Document freshness: 2026-03-12 • v05

This iteration adds a documentation routing rule: future sessions should keep a high-level mental map of what kinds of material belong in which documents, and only reopen and modify the affected documents rather than rereading the full corpus every turn.

Developer Guide

Last updated: 2026-03-12 v04

developer_en.html

This document describes how developers should understand, extend, and maintain the MyAnythingList system.

1. Development Philosophy

Future developers should not need historical chat logs to understand how the system works.

2. Core System Components

Playlist text
   ↓
URL discovery
   ↓
Command parsing
   ↓
Tile model creation
   ↓
Thumbnail / QR resolution
   ↓
Grid rendering

3. Playlist Parser

The parser extracts valid URLs from any UTF-8 text.

Rules: Example:
My research notes:

https://example.com/article

#_ReplaceThumbnailWithImage("cover.jpg")

4. Command System

Commands extend playlist behavior.

Examples:
#_ReplaceThumbnailWithImage("image.jpg")
#_LoadImage("panel.png")
#_LoadImageHideLinks("poster.png")
Commands apply to the most recent URL unless otherwise defined.

5. UI State Architecture

Startup rendering order

Developers should optimize initial work in this order: visible tile shell, visible QR and text geometry, visible thumbnail or URL art, then off-screen work. A visible tile may continue improving after first paint, but it should never begin in an obviously wrong state.

Runtime state is controlled by configuration variables.

Examples: These values may be set through:

6. Thumbnail System

URL art implementation discipline

Thumbnail sources are prioritized:
Command thumbnail
→ Uploaded thumbnail
→ Platform thumbnail
→ Fallback image
Rules:

7. Contribution Guidelines

Documentation routing map

Future sessions do not need to reread every documentation file on every prompt. Instead, they should keep a stable high-level map of document purpose and route each new requirement only to the affected documents.

Selective update rule

When a new prompt introduces a relevant requirement, the assistant should identify the correct document category first, then update only those documents that are actually affected. This avoids unnecessary rereading while preserving full prompt capture in the knowledge base.

The assistant should still maintain continuity with prior requirements, but full-document rereading is not required when the document-purpose map already makes the correct destination clear.

8. Long-Term Goal

The project aims to remain understandable to future developers and AI systems for decades.