Evan Chen

School, the Bay Area, and the discovery of unfamiliar problems

Alongside school, Chen taught himself algebra from a workbook. A seventh-grade math-club tryout in 2008 introduced him to an old MathCounts paper full of unfamiliar questions. He began sustained contest practice in eighth grade; a strong result in the inaugural USA Junior Mathematical Olympiad in 2010 encouraged him to continue.⁶

The attraction was the unfamiliar problem. School exercises usually arrived after an applicable method had been taught; these questions made deciding what to do part of the task. Later Chen would spend much of his teaching time helping students through that uncertainty. At this early stage, however, the evidence describes enthusiasm and practice, not a fully formed educational philosophy.

Zvezda Stankova taught him at the Berkeley Math Circle in 2010–2014 and encouraged university coursework. He also began writing problems and coaching his former middle school’s MathCounts team.⁷ His surviving notes make the coursework concrete. In spring 2012 he took honors abstract algebra with Kelli Talaska at Berkeley, followed by algebraic topology at San José State that fall. In 2013 he took courses in graph theory, real analysis, and algebraic combinatorics. These were opportunities to encounter mathematics beyond the school curriculum before leaving high school.⁸

He attended the US Math Olympiad Program, or MOP, in 2011–2012, under Steve Dunbar and Zuming Feng. A poor 2013 USAMO result excluded him from that summer’s camp; he attended Research Science Institute instead. Under the selection rules then in force, the result also closed his route to the 2014 US team.⁹

Another formative setting was SPARC, the Summer Program for Applied Rationality and Cognition. He attended in 2013 and 2014 and returned as a counselor in 2015. His 2016 account describes a camp interested in applying quantitative reasoning to ordinary life: assessing uncertainty, noticing activities that were not worth continuing, understanding aversion, and protecting time. He also valued peers with whom reflection was socially normal. These were practical experiments, sometimes as simple as trying to climb a tree, rather than only more advanced mathematics. In March 2018 he added an unusually consequential qualification: both his values and the camp’s direction had changed, so the earlier description no longer fitted well. SPARC belongs in the history of his formation without being made into a permanent doctrine explaining his subsequent life.^{10, 11}

Chen remembers doing contest problems when he wanted to, sometimes mentally during everyday journeys, rather than following a rigid training schedule. His retrospective estimate is ten to twenty hours a week, varying considerably. Both parents came from Taiwan, and he held US and Taiwanese citizenship. That citizenship made another national selection process available when the US route closed.¹²

Adults helped open these routes, while peers made them enjoyable. The same institutions also selected and excluded.

Taiwan in 2014: a second route and a new group of peers

In his senior year, Chen flew back and forth to Taiwan rather than abandon the possibility of competing at the IMO. His personal report describes attending the first and third selection camps in spring 2014 and taking the second camp’s tests remotely. The camps mixed lectures, quizzes, selection examinations, shared meals, and informal problem discussion. They were also a language environment. He wrote Chinese handouts on inequalities and complex-number geometry, partly to practice Chinese, and classmates helped identify mistakes. Communication continued on Facebook between meetings.¹³

In the Taiwan camps, he sometimes found the lectures frustrating because they left too little time to try problems. Informal work with other students after class was more satisfying. His report thus records both training and a growing opinion about what made training useful. Writing explanations in Chinese was also a means of joining the group, with classmates helping him correct the language.¹⁴

At the July 2014 IMO in Cape Town, Chen represented Taiwan as TWN2. He scored 36 of 42 points—7 on each of the first five problems and 1 on the sixth—and won gold in joint twelfth place.¹⁵ He had also taken the USAMO from Taiwan, at 12:30–5:00 a.m., becoming a national winner that year.¹⁶

His Cape Town diary contains more than solutions. He served as a photographer and translator for the team, dealt with accommodation and charging problems, played cards, met contestants from other countries, and updated his report on a laptop while waiting. He wrote his examination solutions in English because he thought this would make coordination easier for the team leader. The diary presents mathematical activity embedded in travel, language, friends, inconvenience, and leisure. It also shows an already-established habit of documenting experiences while they were happening.¹⁷

His two national routes also gave him overlapping groups of friends. The Taiwanese experience required travel, testing, and learning to explain himself in Chinese; the US camps and online circuit remained part of his world. His diary’s social and logistical detail makes this a particular experience of Taiwanese American education, rather than a national label that explains everything about him.

