# Emerging house language

Status: working hypotheses, not a final design system. Updated 2026-09-07 by Codex.

## What the owner has actually endorsed

The folio's paper identity, subject-specific illustrations, subtle subject colors, smooth motion and self-drawing plant received positive feedback. The blueprint subject was praised as a surprising but coherent interpretation of paper. The folio itself is not a settled homepage format. The gray surround and redundant explanatory labels were questioned, not ordered removed.

Earlier Claude experiments established useful negative evidence: paper collage can become a campus bulletin board despite good individual details; age spots can read as a dirty screen; uniform dots over everything can destroy paper identity. Ranking an unsuccessful option above another is not endorsement of its direction.

## Vocabulary — ingredients are not recipes

- Substrates: warm opaque stock, tracing paper, colored print stock, technical drawing paper.
- Marks: pencil contour, typed caption, ink fill, selective halftone, registration mark, handwritten annotation.
- Construction: cut edge, displaced sheet, hinge/tape, fixed disc over rotating disc.
- Motion: drawing a mark, lifting a layer, aligning a subject with a reading position.
- Typography: Fraunces for display, Crimson Pro for reading, Courier Prime for instrument labels. Reuse supports family resemblance but is not sufficient to establish it.

## Syntax — the relationships under test

| Relationship | Current evidence | Boundary / failure |
|---|---|---|
| Layer = a distinct reading of shared geometry | Tracing experiment keeps terrain, occupation and memory separately inspectable | Stacked translucency reduces contrast; keep long prose outside the stack |
| Ink = large graphic form | Blue/coral overprint produces a third region, while body text remains clean | Dark overlap can become muddy; do not texture all text |
| Mechanism = navigation | Volvelle aligns an item to a fixed reading position | Center disc initially hid labels; moving parts must not obscure the index |
| Material = local to its object | Texture follows its sheet; removed sheets change visible composite | Screen-fixed stains/texture can look like monitor defects |
| Movement = a change the reader can explain | Lifting removes a layer; rotation selects another subject | Motion should not be the only way to access content |
| Variation = composition as well as palette | These three candidates differ in organizing behavior and graphic weight | A recolored card grid does not establish a different world |

## What belongs, provisionally

A new component should contribute to a coherent material/structural idea, fit the surrounding scale and contrast, remain readable at actual viewing size, and survive a phone-width composition. Physical realism is optional; physical intelligibility is useful. A paper object can be stylized, exaggerated, or animated beyond literal paper behavior.

These are evaluation questions, not admission rules. A surprising experiment may earn a place by contradicting them convincingly.

## How to continue with another designer or AI

1. Read the owner's feedback in the main notebook before copying a visual recipe.
2. Pick a question, not only a legal subject. Legal content can be placeholder material for a visual test; do not invent doctrine and present it as instruction.
3. Find a concrete source and distinguish what it shows from your adaptation.
4. Record the first hypothesis before rendering.
5. Inspect an actual wide and narrow rendering, exercise the interaction, and record discrepancies between intention and result.
6. Change the relevant variable; note the before/after evidence. Do not fabricate a tidy iteration story.
7. Ask the owner a few specific experiential questions. Record their words separately from your interpretation.
8. Preserve useful techniques from rejected compositions, and keep uncertainty visible.

## Open after this round

No owner verdict on the three new experiments yet. We do not know whether tracing paper is too quiet, the ink palette too saturated, or the wheel too antique / space-hungry. Nor do we know whether these techniques improve repeated study. The experiment can be worthwhile artistically without proving that second claim.

Phone testing used a resized desktop browser with an effective content viewport of 354 CSS pixels. It is not physical iPhone or Safari testing. Reduced-motion CSS is present but media emulation was not exercised in this pass. Cross-document transitions are optional enhancement between the experiment pages, with ordinary navigation as fallback; original subject pages are unchanged.

## Owner correction after round 1

The owner rejected the page-level framing: these were conventional websites showcasing isolated features. The diagrams, overlays and rotating object received praise, but did not establish a coherent analog scene. Translucency read as glass. A decorative edge tab looked clickable and was not. The slider effect was unclear. All new research commentary must be plain English, with less poetic copy. Whole pages should be actual legal notes, with sourced concepts and hypothetical examples, not abstract visual filler. This feedback supersedes any suggestion above that the first three pieces were successful overall compositions.

Round 2 tests a continuous annotated contract and a technical-paper civil-procedure note. Both keep research commentary in a secondary English disclosure and make apparent controls functional. They are pending owner review.


### Pending review: open notebook (06)
Facing pages distinguish a problem from its reasoning; one fold-out expands the comparison. Test whether this quieter, less layered composition still belongs to the family. On narrow screens, page relationships survive as sequential leaves. This has not received owner approval.

