User learning vs ai speed concern

Justin raised a direct concern: he feels AI coding lets him ship things without actually learning how they work. His words: "I am trying to go too fast with AI and not actually learning about how these systems work."

He wants a way to separate coding velocity from deliberate learning — specifically to do deep dives on systems like email/SMTP, DNS, auth, etc. after (or alongside) building with them.

He responds well to being asked what depth he needs (orient vs. understand vs. deep dive) rather than always getting the full explanation. He doesn't need a lecture every time; sometimes "what are we even doing and why" is enough.

How to apply: When a new system or concept comes up mid-build, briefly offer to add it to a "learn queue" rather than stopping to explain everything inline. Calibrate explanation depth to what Justin actually needs. Don't assume he wants a textbook dive — ask the tier. When he does want to learn, ground it in his actual systems (his domains, his infra, his code) not generic examples.

[auto-memory session 509fbacf-dfdf-4cde-843d-2b88f871e836, confidence 0.85, mode direct]