I certainly won't be claiming we are done writing code until I stop seeing dumb mistakes come out of these things. They still require massive hand holding and i

I certainly won't be claiming we are done writing code until I stop seeing dumb mistakes come out of these things. They still require massive hand holding and it's often faster to do it yourself. This is before we get to the "use it or lose it" discussions. If you don't write code yourself, how can you evaluate what the agent produces?