Hmm I guess it’s expensive but I don’t see why agents in a Ralph loop under the supervision of a comprehensive test suite can’t deliver top percentile results. You just iterate enough times until you see the desired behavior. Where are the risks?

I disagree, It increases the value of any developer.

Then the question is how do you become an excellent developer?

Could you provide the evidence for this statement please? I'm curious what methodology was used to rate "excellent" and "average", as well as "value", and what those "relative terms" are.