Reading the launch
Most AI-written code "passes all tests" but that doesn't mean the tests actually work. Agents love writing tests that assert nothing. mole checks. It plants realistic bugs in your code on purpose the forgotten await, the off-by-one, the missing null check and sees whether your tests notice. Whatever they miss is a bug your suite would happily ship. Then it writes the missing test for you, and proves it catches the bug before saving it. Useful for anyone shipping code where a silent wrong number is expensive: payments, billing, pricing, permissions. Goated dev.
Reading trades.