You have built with it for two days. Today you stop using it like a beginner. It starts truly working. For you.
Not answering your questions. Reading your files, writing your code, running it, shipping it. Your job: point it at the right target.
Live demo · no rehearsal · LunchRush
Wireframe in. It breaks. One sentence fixes it.
LIVE DEMO · no rehearsal
"Build the screen this wireframe represents." Then, when it breaks: "I expected the station list sorted shortest first; it shows them in fixed order. Sort by current wait time, shortest on top."
Watch the fix land in one pass, no do over.
Just watch
Watch it plan before it writes a line, then build, then break.
Notice: broken is normal, not a failure.
Watch one precise sentence, not frustration, fix it.
You do this exact move next, on your own product.
You've used this move since Day 1
“I expected X. It does Y.” The iterate move.
Describe
Run
Test
State the gap
The one sentence that separates people who ship from people who stall.
You first
Has a vague instruction from a teacher or a parent ever gone badly for you?
▸ click to reveal after they answer
"Clean your room" and "fix the bug" fail for the exact same reason. The other side has to guess what you actually meant, and guesses are usually wrong.
Bug type: logic
The app does the wrong thing. Which report gets it fixed?
The stall
the counter is broken lol
The ship
I expected: clicking +1 twice shows 2. It does: two clicks shows 3, a third click shows 5. The count is doubling somewhere. Find where and fix only that.
Checkpoints
Commit before you experiment.
A commit is a save point for your whole project. Make one right before anything risky, and "I broke it" stops being scary.
LunchRush, live: "save a version: station sorting works" before I try the risky change.
How it actually works
Save, break, roll back. Three moves.
Checkpoint saved
Commit
Try the risky thing
Experiment
It goes wrong
Break
Hate it? Roll back. You're exactly where you were, no damage done.
"Add a photo banner at the top."
"Dark green palette, lime accent."
"Thumbs-up button per station that counts taps."
Growth on LunchRush: commit, then one precise ask. Same loop, new feature each time.
Skip the checkpoint and you get "redo the whole homepage, get creative": one team's real four-hour rebuild, from memory.
One more judgment call
Re-prompt in the same chat, or start fresh?
Bug's still there. What now?
Same chat
Fixing or refining what's already there. It remembers the context.
Fresh conversation
The chat's gone in circles, or you're starting something genuinely new.
A cluttered conversation makes worse guesses.
LunchRush example: chat confused about which screen? Restart fresh with the spec. One wrong detail, like the sort order? Re-prompt in the same chat.
Same bug, five tries in
One of these is the actual fix.
The stall
"no still wrong, try again" ... "still not it" ... "ugh, try again" ... "closer?? try again"
The shipFresh conversation. Here's the current state of my app and the one gap I need closed: [gap statement]. Clean slate, one clear ask.
Copy this prompt · run it now
The gap statement, as a prompt
I expected [X exactly]. It currently does [Y exactly]. The difference matters because [user impact]. Fix the gap without changing anything else that works.
Mastery exerciseFOLLOW ALONG · laptops open
Apply the move to your product, right now.
8:00 on the clock
Non-negotiable: commit your checkpoint before you touch anything risky.
RolesDriver at the keyboard. Navigator writes the gap sentence and judges it against expected-vs-observed. Swap driver at the halfway point.
YOUR TOOLS· Build: Kimi Code (every laptop, no VPN) · v0 (school device backup) · Same prompts, same rules
1 · find the gap
Find the biggest gap between what you meant and what's on screen. Write it: expected / does / matters because.
2 · checkpoint, then close it
Commit a save point. Take your gap sentence into your agent. One precise ask.
3 · stuck? decide
One wrong detail: re-prompt in the same chat. Chat gone in circles: roll back, start fresh with the current state and the one gap.
8:00
Work at your own pace. Post your gap sentence and the fix to the class channel as you land it.
Exit ticket
LEAVE WITH · One mastery move applied to your product + a checkpoint committed
DECIDED · Your re-prompt vs restart instinct, exercised for real
BRING NEXT · Your product open and its gaps named. At 10:35, the landing page assumes it's stable enough to put a front door on.
What's next today
Next: a break, then your front door.
10:25, a short break. 10:35, S13 the landing page, using the name and logo you just picked. 11:00, the build block puts name, logo, and landing page live at once, locked by lunch. This afternoon: the pitch deck, a price, the feature blitz freezing at 14:45, and five actual strangers who touch it while you watch in silence.