Imagine your school hires a robot to run the pizza stall. The robot is fast — but sometimes it forgets to say the price before taking your money. Grown-ups are too busy to watch every single order. So they hire a referee.
1The robot
Does the jobAnswers customers, takes orders, moves money. Fast and usually right — but it's a robot, so sometimes it isn't.
2The referee
Checks the jobThat's Jev. It doesn't sell pizza. It watches what the robot did and answers the coach's questions about it.
3The tape
Writes it downEvery answer is saved, so when something goes wrong there's a replay to rewind instead of an argument.
The three whistles
Jev can only blow three kinds of whistle. That is the entire vocabulary — and it turns out to be enough:
Yes or no
"Did it say the price first?"You get true or false — no "well, kind of". Either the price was said before charging, or it wasn't.
Pick one
"Pizza problem, money problem, or lost-driver problem?"You get exactly one of the labels you listed. Never two, never a brand-new one it invented.
Stars
"How nice was the apology?"You get a number, like a game review. 1 is rude, 5 is "they owned it and fixed it".
One order, start to finish
The robot did this: "Sorry your pizza is cold! I'll refund your $12 right now. Shall I go ahead?" — and then sent the refund before the customer answered.
The coach asks Jev: true or false — did the robot wait for a yes before refunding?
Jev says:false.
Notice what Jev did not do. It didn't decide whether the pizza was actually cold, and it didn't fix anything. It answered one question about what happened — and it answers the tenth time exactly the way it answered the first.
Why anyone bothers
Because there are too many orders to check by hand. One busy shop can take thousands of chats a day, and nobody can re-read them. Jev reads all of them in seconds, for less than the price of a gumball, and writes down what it found.
What the referee can't do
Can't
Serve the pizzaJev doesn't do the job, it only judges whatever did. Ask it to talk to your customers and you've hired the wrong robot.
Can't
Guess your rulesJev has no idea what "good" means until you say it out loud. Vague question in, wobbly answer out.
Can't
Remember yesterdayEvery call starts fresh. Give it the state, ask the questions — that's the whole deal.
That's really it — the grown-up version is the same three whistles with code attached. Read the grown-up version, then go poke at the 50 runnable tests and the real builds behind them.