The author is a CTO/CPTO who explicitly describes himself as a "hands-off manager", who hasn't touched code in over many years. His view that GenAI "only speeds up the typing" is a classic management-layer misconception. It assumes the engineering is the typing. Classical Systems Engineering treats the "development system" as part of the system. Requirements are not "unknowable"; they are derived through rigorous analysis (Functional/Physical architectures). The (post-)Agile software school views rigorous requirements analysis as "Waterfall" and impossible. The author is effectively admitting that the software industry has lost the capability to perform the Requirements Analysis & Allocation tasks defined in standards like IEEE 1220. The author thinks the industry fails because it treats software as "construction" (predictable). I would rather say that industry fails because it abandoned the discipline required to make it predictable. Sending humans to the moon was arguably "Complex", but it was solved by treating it as a determinable engineering challenge. "Users don't know what they want" is a failure of Systems Engineering, not a metaphysical property of software. "Agile" eventually became a "cargo cult" itself, as the pseudo engineering approaches it wanted to replace.