Building with Claude
Going from asking Claude questions to building something that runs — prompts, tools, retrieval, and knowing when the answer cannot be trusted.
The distance between a good conversation with Claude and a working application is mostly engineering, not prompting. This course covers both, in that proportion.
Claude is the model this lab builds on and partners with, so the examples are the ones we actually use.
Every lesson is free and needs no sign-up.
Syllabus
How the model behaves
What it is good at, where it is confidently wrong, and how to design around that rather than hope.
Prompting properly
Instructions, examples, structure, and why "be accurate" does nothing.
Calling the API
Your first request, messages, system prompts, streaming, and what each parameter actually changes.
Giving it tools
Letting the model call your code, and the difference between a tool that reads and one that writes.
Answering from your own documents
Retrieval, chunking, citations, and refusing to answer when the documents do not say.
Shipping it
Evaluating output, cost, latency, failure modes, and what a human still has to approve.