwork into a template. Furniture should recede until needed. It +must never be the most intricate system on a slide. + +## Cover contract + +The cover has one job: cause a qualified reader to stop and open the document. It needs: + +- a four-to-seven-word hook when the story allows; +- one ownable silhouette or spatial event; +- a literal connection to the record; +- a clear value hierarchy under platform compression; +- enough unexplained tension to invite the next frame, without clickbait or ambiguity about + the subject. + +Do not make the cover summarize the article. Make it open a visual question the deck can +answer. + +## Learning after publication + +Record the shipped construction, material world, camera grammar and cover silhouette along +with available document/post analytics. Compare families over multiple posts. Saves, shares, +qualified comments and document engagement are useful signals; none alone defines artistic +quality. A repeated construction that performs is a franchise decision, not automatic license +to use it daily. diff --git a/.claude/skills/carousel-art-studio/references/review-gates.md b/.claude/skills/carousel-art-studio/references/review-gates.md new file mode 100644 index 000000000..a0f5238df --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/review-gates.md @@ -0,0 +1,67 @@ +# Review gates — ambition before polish + +Run these on actual renders, not the treatment. + +## Gate 1 — truth + +- Is each data-bearing position, size, direction and relation sourced? +- Are modeled and illustrative layers identified? +- Does metaphor imply damage, causation, geography or magnitude the record does not support? +- Is source matter represented faithfully rather than fabricated for texture? + +Any failure is a veto. + +## Gate 2 — feed stop + +At 432 pixels, with type hidden for one pass: + +- Can the dominant silhouette be named instantly? +- Is there a real light/dark event? +- Does the crop create energy or consequence? +- Is the focal event still larger than furniture and detail? + +If the frame only becomes interesting after reading, return to composition. + +## Gate 3 — specificity and memory + +- Could another Texas technology or policy article wear this image unchanged? +- Can a reviewer describe it tomorrow with a concrete noun and verb? +- Does the story's own material world determine palette, marks and light? +- Is the direction visibly distinct from the recent-deck ledger? + +Generic competence is a failure at this gate. + +## Gate 4 — space, light and material + +- Can the camera, horizon and depth order be named? +- Do overlaps, occlusion and scale agree with that camera? +- Does a motivated key reveal form, with fill preserving evidence? +- Does every grounded object have a lit ground and readable contact? +- Do adjacent materials differ by response rather than color alone? + +## Gate 5 — type and evidence + +- Does type have a spatial role or a genuinely quiet caption role? +- Does art still work when type is hidden, and does the slide still communicate when art is + temporarily hidden? +- Are labels near the thing they explain without turning the frame into a dashboard? +- Is body text optional to the visual turn and readable at feed size? + +## Gate 6 — sequence + +- Do scale, density, value, vantage and tempo change intentionally? +- Does every slide have a different visual sentence inside the same world? +- Is there pressure and release rather than nine simultaneous peaks? +- Do at least two interior slides equal the cover's craft, rather than merely carry exposition? + +## Gate 7 — finish + +- Does microdetail reward full-size viewing without muddying the thumbnail? +- Do atmosphere and grade unify rather than conceal? +- Is the image optically coherent — one world, one camera, one light logic? +- Are there any “AI luxury” tells: gratuitous glow, generic gold, translucent cards, fake + precision, ornamental data, default serif seriousness? + +Only after these pass should machine QA and panel review begin. Machine green means safe; +panel green means accepted; neither replaces the studio's responsibility to make exceptional +work. diff --git a/.claude/skills/carousel-art-studio/references/styleframe-brief.md b/.claude/skills/carousel-art-studio/references/styleframe-brief.md new file mode 100644 index 000000000..72f40ff47 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/styleframe-brief.md @@ -0,0 +1,108 @@ +# Styleframe competition brief + +Create this artifact before full-deck production. + +## Shared truth + +```text +Article / date: +Cover claim: +Interior-turn claim: +Known visual evidence: +Forbidden implications: +Recent constructions to avoid: +``` + +## Reference cut + +Three to six entries, each with URL/item ID, study-only property, story evidence served, +transformation, do-not-copy note and rights/provenance. + +## Candidate card — repeat at least three times + +```text +NAME — a concrete noun/verb phrase, not an adjective + +Visual thesis: +Memory sentence: +Literal anchor: +Evidence behavior / metaphor boundary: + +Image architecture: +Camera / lens / horizon: +Near-middle-far planes: +Key / fill / practical light: +Hero material and contrast material: +Type's spatial role: +Generative or motion rule: + +Cover shot: +Interior-turn shot: +Why it cannot belong to another story: +Which current AI-design default it rejects: +Hardest technical proof: +``` + +Build both shots. Use shared copy across candidates so copy quality does not bias the visual +choice. Candidate directions must not share image architecture, camera grammar and material +world. + +## Render board + +For every candidate include: + +- 1080×1350 cover; +- 432-pixel cover thumbnail; +- 1080×1350 interior turn; +- 432-pixel interior thumbnail; +- grayscale thumbnails; +- notes on what survived or disappeared. + +Do not compare browser screenshots at arbitrary zoom. Use engine output and actual thumbnails. + +## Selection table + +Score 1–10 without averaging away a veto: + +| Dimension | Candidate A | Candidate B | Candidate C | +|---|---:|---:|---:| +| Feed stop | | | | +| Story specificity | | | | +| Memorable silhouette | | | | +| Camera/depth | | | | +| Light/material | | | | +| Type/art relationship | | | | +| Evidence integrity | | | | +| Interior range | | | | +| Novelty versus recent decks | | | | +| Feasible without compromising the idea | | | | + +Then write: + +```text +SELECTED: +Decisive pixel evidence: +What must be fixed before production: +Discard A because: +Discard B because: +Discard C because: +``` + +## Production shot card — repeat per slide + +```text +SLIDE / visual job / visual sentence: +Claim and source locator: +Shot and camera: +Dominant silhouette: +Depth order: +Light/material event: +Evidence geometry: +Type role and protected reserve: +Continuity from previous / invitation to next: +Technique and why this technique: +Feed-size survival requirement: +Pass-specific risks: +``` + +If three cards could be merged by swapping their copy, return to deck direction. diff --git a/.claude/skills/carousel-art-studio/references/visual-thesis.md b/.claude/skills/carousel-art-studio/references/visual-thesis.md new file mode 100644 index 000000000..4e5c55702 --- /dev/null +++ b/.claude/skills/carousel-art-studio/references/visual-thesis.md @@ -0,0 +1,78 @@ +# Visual thesis — from record to image + +## Start with a claim card + +Write this before metaphor or style: + +```text +CLAIM +One sentence: +Source / locator: +Place and time: +Known quantities: +Unknown or withheld: +What the image may assert: +What the image must not imply: +``` + +The last two lines are art direction. A line between two nodes asserts relation. A larger form +asserts magnitude or importance. A map anchor asserts place. A broken object asserts damage. +An empty chair asserts absence. If the record does not support the sentence, the geometry may +not say it. + +## Find the story's native material + +List the actual nouns and verbs in the record before reaching for abstract graphics: + +- nouns: site, machine, filing, material, landscape, institution, network, measurement; +- verbs: accumulate, split, queue, harden, migrate, expose, withhold, overlap, accelerate; +- forces: pressure, delay, heat, friction, scarcity, distance, load, uncertainty; +- scales: microscopic, human, architectural, regional, systemic, temporal. + +The strongest world usually joins one literal anchor with one explanatory transformation. A +literal transformer may contain a field of queued records. A filing may become architectural +space. The literal anchor keeps the metaphor accountable; the transformation makes it art. + +## Thesis test + +Use this sentence: + +> The record becomes **[world/object/system]**, seen from **[vantage]**, where **[verified +> tension]** appears as **[behavior]**, illuminated by **[motivated source]**. + +It passes only if each blank changes the pixels. “A cinematic data landscape showing change” +fails: no object, camera, behavior or source has been chosen. + +Then write: + +- **memory sentence:** what someone can concretely describe tomorrow; +- **literal anchor:** what prevents the image floating free of the story; +- **evidence behavior:** which geometry carries verified information; +- **metaphor boundary:** what is illustrative and cannot be read quantitatively; +- **anti-direction:** the obvious treatment this will refuse. + +## Transformation ladder + +If the first idea is generic, move one rung at a time: + +1. **depict** — show the real noun; +2. **reframe** — choose a consequential vantage or crop; +3. **expose** — reveal internal structure, time or hidden relation; +4. **materialize** — give an abstract process physical behavior; +5. **systematize** — let verified data drive form; +6. **stage** — use light, scale and sequence to create the visual event. + +Do not skip directly to spectacle. Every rung must preserve the truth of the previous one. + +## Three concept families worth testing + +These are prompts for divergence, not reusable styles: + +- **Evidence object:** one story-native object becomes monumental; record fragments, measured + geometry or institutional marks are physically integrated into it. +- **Impossible evidence camera:** a truthful vantage no physical camera can occupy — cutaway, + multi-scale section, time stack — reveals the relation the article explains. +- **Behavioral field:** a deterministic system enacts the verified process; a literal anchor, + legend and metaphor boundary keep it honest. + +If all three yield the same centered composition, divergence has not happened. diff --git a/examples/art-atelier/README.md b/examples/art-atelier/README.md new file mode 100644 index 000000000..626dabead --- /dev/null +++ b/examples/art-atelier/README.md @@ -0,0 +1,29 @@ +# Carousel art atelier + +Six executable construction studies for the `carousel-art-studio` skill. They prove larger +visual ideas — camera, depth, light, material, evidence and typography — with the same offline +browser stack used in production. + +They are **not slide templates** and do not contain reusable story claims. Copy a mechanism, +never a composition, palette, object or lockup. + +## Render + +```bash +python3 .claude/skills/carousel-engine/render.py \ + --slides-dir examples/art-atelier/slides \ + --out-dir out/art-atelier/tmp/render + +python3 .claude/skills/carousel-engine/qa.py \ + --render-dir out/art-atelier/tmp/render + +python3 .claude/skills/carousel-engine/assemble.py \ + --slides-dir examples/art-atelier/slides \ + --render-dir out/art-atelier/tmp/render \ + --out-dir out/art-atelier/tmp/final \ + --title "Texas AI Docket — code atelier" +``` + +Inspect the 1080×1350 renders, the 432-pixel thumbnails and the contact sheet. See +`.claude/skills/carousel-art-studio/references/atelier.md` for the construction index and the +rules for learning from these studies. diff --git a/examples/art-atelier/slides/slide-01.html b/examples/art-atelier/slides/slide-01.html new file mode 100644 index 000000000..a46f9040a --- /dev/null +++ b/examples/art-atelier/slides/slide-01.html @@ -0,0 +1,142 @@ + + + + + + + + + + + +
Code atelier · evidence object
+
01
+

