@abucci You are 100% correct, I'm not dismissing the importance of essential complexity or the fact that brute-force generation is no substitute for intelligent design.However, I don't feel that the 'net negative' conclusion of using LLMs in coding is an inevitability.If we build agentic loops that provide exhaustive verification previously unthinkable, the LLM's inherent "weakness" as a programmer becomes less relevant; it is no longer a source of truth, but a high-velocity engine for exploring the state space of a specification.The 'essential complexity' shifts from implementation to specification, leaving a place for LLMs to positively contribute. This shift was unimaginable a few years ago, but is now accessible to play with as a concept to anyone with a VPS and a harness.