issues/15-progress.md
Phase 15 Progress Report
Phase 15 Goals
"Chromatic Encoding — Text as Color, Collection as Palette, Arrangement as Art"
Phase 15 transforms semantic embeddings into colors, turning each poem/passage into a pixel. Artists arrange these pixels into images, and the arrangement itself becomes a reading order — the image IS the navigation sequence, frozen in color.
From Previous Phases
- 768-dimensional embeddings for all poems
- Cosine similarity infrastructure
- Semantic color assignments (limited, to be expanded)
- Static HTML generation pipeline
Phase 15 Objectives
- Design embedding→color mapping algorithms (PCA, anchor-based, modular overflow)
- Generate color palettes representing entire collections
- Create default arrangements (linear, similarity-chain)
- Implement bidirectional encoding: image↔reading sequence
- Enable artists to compose images from poem-pixels
The Core Inversion
| Traditional | Chromatic Encoding |
|---|---|
| Images illustrate text | Text constitutes the image |
| Reading is temporal | Reading is spatial |
| The author sequences | The artist arranges |
| Navigation is abstract | Navigation is visible |
Phase 15 Issues
Active Issues
| Issue | Description | Status | Priority |
|---|---|---|---|
| 15-001 | Chromatic Text Encoding System (Vision) | Open | High |
Sub-Issues (To Be Created)
| Sub-Issue | Description | Status |
|---|---|---|
| 15-001a | Design embedding-to-color mapping algorithms | Planned |
| 15-001b | Generate poem color palette | Planned |
| 15-001c | Create default arrangements | Planned |
| 15-001d | Implement image→sequence decoder | Planned |
| 15-001e | Build encoded reading page generator | Planned |
| 15-001f | Interactive palette arrangement tool | Future |
Completed Issues
None yet.
Key Concepts
Embedding → Color
Each 768-dimensional vector becomes RGB:
poem "silence" → [0.23, -0.15, 0.87, ...] → RGB(142, 87, 203) → █Mapping Strategies
| Strategy | Method | Character |
|---|---|---|
| Naive | First 3 dims | Fast, ignores most data |
| Folded | Sum groups of 256 | Uses all dims, loses structure |
| PCA | Project to 3D | Preserves variance, similar = similar color |
| Anchored | Similarity to R/G/B concepts | Semantic meaning → color meaning |
| Modular | Overflow arithmetic | Unexpected relationships |
The Palette
~7,800 poems = ~7,800 pixels = a 89×88 image
Each pixel is a poem. The collection is a visual artifact.
Encoded Reading Order
The spatial arrangement of pixels defines a reading sequence:
Image pixels: [P42, P17, P89, P3, ...]
Reading: silence → fire → memory → ...The image IS the navigation. Share the image, share the reading experience.
Completion Criteria
- [ ] At least one embedding→color mapping algorithm implemented
- [ ] Poem color palette generated (
palette.json,palette.png) - [ ] Default arrangements created (linear, similarity-chain)
- [ ] Image→sequence decoder working
- [ ] Encoded reading pages generated
- [ ] Round-trip verified: poems → image → reading sequence
Phase Status: OPEN
Created: 2026-01-28
Cross-Phase Dependencies
Depends on:
- Phase 1-8: Embedding infrastructure
- Optionally Phase 14: Book passages could also become pixels
Enables:
- Visual representation of literary collections
- Artist-curated reading experiences
- A new form of literary art: the chromatic arrangement
Related Documents
- 15-001: Chromatic Text Encoding System (vision issue)
assets/embeddings/embeddinggemma_latest/poem_colors.json— Existing color assignmentssrc/similarity-engine.lua— Cosine similarity for anchor mapping- Phase 14: Narrative Tapestry (complementary approach)
Philosophical Note
The artist who arranges the pixels is not illustrating the work. They are editing it — choosing which poem follows which, creating a new path through meaning. The image is their editorial signature, frozen in color.