0.3 The roadmap
Three routes depending on your time. All use the same chapters.
Route A — "I want to understand LLMs" (2 weeks, reading only, ~1 h/day)
| Day | Read | Goal |
|---|---|---|
| 1 | Part 0, Part 1 | Vocabulary, the family tree |
| 2 | Part 2 (ch 5–7) | Why hand-written rules failed |
| 3–4 | Part 3 (ch 8–13) | Learning from examples; n-grams |
| 5–6 | Part 4 (ch 14–17) | Neural networks, gently |
| 7 | Part 5 (ch 18–21) | Word meaning as vectors |
| 8–9 | Part 6 (ch 22–26) | The Transformer |
| 10 | Part 7 (ch 27–30) | Tokens |
| 11–12 | Part 8 (ch 31–35) | What an LLM is |
| 13 | Part 9 (ch 36–39) | How it answers you |
| 14 | Part 11 (ch 44–49) | Using LLMs well |
Route B — "I want to build with LLMs" (6 weeks, ~1–2 h/day)
Weeks 1–2: Route A, but do every Practice section. Week 3: Part 7 + Part 11 (tokens, prompting, context) + Project 3 (RAG). Week 4: Part 10 ch 40–41 — Project 1: train a tiny LLM. Week 5: Part 10 ch 42 — Project 2: fine-tune with LoRA. Week 6: Part 11 ch 48–49 + Project 4: your own app with tests.
Route C — "I want to go deep / change careers" (12 weeks, ~2 h/day)
| Week | Focus | Deliverable |
|---|---|---|
| 1 | Part 0–1 + Python/maths warm-up (§0.5) | Environment working, notes.md started |
| 2 | Part 2–3 (rules, classical ML) | A spam classifier you built |
| 3 | Part 3 (ch 10–13) | An n-gram text generator |
| 4 | Part 4 (neural nets) | A neural net trained from scratch in NumPy |
| 5 | Part 5 (word2vec) | Word vectors trained + explored |
| 6–7 | Part 6 (Transformer) | Attention implemented by hand; tiny transformer trained |
| 8 | Part 7 (tokens) | Your own BPE tokenizer |
| 9 | Part 8–9 (LLM internals + inference) | A written explainer you publish |
| 10 | Part 10 (build) | A tiny LLM trained on a book |
| 11 | Part 10 (fine-tune) | A LoRA fine-tune of a small model |
| 12 | Part 11–12 | A deployed RAG app with an eval suite |
Then: read the companion reference guide, read the original papers (Appendix E), and pick a specialisation (training, serving/infra, applications, evaluation, safety).
Beyond all three routes: Parts 12–16 continue the same core arc — agents, MCP, guardrails, and an end-to-end capstone project — budget another 3–4 weeks if you want to go all the way to a working, guarded agent. Parts 17–19 are optional deep-dives, not a fourth route: model compression, agent frameworks and governance, and the internals (AdamW, GQA, quantization) behind things you'll have already used by then. Read whichever one is relevant, in any order, whenever you need it — none of them are required to say you've finished this guide.