Suppose you are learning how to set up a portable sound recorder for an interview. An AI assistant turns the manual into a lesson and quizzes you at the end. You get every answer right. Then three weeks pass, you arrive at a venue, and you cannot remember where to check that both microphones are being recorded. Your first quiz told you something useful about that afternoon. It did not tell you what you would know on the day you needed it.
A study plan has two clocks. One measures the gap between learning and the next review. The other measures how long the knowledge needs to last. If you ask AI for a revision calendar, give it a reason for the dates. A plan for tomorrow’s interview can look quite different from a plan for a job you will do again next month.
What spacing research can tell you
In a study published in November 2008, Nicholas Cepeda and colleagues taught participants facts, varied the gap before a review, and then tested what participants remembered after another delay. For a given test delay, a longer gap before review initially improved performance, but extending the gap further eventually reduced it. The review gap associated with the best results was longer when the final test was further away.
That result gives you a reason to think about when you will use what you learn. It does not give you a universal calendar. The experiment tested factual material under specified conditions. Setting up a recorder includes physical actions and troubleshooting as well as facts, and the study did not test AI tutors. Treat the calendar as a starting point and use your own attempts to see what needs another review.
Put a real task at the end of the calendar
Start by defining the performance you care about. For the recorder, the goal might be to connect two microphones, record a short sample and play it back to confirm both voices are present. That is more concrete than ‘know how to use the recorder’. It also tells you which questions are worth asking. If your AI quiz is mostly about names for different cables, it might miss the one check that would save an interview.
If the interview is tomorrow, try a short unaided setup after a break today and again before you leave, if time allows. If the job is three weeks away, spread a few attempts across the intervening period rather than spending the entire practice budget on the first day. These are examples of how to plan around a deadline. The exact intervals are suggestions to test, not timings established by the study.
For a skill with no fixed date, pick a plausible next use. A creator who runs a workshop once a quarter can set a trial before the next session. A shop owner training a colleague to open alone can use the first solo opening as the target. When the date changes, move the practice that prepares for it. There is little point in preserving a calendar that no longer matches the job.
Make the review reveal what you can do
Ask AI to prepare short practice tasks using the source material, then put the instructions aside during the attempt. A quiz might ask how you would check that two inputs are recording. A stronger rehearsal asks you to demonstrate the check with the actual recorder and play the file back. If the model describes a menu your recorder does not have, verify the instruction against the manufacturer’s manual and revise the exercise.
When you get stuck, write down where. Perhaps you remember that a playback check matters but cannot find the control. Practise that step, then do a fresh attempt after another gap. If you can complete it unaided and explain what you checked, try again after a longer interval or in a slightly different setup. Our guide to gradually removing steps from a worked example can help if the jump to an unaided attempt is too large. It changes how much support you get; spacing changes when you return.
Keep the practice small enough to repeat. If every review means rebuilding your entire recording kit, a busy week can erase the plan. You might rehearse the input check on its own between full setups, then use the next complete attempt to see whether you remember to perform it without a reminder. The short drill and the full rehearsal answer different questions.
Give AI enough information to draft a useful plan
- Name the date or situation in which you will use the skill, and say what successful performance looks like.
- Give the assistant your available time and ask it to spread short attempts across that window. Ask it to keep the answer or checklist separate from each task.
- Log what you could do unaided, what needed a hint and what you still could not do. Use those observations to bring forward or postpone the next review.
- Keep a verified checklist for the actual job if a mistake would matter. Learning a procedure and having a backup on the day are compatible goals.
After the job, compare the rehearsal with what actually happened. Maybe setup was easy but you did not practise recognising a disconnected microphone. Add that case for next time. An AI can draft another calendar quickly. The value comes from telling it what you need to be able to do, then checking whether the practice holds up when the lesson is no longer in front of you.