Hey, it's Cindy 🌱 The setup I use to actually learn things with Claude instead of just getting the answer, so it sticks.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
What the research actually says, because the headline oversimplifies it. The Australian report behind the "AI is making students lazy" coverage is Artificial intelligence, cognitive offloading and implications for education (March 2026), by Professor Leslie Loble at UTS and Professor Jason Lodge at UQ, for the Australian Network for Quality Digital Education. Its argument is more useful than the headline: offloading work to AI is not automatically bad. Handing over lower-order tasks is fine and often smart. The risk is unstructured offloading of the thinking you were supposed to be doing yourself, and the report is blunt that this hits hardest for people who don't yet have the background knowledge to judge the answer. That's the actual line to stay on the right side of.
the setting
Turn on Learning style first
Claude has a built-in style that stops it handing you finished answers and makes it work through guiding questions instead. It started inside Claude for Education and has been available to everyone since August 2025.
Find the style selector in the chat box, near where you type.
Choose "Learning" from the dropdown.
Check it took by asking something you already know. If it answers with a question back, you're in the right mode.
If you can't find the selector, you can get most of the same behaviour from the prompt below. The style is a convenience, not the mechanism. The mechanism is that you do the retrieving.
the setup
The prompt that makes Claude a coach
Paste this at the top of a study session. It sets the rules once so you're not fighting for them every message.
🧠 the coach setup
You're my tutor for [topic]. My goal is [what you want to be able to DO afterwards, e.g. "explain how Meta's ad auction picks a winner, without notes"]. My current level: [beginner / I know X but not Y].
Rules for this whole conversation:
1. Never give me a finished answer to a question I could work out. Ask me a guiding question instead.
2. When I'm stuck, give me the smallest possible hint, not the solution. Make me take the next step.
3. After each concept, quiz me on it before moving on. Don't move on until I can explain it back in my own words.
4. If my explanation is vague or I'm using words I can't define, call it out. Don't accept fluent-sounding nonsense.
5. Keep a running list of what I've got solid and what I'm still shaky on.
Start by asking me what I already think I know about [topic], so you can find the gaps.
Rule 4 is the one that earns its place. The failure mode with AI tutoring isn't getting things wrong, it's the comfortable feeling of understanding something because you just read a clear explanation of it. Being asked to define your own words breaks that illusion fast.
the loop
The four moves, in order
1. Quiz me until I can explain it back Paste your notes, lecture slides or the article and ask Claude to question you on them. Explaining it back in your own words is the test, not recognising the right answer.
2. One worked example, then you Ask for a single fully worked example, then do the next one yourself and have Claude mark your reasoning, not just your answer. Ask where your logic went wrong, not whether you got it right.
3. Ask what you're still getting wrong End every session with it. You get a specific gap list rather than a vague good feeling, and it's the input for your next session.
4. Come back tomorrow, not tonight Ask Claude to re-quiz you on yesterday's gaps at the start of the next session. Recalling something after you've half-forgotten it is where the memory actually forms.
💡 the end-of-session prompt
We're done for today. Three things:
1. What am I still getting wrong, specifically? Quote back anything I explained badly or got fluent-sounding but hollow.
2. Rank what I covered today from most solid to shakiest.
3. Write me 5 recall questions on the shakiest items, and don't include the answers. I want to try them cold tomorrow before we start.
why it works
Retrieval is the thing that builds memory
The reason "quiz me" beats "explain it to me again" isn't motivational, it's one of the most replicated findings in learning research. In Roediger and Karpicke (2006), students read prose passages and then either took recall tests on them, with no feedback at all, or simply reread the same passages. Here's the part worth sitting with: when the final test came five minutes later, the rereaders did better. At two days and at one week, the tested group won, and the gap grew with time.
That last detail is the whole point. Rereading feels like learning and testing feels like failing, and the feelings are backwards. If Claude does the retrieving for you, you get the comfortable feeling and none of the memory. Every move above exists to keep the retrieving on your side of the table.
the honest part
When to just let it answer
Coach mode is not always the right mode, and pretending otherwise wastes your time.
Let it answer when the task is genuinely below you: formatting, boilerplate, converting units, summarising something you'll never need to recall, or looking up a fact you'd have googled anyway.
Coach mode when you'll need to do this again without help, when someone will question you on it, or when it's foundational to something you're building toward.
The report's own framing is the useful test: offload the lower-order work, keep the thinking. If you'd be embarrassed to be asked "so how does that actually work?" in a meeting next month, that's a coach-mode topic.
the waitlist is now open
AI System Club is launching soon
Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.
btw this list gets first dibs the moment doors open :)