Therecordhasweight

+

A filing becomes a monumental optical object: cropped, dimensional and + physically marked by its own evidence layer.

+
Texas AI DocketCONSTRUCTION STUDY · NOT DATA
+ + + + + diff --git a/examples/art-atelier/slides/slide-02.html b/examples/art-atelier/slides/slide-02.html new file mode 100644 index 000000000..5bbd25ee5 --- /dev/null +++ b/examples/art-atelier/slides/slide-02.html @@ -0,0 +1,122 @@ + + + + + + + + + + +
Code atelier · impossible camera
+
02
+

What the
surface hides

+

Outer mass, layered section, internal path — and the light that makes the + reveal legible.

+ + + +
Texas AI DocketCUTAWAY · DEPTH BEFORE TEXTURE
+ + diff --git a/examples/art-atelier/slides/slide-03.html b/examples/art-atelier/slides/slide-03.html new file mode 100644 index 000000000..e8b179266 --- /dev/null +++ b/examples/art-atelier/slides/slide-03.html @@ -0,0 +1,129 @@ + + + + + + + + + + + +
+
+
Code atelier · behavioral field
+
03
+

The threshold
moves

+
fieldthreshold
+
Texas AI DocketILLUSTRATIVE SYSTEM · SEE DOSSIER
+ + + + + diff --git a/examples/art-atelier/slides/slide-04.html b/examples/art-atelier/slides/slide-04.html new file mode 100644 index 000000000..3ea9f6bc1 --- /dev/null +++ b/examples/art-atelier/slides/slide-04.html @@ -0,0 +1,145 @@ + + + + + + + + + + +
Code atelier · bound geometry
+
04
+

Make the
number physical

+
Illustrative study values
18 · 32 · 27 · 47 · 39
+ + + +
Texas AI DocketVALUES DECLARED · FORM BOUND
+ + diff --git a/examples/art-atelier/slides/slide-05.html b/examples/art-atelier/slides/slide-05.html new file mode 100644 index 000000000..c9aed0eb5 --- /dev/null +++ b/examples/art-atelier/slides/slide-05.html @@ -0,0 +1,131 @@ + + + + + + + + + + +
Code atelier · evidence archaeology
+
05
+

What the
margin reveals

+

Layered sheets, a physical edge, selective reveal and one magnified + evidence zone.

