The idea behind this lab
How do you know a request is clear? Not by reading it. You wrote it, so it is clear to you.
Here is a real test:
Give the same request to two people who cannot talk to each other. If they build the same
thing, the request was clear. Wherever they differ, you left a decision open.
Two fresh chat windows are your two people. They cannot see each other. They will not warn
you when they guess.
What you need
Two separate, fresh chats. Two different tools is even better, if you have them. No agent,
no installation.
Round 1 — the vague version
Send this to both:
Write a function that decides when a flashcard should next be reviewed.
Now compare the answers.
| Chat A | Chat B | |
|---|---|---|
| What does the function take as input? | ||
| What does it return? A date? A number of days? | ||
| How many difficulty choices does it allow? | ||
| What does it do with a brand new card? | ||
| Is there any smallest or largest gap? | ||
| Could you swap one for the other in a project? |
They will not match. They may not even be the same shape of thing.
Round 2 — the clear version
Now send both chats the SCHEDULING RULES and EXAMPLES sections from your plan in Lesson 2.5,
with this line above them:
Write a TypeScript function that follows these rules exactly.
Fill in the same table again.
What you should see
The second round matches closely. Not word for word, but the shape, the inputs, the outputs
and the behaviour all line up. Either one could be used in the project.
That is what a good request does. It does not just get you a better answer. It gets you an
answer that does not depend on which tool you happened to use, or which day you asked.
The part that matters most
Look carefully at where round 2 still differs.
Maybe one rounds 2.5 days to 3 and the other to 2. Maybe one treats a card due today as due
and the other does not. Whatever it is, you have found a real hole in your plan, and you
found it in five minutes instead of in Module 4.
Fix your plan document now. Add the line that closes the gap. This is the loop:
write it → give it to two chats → find where they differ → close that gap → repeat
You will use this loop for the rest of your working life. It works on specifications for
people, too.
Finish this before Module 3
- Run both rounds and fill in both tables.
- Find at least one difference in round 2 and fix your plan to remove it.
- Save your updated plan. Module 3 begins with an agent, a real project folder, and this
document.