bottlenecks I see with AI coding agents:
-your harness/loop is not the only thing that needs to self improve, you do too
-over engineering has made its way to the practices being defined now, simplicity that fits your use case will always win adoption
@elorm_elom Love this! Suggestion - would be nice to see examples of features delivered/code generated by the loops. Adding some mitigation in the loop to prevent bias from the agent taking a shortcut to achieve the success criteria i.e writing a bad test, would also be great