+ + + +
+
Texas AI DocketSOURCE MATTER · LIGHT AS REVEAL
+ + diff --git a/examples/art-atelier/slides/slide-06.html b/examples/art-atelier/slides/slide-06.html new file mode 100644 index 000000000..2ecf1fd5d --- /dev/null +++ b/examples/art-atelier/slides/slide-06.html @@ -0,0 +1,143 @@ + + + + + + + + + + +
Code atelier · type as structure
+
06
+

Language does
the work

+
glyph = chassis
baseline = rail
counter = aperture
+ + + +
Texas AI DocketTYPE PARTICIPATES · COPY STAYS CLEAR
+ + + + diff --git a/knowledge/carousel/REFERENCE_ATLAS.md b/knowledge/carousel/REFERENCE_ATLAS.md new file mode 100644 index 000000000..8837d3861 --- /dev/null +++ b/knowledge/carousel/REFERENCE_ATLAS.md @@ -0,0 +1,120 @@ +# Reference atlas — computational editorial art + +This atlas is a research instrument for the artwork lane, not a mood-board or a list of styles +to imitate. Open only the sources relevant to the day's visual problem. For every selected +reference, record: + +1. the precise craft property being studied; +2. the story evidence it could help express; +3. the transformation that makes the result specific to this record; +4. what must not be copied. + +Prefer primary sources, original project pages and technical documentation. Do not download or +embed copyrighted reference imagery in the repository. Public-domain archival material still +needs source and rights notes in the day's dossier. + +## Anti-default art direction + +| Source | Study for | Translation rule | +|---|---|---| +| [Anthropic frontend-design skill](https://github.com/anthropics/skills/blob/main/skills/frontend-design/SKILL.md) | Explicit recognition of common AI design defaults and the demand for committed aesthetic choices | Use its anti-generic diagnostic, not its example aesthetics. Reject cream/serif, dark/accent and broadsheet defaults unless the claim uniquely demands one | +| [Anthropic canvas-design skill](https://github.com/anthropics/skills/blob/main/skills/canvas-design/SKILL.md) | Visual hierarchy, minimal essential text and art-first composition | Carry the 90/10 visual-to-essential-text instinct into the cover; preserve Docket evidence and accessibility requirements | +| [Anthropic algorithmic-art skill](https://github.com/anthropics/skills/blob/main/skills/algorithmic-art/SKILL.md) | A generative philosophy before parameters, seeded variation and process-driven form | Write a claim-specific visual thesis before code. Never use generative complexity as decoration | + +## Editorial systems and data storytelling + +| Source | Study for | Translation rule | +|---|---|---| +| [The Pudding — how to make dope shit, part 2](https://pudding.cool/process/how-to-make-dope-shit-part-2/) | Iteration, visual form finding and the distance between an idea and a finished explanatory image | Borrow the iterative behavior. Produce real styleframes instead of defending the first concept in prose | +| [The Pudding — storytelling, part 3](https://pudding.cool/process/how-to-make-dope-shit-part-3/) | Narrative focus and how interaction/visual form serves a specific reader takeaway | Reduce each slide to one visual sentence; do not compress the article into decorative facts | +| [The Pudding — What does AI see?](https://pudding.cool/2024/07/ai/) | A technical subject turned into a legible, art-directed visual experience | Study the coordination of explanation, sequencing and custom imagery; do not borrow its visual identity | +| [Pentagram — Data Driven Experiences](https://www.pentagram.com/work/data-driven-experiences) | Data as spatial, material and experiential form | Ask how the record can become environment, object or motion instead of a conventional chart | +| [Pentagram — La Lettura](https://www.pentagram.com/work/la-lettura) | Commissioned data visualization as a recurring editorial institution | Study range without losing publication authority. Do not import a house style from another publication | +| [Pentagram — Deloitte Insights](https://www.pentagram.com/work/deloitte-insights) | A coherent editorial system that can hold varied content | Separate enduring Docket furniture from the bespoke image so coherence does not force template art | +| [Stamen](https://stamen.com/) | Expressive cartography and data-driven visual systems | Let real geography and data shape the image. Never use map texture when the claim is not geographic | + +## Motion, title design and dimensional information + +| Source | Study for | Translation rule | +|---|---|---| +| [FIELD.IO](https://field.io/) | Computational systems with strong motion logic, materiality and spatial composition | Freeze the most consequential instant into a carousel frame; define what force or rule produced the form | +| [Territory Studio — American Assassin](https://territorystudio.com/project/american-assassin/) | Dense information integrated into cinematic image-making | Study hierarchy, parallax and luminous depth. Remove fictional surveillance language and use only record-grounded signals | +| [Territory