Suggested Reading Order

Three paths through the primer. Pick the one that matches your role and time budget, then use the tables below to navigate. All paths start at the Welcome page.

---

PM / Product Path (~3–4 hours)

For product managers, product strategists, and technical leaders who need fluent understanding without deep implementation knowledge.

#	Page	Time	Notes
1	Welcome	5 min	Start here
2	What Is an LLM?	15 min	Foundation
3	Tokens & Tokenization	15 min	Explains billing
4	Embeddings	15 min	Powers semantic search
5	Transformer Architecture	20 min	How it works
6	Attention Mechanism	15 min	Why context matters
7	Training Pipeline	25 min	SFT, RLHF explained
8	Context Windows	20 min	Memory & cost
9	Mixture of Experts	15 min	Why MoE matters
10	RAG	20 min	Private knowledge
11	Agents & Tool Use	20 min	Agentic systems
12	Cost & Latency	20 min	Budgeting AI
13	Evaluation & Benchmarks	15 min	Reading leaderboards
14	Agentic Vocabulary	25 min	Terminology
15	Tools Landscape	30 min	Reference

---

Engineer Path (~6–8 hours)

For software engineers, platform engineers, and ML engineers who want deep conceptual grounding plus implementation context.

#	Page	Time	Notes
1	Welcome	5 min
2	What Is an LLM?	15 min
3	Tokens & Tokenization	15 min
4	Embeddings	15 min
5	Transformer Architecture	20 min	Read full paper abstract
6	Attention Mechanism	15 min
7	Model Architecture Types	15 min
8	Training Pipeline	25 min	Read InstructGPT abstract
9	Context Windows	20 min
10	Mixture of Experts	15 min
11	Scaling Laws	15 min
12	RAG	20 min
13	Agents & Tool Use	20 min
14	Multimodal	15 min
15	Evaluation & Benchmarks	15 min
16	Cost & Latency	20 min
17	Prompt Engineering	20 min
18	Agentic Vocabulary	25 min
19	Tools Landscape	30 min	Reference

---

Researcher Path (~12–15 hours)

All pages in sidebar order plus full paper reading. For each paper cited in a page, read the full paper — not just the abstract. Use the Papers reference as your guide.

Work through every page in the order they appear in the sidebar: Basics → Intermediate → Advanced → Use Cases. When a page cites a paper, follow the link and read according to the PM reading guide at minimum; for the researcher path, read the full paper.

Estimated total: 12–15 hours depending on paper reading depth. The scaling laws papers and the transformer paper in particular reward slow reading — both are dense with implications that take time to absorb.