Open slides full screen · Download PDF (click the slides once, then use the arrow keys)
What we covered
- How the course works: 8 to 12 weekly sessions of one hour, homework, discussion afterwards. What you will take away is listed on the About page.
- AI today is a jungle of names. Google, OpenAI, Anthropic, xAI and Meta make the big models: Gemini, ChatGPT, Claude, Grok and Muse.
- Why we picked Claude Code: it is an agent with a strong harness and many tools, and most of us already have access to it. One caveat: it does not generate images.
- Chat, Work and Code: three ways to use the same AI. Chat is an advisor on the phone. Work is a contractor you hand a folder. Code is a colleague sitting next to you with access to everything you have.
- The Roadmap for the coming weeks.
- Demo: Claude Code in different windows, folders and projects, tools, models, context and tokens.
Key takeaway
Once you are comfortable with the “Code” level of working with AI, every other AI product feels simple, because they are all built from the same parts: a model, a harness and tools. See Mental models.
Homework
Install Claude Code and try it. Do not be afraid of the terminal or of the VS Code extension; they are the same agent in a different window, and you can switch any time.
- Read Choosing a client and pick one. Undecided? Windows: the terminal via the Terminal app, or VS Code if you like seeing your files. Nervous: the desktop app.
- Install it: Install Claude Code on Windows, Install Claude Code on Mac, Claude Code in VS Code or Claude Code in the desktop app.
- Do Your first session: a folder, three files, three questions, one small change. Twenty minutes.
- Bring one answer to the next session: what surprised you?
Stuck? Each guide ends with the errors people actually hit. Otherwise bring the laptop.
If Claude stops and mentions a limit
Nothing is broken and nothing is lost. Paid plans include an amount of work per five hours; see Usage limits for what resets when, and start a fresh conversation with
/clearbetween unrelated tasks.
Terms from this lecture
Model · Agent · Harness · Tool · Context window · Token · Claude Code · Client · Terminal · Chat, Work and Code · Project · Permission mode · Usage limits