Studio — Expo 2020 Dubai](https://territorystudio.com/project/expo-2020-dubai/) | Large-scale data and architectural media as one visual system | Translate scale, choreography and system logic into an editorial still rather than mimicking spectacle | + +## Generative composition and natural systems + +| Source | Study for | Translation rule | +|---|---|---| +| [Art Blocks](https://www.artblocks.io/) and [Curated collection](https://www.artblocks.io/discover/curated) | The range possible when a work is a coherent algorithm, not an effect preset | Study system-level originality and parameter discipline. Do not clone a project, artist signature or token aesthetic | +| [Art Blocks protocol philosophy](https://docs.artblocks.io/protocol/overview/) | Deterministic output and the relationship between authored system and generated instance | Treat the day's seed as reproducibility, not novelty by roulette; art direction still owns the result | +| [The Nature of Code](https://natureofcode.com/) | Vectors, forces, agents, particles, oscillation and emergent systems | Choose a physical model only when its behavior explains the claim; label metaphorical simulation as metaphor | +| [p5.js examples](https://p5js.org/examples/) | Small, inspectable creative-coding patterns | Use examples to learn a mechanism, then rebuild it within the committed offline engine and story palette | +| [The Book of Shaders](https://thebookofshaders.com/), [noise](https://thebookofshaders.com/11/) and [fBm](https://thebookofshaders.com/13/) | Coordinate thinking, procedural texture, noise and layered fields | Design scale, anisotropy and domain warping from the material being depicted. Generic multicolor noise is not a visual thesis | + +## Camera, light, material and finish + +| Source | Study for | Translation rule | +|---|---|---| +| [Three.js examples](https://threejs.org/examples/) | Production patterns across geometry, cameras, light, materials and post | Identify one relevant mechanism; never treat the examples gallery as a menu of looks | +| [Three.js post-processing manual](https://threejs.org/manual/en/post-processing.html) | Ordered image finishing and the purpose of individual passes | Grade a resolved image. Bloom, contrast and grain may unify; they cannot repair composition | +| [Three.js shadow manual](https://threejs.org/manual/en/shadows.html) | Shadow-map tradeoffs and alternatives | Budget shadows for the hero form and preserve a readable lit ground/contact relationship | +| [WebGL Fundamentals — perspective](https://webglfundamentals.org/webgl/lessons/webgl-3d-perspective.html) and [camera](https://webglfundamentals.org/webgl/lessons/webgl-3d-camera.html) | How field of view, camera transform and clipping create spatial meaning | Put lens, camera height, target and horizon in the styleframe brief before surface polish | +| [WebGL Fundamentals — directional light](https://webglfundamentals.org/webgl/lessons/webgl-3d-lighting-directional.html) and [point light](https://webglfundamentals.org/webgl/lessons/webgl-3d-lighting-point.html) | Normals, direction, falloff and readable form | Name key, fill and practical light. If the light has no story-world source, it is decoration | +| [WebGL Fundamentals — shadows](https://webglfundamentals.org/webgl/lessons/webgl-shadows.html) | Projective shadow construction | Use cast shadow to prove spatial relation, not as a generic drop-shadow effect | +| MDN — [`feDiffuseLighting`](https://developer.mozilla.org/en-US/docs/Web/SVG/Reference/Element/feDiffuseLighting), [`feSpecularLighting`](https://developer.mozilla.org/en-US/docs/Web/SVG/Reference/Element/feSpecularLighting), [`feTurbulence`](https://developer.mozilla.org/en-US/docs/Web/SVG/Reference/Element/feTurbulence), [`feDisplacementMap`](https://developer.mozilla.org/en-US/docs/Web/SVG/Reference/Element/feDisplacementMap) | Browser-native relief, surface response and distortion | Use SVG filters for physically coherent surface cues; test at the final raster size because filter scale is part of the image | + +## Scientific visualization and honest wonder + +| Source | Study for | Translation rule | +|---|---|---| +| [NASA Scientific Visualization Studio — principles](https://svs.gsfc.nasa.gov/gallery/visualization-principles/) | Accuracy, clarity, provenance and visual explanation at spectacular scale | Adopt its evidence discipline: distinguish measured, modeled, illustrative and unknown content in the dossier | +| [NASA Scientific Visualization Studio](https://svs.gsfc.nasa.gov/) | Camera paths, scale transitions, atmospheric phenomena and data-rich imagery | Study how real phenomena generate beauty. Do not borrow a space aesthetic for an earthbound administrative claim | +| [NASA SVS — Dust in the Wind](https://svs.gsfc.nasa.gov/4362) | Technical-art collaboration around particulate flow and atmospheric data | Use it as a model for joining subject expertise with rendering craft, not as a dust-effect recipe | +| [d3](https://d3js.org/), [d3-contour](https://d3js.org/d3-contour/contour) and [d3-geo](https://d3js.org/d3-geo) | Data-bound geometry, isolines and truthful projections | Bind visible form to verified values and real projection. Decorative geometry must never masquerade as data | + +## Texas material and cartographic ground + +| Source | Study for | Translation rule | +|---|---|---| +| [Texas General Land Office archives](https://www.glo.texas.gov/archives-and-heritage) | Historic maps, land records and Texas-specific material culture | Search for place- and period-specific evidence. Record item-level rights and citation before use | +| [Texas State Library and Archives collections](https://www.tsl.texas.gov/collections) | State records, photographs, maps and institutional visual language | Let genuine record material inform marks, scale and palette; avoid faux-aged archival styling | +| [USGS geologic map symbol standard](https://pubs.usgs.gov/of/1999/of99-430/) | A rigorous grammar for geological pattern, line and annotation | Use the visual syntax when geology is materially relevant and explain any simplification | +| [Library of Congress Sanborn data package](https://data.labs.loc.gov/sanborn/) | Public-domain fire-insurance map geometry and urban layers | Use real place geometry and retain source metadata; do not paste a map merely to signal “history” | +| [Library of Congress map collections](https://www.loc.gov/maps/collections/) | Primary-source cartographic composition and regional detail | Verify item rights and transform source material around the day's claim rather than reproducing a decorative crop | + +## LinkedIn-native viewing + +| Source | Study for | Translation rule | +|---|---|---| +| [LinkedIn document upload support](https://www.linkedin.com/help/linkedin/answer/a518909) and [document FAQ](https://www.linkedin.com/help/linkedin/answer/a523054) | Current document constraints and platform behavior | Treat the 1080×1350 render and vector PDF as one product; verify platform requirements when they may have changed | +| [LinkedIn thought-leadership planning](https://www.linkedin.com/business/marketing/blog/content-marketing/creating-a-thought-leadership-marketing-plan) | Accurate, relevant, digestible, shareable and actionable communication on mobile | Make the image create curiosity, then make the slide reward it with one useful idea; authority cannot compensate for friction | +| [LinkedIn post analytics](https://www.linkedin.com/help/linkedin/answer/a525196) | The metrics available for learning after publication | Evaluate construction families over time. Do not optimize a single post or confuse reach with visual quality | + +## Reference-set recipe for one deck + +Use **three to six references**, enough to triangulate and few enough to influence decisions: + +- one story-domain source: the actual place, material, machine, institution or record; +- one composition/camera source; +- one light/material or mark-making source; +- one data/scientific source when geometry carries evidence; +- optionally one sequencing or LinkedIn-native source. + +At least two references must come from outside the current AI-design aesthetic. At least one +must be a primary story-domain source. If all references look alike, the set is permission to +imitate, not research. + +Use this note shape in the day's visual research artifact: + +```text +REFERENCE +URL / item ID: +Study only: +Story evidence it serves: +Transform into: +Do not copy: +Rights / provenance: +``` + +The final image should make the references difficult to reverse-engineer. What remains visible +is the Docket's judgment about this particular record. diff --git a/knowledge/carousel/VISUAL_CEILING_AUDIT.md b/knowledge/carousel/VISUAL_CEILING_AUDIT.md new file mode 100644 index 000000000..7868d0b18 --- /dev/null +++ b/knowledge/carousel/VISUAL_CEILING_AUDIT.md @@ -0,0 +1,182 @@ +# Visual ceiling audit — replacing the V1 artwork lane + +## Mandate + +This is not a regression repair. The entire daily artwork lane was a competent V1: it proved +that a researched article could become a coherent, deterministic, hand-coded LinkedIn carousel. +The next lane must prove something much more ambitious: that every day's record can generate a +piece of computational editorial art people stop for, remember and want to share. + +The target is not “raise a six to a seven.” It is a new ceiling: + +- slide 1 stops the feed before the headline has been read; +- every later slide has an authored visual idea, including deliberately quiet slides; +- the image and the claim are inseparable — another story could not wear the same frame; +- light, material, camera, scale and motion are designed, not added as finish; +- the deck has cinematic rhythm rather than nine variations of one plate; +- the pixels hold up beside elite editorial illustration, data storytelling, title design and + generative art while remaining accurate to the public record. + +Scores remain useful diagnostics. They are not the