Practical AI training for everyone
Module 1
7 min read
The most common way students use AI is also the least effective: paste in a reading, ask for a summary, read the summary, feel prepared. The feeling is real and the preparation is not. You have just read a fluent piece of prose about a topic, which produces a strong sense of familiarity and almost no ability to reproduce the argument three weeks later under exam conditions.
This is not a reason to avoid the tool. It is a reason to use it for the part that is genuinely hard — building understanding you can retrieve — rather than the part that only feels productive.
Re-reading and summary-reading both produce fluency: the material goes down easily, so you conclude you know it. Retrieval — pulling something out of your head with nothing in front of you — is the only thing that both builds and honestly measures memory. Every technique in this module is a way of getting AI to force retrieval instead of substituting for it.
The single highest-value change: stop asking for explanations and start asking to be tested. A model is a tireless, infinitely patient examiner that will generate a hundred questions on a topic and never get bored of your wrong answers.
Passive vs. retrieval prompt
Passive: 'Summarise this chapter on price elasticity.' Retrieval: 'I have read this chapter on price elasticity. Ask me eight short questions on it, one at a time, waiting for my answer before the next. Two should be definitions, four should be applications to a scenario not in the chapter, and two should be about the limits of the concept, when it stops being a useful model. Tell me what I got wrong and what I was vague about.' The second one is uncomfortable, which is the point.
Explaining a concept in your own words exposes exactly the parts you have only half-absorbed, because vagueness is invisible when you read and obvious when you write. AI is unusually good here: it can read your explanation, find the place where you have used a technical term as a substitute for understanding it, and ask the follow-up question your tutor would have asked.
Explain-back prompt
'Here is my explanation of why the central limit theorem matters, in my own words. Do not correct it yet. First, tell me which sentences are precise, which are vague, and which are wrong. Then ask me two questions that would only be answerable by someone who actually understands it, and I will answer before you explain anything.'
There is a legitimate and very good use for summarising: a dense paper you cannot get a foothold in. Ask for the argument's shape first — what is being claimed, what is the evidence, what would have to be true — then go and read the actual paper with that map in hand. The map makes the paper readable. It does not replace the paper, and a marker can always tell the difference between someone who read the map and someone who read the source.
Try it now
Take a topic you have a lecture on this week and have already read. Ask AI to test you on it with eight questions, one at a time, without giving you the answers first. Notice how much you thought you knew versus how much came out.
A few quick questions. You'll see an explanation after each one.
Question 1 of 4
Why does reading an AI summary feel like effective revision when it usually is not?