From curious to capable.

What the model actually does. Two hours that replace a year of magical thinking. You leave with the right mental model and your first working API call.

1. § 01.01What an LLM actually isNext-token prediction. Why pattern-plausible isn’t the same as true.Read + lab · 18 min
2. § 01.02The context windowWorking memory. What falls out the back of a long conversation. Interactive slider.Read + lab · 18 min
3. § 01.03HallucinationsWhy they happen. Three defenses that work. When to trust a confident answer.Read + lab · 16 min
4. § 01.04Tokens & temperatureThe two technical knobs you need to understand exactly once.Read + lab · 15 min
5. § 01.05Models vs productsClaude, Code, Cowork, Design, the API — which to reach for, when.Read + lab · 15 min
6. § 01.06Anatomy of a promptContext, goal, constraint. Three ingredients. Most disappointing prompts are missing one.Read + lab · 22 min
7. § 01.07Your first API callInstall the SDK, set your key, run hello_claude.py. Fifteen minutes.Hands-on · 25 min

Prompting as a craft. Seven patterns that compound. Climb the ladder only as high as the task requires.

1. § 02.01The ladder of promptsSix rungs from one-shot to iteration.Read + lab · 18 min
2. § 02.02The Role patternTell the model who to be, and its outputs sharpen.Read + lab · 14 min
3. § 02.03The Examples patternShow, don’t tell. One edge example beats five typical ones.Read + lab · 14 min
4. § 02.04The Format patternControl the shape, control half the usefulness.Read + lab · 12 min
5. § 02.05Step-by-stepAsk the model to think before answering.Read + lab · 12 min
6. § 02.06The Persona patternTeach the model your voice with samples.Read + lab · 14 min
7. § 02.07The Critique patternAsk the model to disagree with you.Read + lab · 16 min
8. § 02.08The Constraint patternTelling it what NOT to do is often more powerful.Read + lab · 16 min
9. § 02.09The prompt labCompose a real prompt with all six ingredients live.Hands-on · 20 min
10. § 02.10Failure modes & their repairSix predictable failures, one-line fixes.Read + lab · 14 min
11. § 02.11Your prompt libraryThe compound asset.Hands-on · 12 min

Tools turn talk into action. The shift from chatbot to agent in one capability.

1. § 03.01What a tool isA function the model can ask you to run. The whole game in one diagram.Read · 6 min
2. § 03.02Declaring a toolName, description, JSON Schema.Read · 5 min
3. § 03.03The tool-use loopEighty lines of Python. The shape of every agent in the world.Hands-on · 25 min
4. § 03.04Structured outputsWhen you want JSON, not prose.Read · 5 min
5. § 03.05The agentic disciplineScope. Preview. Reversibility.Read · 5 min

You walk out with a working agent. Plan, Act, Observe, Reflect — the full loop, end to end.

1. § 04.01The agent loopPlan → Act → Observe → Reflect.Read · 8 min
2. § 04.02A real trace22 seconds, six tool calls, one polished briefing.Read · 6 min
3. § 04.03Build your research agentThe capstone. ~250 lines.Hands-on · 40 min
4. § 04.04Reversibility, firstDrafts before sends. Copies before moves.Read · 4 min
5. § 04.05Where the craft goes from hereMCP. Multi-agent systems. Evaluation. Production.Read · 5 min