What you will be able to do
Explain Claude in plain termsModel, tools, and context. Spot when a miss is wrong context, wrong tool, or missing verification.
Open Claude with intentUse modes, the model picker, file attach, connectors, and research without hunting for controls.
Route work to the right surfaceChoose web chat when you need outside context, and Desktop Cowork when the job lives in a local folder.
Pick the right model for the jobCompare Haiku, Sonnet, and Opus on one brief, and keep the cheapest model that clears your quality bar.
Make work compoundUse Projects, Artifacts, and reusable briefs so the next chat starts smarter than the last one.
Connect tools safelyWire connectors and Cowork with least privilege, and recognize plumbing vs judgment failures.
Ship a skillTurn a repeatable workflow into a skill you can trigger again next week.
Decide what not to shareApply permission, stakes, verifiability, and control before a sensitive task enters the chat.
Proof points from the labs
These are the concrete checkpoints we use during the session. If you can do these, the goals above are not abstract.What “done” looks like tomorrow
You are done with this session when you can sit down on Monday and do all four without opening a cheat sheet:- Pick web chat or Desktop Cowork for the task in front of you.
- Choose Haiku, Sonnet, or Opus on purpose.
- Attach the right files or folder, and say what Claude still cannot see.
- Check the output against a real standard before it leaves your desk.
Stretch, not required today
If you finish early or come back later, push into Projects, Skills, Connectors, MCP, Agents, and the Resources shelf. Those deepen the same operator loop. They are not required to call today a success.Start with foundations
Begin with the mental model, then open Claude and do the first lab.