ambition. + +## Corpus reviewed + +The audit covers every shipped carousel contact sheet available in the repository at the time +of review, its artwork-panel median and the rendered article for the three newest decks. The +eleven-deck median is **6.9** and the mean is **6.8**. More importantly, the same ceiling is +visible in high- and low-scoring decks: coherent art direction, but limited spatial drama and a +small family of repeated constructions. + +| Date | Artwork median | Dominant construction | Ceiling observed | +|---|---:|---|---| +| 2026-08-16 | 6.8 | Dark editorial atmosphere | Polished title-card language carries too much of the deck | +| 2026-08-18 | 6.0 | Cream record plates | Documents and text become layout furniture rather than image-making | +| 2026-08-19 | 7.8 | Dark technical editorial | Strongest V1 deck, but still largely diagram-plus-type rather than cinematic scenes | +| 2026-08-20 | 7.0 | Restrained cream editorial | Consistent and readable; limited depth, scale change and material surprise | +| 2026-08-21 | 7.0 | Dark data/record graphics | Information is designed, but frames share a similar planar grammar | +| 2026-08-22 | 7.0 | Cream archival editorial | Evidence is present; the document surface often becomes the composition | +| 2026-08-25 | 7.1 | Dark technical plates | Good system coherence; few images have a singular silhouette at feed size | +| 2026-08-26 | 6.9 | Cream serif editorial | Publication polish without a sufficiently new visual world | +| 2026-08-27 | 6.0 | Black industrial object plates | Repeated headline/object/footer staging and under-realized techniques | +| 2026-08-28 | 6.4 | Geological strata/data plates | Several slides reuse the same layered-section construction | +| 2026-08-29 | 6.5 | Institutional room/material plates | Planar centered compositions; type on muted rectangles dominates the image | + +The last three scores are evidence, not the scope. A return to the earlier median would only +restore the V1 ceiling. + +## What V1 already gets right + +The replacement should preserve these gains: + +- a real article claim and evidence dossier precede the art; +- slide HTML is bespoke, deterministic and reconstructable; +- the render engine already supports Canvas, SVG, d3, software 3D, Three.js, signed-distance + fields, relief, engraving, cartography and film-grade post-processing; +- type remains vector and objective QA catches many silent production failures; +- the decks maintain publication coherence and avoid generic stock imagery. + +This is why buying another rendering primitive is not the main move. The machinery is already +capable of more than the lane routinely asks it to make. + +## The six structural limits + +### 1. Treatment is prose, not pixels + +The treatment directors can read and describe, but they cannot make competing styleframes. +The showrunner chooses among promises about images instead of images. A paragraph can sound +cinematic while collapsing into a dark rectangle, a centered object and a footer at 432 pixels. + +**Replacement:** require divergent rendered styleframes — at minimum the cover and one interior +turn — before a deck language is chosen. Judge silhouette, value structure and legibility at +feed size, not rhetorical confidence. + +### 2. One generalist crosses too many crafts + +Phase 11 asks one showrunner to translate research, direct a visual system, model scenes, light +materials, compose type, code every slide and finish the deck. That encourages the cheapest +reliable construction and makes difficult art the first thing traded away under time pressure. + +**Replacement:** give the art maker a studio operating system: visual thesis, reference +research, styleframe, graybox, camera, light, material, type integration and finish. Later, +wire the routine so those artifacts exist before full production. + +### 3. The review loop is plan-relative + +Pixel critics compare the render with the chosen dossier. They can catch poor execution of a +good idea, but a thin idea faithfully executed can pass. Judges arrive after the expensive +decisions and are deliberately not an art-direction loop. + +**Replacement:** add a ceiling review before production: “Is this image singular enough to +deserve nine frames?” Keep later critics, but make them compare against both the plan and the +north-star scorecard below. + +### 4. Research is counted, not converted into craft + +The routine requests searches, but does not persist an annotated visual research artifact. +Search activity therefore has no reliable path into camera, material, palette, typography or +visual metaphor. + +**Replacement:** a small, cited reference set per deck. Every reference must name what is being +studied and the transformation into this story. Reference is a springboard, never a style to +copy. + +### 5. The technique library names tools more than shots + +The existing library is strong on primitives and failure modes. It is weaker on the larger +construction: what