Session 1 gave you the loop. Today you get precise about it — and you pitch the project you’ll be graded on by your peers at the showcase.
Three things:
Important scheduling note: your capstone has to be finished by the end of Session 3. Session 4 is 60 minutes and it’s entirely demos — there is no build time. Plan accordingly, which mostly means picking something smaller than you want to.
| Block | |
|---|---|
| 1 | Warm-up: four ways to prompt the same app |
| 2 | How much to ask for at once |
| 3 | Prompt lab — diagnosing and rewriting bad prompts |
| 4 | Break |
| 5 | How to check the AI’s work — the four checks |
| 6 | Build a practice project, spec written first |
| 7 | Swap machines, break your partner’s app |
| 8 | Capstone pitch + scope check |
| 9 | Capstone: plan and first build |
cd $HOME\Documents\Projects
cd <Name>
mkdir myproject
cd myproject
claude
<Name> is your name — the folder you made in Session 1. Your name folder is
where all your projects live. One command per line — 5.1 doesn’t support &&.
Run it: python app.py
A: make a study timer
B: make a pomodoro timer in python
C: Build a Pomodoro timer in Python with tkinter. 25-minute countdown,
big digits, Start/Pause/Reset buttons. Beep when it hits zero.
Save as pomodoro.py
D: Build a Pomodoro timer in Python with tkinter using a class-based
architecture. Use threading.Timer for the countdown, not time.sleep.
Store state in an enum. Label in monospace at size 48 using grid
geometry, buttons in a frame below using pack. Beep with
winsound.Beep at 1000Hz. Type hints and docstrings on every method.
C is right. But understand why D is wrong, because it’s the mistake people make once they get comfortable:
Every implementation detail in D is a decision you now own. If threading.Timer is
the wrong call for this, that’s on you, not the AI. You also spent five minutes writing a
prompt to save two minutes of work.
Specify what you care about. Let the AI decide the rest.
D becomes correct when you have a real constraint — matching existing code, a required library, a rubric. Not before.
Starting from nothing → describe the whole app. You want a coherent first version where the pieces fit together.
Changing working code → one thing at a time.
5 changes → run → broken → which one? could be any. could be two
interacting. you're debugging blind.
1 change → run → broken → it's that one. 30 seconds.
Not caution — speed. Every batch of untested changes is a debugging session you’re saving up for later, with interest.
Size test: can you describe what should be different in one sentence?
| Too big | Right size |
|---|---|
| “Add multiplayer” | “Add a second score counter labeled Player 2” |
| “Make it look better” | “Make the background dark gray, text white” |
| “Add saving” | “When I click Save, write the list to tasks.txt” |
Some changes genuinely come as a set. Adding a second player needs a score and a turn indicator and turn-switching. Fine — but name the pieces explicitly:
Add two-player mode. That means: a second score counter, a label showing whose turn it is, and turns alternating after each roll. Those three things only.
Naming them yourself gives you a checklist to verify against afterward.
| Level | What you did | What it proves |
|---|---|---|
| 0 | It generated without an error | Nothing |
| 1 | It ran without crashing | The syntax is valid |
| 2 | You used it and it looked right | Works for the happy path |
| 3 | You tried to break it and couldn’t | Probably holds up |
| 4 | You read the code and it does what you asked | You actually know |
Most people stop at level 1 and believe they’re at level 4. That gap is where every disaster lives.
1. Does it do the thing? Not “an answer” — the right answer. Do the math yourself and compare.
2. Does it survive the wrong thing? Empty input. Wrong types. Negatives. Zero items. A thousand items. Double-clicks. This is where AI-written code fails most, because the AI optimizes for the case you described.
3. Did it change anything you didn’t ask for?
What did you just change? List every change.
Then check that list against what you asked for. Real failure mode: you request a color change, it also rewrites your scoring logic.
4. Can you explain it?
Explain what the function on line 40 does, like I'm 12.
If you can’t say what a chunk of your own code does, you’re at level 1 pretending to be level 4. This is also the fastest way to actually learn to code while building.
What could go wrong with this? What inputs would break it?
I expected X but got Y. Why?
Is there a simpler way to do this?
The catch: the AI checking its own work is a hint, not a guarantee. It’s the same guesser that wrote the bug. Use it to surface candidates, never to prove absence.
def average(scores):
return sum(scores) / len(scores)
Check 1: passes. Check 2: empty list → ZeroDivisionError.
It didn’t make a mistake. It answered exactly what you asked. You left out the question.
Before typing anything:
APP NAME:
WHAT IT DOES (one sentence):
MUST HAVE (3, no more):
1.
2.
3.
WHAT IT LOOKS LIKE:
HOW I'LL KNOW IT WORKS (3 specific tests):
1. If I ______, it should ______
2. If I ______, it should ______
3. If I ______ (something wrong), it should ______
You’re writing the tests before the code exists. Test 3 must be a wrong-input test.
Keep this small — it’s practice, not the capstone.
Mark your tests honestly at the end. A failed test you can name beats a passed test you didn’t really check.
Switch machines. Break your partner’s app. No reading their code — use the app only.
Empty everything · wrong types everywhere · 0, negatives, 999999, paragraphs · everything clicked twice and out of order · window shrunk to nothing.
Write each finding as a proper bug report:
What I did · What I expected · What actually happened
Then swap back and fix the top one.
Now notice something: paste that bug report straight into the AI as your prompt. It works. A well-written bug report and a well-written prompt are the same document, and that’s not a coincidence.
Your partner finds in five minutes what you missed in forty — because you were trying to make it work and they were trying to make it fail. Both are necessary. Companies hire people whose entire job is the second one.
I'm building ______.
You use it to ______.
The three things it has to do:
1.
2.
3.
I'll know it works when:
- If I ______, it ______
- If I ______ (wrong input), it ______
Your partner asks you two questions:
Almost everyone should cut one. Then an instructor signs off.
You have the rest of today plus Session 3. That’s it.
| If you pitched | Build this instead |
|---|---|
| Multiplayer online game | Two players, same keyboard |
| Anything with user accounts | One user, saves to a text file |
| Website with a database | One page, data in a JSON file |
| “A whole RPG” | One battle, three enemies |
| Machine learning anything | Rules-based version of the same idea |
| A Discord/Instagram bot | Same logic, local app, no API |
Cutting scope isn’t giving up. It’s the actual skill. A small thing that works beats a big thing that doesn’t — today, and in every job you will ever have.
cd $HOME\Documents\Projects
cd <Name>
mkdir capstone
cd capstone
claude
Four steps, in order:
1. Ask for a plan, not code.
I want to build [your app]. Don't write any code yet. Tell me what
files you'd create, what goes in each one, and what order you'd
build them in.
2. Argue with the plan. Change at least two things. Combine files, drop a feature, reorder the build. Changing a plan is free. Changing code is not.
3. Save it as PLAN.md in your project folder.
4. Build component 1 only. Run it. Check it.
cd $HOME\Documents\Projects
cd <Name>
Copy-Item -Recurse capstone capstone-working
That’s a complete copy of your folder, frozen at a moment it worked. You’ll learn the full system next session. For now: it works, so copy it.
How much to ask for:
New app → the whole thing. Existing app → one change at a time. If you can’t say what should be different in one sentence, it’s too big.
How to check:
- Does it do the thing?
- Does it survive the wrong thing?
- Did it change something I didn’t ask for?
- Can I explain it?
Your project is on this computer. Save a copy to your Google Drive folder:
Your project gets too big for one file, and you finish your capstone. Session 4 is demos only.
This Vibe Coding Curriculum by Brian Bird, created in with AI assistance, is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.