How it works

1. 1

   Read

   We look through the code files related to the bug.

2. 2

   Prove

   We write a test that fails, showing the bug is real.

3. 3

   Try many

   Six different fixes are written at the same time.

4. 4

   Check

   Every fix goes through the same four checks.

5. 5

   Choose

   Of the fixes that pass, the smallest change wins.

The four checks

• Runs — The program still starts without errors.
• Stops the bug — The failing test now passes.
• Old tests pass — Nothing that worked before is broken.
• Tests untouched — The fix didn't cheat by changing the tests.