Notes and Revision That Actually Stick

Most notes are written for the wrong person

Think about who your notes are for. Not you now, sitting in the lecture or in front of the chapter with everything fresh. You in five weeks, who has forgotten the context, the example the lecturer used and half the vocabulary.

Almost all note-taking fails that test. A tidy summary of a chapter, written while the chapter is still open, tends to be a compression of text you understood at the time. Five weeks later it reads like a list of headings for something you once knew.

AI makes this worse before it makes it better, because it makes producing summaries effortless. You can generate an immaculate summary of anything in seconds, feel productive, save it, and never think about it again. A generated summary you did not engage with is not a note. It is a second copy of the material, slightly shorter.

What makes a note useful later

Three things, and you can build all three quickly.

Your own words. The moment of translating something into your own phrasing is where a surprising amount of the learning happens. If a summary arrives already formed, you skip that moment. So use AI to check your summary rather than to write it: write four sentences from memory, then paste both your version and the source and ask what you got wrong, what you left out, and what you have subtly overstated. That is a genuinely excellent use, and it is far more useful than a perfect summary you did not write.

A question, not a statement. A note that says "three causes of the 1929 crash" is a heading. A note that says "what were the three causes, and which one do historians disagree about?" is something you can use. Statements can only be reread. Questions can be answered, and answering is the thing that works.

A link to something else. Isolated facts are hard to keep. Notes that say how this connects to the thing from two weeks ago hold on much better, and asking for those connections explicitly is something AI does well.

Testing yourself beats rereading

If there is one piece of study advice worth building a routine around, it is this: retrieving something from memory does far more for retention than looking at it again. It is one of the most consistently repeated findings in the study skills literature, and you do not have to take anyone's word for it. Try both for a week and see which one you can still do on Friday.

The reason rereading is so popular despite this is that rereading feels good. The material looks familiar, the pages go by, and familiarity is very easily mistaken for knowledge. Testing yourself feels bad, because you get things wrong, and getting things wrong in your own bedroom is unpleasant in a way that getting things wrong in an exam is worse.

That is the trade. The uncomfortable method is the one that works. This is why the practical value of AI here is not that it explains things. It is that it removes the excuse: writing your own practice questions is tedious, and now it takes ten seconds.

The practice question prompt

This is the one to keep. It works for a lecture, a chapter, a set of notes, a report you need to know inside out for a meeting.

Here is material I am revising. Write me practice questions on it.
Rules:

Questions only. Do not give me any answers yet.
10 questions, mixed difficulty, ordered easiest first.
At least three should ask me to explain WHY, not just recall a fact.
At least two should apply the idea to a situation not in the material.
One should ask me to state the strongest objection or limitation.
Do not quote the material back to me in the question, or the
wording will give the answer away.
I will answer them, then paste my answers back for you to mark
against this material and tell me what I have missed.
Material:
[paste your notes or the chapter]

The last line is what makes it a loop rather than a worksheet. Answer everything, badly if necessary, then paste your answers back and ask what you got wrong. Ask it to be blunt and not to praise you.

Spacing, without any complicated system

You do not need an elaborate schedule. The principle is that a gap before you test yourself does more good than testing immediately, and the gap should get longer as the material sticks.

A version that fits on a sticky note: attempt the questions the day you make them, again about three days later, and again a week or two after that. Anything you get wrong goes back into the earlier slot. Anything you get right twice in a row moves to the back of the pile. That is the entire system, and it beats any app you would abandon in a fortnight.

Here is the material, and here are my answers to the questions
you set. Mark them against the material only.
For each answer tell me: right, partly right, or wrong, and why.
Where I was partly right, say exactly which bit was missing.
Point out anywhere I used the right words without showing that
I understand the idea.
Then list the two topics I should go back to first.
Do not be encouraging. I want to know what is wrong.
Material:
[paste it]
My answers:
[paste them]