Academics
How LLMs actually work
A free series written for students rather than buyers: what a model does with your text, why context costs what it costs, and where the engineering decisions behind this product come from. Each lesson ends with a short quiz.
- Lessons published
- 6
- Reading, all lessons
- 46 min
- Planned
- 10
- Cost
- Free
Lessons
quiz progress saves on this device only
- 01Tokens & tokenizationHow raw text becomes token IDs, and why that shapes context limits and cost.7 minQuiz
- 02How transformers readEmbeddings, attention, and the forward pass, explained without assuming an ML background.9 minQuiz
- 03Training at a glanceWhat pre-training and fine-tuning do — and what they can't fix.8 minQuiz
- 04Inference & the KV cacheWhat happens per token at serving time, and why session shape matters.8 minQuiz
- 05Prompt caching in practiceStable prefixes, append-only sessions, and reading hit-rate headers on our API.7 minQuiz
- 06Evals, safety & limitsHow models are measured, where they fail, and how to use them responsibly.7 minQuiz
Planned modules
Titles and one-line scope only. Nothing here is written yet, and we would rather show you an empty shelf than a stub.
- 07Prefill, decode & serving economicsQuantization, batching, and speculative decoding — where serving cost really goes.Not written
- 08Agents that actTool calling, RAG, and multi-agent patterns — plus supervising agents with real permissions.Not written
- 09Reasoning models & distillationWhat test-time reasoning buys, and how small models inherit big ones' skills.Not written
- 10Fine-tuning for your domainWhen prompting stops being enough: SFT data, evals, and honest expectations.Not written
Here to ship rather than to study?
The API docs cover the fastest path to a working request. The caching rules in there are the same ones these lessons explain.