Will we approach AI generated code with an industrial mindset that treats humans as expensive code writers to be replaced with AI, or with a knowledge-work mindset that believes that working software requires ongoing human understanding?

This talk will offer a Systems Thinking view of goals, feedback loops and leverage points to explore two conflicting goals of AI adoption: "Reduce Development Costs" vs "Enhance Human Understanding".

Donella Meadows, in her influential book “Systems Thinking, a Primer” says that you can’t optimize for contradictory goals simultaneously. Applied to AI, can we chase reducing costs AND preserving understanding, or will we only fool ourselves that we are doing both?