Study a new topic
~ min read
30-second summary
- Formal study is a different game from personal curiosity: there’s a textbook, a professor, a final test. The AI knows none of the three, so you have to anchor it.
- Three techniques that change the quality of a study session: paste the textbook paragraph before asking for the explanation, ask exam-style questions on that chapter, calibrate the AI to your professor’s style using notes or slides.
- “Holding a nice conversation with the AI” is not “being able to answer at the exam”. Final check: explain it out loud with the chat closed.
- Your professor’s material (lecture PDFs, slides) can be covered by academic copyright: be careful with public chats.
This module isn’t the “do my homework for me” manual. It’s the “study better” manual. They’re two different things, not because one is moral and the other isn’t, but because the first doesn’t actually teach you anything and falls apart at the first oral exam. AI is a study tool when it explains things, quizzes you, helps you review; it’s a problem when it produces, in your place, the text you’re supposed to hand in. This module gives you techniques for the first use; the final lesson, Ethics of studying, comes back explicitly to where the line falls.
From curiosity to formal study
Section titled “From curiosity to formal study”In Learn something new you saw how to use the AI to understand a topic that interests you: blockchain, inflation, bacterial genetics. There the stakes were “did I get it or not?”, and you alone were enough to judge.
Formal study is another story. Three things change:
- There’s a reference material, picked by the course. Often it’s not a single textbook: it’s a mix of chapters from different books, the professor’s handouts, slides, articles. What matters is what that material says, not how the topic is told in general on the internet.
- There’s a professor who decides what counts as “right”, both through what they put in the material and through what they add out loud in class. Definitions and the boundaries of a concept can shift between authors and between professors; in your course the professor’s version wins, even when it’s in tension with the textbook.
- There’s a test at the end: oral, written, paper. Someone will ask you precise questions and expects precise answers, with the vocabulary of the course.
The AI knows none of these three things. Left to its defaults, it fishes from Wikipedia, popular articles, English-language textbooks, and gives you back a mix that’s almost-right but doesn’t match your syllabus. To make it useful inside your course, you have to anchor it to your material and calibrate it to your professor. That’s the leap compared to the free use you saw in Learn something new.
Three techniques to anchor the AI to your syllabus
Section titled “Three techniques to anchor the AI to your syllabus”The three techniques work on different levels: the first anchors the AI to the content of your syllabus, the second to the type of questions you expect, the third to the professor’s style that questions you. They combine: the most complete version is pasting the paragraph and asking exam-style questions on that paragraph and attaching style notes.
1. Paste the textbook paragraph
Section titled “1. Paste the textbook paragraph”The simplest move and the most underrated. When you need to understand a concept from the course, don’t ask “explain X to me”: open the book, copy the paragraph that introduces it (or the bullet point from the slide), paste it into the chat, and then ask for the explanation while staying inside that frame.
The prompt looks different. From:
“Explain the dissolution powers of the President of the Republic to me.”
To:
“[textbook paragraph pasted]. Explain this paragraph to me in simpler words, but stay inside the frame the text gives. If you add something that isn’t here, flag it for me.”
The difference is huge. In the first version the AI gives you back an averaged summary: definitions floating around the web, some reference to the Constitution, some common sense. In the second version the AI is forced to reformulate what’s in front of you, nothing more. If the textbook insists on a particular distinction (for example between “formally presidential acts” and “substantively presidential acts”), the explanation that comes out keeps that distinction. If you ask in general instead, that nuance disappears and at the exam you’re missing it.
The same goes for slides: if the professor put a definition on a slide, that definition is the reference. Paste it. If the slide is just three-word bullets, add a line at the end of the prompt with what the professor said out loud on that point, even just a few words: the slide alone is too sparse to anchor the AI, the slide plus your verbal summary is enough.
2. Ask exam-style questions on that chapter
Section titled “2. Ask exam-style questions on that chapter”Once you’ve understood the concept, flip the roles (the “return questioning” technique you saw in Learn something new), with one extra spec: force the level of the exam.
“Based on the chapter I pasted above, give me three oral-style questions and three written-style questions, at the level of precision expected by a constitutional law exam in the second year of law school. One at a time, wait for my answer before moving to the next. After each answer tell me if it’s correct and what’s missing to get a top grade.”
The AI changes tone. Without the level spec, the questions come out generic (“what is the President of the Republic for?”); with the spec, they come out more precise (“distinguish between presidential acts that require ministerial countersignature and those that don’t, citing one example for each category”).
The distinction between oral and written matters because they’re different tests. The oral rewards your ability to synthesize and reason out loud; the written rewards precision and completeness. Practicing only the format you’ll get at the exam is less useful than practicing both: it gives you different angles on the same material.
