AI Guides › Claude Mastery
Claude For Learning A New Subject
By Nigel Guy · 2 min read
The default way people use Claude to learn something new is to ask it to
explain the topic, read the explanation once, and consider themselves
briefed. That's absorbing an explanation, not learning — it feels like
progress because the explanation made sense while you were reading it, which
is a much lower bar than being able to use the material without it in front
of you.
The rule: Claude is far more useful as a tutor that questions you than as
an encyclopaedia that explains at you — the mode most people miss is the one
where it checks whether you actually understood, rather than just telling
you again.
The mechanism: the tutor loop
- Ask for the explanation first, at the level you're actually at. Say
plainly what you already know, so it isn't guessing your starting point.
- Then ask it to test you, not just tell you. "Ask me three questions
that would reveal whether I actually understood that" turns a passive
read into something you have to produce an answer to.
- Answer without looking back at the explanation. This is the part
that actually does the learning — the discomfort of not immediately
recalling something is the signal of where the gap actually is.
- Ask it to identify specifically what you got wrong or missed, not just
whether you passed. "You got the mechanism right but mixed up which
direction it goes" is useful; "close, but not quite" is not.
- Ask for one worked example applying the idea, then try a second one
yourself. Watching one worked example and doing the next one yourself
is where an explanation turns into something you can actually use.
- Come back later and ask it to test you again, cold. Understanding
something in the same conversation you learned it in is a weaker
signal than recalling it accurately a day or a week later.
What to skip
Skip this full loop for something you just need a quick answer to, not
genuine understanding of — a one-off fact doesn't need a tutor cycle built
around it. And skip moving on to the next topic once you've read an
explanation you followed easily; "I followed it" and "I could explain it
back without help" are different states, and only the second one is
actually learning.
Guardrails
- Claude can explain a topic fluently and still get a specific detail wrong,
particularly on anything niche, fast-moving, or genuinely contested among
experts — cross-check anything you intend to rely on, especially before a
test, exam, or professional context.
- Being quizzed by Claude is a study aid, not a substitute for a real
assessment where one is required — a qualification, a certification, a
formal course still needs its own actual evaluation.
- The tutor loop works best on subjects with a checkable right answer.
For genuinely contested or opinion-driven material, ask it to represent
more than one credible view rather than treating any single explanation as
settled.
All 751 AI guides · JulieMango plans from £17/mo