How Fields Medalists Study Math, Habit by Habit

If you searched for how Fields Medalists study math, the honest answer is probably less glamorous than the search result promised. There is no citable medalist method that turns ordinary students into research prodigies. There are, however, documented work habits that survive contact with a normal exam calendar: doing problems before reading solutions, staying with difficulty long enough to learn from it, checking understanding by re-deriving and explaining, choosing work just beyond the current range, and protecting a few serious blocks instead of pretending every study hour is equal.

The map below separates source evidence from article synthesis. A documented case means a mathematician said or did the thing in the cited source. Article synthesis means the habit category is being constructed here from several sources and translated into GRE quant, SAT/ACT math, or MCAT chemical and physical foundations practice. If you like that kind of evidence audit, the same habit-versus-outcome caution appears in this Bryan Johnson habits audit for students.

The evidence-labeled habit map

A caution about recency: the Tsimerman and Deng profiles cited here were published on July 23, 2026, only days before August 1, 2026. The quotes are useful, but anyone reusing the 2026 prize details later should re-check the source pages rather than treating a fresh profile as settled archival context. [3][5]

Start with the problem, not the explanation

The most portable habit is also the least cinematic: work the problem. Tao’s “Work hard” essay is useful here because it refuses the clean myth. He writes that mathematical breakthroughs “are not powered solely (or even primarily) by Eureka moments of genius, but are in fact largely a product of hard work,” and he includes “a serious amount of reading and writing, and not just thinking” in that work. [1]

For an exam student, the relevant move is not to imitate a research mathematician’s reading list. It is to reverse the order that anxious studying often takes. The weak version is: read the chapter, watch an explanation, nod along, then try one or two familiar examples. The stronger version is: attempt questions first, record where the solution path broke, then use explanations to repair that break.

That applies differently by exam, but the behavior is the same. In GRE quant, start a topic session with problems that force you to choose a method, not just apply the last formula you saw. In SAT or ACT math, do enough mixed questions that you have to recognize the type under time pressure. In MCAT C/P, work passage-based questions before rereading the content summary, because the exam often tests whether you can use a principle inside a passage rather than recite it in isolation.

This is also where AI tools can either help or quietly ruin the session. A solution generator is useful after you have made an attempt and can compare its reasoning against your own error. It is much less useful as a substitute for the first attempt. The same practice-first and error-review framing is the point of this guide to using ChatGPT for exam prep.

A practical version for tonight

• Pick one narrow target: inequalities for GRE quant, functions for SAT/ACT math, fluids or electrochemistry for MCAT C/P.
• Attempt a short set before opening the explanation.
• Mark each miss by failure point: concept not known, setup wrong, algebra or arithmetic slip, time pressure, or misread wording.
• Read the explanation only after you can name what you tried.
• Redo the missed problem later without looking, because recognition is not the same thing as retrieval.

Stay with one hard problem long enough to learn from the obstruction

Tsimerman’s line about competition math is worth using carefully. He says it teaches you “to sit with one problem for many hours, even though you’re almost certainly stuck.” [3] That is a research and competition-math habit, not a command to spend many hours on every GRE, SAT, ACT, or MCAT question. Timed exams punish that literally. The transferable part is the tolerance for being stuck long enough to inspect the blockage.

Many test-takers leave hard problems too quickly during review and too slowly during the test. During the test, you need triage. During study, you need friction. If every wrong answer is immediately converted into “I watched the explanation,” the mistake has not done its job. You have only outsourced the uncomfortable part.

Schwartz gives a useful antidote to the shame around slowness. In the quote collected by Almossawi, he says, “I was, and still am, rather slow. I need time to seize things because I always need to understand them fully.” [4] For exam prep, that does not mean slow work is always better. It means slowness during review can be productive when it produces a sharper understanding of the exact step you missed.

A good review block on one hard problem might look like this: first, rewrite the question in your own words; then identify what information is given and what is being asked; then try a different representation, such as a diagram, equation, table, or units analysis; then compare your path with the official explanation. If the official solution uses a move you never considered, the review target is that move, not the answer choice.

This matters most for GRE quant comparison traps, SAT/ACT function and geometry questions, and MCAT C/P passages where the equation is familiar but the setup is not. The useful lesson is often not “learn this formula.” It is “notice that the problem was asking for a ratio, not an absolute value,” or “the passage gave the proportional relationship, so plugging numbers was optional.”

Verification is where solved starts becoming learned

Deng’s profile supplies the most physical image of verification. As reported by Quanta, he read a 75-page technical PDE paper line by line, 8 hours per day for a week, and summarized the work this way: “I just checked every single detail.” [5] That is not an exam-prep schedule. It is a case showing what full verification can mean when the work is genuinely technical.

Tao’s problem-solving advice makes the exam translation clearer. One of his checks for understanding is whether you can “explain the solution to a classmate.” He also suggests playing with a solved problem by removing hypotheses, strengthening conclusions, or changing conditions. [2] That is a better standard than “I understood it when I read it.”

For standardized tests, verification has to be smaller and more repeatable. After a missed GRE quant problem, re-derive the solution with the explanation closed. After a missed SAT or ACT math item, explain why the tempting wrong answer was tempting. After a missed MCAT C/P question, state the principle, the passage clue, and the calculation or reasoning step that connected them. If you cannot say those three things, the problem is not finished.

The strongest version of an error log is not a museum of wrong answers. It is a re-derivation queue. A useful entry records the failed move, the corrected move, and a future trigger: “When I see a percent-change comparison, define the original value first,” or “When units contain seconds squared, check whether acceleration is hiding in the passage.”

