How Do You Use AI?

A self-assessment based on Anthropic's skill formation research

Anthropic studied 52 developers learning a new programming library with and without AI assistance. The finding was not that AI helps or hurts -- it was that how you interact with AI matters far more than whether you use it. Six distinct patterns emerged, each with dramatically different outcomes for learning and productivity.

This is not a quiz with right answers. It is a mirror. Answer honestly about how you actually work with AI, not how you think you should.

Your Pattern

The learning-productivity tradeoff space

These patterns are not fixed. Most people shift between them depending on context -- whether they are learning or shipping, under time pressure or exploring, working in familiar territory or new ground. The value is knowing which pattern you default to, so you can choose deliberately.

The Research

Key findings from the study of 52 developers learning Trio, a Python async library.

Reflection

These are not questions to submit. They are questions to sit with.

When do you need to learn, and when do you need to finish?
The study found no single "best" pattern. Delegation is rational when you need output, not understanding. The danger is when delegation becomes your default even when understanding matters.
What is your default pattern? Is it different from what you would choose deliberately?
Defaults tend to emerge from convenience, not intention. The developers in the study did not choose their patterns consciously -- they drifted into them. Awareness is the prerequisite to choice.
How does your pattern change under time pressure?
Progressive Reliance -- starting with questions but escalating to delegation -- was one of the lowest-scoring patterns. Time pressure tends to push people toward cognitive offloading precisely when building understanding might matter most.