The task
You just finished reading the 40-page calculus chapter on definite integrals. The end-of-chapter summary in the textbook is the standard generic version — “in this chapter we explored the fundamental theorem and several integration techniques” — which doesn’t help you know what your specific professor will test or which worked example pattern shows up in exams every year. You want a 3-level summary: a plain-language gist that survives a week of distance from the chapter, 3 worked examples chosen for the patterns the exam loves, and 5 ranked “test bait” items grounded in past exam questions from this course — each with the trap that catches everyone.
Where AI helps — and where it does not
AI is genuinely good at compressing dense academic content into a plain-language gist, identifying canonical worked examples (the patterns that show up in every textbook), and reorganizing chapter content by “likely exam topic” instead of by section order. For STEM, it can name the algebraic moves or substitutions that distinguish a hard problem from an easy one.
What AI cannot do reliably: predict your specific course’s exam style without examples. Different professors test different patterns from the same chapter — one will hammer the chain-rule edge cases, another will only ask straightforward applications. Feed 2-3 past exam questions from your course so the model can calibrate. AI is also poor at humanities argument structure when given only the source material; for argument-heavy chapters, feed it the seminar discussion or the professor’s lecture notes too.
A specific failure mode: AI tends to write a “gist” that’s just a chapter paraphrase (“This chapter covers definite integrals, applications, and…”). Tell it: “the gist should be how you’d explain this to a smart friend who never took calculus — one core analogy, no jargon, and a sentence on why anyone should care.”
What to feed the AI
- The chapter text (paste in chunks if long), or your existing notes if you already digested it
- 2-3 past exam questions from your course covering similar chapters (this is the calibration step that makes test-bait useful)
- Your current stage: first read / second read / revising for exam (changes depth)
- Your professor’s known testing pattern, if you have one (algebra-heavy, conceptual, application-heavy, proof-based)
- 1-2 specific worked examples from the chapter you already half-understood (so the model knows your baseline)
- The chapter that came before and the chapter coming next (chapter-to-chapter connection is half the value)
- Topics you skimmed because they felt dense (the model should explicitly cover those at higher depth)
- Whether you have problem sets done already and which problems you got wrong (those are your real test-bait)
Copy-ready prompt
Summarize this textbook chapter at three levels.
Chapter content (or my notes): {paste}
Stage: {first read / second read / revising for exam}
Past exam questions from this course (2-3, to calibrate exam style): {paste}
Professor's known testing pattern: {paste or "unknown — infer from past exams"}
Topics in this chapter I skimmed: {paste}
Prior chapter + next chapter (for context bridges): {paste}
Problems I already got wrong (these are real test-bait): {paste or "none yet"}
Return:
1) Gist (under 120 words). Write as if explaining to a smart friend who never took this subject — one core analogy, no jargon, and a sentence on why the chapter matters. Not a paraphrase of the textbook intro.
2) Worked examples — 3 examples chosen for the patterns the past exams show. For each:
- The setup ("given X, find Y")
- The trick or the move that opens the problem (the part students miss)
- The solution, step-by-step but compressed
- The 1-sentence "why this pattern matters"
3) Test bait — 5 things this professor will likely test, ranked by frequency in the past exam questions I gave you. For each:
- The pattern in 1 sentence
- The trap or common mistake (the part graders dock points for)
- One sentence on how to recognize the pattern in 20 seconds when the exam clock is running
4) Notes to self — connection to the prior chapter (what this chapter assumes) and the next chapter (what this chapter sets up). 2-3 sentences.
5) Honest gaps — if any topic in this chapter is unusually under-represented in past exams (i.e., the textbook covers it but the exam doesn't), say so explicitly. Studying everything equally is a study strategy mistake.Shorter variant — exam-week compression
Compress this chapter to the 3 things the exam will test. Past exam style: {paste}. For each, give me the pattern, the trap, and one practice problem I should redo from the chapter. Skip the gist; I've read the chapter.Sample output
A useful gist (calculus, definite integrals): “Definite integrals are ‘how much accumulated’ — if a velocity tells you speed at each moment, the integral tells you total distance over a time range. Most chapter content is just techniques to compute the answer when the function is too messy for the obvious shortcut. The big idea most students miss: when the function changes sign, the integral cancels the negative part — which is why an absolute-value integrand changes the whole problem.”