Writing before college, and the decision to transfer

At Irvington High School, Chen began turning geometry-club notes into Euclidean Geometry in Mathematical Olympiads. In fall 2013 he had only two regular classes. School officials required three office-assistant periods, but the work left long intervals in which to write. He submitted the manuscript in April 2014; the digital book appeared in March 2016, with print later. Writing therefore began before his IMO medal and continued through his first two university years.¹⁸

Writing had its own place in his formation. In a March 2015 essay, he describes disliking English classes but discovering that writing helped clarify thoughts. Old notes could expose recurring mathematical ideas that had not been visible while he was solving individual problems. He also describes writing as a place to revise beliefs rather than preserve a performance of certainty. Yan Zhang and the SPARC environment appear in this account as influences on thinking about disagreement; Paul Graham’s essays provided another language for revision and composition. These are declared intellectual connections, not proof that Chen adopted an entire philosophical school.¹⁹

The passage into college had already begun experimentally. In February 2014, while visiting Boston for the Harvard-MIT Math Tournament, he sat in three Harvard mathematics lectures. He followed two, could not follow the third, and posted the notes anyway.²⁰ His first Harvard year included Dennis Gaitsgory’s Math 55 sequence, number fields with Mark Kisin, and courses in algebraic geometry and set theory. He was learning advanced material and recording it as he went.²¹

He entered Harvard in 2014 and transferred to MIT in 2015. His contemporaneous explanation rejects the easy prestige story. He praised Harvard’s mathematics department and its curricular freedom. The decisive issue, he said, was housing and social life. Taking advanced classes placed him among older students who lived and ate elsewhere, while first-year living and dining arrangements left him isolated. He spent part of the year living in an MIT dorm before formally moving. MIT’s ability to choose a living community mattered more to him than the abstract comparison between departments.²²

The transfer changed Chen’s daily company: whom he could live and eat with. His account places that experience beside the courses in explaining why he moved.

He completed the MIT mathematics bachelor’s degree in 2018 and continued into doctoral study.²³ In a 2023 reassessment, he regretted how narrowly academic his undergraduate life had been: he wanted more time with friends and more music, linguistics, humanities, and teaching courses. He remembered much of his mathematical coursework as broad, shallow, and subsequently forgotten. That regret sits alongside the earlier pleasure in unfamiliar lectures. It records a change in his sense of what college time was for.²⁴

The transfer also brackets the beginnings of two long projects. He started the exposition project An Infinitely Large Napkin in December 2014 and released its first public drafts in July 2016. OTIS began in 2015. Books and teaching were developing while he was still finding an undergraduate home, not being added only after a research career had been established.^{25, 26}

He also worked for educational organizations. At Expii in 2013–2015 he developed the platform’s markup language; at Art of Problem Solving in 2014–2020 he assisted teaching and designed material, especially for Worldwide Online Olympiad Training.²⁷ Content and software were already linked in his work before OTIS became his principal independent operation.

Learning to do research, and the long MIT period

Emory’s 2015 REU, supervised by Ken Ono, introduced Chen to undergraduate research.²⁸ He then worked with Joe Gallian at the Duluth REU in 2016–2017, on number theory and combinatorics.²⁹

At Emory he collaborated with Peter S. Park and Ashvin Swaminathan on a question about elliptic curves, equations whose solutions can be counted modulo a prime. Their paper sought bounds for the smallest prime satisfying a specified condition on that count, in the special case of curves with complex multiplication. The acknowledgments name Ono’s supervision, Jesse Thorner’s advice, and a useful reference supplied by Jean-Pierre Serre; NSF funding supported the REU. This was a small research group using a mentor’s guidance and an existing literature, rather than a contestant working alone toward a known answer.³⁰

His July 2015 exposition shows what learning the project involved. He moves from arithmetic progressions and Linnik’s theorem about least primes to the behavior of elliptic curves, using characters, L-functions, and complex analysis to connect the questions. He explains which older results their argument needs, and frankly says that some details are still beyond his knowledge. Writing this early paper was therefore also an apprenticeship in learning enough background to adapt a proof, recognizing a special case, and explaining a result before feeling expert in every ingredient.³¹

Another undergraduate project, published in the Electronic Journal of Combinatorics in 2018, constructed partially ordered sets with unusually small balance constants. In everyday terms, a partial order specifies some precedences without completely ranking every pair. Researchers ask how balanced the possible complete rankings can be. Chen’s family approached a particular numerical bound and answered a question in that area. It made progress around the one-third–two-thirds conjecture; it did not prove that general conjecture. This combinatorics work sits beside the number theory, showing that his undergraduate mathematical formation was broader than his later thesis specialty.³²

He mentored PRIMES/CrowdMath research in 2016–2017 and designed entrance examinations in 2016–2021.³³ A research mentor had to help organize work with an unknown outcome, alongside teaching established results.

