There’s increasing evidence from AI productivity studies that we’re not actually that productive with AI, and mounting potential concerns grow concerning the long-term impact of skill building and knowledge. Productivity can’t be measured at times because output can’t be.
Catch 22
AI tools often require deep, specific knowledge to be used effectively. But this knowledge makes AI unnecessary in the very first place for those individuals.
Effective prompting requires the kind of specific, detailed knowledge that comes from experience — but if you had that experience, you probably wouldn’t need the AI to solve the problem. It’s recursive incompetence all the way down, like an Escherian nightmare where every level of abstraction depends on understanding the level below it, but nobody remembers how to get to the bottom floor.
Systemic erosion of judgement and agency
The speed and convenience gained from AI usage also erodes judgement due to the knowledge built through friction. Seniors don’t see the need to pass useful skills onto juniors because comprehension can be done by AI.
Technology amplifies everything except judgement. AI can’t replace the understanding that comes from wrestling with problems. Automation is mistaken for comprehension. If we let AI automate decisions, we are at the risk of losing the practice and experience from independent thinking.
Real [../thoughts/Agency] is acting without external influence. AI shapes our choices or nudges our behaviour by showing the path of least resistance.
We’re trading the friction that builds judgement for the speed that erodes it. - Nikunj Kothari
Loss of intrinsic motivation
AI increases user output but its users report feeling loss of motivation and boredom when doing tasks without AI.
Perhaps most tellingly, Harvard Business Review’s 2025 research reveals that while AI makes people more productive, it also makes them “less motivated and more bored when they had to work on other tasks, without the use of AI”.
The cycle repeats itself
AI is trained on the artifacts we produce. The next generation of AI will be trained on code written by humans who learned to program by copying AI output.
Related: Tools for Conviviality, Taste, Agency, Meatware