Timed benchmarks still matter, because exams are timed even when review is slow. The Part 107 drone exam study plan is a useful comparison point for building a plan around timed practice and review checkpoints rather than vague “study more” intentions. The exam is different, but the planning problem is familiar: practice has to produce evidence about timing, accuracy, and repeat mistakes.

Choose difficulty that is slightly uncomfortable, not theatrical

Calibrated difficulty is the easiest habit to distort. Too easy, and you are polishing fluency you already have. Too hard, and you are mostly collecting discouragement. The sources do not give a neat formula for the perfect difficulty level. The category is an article synthesis from the way these mathematicians talk about rebuilding understanding, staying patient, and working past the first smooth explanation.

Tao’s “Learn and relearn your field” advice is useful because it treats knowledge as something that must be revisited and reorganized, not simply acquired once. [6] Mirzakhani’s Guardian interview adds the emotional texture: “the beauty of mathematics only shows itself to more patient followers.” [7] Neither source is telling a student to make an MCAT physics set impossibly hard. They do support a less fragile view of difficulty: confusion is not automatically a sign that the session is failing.

For GRE quant, calibrated difficulty might mean moving from single-topic drills into mixed quantitative comparison sets once the basics are stable. For SAT/ACT math, it might mean adding questions where the tested idea is disguised by wording or geometry. For MCAT C/P, it often means passage-based practice after content review, because knowing the equation is easier than choosing and adapting it under passage constraints.

Tao’s suggestion to play with a solved problem is especially useful here. [2] Once you solve a question, ask a small variation: What if the quantity were doubled? What if the answer choices were removed? What if the graph were described in words instead? These are hypothetical variations for practice, not claims about actual exam items. The point is to make the solved problem less brittle.

Difficulty should change what you do next. If errors are mostly content gaps, return to targeted review. If errors are mostly setup failures, do more mixed problems and write the first line of reasoning before calculating. If errors are mostly timing errors, use shorter timed sets and review the decision to skip, estimate, or compute. One kind of “hard” does not call for one universal remedy.

Protect deep work without worshiping the clock

Huh’s work routine is useful precisely because it does not sound heroic. In his Quanta profile, he describes about 3 hours of focused work per day and then says, “Then I’m exhausted.” He also says, “intention and willpower are highly overrated.” [8] That is one self-report from one mathematician, not a study-time prescription. It does puncture the fantasy that serious mathematical work always means clean, all-day intensity.

For exam prep, the lesson is to protect the part of the day when actual thinking can happen. A serious block is not just time near a book. It is a block with a defined task, a working surface, problems attempted before explanations, and enough quiet to notice why a solution path is failing. Some students can do that after dinner. Some need morning. Some need shorter blocks because work, school, or caregiving leaves no pristine schedule.

Hour totals still matter at the planning level. A student who never opens a practice set will not be rescued by elegant habits. But hour-count worship creates its own bad data. Three distracted hours of rereading can feel morally superior to one concentrated block of mixed problems and error review, while producing less evidence about what will happen on test day.

A weekly planner should therefore reserve a few blocks for the work that cannot be multitasked: timed sets, hard-problem review, and re-derivation. Lighter tasks can sit elsewhere: formula review, flashcards, organizing the error log, or watching an explanation after the attempt. If you need a planning shell, put the routine into a study-planners template rather than trusting the plan to remain in your head.

A modest routine built from the habits

The useful exam routine is small enough to survive an ordinary week. It does not require pretending to be Tao, Huh, Deng, Tsimerman, Schwartz, or Mirzakhani. It borrows the structure of the habits and leaves the mythology alone.

1. Attempt before reading. Start each serious GRE quant, SAT/ACT math, or MCAT C/P block with problems, not explanations.
2. Stay with one hard problem during review. Do not spend unlimited test-day time on it, but do spend review time finding the exact obstruction.
3. Re-derive and explain. Close the solution, redo the problem, and explain the decisive step as if a classmate had asked why it works.
4. Adjust difficulty from evidence. If everything is right, mix the topics or add timing. If everything is wrong, narrow the topic and rebuild the prerequisite.
5. Protect a few real blocks. Put the most demanding practice where your attention is best, and move administrative study tasks to lower-energy time.

That routine is not glamorous, but it gives tonight’s study session a job. The first job is not to feel inspired. It is to produce a more accurate next attempt.

Where the transfer stops

Research math and standardized-test math are not the same activity. Research is open-ended, untimed, and creative in a way that GRE quant, SAT/ACT math, and MCAT C/P are not. Exam math is timed, closed, scored, and format-bound. A Fields Medalist habit does not tell you which MCAT equations are high yield or which SAT question types deserve the most practice.

The habits transfer as practice structure: attempt first, tolerate difficulty during review, verify by re-deriving and explaining, calibrate the next set, and protect real thinking time. They do not transfer as a secret content strategy or a promise of genius.

References

1. Work hard — Terry Tao’s blog
2. Solving mathematical problems — Terry Tao’s blog
3. Jacob Tsimerman Wins 2026 Fields Medal for Andre-Oort Conjecture Proof — Quanta Magazine — July 23, 2026
4. Slow and Fast Learners: 3 Quotes — Almossawi
5. Yu Deng Wins the Fields Medal 2026 for His Work on the Random Data Problem — Quanta Magazine — July 23, 2026
6. Learn and relearn your field — Terry Tao’s blog
7. Interview: Maryam Mirzakhani, Fields medal winner — The Guardian — Aug. 13, 2014
8. June Huh, High School Dropout, Wins the Fields Medal — Quanta Magazine — July 5, 2022