## Owner correction — 2026-09-09: reinterpret the information

The first full course was judged visually good but still a styled container for the pack. Source fidelity and comfortable reading are infrastructure, not the intended artistic outcome. The owner explicitly asks to reconsider prose and tables through the medium: relationships may call for drawings, symbols, spatial arrangements, schematics, physical interactions or text, chosen case by case.

Desktop may have a richer composition than mobile; mobile must remain good, but does not dictate a reduced desktop experience. Physical work must be perceptible: local paper textures, ink overprints, pencil/pen, ruler marks, actual layering and attachment. These are a vocabulary to compose with, not a checklist to put on every page. Earlier rejections of dirt-like stains, uniform dotted overlays and incoherent campus-pamphlet collage still apply.

The owner questions the search bar as an assumed website convention. Cross-chapter search does have a different scope from browser find, but that is not sufficient reason to make it a compositional centerpiece. Decide navigation from the study object rather than default website components.

Next design hypothesis (not yet built or approved): unit 08's temporal and causal comparisons can become a single ruled reconstruction with alternative continuations. Position encodes phase; a branch encodes a different cause of non-consummation; annotations preserve conditions and exceptions. Do not imply that elapsed time alone determines commencement of execution. Preserve access to the full original reading, and document which sections the authored visual interpretation covers.

Before implementation, inspect Claude's new parallel work as requested. Remote main has no commits after 6fb5893 at this check; local discovery is underway.

### Inspection and specimen, 2026-09-09

Located Claude's independent work in a temporary scratchpad. Its v3 was inspected in a live browser (unit 10 opening and decision matrix); the v4 build log describes an apostila material vocabulary, but its HTML still referenced the older styles at inspection. Do not report unfinished v4 as a reviewed finished design. No files in Claude's project were changed.

The new `courses/teoria-do-delito/atelier/iter.html` is an authored visual reading of unit 08's selected temporal/causal relationships. It is not a full chapter replacement. Sequence becomes a ruled line; the execution boundary is marked as an argument to establish, not an automatic time threshold. Alternative cases replace an intervention drawing and its legal consequence. A fold-out exposes all comparisons without requiring memory of a previous selection. Separate attached papers hold conditions which must survive the simplification.

Actual corrections: the first result slip obscured the last event and was moved below the diagram; the initial narrow rendering would have shrunk labels too far, so the drawing now retains scale inside an explicitly labeled horizontal scroll area. The mobile phase line reflows vertically. A weapon illustration was replaced with poison/antidote objects in the corresponding hypothetical. Body/diagram text remains in the DOM with a dynamically updated SVG description. Keyboard Enter changed cases and opened/closed the comparison. At 390 CSS px, whole-page scrollWidth equaled viewport width. Desktop and mobile screenshots inspected; physical iPhone/Safari untested.

Bounded source review caught the need to distinguish definitive abandonment from a temporary pause, clarify the agent's causal contribution to preventing death, and qualify the attempted-crime penalty. Those points were added. The legal basis was checked against the official compiled Penal Code (arts. 14–17): https://www.planalto.gov.br/ccivil_03/decreto-lei/del2848compilado.htm . The underlying complete reading is linked; original source files unchanged.

Open for owner review: whether this composition feels materially convincing and whether changing the drawn cause is more useful/expressive than the original table. Do not record the specimen as endorsed. The current richer desktop spread is intentional; mobile rearranges it without determining the desktop layout.

## Material-depth follow-up — 2026-09-10

Owner verdict: the causal reconstruction was a much-needed improvement, but individual components lacked texture, depth and distinct identity; the background remained unresolved. That is approval of the direction, not of every component or a final house style.

This focused pass gives the whole object a bounded blue bookcloth backing on warm white. This replaces the undifferentiated gray-beige surround with a material that supports the paper stack. The blue field is confined to the backing, with fine woven marks and a seam; it is not a new dark theme for the reading.

Material identities: the timeline uses warm drawing stock; the causal diagram has green engineering stock, two grid scales and a quiet fold; the consequence slip has rougher grain, a torn lower edge and transparent tape. The correction uses yellow adhesive paper with a darker attachment band and a lifted edge. The legal interleaf is finer ivory paper with a double blue rule and a second leaf. The exception note is rose ruled stock with separate ink color. Shadow spread follows apparent attachment and distance instead of applying one identical shadow to every rectangle.

Actual corrections: first-render timeline grain was too coarse, so its contrast was reduced. The initial medium-width title was squeezed into awkward lines, so the spread becomes a full-width title at that size. Fixed intermediate timeline column sizing and constrained the diagram to its own scrolling region. At 390 and 780 CSS px the page did not overflow horizontally. Wide rendering inspected at a requested 1440 px. Keyboard Enter changed the case and opened the comparison; the selected label and paper color agree. The full study text and causal script are unchanged; extracted HTML text was compared to the previous commit and matched exactly.

The background is still a design hypothesis awaiting owner judgment. Physical-phone/Safari rendering remains untested. No claim is made that stronger grain automatically improves a component; the differences must remain legible without making the screen appear dirty.