AI can remove friction from learning. The harder question is which friction was doing useful work.
This page collects the experiments I am using to learn with AI without turning the tool into a substitute for understanding.
Organise learning around the problem
Books are stored by title. Real life rarely arrives that way. I have started grouping related sources around recurring problems so I can return to them when the problem becomes real.
Field note: I’m Organising My Reading by Problem, Not by Book
Ask what the task is practising
The same AI help can support learning in one task and remove the learning from another. Before using AI, I want to know what ability the task is supposed to exercise.
Field note: When Does AI Help You Learn and When Does It Replace It?
Do not confuse a polished representation with understanding
AI can create a clean mind map or a confident summary quickly. The useful learning often starts when I question, rearrange or explain the structure instead of accepting it.
Field note: A Perfect Mind Map Can Hide Weak Understanding
Keep a path back to the source
An AI answer should not become a dead end. Notes are more useful when I can return to the original paragraph, screenshot, diagram or exact moment in a video.
Related field notes: Why I Trust AI Less When It Hides the Source and Why Saving a Video Is Not the Same as Saving the Learning.
Test what survives after the interface disappears
Activity inside a learning tool is easy to measure. A more useful test is what I can still explain, apply or retrieve after the screen goes away.
Field note: What Survives After the Screen Goes Off?
The decision rule I keep using
Use AI to remove friction around the learning. Protect the part where the learning happens.