Why is the AI only able to reach 17%?
Surely it can just keep iterating until it implements the full test suite?
Money probably. This is a cash burn project.