Claude TutorialA Claude case study
Same model. Four places it works.
AI products are no longer one chat box. Education has to teach the map. This lab uses Claude — Chat, Cowork, Code, Chrome — as the worked example, and one Flappy Bird build as the thread. Twenty minutes if you take it. Five if you read how it was designed. Turn on Voice in the header, or hit Listen on a section, and the lab talks you through.
Chat
Think with you
Use Chat when the work is still thinking. You stay in the conversation: draft, debate, rewrite, decide. Artifacts hold a version beside the thread. Projects and memory keep context for the next chat. If the only thing you need back is a sharper thought or a draft you will finish yourself, stay here.
- Output: a decision, outline, or draft
- You want to steer every turn
Cowork
Delegate the work
Use Cowork when you can describe a finished piece of knowledge work and walk away. Grant a folder. Ask for a one-pager, a deck, a spreadsheet with working formulas, or a research brief. Claude plans, runs, and leaves files. Same agentic engine as Code — no terminal, not a git repo.
- Output: a document, deck, or spreadsheet
- Files in, a deliverable out
Code
Change the repo
Use Code when the output has to live in a repository. Open the terminal (`claude`), the desktop Code tab, or the IDE extension. Claude reads the repo, edits files, runs tests, commits, and opens a pull request. If you would not put the result in git, do not open Code.
- Output: a commit, fix, or pull request
- The work already lives in a repo
Chrome extension
Act on the live web
Use Claude in Chrome when the work is on a live page. The side panel sits beside the tab: read, click, type, draft a reply. Cowork and Code can drive the same extension — from the terminal, `claude --chrome` or `/chrome`. If there is no URL, Chrome is the wrong surface.
- Output: something that happened on a page
- A live tab or localhost has to be open