What 'real' means

1. Principle

Most things you ship with AI are not real. They are demos that survive a refresh.

Real means consequences leave the session:

• Data persists — what someone typed is still there tomorrow, on someone else's device
• Money moved, OR a decision someone acted on, OR a record written somewhere that matters
• The system changes state that did not exist before the user touched it

A page that calls a model and shows a result is not real. A page that calls a model, stores the result, and lets someone return to it tomorrow, is real.

2. Do (you try)

Take your project list from unit 0.1. For each row that is "shipped," apply the refresh test.

3. Verify (how you know)

The "real" count is almost always smaller than the "shipped" count. Often by a lot.

Two questions to ask the rows where you wrote "real":

1. If your laptop fell into the sea right now, does the data still exist?
2. If a user opens the project on their phone in a week, can they pick up where they left off?

Two "yes" answers and the row is real. One "yes" and the row is a demo that happens to persist briefly.

4. Reference