Chen’s 2018–2025 MIT doctorate was supervised by Wei Zhang. His January 2025 thesis develops explicit formulas in particular cases of conjectural identities between analytic expressions and arithmetic-geometric quantities.³⁴ A related paper formulates a semi-Lie extension of the arithmetic fundamental lemma and proves its first nontrivial case. These are technical names for comparisons between mathematical objects described in different languages; the achievement is a bounded case of a larger program.³⁵

Chen thanks Zhang for patient, prompt help through the qualifying exams and thesis. Ryan C. Chen’s reading and discussions helped remove theorem hypotheses; a fall 2024 seminar improved the work, and Mark Sellke proofread it. Departmental encouragement, administrative assistance, and NSF graduate fellowship support accompanied the research. He calls MIT his home of ten years.³⁶

The economics of graduate study were visible to him before entering the doctorate. In November 2017 he opposed a proposed tax treatment of tuition waivers, arguing that a nominal tuition benefit could not be treated like spendable stipend income. His illustrative figures describe a hypothetical MIT student at that date, not his own earnings. The intervention shows his attention to the arrangements that made prolonged study affordable.³⁷

Nor did Chen regard research as the inevitable destination of every successful contestant. In a 2016 essay, he objected to treating software, finance, teaching, or other careers as defections from the supposedly proper academic route. He argued that contests could have value in themselves, through demanding work, interesting problems, and friendships. His addendum conceded two complications: the distinction between contests and research could encourage people with weak contest results, and the olympiad community itself could be elitist and closed. This was an ongoing dispute about what his community valued, while he was doing research himself.³⁸

Returning to camps and becoming responsible for examinations

His early classroom experience involved learning how much a teacher could misjudge. In December 2014 he taught five students at an A-Star winter olympiad camp. Four mornings, one for each main contest subject, gave him three-hour blocks: theory and examples, time to try problems, individual conversations, and student presentations. He wanted each student to explain a solution rather than merely listen. Afterwards he admitted that familiar material had bored the class, practice problems needed too many hints, and the morning workload had tired students before their afternoon sessions. A student’s silence could mean prior knowledge rather than confusion. This account, written in February 2015, makes the young instructor’s enthusiasm and mistaken estimates equally visible.³⁹

Chen’s relationship to the American olympiad system changed from participant to worker. His MOP archive distinguishes organizing from title: he was an organizer from 2016 through 2025 and assistant academic director from 2018 through 2025. The responsibilities included teaching, grading, examinations, materials, and the running of a temporary student community. MOP operated at Carnegie Mellon under Po-Shen Loh during much of this period; the 2020 and 2021 camps moved to Discord during the pandemic, and the 2025 camp was hosted by the Illinois Mathematics and Science Academy.⁴⁰

His 2025 advice to a MOP student emphasizes the unusual value of living with teachers and peers. Three weeks, he argues, cannot substitute for steady work during the year. The benefit is seeing someone make a move one would never have considered, then using that surprise to guide later practice. He remembers a classmate’s unexpected use of quadratic reciprocity while solving a problem together in 2011. That encounter was still useful almost fifteen years later. Camp time also included leisure: he reports eleven K-pop rehearsals against eight practice exams in 2025. The social environment and the mathematical examples were both part of what he wanted students to notice.⁴¹

Responsibility for contests also produced public failures. In a 2018 apology, Chen recalls setting HMMT’s February 2016 problems so hard that more than a hundred students solved at most two of thirty individual questions. He also acknowledges a major 2017 scoring error that changed the leading results. Coaches’ replies reminded him that most participants would never place highly, but their experience still mattered. He adopted ARML’s distinction between a problem that nearly everyone might solve on a good day and an insultingly trivial giveaway. Several attainable problems, plus attention to the whole exam’s time limit, mattered more than one easy opening question. His concern for access thus included correcting examinations he had helped make discouraging.⁴²

In a 2019 text framed as a speech he would like to give, he argues that a camp cannot be justified solely by the six students selected for the IMO. Most attendees will not be on that team, and their learning, friendships, and experience still matter. The qualification is essential: this was a proposed speech, not one he delivered as camp director. Another essay that year criticized the lack of supportive feedback around repeated selection tests. Together they document a staff member questioning how a selective institution treats the people it does not finally select.^{43, 44}

Chen was USAMO editor-in-chief in 2019–2022 and helped design and grade US team-selection exams.⁴⁵ His later international IMO coordination was a separate responsibility from coaching a national team.