3. Calibrate the AI to your professor’s style
Section titled “3. Calibrate the AI to your professor’s style”This is the move that matters when the exam gets close. If you have lecture notes, the professor’s slides, or even just an outline of a past oral exam circulating among students, paste it in as a style reference.
“These are notes from three lectures by my professor. They tend to insist on X and Y, and often ask questions like Z. When you quiz me on this chapter, simulate their way of asking, not a generic quiz.”
The AI doesn’t know your professor by name, but it picks up on the pattern: if the professor tends to ask for practical applications, it gets that from the notes and gives you those; if instead they ask for word-for-word definitions, that comes out too. The simulation isn’t perfect, but it gets you much closer to what awaits you in the classroom than a random quiz from the internet.
A worked example. You’re preparing for a modern history exam, and from the notes of the last three lectures your professor has insisted on the role of the press in shaping public opinion between the two world wars, often citing two specific authors. The prompt becomes:
“I’m studying for a modern history exam. In the notes I’m pasting in, my professor insists a lot on two things: the role of the press between the two world wars and the authors Mosse and Gentile. When you quiz me on the chapter I’m passing you right after, simulate their way of asking: give me questions that connect the general theme to these two authors, one at a time, waiting for my answers. Notes: [paste 1-2 pages of notes].”
On the lead of those two authors, the AI pulls out questions like “how would Mosse explain this passage from the textbook?” instead of asking for the definition of a term. That’s the leap between a generic quiz and a quiz from your professor.
The difference in practice
Section titled “The difference in practice”A constitutional law student, Bin/Pitruzzella textbook, is studying the powers of the President of the Republic. First try, asking in general.
The answer is honest, but there’s a problem: the closing comparison with the United States is side material the textbook doesn’t cover in that section, and at the exam it costs you time. On top of that, it’s missing a distinction that Bin/Pitruzzella develop with care, on the substantively or formally presidential nature of the dissolution act, which is exactly the kind of nuance constitutional law professors insist on.
Now the same question, anchored to the textbook paragraph.
Three differences compared to the general answer. First, the vocabulary is the textbook’s (“formally presidential but substantively complex act”, “check on legality and timing”), which is what the professor expects to hear. Second, no comparison with the United States, which isn’t in the paragraph. Third, the AI honestly flags the point the paragraph leaves out, so you know where to go look for the rest.
A quick test if you suspect it’s drifting off: ask the AI “which part of the paragraph I pasted is this sentence based on?”. If it answers by pointing you to the exact spot in the text, you’re anchored; if it talks around it or cites general knowledge, you’re back in open water and you need to re-paste.
”I get it” is not “I know what the professor expects”
Section titled “”I get it” is not “I know what the professor expects””There’s a specific risk in using the AI as a study tutor: you hold a nice conversation, the AI compliments you, you leave the session convinced you know the topic. Then the professor asks their question and you freeze.
The reason is that holding a conversation is one thing, answering a precise question is another. In a conversation you can wander, reformulate, let the chat context feed you the words you’re missing; at the oral exam you’re alone, with a professor waiting for an answer that uses the textbook’s words.
A quick check that always works, at the end of a session: close the chat. Open the book to the chapter. Close that too. Explain the topic out loud, alone, as if you were at the oral. If you stumble, mark the exact point where you stumble: usually it’s the vocabulary (you’re missing the textbook’s technical word) or the connection (you know the individual facts but not how they fit together). For the vocabulary you go back to the textbook and write the definitions out by hand; for the connection you go back to the AI and ask it to give you three connection questions on that chapter, no more definitions. It’s a variant of the Study method: you explain a topic out loud as if you were teaching it to a child. Where you stumble, that's where you don't really know it. , applied to the course syllabus.
What NOT to do
Section titled “What NOT to do”Don’t ask “explain it to me easy” and stop there. The AI simplifies happily, but the exam asks for the level of the course, not the bar-stool level. The easy explanation is the first step, not the last: after you’ve understood, go back to the textbook’s technical vocabulary.
Don’t use the AI’s voice instead of your own at the oral. Answers built with the AI sound polished, and at the oral they sound fake. The professor immediately tells the difference between an answer you’ve chewed through yourself (even with some stumbles) and one recited from memory with a chatbot cadence. Study with the AI, answer with your own head.
Check what you’ve understood
Section titled “Check what you’ve understood”What comes next
Section titled “What comes next”Once you’ve understood a topic, you need to turn the material you have into something studyable: notes taken in class, lecture audio, the professor’s slides, textbook chapters. The next lesson takes this step in hand.