Wavemind

/wavemind visualize <artifact-id>

Generate a living memory document from a stored thinking artifact.

Steps:

1. Read the artifact from data/artifacts/<id>.md
2. Analyze the thinking artifact. Your job is editorial, not generative:
   • Identify distinct rounds/sections of the conversation
   • Extract punchline quotes (memorable original words from each speaker)
   • Mark pivoting moments (where thinking shifted direction)
   • Clean up the transcript: remove filler, fix noise, preserve original words
   • Do NOT summarize, generate insights, or create new content
3. Generate a structured analysis as JSON (see Analysis Format below)
4. Generate a self-contained HTML file following the NoteBlock editorial layout:
   • Start with "What Came Out of This" actionables section right after the header (why it matters + action items)
   • Then the timeline with each section: header, dialogue bubbles, punchline quote callout, expandable transcript
   • First-person perspective. It's the user's living memory, not a third-party report
   • Dialogue format with speech bubbles (user = white left-aligned, other speakers = dark right-aligned)
   • Progressive disclosure: punchline visible, full transcript behind "Read original" toggle
   • The HTML must be fully self-contained (inline CSS/JS, no external dependencies)
5. Save to data/visuals/<id>.html
6. Report the file path

/wavemind list

Browse all stored thinking artifacts and their visualization status.

Steps:

1. Read data/index.json
2. Display a formatted table:

ID                          | Title                  | Rounds | Date       | Status
20260327-zai-prep           | ZAI Ambassador Prep    | 11     | 2026-03-27 | visualized
20260329-design-evolution   | Design Evolution       | 6      | 2026-03-29 | not visualized

3. If no artifacts exist, say "No artifacts captured yet. Run /wavemind capture to start."

Analysis Format

When analyzing a thinking artifact, produce this structure:

{
  "title": "Title from the user's perspective",
  "date": "2026-03-27",
  "participants": "Lily + Growth",
  "sections": [
    {
      "round": 1,
      "header": "High-signal title from user's perspective",
      "quote": "Memorable speaker quote, max 10 words",
      "dialogue": [
        {"speaker": "lily", "text": "One thought per bubble. Keep it short."},
        {"speaker": "lily", "text": "", "image": "screenshot.png", "image_alt": "Terminal color comparison"},
        {"speaker": "lily", "text": "Here's what it looked like after the fix:", "image": "after-fix.png", "image_alt": "Terminal after color fix"},
        {"speaker": "growth", "text": "Response in their original words."}
      ],
      "is_pivoting_moment": false
    }
  ],
  "actionables": {
    "why_it_matters": "One paragraph explaining why this conversation mattered and what shifted.",
    "items": [
      "Concrete next step 1",
      "Concrete next step 2",
      "Concrete next step 3"
    ]
  }
}

Key rules:

• header: High-signal, from the user's perspective. BAD: "Discussion of Opportunity." GOOD: "I'm the Orchestrator, Not the Promoter."
• quote: A real speaker quote, not AI-generated. The "memory hook," what you'd remember a week later.
• dialogue: The actual back-and-forth, one thought per bubble. Use original words (including mixed Chinese/English). Clean filler but don't rewrite. Entries with an image field render as an <img> tag inside the bubble instead of (or alongside) text.
• is_pivoting_moment: True only when thinking genuinely shifted direction. Not every round is a pivot.
• actionables: Always include. why_it_matters is a single paragraph explaining significance. items are concrete, specific next steps. Not vague ("think about X") but actionable ("build X", "pair Y with Z", "test A").

HTML Visual Guidelines

When generating the HTML visualization:

• Style: Clean, editorial. Think newspaper or literary journal, not dashboard.
• Color palette: Cream background (#F9F8F6), charcoal text (#1A1918), gold accents (#CBA16E).
• Typography: Serif for headers and quotes (Playfair Display or Georgia), sans-serif for UI (Inter or system). Generous whitespace.
• Layout (NoteBlock model):
  • Header: Title (large serif), date, participants, "Living Memory" label with gold dot
  • Actionables (right after header): "What Came Out of This" with two-column grid: "Why It Matters" + "Actionables" list with gold bullet dots. This goes at the top so readers get the takeaway first.
  • Timeline: Vertical gold line at ~28% width. Each section is a grid row.
  • Left column (28%): Round label, punchline quote callout (large gold open-quote mark, bold italic serif)
  • Right column: Section header (bold serif), dialogue bubbles, "Read original" toggle
  • Dialogue bubbles: User = white with light border, left-aligned. Other speakers = dark (#2C2C2C), right-aligned. One thought per bubble.
  • Images in bubbles: When a dialogue entry has an image, render it as <img src="../artifacts/filename.png"> inside the bubble. Style with max-width: 100%; border-radius: 8px; margin-top: 8px; cursor: pointer. Add a lightbox overlay so clicking the image shows it fullscreen with an Escape-to-close handler. Use direct file paths (same approach as frontend-slides), not base64 embedding.
  • Pivoting moments: Gold-filled timeline dot + "Pivoting Moment" badge
  • Progressive disclosure: "Read original" expands to full clean transcript
  • Footer: "WaveMind · Captured [date] · Visualized [date]"
• Self-contained: All CSS and JS must be inline. No external dependencies (except images, which are referenced by relative path).
• Responsive: Hide quote callouts on mobile, switch to single-column layout.

Reference: See data/visuals/glm-benchmark.html for an approved example (GLM Aesthetic Benchmark).

Data Directory

Runtime data lives in data/. New artifacts and visuals are gitignored, but seeded examples are checked in for reference.

data/
├── .gitignore       # Ignores runtime data, keeps seeded examples
├── index.json       # Artifact registry (created at runtime)
├── artifacts/       # Raw markdown files (glm-benchmark.md seeded)
└── visuals/         # Generated HTML files (glm-benchmark.html seeded)

Index Format

index.json is an array of artifact entries:

[
  {
    "id": "20260327-zai-prep",
    "title": "ZAI Ambassador Prep",
    "source": "brainstorm",
    "tags": ["zai", "ambassador", "strategy"],
    "rounds": 11,
    "word_count": 3200,
    "created_at": "2026-03-27T00:00:00Z",
    "file": "artifacts/20260327-zai-prep.md",
    "visualized": true
  }
]