Spot the mistakes
Guess first, no penalty. An AI wrote average_order_value(). Customer A has orders of 10 and 20, customer B one order of 100. A's result comes out as 65.0, not 15.0. Tap the line that causes it.
The function, line by line
0 of 1 found
The worst AI bugs don't crash. The code runs and returns tidy numbers, and they're wrong. A quick read misses them because each line looks fine on its own.
Two habits catch them:
- Trace one small example by hand. Pick inputs you can work out in your head and compare with what the code returns.
- Write a test with mixed input. One customer hides this bug, because the shared total is that customer's total. Two customers expose it.
The expected value in a test comes from your hand trace or the requirement. Never paste in what the code returned, or the test locks the bug in.
A worked example
Hand trace: A = (10 + 20) / 2 = 15.0 B = 100 / 1 = 100.0
Code says: A = 65.0 B = 130.0 -> bug
Fix: keep one total per customer:
totals[c] = totals.get(c, 0) + order["amount"]
Test I keep:
orders = [{"customer": "A", "amount": 10.0},
{"customer": "A", "amount": 20.0},
{"customer": "B", "amount": 100.0}]
assert average_order_value(orders) == {"A": 15.0, "B": 100.0}Fill the gaps
Fill the gaps.
Test plan
I use so a shared total shows. The expected value comes from , and the assert checks .
Try it
Your steps, about 2 minutes
- 1.Trace a small, mixed example by hand for one function an AI wrote for you, and turn it into a test. Add "trace one mixed example" to your kit.2 min