Another account concerns developing the US online olympiad USEMO. Problems pass through proposals, review, and test-solving over several stages. Reviewers assess attractiveness as well as difficulty; test-solvers expose alternate approaches and errors. Chen describes shuffling packets and withholding labels to reduce preconceptions, combining ratings with written comments, and balancing a complete examination rather than choosing the highest-rated problems independently. The labor is distributed, and a polished six-problem paper hides a much larger pool of mathematical work.⁴⁶

The international grading work required a very different rhythm. For IMO 2024’s “Turbo the snail” problem, Chen and his colleagues first solved the question, classified possible optimal strategies, and spent a day preparing a marking scheme. He made its illustrated appendix, then read scripts from roughly 3 p.m. to 1 a.m. before coordination began the next morning. Country leaders met coordinators to agree scores, supplying translation where needed. Explaining the rubric repeatedly was part of maintaining fairness. In 2025 he handled around a hundred scripts at his table, finding handwriting sometimes harder than language and wishing the scheme had anticipated particular near-solutions. These reports show diplomacy, fatigue, interpretation, and teamwork within a job often summarized merely as “grading.”⁴⁷

His September 2026 update records a third year of on-site coordination, at that summer’s IMO.⁴⁸

The resulting craft involves more than inventing something hard. His 2021 problem-writing essay separates aesthetic qualities from the problem’s ability to distinguish performances. A short, natural statement may invite students to begin; an unexpected or elementary solution can make the experience memorable. Difficulty alone cannot supply these qualities. Conversely, an attractive problem may fit poorly into a particular exam if its placement or prerequisite knowledge produces an unhelpful distribution of results.⁴⁹

His 2024 critique of grading rubrics adds a disagreement that a celebratory account would miss. He worries that additive partial-credit schemes can reward students for accumulating predictable fragments instead of pursuing complete solutions. He favors rules that encourage genuinely solving the problem and opposes turning grading knowledge into an additional contest. This is his position within a professional debate, not a universal account of how all olympiad scripts ought to be judged. It also puts pressure on his educational ideal: broad encouragement and rigorous assessment do not always call for the same institutional decisions.⁵⁰

OTIS: from a small class to an independent teaching organization

OTIS began in 2015 with five students and conventional lectures. By Chen’s November 2023 account, it served approximately four hundred students, and its center of gravity had shifted to written material, a course website, and asynchronous conversation. He came to regard writing as something that could be corrected and reused more effectively than a performance delivered once. Discord, introduced to the program in 2019, made timely exchanges possible without requiring everyone to attend the same meeting. This was a development over time: a 2016 essay had still defended the indispensability of in-person contact, a position he later explicitly reconsidered.^{51, 52}

The full name, Olympiad Training for Individual Study, describes its intended arrangement. Students work independently on proof-based problems; the organization provides selected material, ways to ask questions, and a continuing community. By 2026 its public name was OTIS Math LLC. It admitted students internationally and without a fixed age band. Admission checked mathematical and English-language readiness, written work, and suitability for this form of learning rather than ranking applicants in a merit competition. Chen also published a list of authorized outside instructors who could use the platform under their own arrangements.⁵³

This should not be confused with a beginner’s introduction to all mathematics. A student needs enough background to start writing proofs and enough independence to benefit from written explanations. Those requirements are practical limits on whom the model serves. Removing age and national restrictions broadens access, but it does not eliminate differences in language, time, confidence, or prior opportunities to encounter proof-based work.

For 2026–2027, tuition was $240 per semester, with need-blind admission and full demonstrated-need aid. Late payment cost $60; late entry did not prorate fees. The September–December and January–April semesters had no core synchronous meetings. Students chose units across three tracks, submitted outlines for sanity-check feedback, and used hints, discussions, solutions, and optional tests.⁵⁴

In 2016 he had already questioned hourly tutoring prices: families paid for meetings while preparation, grading, correspondence, and exam writing appeared free. He argued that those tasks sustained students’ independent practice. The complaint explains why a visible teaching hour can misrepresent what a course provides.⁵⁵

His July 2026 CV reports over seven hundred students annually worldwide.⁵⁶ This later self-report follows approximately four hundred in 2023. Neither enrollment figure, combined with tuition, yields his income: aid, outside-instructor arrangements, staffing, and costs intervene. The public record establishes a tuition-supported teaching organization with subsequent sponsorship, while its accounts and his household earnings remain unavailable.