A useful worked example: “Setup: ∫ from -2 to 3 of |x² - 1| dx. Trick: don’t integrate |x² - 1| directly. Split the interval at the zeros of (x² - 1), which are x = -1 and x = 1, then flip the sign of the integrand on the interval where (x² - 1) is negative. Why this matters: every exam in this course has had at least one absolute-value-inside-integral problem; the cancellation trap is the #1 docked-points error.”
A useful test-bait line: “Pattern: absolute value or square-root inside a definite integral with a sign change in the interval. Rank: #1 (appeared on 4 of the 5 past exams). Trap: forgetting to split the integral at the zeros of the inner function — students integrate as if the function were positive throughout. Recognition: see absolute value or even-root and check whether the inner function has a zero inside the bounds; if yes, split.”
A useful note-to-self: “This chapter assumes you can find zeros of polynomials by inspection (Ch 2) and that you remember the chain rule (Ch 4). Next chapter builds on this with techniques like substitution and integration by parts; the absolute-value pattern from this chapter shows up there as a subproblem inside u-substitution.”
How to refine
- Force a real gist: “Re-read your gist. If it sounds like a paraphrase of the textbook intro, rewrite as if explaining to a smart friend who never took this subject. One core analogy. No jargon. Include why anyone should care.”
- Ground test-bait in past exams: “For each test-bait item, cite the past exam question(s) it maps to. If you cannot cite a past exam, label as ‘pattern from textbook, not yet observed in exams’ and rank lower.”
- Add the 20-second recognition cue: “For each test-bait pattern, add one sentence: how do I recognize this pattern in 20 seconds when the exam clock is running? Recognition is what saves points; technique without recognition fails.”
- Surface the under-tested topics: “Re-read the chapter against the past exams. If any topic gets significant pages in the textbook but never appears on exams, flag it explicitly so I don’t over-study it.”
- Match depth to stage: “If I said ‘first read,’ lean heavier on the gist and the connections to prior chapters. If I said ‘revising for exam,’ compress the gist and double the test-bait section. Stage drives the weighting.”
Common mistakes
- Reading the summary instead of reading the chapter — the summary is a study tool, not a substitute; AI summaries miss the nuance that builds real understanding
- Trusting “test bait” without past-exam grounding — every professor tests differently from the same chapter; without past exams, the test-bait is generic and likely wrong for your course
- Skipping “notes to self” — chapter-to-chapter context is half the value of the summary; an isolated chapter summary misses why the chapter exists
- Treating worked examples as the only practice — the worked examples show patterns; you still need to do problem sets to encode the moves
- Asking AI to “make it easier to understand” without specifying confusion — “easier” produces noise; “explain step 4 of the worked example without using the chain rule” produces signal
- Letting AI invent past exam patterns — if you cannot give past exams, ask for “patterns common in textbooks at this level” and label them as such, not as “this exam will test”
- Studying every topic equally because the textbook covers them all — past exams reveal weighting; let the model rank
- Using the summary the day of the exam — the value is in pre-exam study; the day of, redo problems, don’t re-read summaries
FAQ
- For STEM vs. humanities chapters — does this work differently?: STEM benefits from the worked-examples layer because exams test moves; humanities benefits from an “argument structure” layer instead — what’s the thesis, what counter-arguments does the chapter address, what evidence does it use. Swap “worked examples” for “argument map” in the prompt.
- Should I show this to my professor as proof of study?: Use it to check your own understanding, not as proof. Most professors will recognize AI-summary writing in 10 seconds; if they ask you to explain a worked example you “summarized,” you need to be able to do the math live.
- What if I don’t have past exams from this course?: Ask the professor’s office, a TA, or upperclassmen. Lacking that, use exams from a similar-level course at another school as a rough calibration — note in the prompt that the exam style is approximate, and treat test-bait as directional.
- The model keeps writing a gist that paraphrases the textbook — what changes?: Add: “The gist must use a core analogy or metaphor from outside the field. If the gist contains any phrase that could appear in the textbook itself, rewrite. Imagine explaining to a friend who has never studied this subject; the gist should leave them with one mental image and one reason to care.”
- How long should this take?: 10-20 minutes after a 40-page chapter. If you’re spending 45 minutes on the summary, you’re using it as study-avoidance — get back to problem sets.
Related
- AI lecture note cleanup
- AI explain difficult concept
- Research Summary Prompts: From Source to Brief
- AI Socratic Study Buddy
- AI Helps You Run a Weekly Study Reflection
Tags: #AI writing #Learning #Workflow #Study