Most work on AI asks which jobs survive. This project asks a different question: what happens to judgment, attention and expertise when the work that used to build them is delegated? The evidence says the answer depends almost entirely on how the tools are used.
What predicts whether you keep developing is not how much AI you use. It is how much of your own reasoning survives using it.
A 2026 controlled experiment separated two variables everyone had been conflating — frequency of AI requests, and the independent reasoning preserved while making them. Skill development tracked the second, not the first. Research on programmers found the same shape: automating a task wholesale accelerates decay, while iterative use — forming a view, evaluating output, correcting it — can preserve or even scaffold skill.
The studies this project rests on, in plain language. Where findings are contested, they are labelled contested — because the honest state of this field is that some of it is settled and some of it is not.
The full argument, free to read, with your place saved as you go.
What erodes, why it is invisible, and the evidence that it is real.
Friction removed, originality converging, mentorship dissolving, presence traded away.
The recurring decisions — delegation, verification, apprenticeship, judgment.
The intentional-use framework, six practices, and the measurement instrument.
The counter-movement already underway inside the most technical organisations on earth.
A structured self-assessment across five dimensions. Twenty questions, ten minutes, a score you can track quarterly — and an organisational version that surfaces the gap between what leadership believes and what staff experience.
Take the assessmentNo paywall, no account, no email required. Your place and bookmarks save on your own device.