Hey there,
Every change an agent writes for you either builds your understanding or spends it. Spending it means you ship something without learning why it works. From the inside, the two feel exactly the same.
The tests pass, the page loads, you move on. Nothing tells you which one just happened, and there is no error message for "shipped three features you cannot explain." You find out later, when something breaks and the code is yours in name only.
There is a study worth knowing about. Anthropic ran a randomized trial with 52 professional developers, most with four or more years of experience, half coding with an AI assistant, half without. The ones using AI scored 17 percentage points lower on a quiz about the code they had just written, roughly two letter grades. The ones who asked the assistant to explain, and asked why, scored higher than the ones who handed the work over.
(The paper is Shen & Tamkin, "How AI Impacts Skill Formation," 2026.)
The tool was the same. What differed was what each developer did with the answers. So the fix is not to use agents less. The fix is to ask why before you accept, so that what you ship is something you could defend if a teammate questioned it.
Here is the practice version. Pick something you built recently, at work or in a class, and explain out loud why it is built that way, not just what it does, as if to a teammate.
Wherever you stall is where the gap is.
If you want help closing it, Course 0 is where I would start. It is a self-paced online course you take on your own schedule: eight lessons, two hands-on labs, a ten-question written assessment with feedback, and three challenges on a chat app running on your own computer. It teaches one workflow end to end: before you accept a plan from your AI agent, ask for its reasons, make it show a test that backs them, then decide.
I built it, and I lecture at UC Berkeley. It is for people who build with agents every day and want to defend what they ship. You keep the speed and you keep your understanding. It is $99 once, with a 30-day money-back guarantee.
Two ways in, whichever suits you. The course:
https://systemthinkinglab.ai/course-0
Or, if you would rather start free, the five-day email course: