writing-learning-roadmaplisted

# Writing a Learning Roadmap ## Overview A learning roadmap is a numbered sequence of atomic docs (`00-overview.md`, `01-*.md`, `02-*.md`, ...), each covering exactly one concept, that takes a reader from foundations to a working end-to-end system in a topic. Every doc follows the same fixed template so readers know what to expect and authors can't skip a required part. ## When to use - User asks for a "roadmap", "curriculum", "learning path", or "step by step guide" for a technical topic (a library, a technique, a subfield). - User references an existing roadmap (e.g. `gpt/docs/roadmap/`) and wants the same structure for a new topic. - Not for: a single tutorial/README, an API reference, or a one-off explanation — this skill is specifically for a *multi-doc, ordered, atomic* curriculum. ## Process 1. **Find the audience's starting point.** Ask (or infer from context) what the reader already knows and what they're missing — the roadmap bridges exactly that gap, no more. State it in one line at the top of the overview doc ("Audience: ... but haven't ..."). 2. **If an existing roadmap is referenced as a model, read 2-3 of its docs first** (the overview + one early + one late doc) to match its exact template shape, tone, and section headers — don't reinvent the template. 3. **Decompose the topic into phases**, each phase a coherent theme (3-5 docs), ordered so each phase's docs only depend on concepts from earlier phases. Write this as a table in