The course’s assignments also changed. In a 2018 essay, Chen describes moving away from short sheets on which nearly everything was mandatory toward larger banks of weighted problems. Students could choose a substantial portion while still encountering required material. The change emerged from a concrete dissatisfaction: students struggled with sheets he had made, and he wanted them to make useful decisions rather than experience unfinished work as a personal failure. The deeper problem bank made choice feasible; when every good problem was scarce and costly to vet, he had felt compelled to assign them all.⁵⁷

A larger archive made choice possible, but cost work to assemble. Chen explicitly considered the student’s hours, including time that might be spent on music or another activity. The redesigned sheet gave a student ways to stop or change direction while still making substantial progress. This also made the selection of material, rather than merely the quantity assigned, part of the teacher’s responsibility.

The course uses a game-like interface with progress, statistics, and hidden elements. Chen explained this design publicly in 2021, and the code is visible in the repository. Gamification can help represent progress and invite exploration, but it would be wrong to assume that an elaborate interface proves better mathematical learning. The more concrete question is what the game rewards: completing work, trying different topics, participating, and learning how to use the environment, rather than only placing above classmates.^{58, 59}

In September 2026, CJ Quines became OTIS’s first staff member. Quines had studied mathematics and computer science at MIT, grew up in the Philippines, and shared puzzle-hunt work with Chen, including OPAL 3. Chen simultaneously announced a Global Talent Fund STEM Talent Fellowship and credited its support with helping make the hire possible. The course page also associates that support with financial aid. After years of operating alone, he was beginning to distribute the work among a small team.^{60, 61}

What he tries to teach when he teaches problem-solving

Chen’s teaching does not reduce to a list of tricks, even though it includes techniques. In a 2017 essay about course design, he argues for organizing material around the main work a solver must learn to do. His combinatorics categories include recognizing when a global aggregate is more informative than local detail, and understanding the rigid structure of an object before attacking the requested conclusion. Topic names are useful only when they help students recognize a type of mathematical action. He also admits areas, such as functional equations, where he has not found a comparably satisfying classification.⁶²

For an unfamiliar reader, the distinction can be pictured without reproducing a contest solution. One student starts calculating the quantity the question asks for. Another first asks what sort of object could possibly meet the conditions, constructs examples, and learns which features are fixed. The second student may postpone the advertised target while acquiring the information needed to reach it. That is a teachable habit, although it cannot be supplied as a universal recipe.

A 2025 email to a student illustrates the habit in ordinary coaching. The student had tested possible sizes of a set and then run out of ideas. Chen urged them to explore what the set’s defining conditions meant, what examples existed, and how bigger examples might be built. His proposed analogy was a conversation driven by curiosity rather than an interrogation run from a checklist. He acknowledged that this described his instinct more readily than it yielded actionable instructions. The admission preserves an important limit: an expert’s natural orientation is not automatically a beginner’s usable method.⁶³

Another 2025 essay distinguishes a correct proof from a well-understood one. Two apparently different olympiad problems shared an underlying idea. A student could write a technically sufficient solution to the first while obscuring that idea so thoroughly that it would not transfer to the second. Chen therefore presses students to write clearly, not merely to satisfy a scoring threshold. Exposition becomes a test of what they have recognized, rather than decoration applied after the mathematics is finished.⁶⁴

He has also had to reconsider the way difficulty is discussed. His Math Olympiad Hardness Scale, introduced in 2019, supplied a rough vocabulary for challenge. By 2024 he complained that students were turning such numbers into status comparisons and unproductive arguments. Ratings could help someone choose beginner practice, but obsession with ranking replaced the more useful work of identifying the difficult step and learning from it. The teacher here is responding to an unintended use of his own tool, not simply criticizing a habit from outside.⁶⁵

These positions leave room for both technical expertise and the student’s independence. They require a teacher to know enough to curate good examples, explain what is happening, and notice a misleading explanation; they also require the student to try things whose outcome is not yet known. In Chen’s writing, successful instruction can therefore be hard to summarize as an immediate transfer of information. It often means changing what someone attends to when the obvious next step runs out.

