I wonder if one consequence of agentic workflow is a gradual normalization of heuristic correctness. Not wrong, not careless, just increasingly comfortable with good enough for now because another iteration is always available?