Use AI coding tools without skipping the learning.

A free, open-source project for junior developers. Practical habits, backed by research, in your inbox.

The problem

Debugging by pasting errors back

In an Anthropic study of developers learning a new library, people who looped errors back to the AI without forming a hypothesis scored lowest on the later quiz.

RCT, post hoc. The patterns weren't randomized and the groups were small (n = 4), so this is an association.

Assisted gains, unaided losses

In a math study with high-school students, unrestricted AI help raised practice scores, but the same students did worse on a closed-book exam later that session. A tutor with safeguards mostly avoided the drop.

RCT, math, not coding. Applying it to coding is our extrapolation.

A self-check

Could you explain the last diff you approved?

A question for you, not a statistic.

Three ways to use AI

Learning: You do the thinking; the AI questions, hints and checks.
Onboarding: Using AI to find your way around a codebase that's new to you.
Delivery: Shipping work with AI while still understanding what changed and why.

What you get

One practice to try
One tool or mode tip
One "stuck of the week" theme
What changed in the project

Built in the open

MIT-licensed, with the research and plans on GitHub. When we say a toolenforces something, we mean a specific, tested block. Everything else is labelled as advice.