Lesson 1 · Decision skill
Chat is a surface.
Work is a loop.
Choose the smallest AI environment that can complete the real job reliably.
What changed?
ChatGPT and Gemini made AI useful through a familiar rhythm: ask, receive, refine. That rhythm remains excellent for explanation, drafting, brainstorming, and contained analysis.
Agentic environments such as Codex and Claude Code add a different capability: they can work inside a chosen project, read files, use tools, make changes, and run checks. The important shift is not “a smarter answer.” It is a longer, supervised loop.
Use chat for a contained answer. Use a workspace when the result depends on context, tools, multiple steps, or proof.
Focused chat
- Explain a concept
- Draft from supplied text
- Brainstorm options
- Transform one artifact
Agentic workspace
- Read several project files
- Use tools or connected systems
- Create and revise outputs
- Run a check and iterate
Exercise · Route the work
Which surface fits each task?
Choose based on the shape of the job—not the brand you prefer.
No files, tools, or external action are required.
The output must preserve approved figures and reconcile conflicts.
You can evaluate the options yourself in one pass.
The job involves a file, repeated operations, and a checkable result.
Have a question? Ask your AI teacher to compare one of your real tasks with the four scenarios above.