the viewer sees, where the camera is, which material owns the highlights, +how evidence enters the image and why the frame has a memorable silhouette. + +**Replacement:** pair the engine with a code atelier of finished constructions. Agents should +learn how primitives combine into a shot, not treat “flow field” or “SDF” as an art direction. + +### 6. The lane has converged on familiar AI-design defaults + +The corpus repeatedly reaches for restrained serif editorialism, cream archival plates, or +near-black technical plates with one warm accent. These are coherent, but they are also among +the most common current AI design defaults. Consistency has become similarity. + +**Replacement:** use an explicit anti-default pass. If the concept can be summarized as “dark +field + floating object + elegant serif” or “cream paper + rules + document fragment,” it must +earn that choice against two materially different alternatives. + +## The 10x scorecard + +Evaluate the full-resolution render and the 432-pixel feed thumbnail. A showstopper should land +at **9 or better in every category**; one category cannot average away another. Accuracy and +legibility remain hard gates. + +| Dimension | 6-level V1 | 9-level target | Review question | +|---|---|---|---| +| Feed stop | Tasteful layout | Immediate visual event | Would the silhouette interrupt a fast scroll with type hidden? | +| Story specificity | Topic-adjacent mood | Claim embodied in form | Could this exact image belong to another article? | +| Composition | Balanced arrangement | Directed tension and hierarchy | Is there one unmistakable entry point and a controlled eye path? | +| Depth and camera | Layered flatness | Intentional space and scale | Can the reviewer name lens, vantage, horizon and depth order? | +| Light and material | Palette plus texture | Light reveals story material | Do highlights, occlusion and shadow explain what things are made of? | +| Type relationship | Text placed over/around art | Type participates in the scene | Does type affect scale, rhythm or meaning without becoming a plate? | +| Evidence integrity | Accurate labels and numbers | Evidence is the visual engine | Is every data-bearing geometry traceable and every metaphor bounded? | +| Craft density | One technique executed cleanly | Multiple systems resolve as one image | Does close inspection reward attention without muddying the thumb? | +| Deck rhythm | Variations on one system | Designed sequence of pressure and release | Do scale, density, perspective and tempo change across the swipe? | +| Memorability | Professionally on-brand | Ownable image | Can someone describe the frame tomorrow in one concrete sentence? | + +### Hard vetoes + +Reject a direction before production when any of these are true: + +- the cover depends on reading more than seeing; +- the main image has no recognizable silhouette at 432 pixels; +- the reference set contains only other AI-made work or only one aesthetic family; +- the visual metaphor implies an unsupported fact, magnitude, location or causal relationship; +- three or more slides could be produced by changing the text inside the same composition; +- “premium,” “cinematic,” “editorial,” “bold” or a named technique substitutes for a shot; +- polish is expected to rescue an unresolved graybox; +- the deck repeats a recent material world, palette, camera grammar or signature construction + without a claim-specific reason. + +## The new capability contract + +Before the daily routine is rewired, the artwork layer needs four durable assets: + +1. **This audit** — a shared definition of the ceiling and the gap. +2. **A reference atlas** — primary, craft-specific sources with translation rules. +3. **A carousel art-studio skill** — a progressive workflow that turns evidence into visual + thesis, divergent pixels and production-ready shot plans. +4. **A code atelier** — executable styleframes proving that the committed engine can produce + dimensional, authored images rather than merely naming advanced techniques. + +Only after those exist should the routine decide which agents make which artifacts. Otherwise +automation would merely make the V1 process more elaborate. + +## Definition of success for the later workflow phase + +When the new capability is wired into the routine, it should be measurable beyond panel score: + +- compare styleframe candidates at feed size before committing the deck; +- record which visual construction, material world and camera grammar shipped; +- track cover dwell/stop proxies, document opens, completion/swipe behavior where LinkedIn + exposes them, saves, shares and qualified comments; +- run periodic blind side-by-side reviews against both prior Docket decks and current + best-in-class references; +- treat repeated visual language as creative debt even when it scores well. + +The north star is simple: a reader should recognize the Texas AI Docket because the work is +specific and excellent, not because every deck looks the same.