A free preview, generated by MasteryLoop HQ.
You've seen the screenshot. A long prompt, five numbered stages, a claim that it turns a chatbot into a real tutor. You saved it. Maybe you ran it once. The explanation you got back was good — better than what a chatbot normally hands you. Then you closed the tab and never went back. That's how it usually goes. The prompt wasn't the problem. These threads spread because the thing worked: run all five in order and you get a teaching sequence instead of an answer.
Strip the formatting away and the sequence those prompts describe is a defensible one: activate what you already know, read actively rather than passively, practise retrieval instead of rereading, explain the idea back in your own words, then revisit it on a spacing schedule. Every step there has decades of education research behind it. The teach-back step in particular is the Feynman technique, and the reason it works is that recognising a good explanation and constructing one are different cognitive acts. The prompts weren't snake oil. They were a real curriculum, compressed into something you could paste.
The threads usually framed it as a hidden capability — five prompts to access a mode the model was otherwise keeping from you. That part was overstated. Claude has a real Learning Mode, built by Anthropic: it shipped with Claude for Education in April 2025 and was extended to all users that August. It works through Projects and output styles, and it's Socratic by design, answering a question with a question rather than handing over the solution. Nothing was hidden. A well-written prompt made the model behave more like a tutor, which is worth doing — but it wasn't unlocking a secret.
The obvious complaint about learning inside a chat window used to be that it forgot you. Close the tab and the model had no idea what you'd covered, what you'd struggled with, or what you were meant to revisit. That complaint is out of date. Anthropic rolled persistent memory out to every Claude account, free tier included, in early March 2026. It saves what you're working on as you go and carries it across sessions, with separate memory for each Project. Study something on a Monday and it can still know that on Friday.
Remembering what you studied isn't the same as knowing when to make you prove you still know it. Memory is passive — it recalls context when a conversation calls for it. Retention needs something active: a system that decides a concept is due today rather than next week, asks you about it without being prompted, scores how well you actually answered, and moves the next interval out or pulls it back in based on that score. That's a scheduler, and it's a different kind of thing from a memory. It also can't be a prompt, because a prompt only runs when you remember to run it — which is precisely the decision a spacing schedule exists to take out of your hands.
The five-stage learning prompts that circulate online describe a genuinely good method. The two things people usually blame for them not sticking have both stopped being true: Learning Mode is a built-in Claude feature rather than something you unlock, and Claude has remembered across sessions since March 2026. What's still missing is scheduling — deciding when a concept comes back, testing it, scoring the answer, and adjusting the next interval accordingly. That's the part a prompt can't do, because a prompt only runs when you remember to run it.
Think of something you read this year that you'd have said you understood at the time. Could you explain it to someone now, without looking — and if not, was the problem the explanation, or that nothing ever asked you again?
Reading a good explanation feels like understanding it. Usually it isn't the same thing — and you don't find out which one you've got until someone asks you to explain it back.