[lbalazscs]

It's naive to hope that automatic tests will find all problems. There are several types of problems that are hard to detect automatically: concurrency problems, resource management errors, security vulnerabilities, etc.

An even more important question: who tests the tests themselves? In traditional development, every piece of logic is implemented twice: once in the code and once in the tests. The tests checks the code, and in turn, the code implicitly checks the tests. It's quite common to find that a bug was actually in the tests, not the app code. You can't just blindly trust the tests, and wait until your agent finds a way to replicate a test bug in the code.