Books, drafts, and the practical work of exposition

Euclidean Geometry in Mathematical Olympiads, often abbreviated EGMO, is a substantial book about solving proof-based geometry problems, with roughly three hundred problems and two hundred fifty diagrams. It combines synthetic arguments—reasoning directly with geometric relations—with algebraic methods such as coordinates and complex numbers. Its public page describes the necessary proof experience, supplies errata, and links translations into Chinese and Japanese. Those editions establish the book’s movement into other language communities, without establishing how many readers used it or what results it caused.⁶⁶

For EGMO’s revisions, Chen paid friends to check chapters and learned that editorial review and copyediting did not amount to complete mathematical verification. The MAA supplied a copyeditor, cover, production format, and marketing. He retrospectively allocated half the thirty-month process to review and valued the publisher’s practical support and prestige despite the wait.⁶⁷ The book entered the MAA Problem Books series as volume 27, a 311-page commercial publication, rather than remaining only his downloadable notes.⁶⁸

The continuing errata are part of the book’s professional life. A mathematical text must remain accountable after publication, because a small error can obstruct a reader for much longer than the author spent making it. Chen’s public corrections allow the reader to see a work that is maintained rather than a finished object protected by its author’s reputation. This is especially relevant to a book acquired on the strength of its author’s contest credentials.

The publisher’s description emphasizes worked examples that explain how a solution might be found, then hints and selected solutions for practice. Its selected excerpt from Mark Hunacek’s MAA review recommends the book to competition students and coaches. That is evidence of reception within an established mathematical publishing network, with the limitation that the complete review was unavailable. Chen’s own page is more explicit about the proof experience a reader needs.^{69, 70}

An Infinitely Large Napkin followed a different publication model. Its first public release in 2016 gathered explanations for readers comfortable with proofs but unfamiliar with higher mathematics. The current project ranges across undergraduate and some beginning graduate topics, with formal definitions accompanied by intuition and connections. It is not intended as a complete specialist reference. Its changing contents, missing sections, and requested corrections remain visible.^{71, 72}

The software repository makes some production mechanics explicit. The source uses LaTeX and Asymptote; contributors can propose corrections, and an automated workflow helps produce the downloadable PDF. The repository credits outside work on deployment. That division of labor is small but real: making prose, diagrams, source files, and public releases behave as one maintained publication is a different task from having written the chapters. A link to a community project using Typst and Lean does not mean Chen personally formalized the entire book in a proof assistant.^{73, 74}

The two books therefore occupy different institutions. EGMO passed through a publisher and commercial book distribution; Napkin operates as a continuing public project with community corrections. Both let readers work without being in Chen’s course. They also expose how much educational writing depends on selection, notation, illustration, and revision—work that cannot be inferred from knowing the underlying theorem.

An ordinary workday: students, software, and corrections

A September 2023 diary is one of the clearest descriptions of how Chen’s course operated. In the span from Sunday afternoon to Monday, he processed aid and registration, fixed website bugs, read submissions, answered mathematical questions, and rewrote solutions. A student’s question about a geometry diagram led him to change the explanation, redraw circles, move labels, and enlarge the figure. Another exchange revealed that his own answer was wrong; he asked the course community and Art of Problem Solving for help. This is one recorded interval, not a statistically representative timetable, but it shows the mix of work a lecture-based account would leave out.⁷⁵

The diary’s importance is not that it proves exceptional busyness. It shows how a student’s difficulty can reach several different layers of a course. The problem may be correct but the explanation unclear; the explanation may be adequate but the diagram hard to read; the content may work but a login prevent access. Improving the service may require mathematics, writing, drawing, customer communication, or code. None can safely be dismissed as peripheral if it is the thing stopping someone from learning.

OTIS-WEB is a public Python/Django course-management project, with a production server described as hosted by PythonAnywhere. Its repository includes registration, payments, rosters, assignments, examinations, and the game interface. It offers a path for students to contribute improvements and makes clear that beginners can ask for guidance. The public repository documents the software and development process; it is not a release of students’ private submissions or the live production database.⁷⁶

Chen also developed VON, a system for managing mathematical problems and rendering them into TeX documents. Its documentation describes storing such information as sources, tags, difficulty, and authorship. Organizing a reusable problem collection is substantial infrastructure: the same item may later enter a handout, a solution, or an examination. Public code for that system should not be confused with public access to every problem in Chen’s own database.⁷⁷

The invitation to students to write code is related to a wider preference for learning by attempting a useful project. In a 2023 essay, Chen explains how a modest contribution can make programming less abstract: learn enough to fix something, ask for help, and expand from there. He cites his books and course as projects begun because they seemed worth trying; he also acknowledges abandoned projects. Agency here is a practice acquired through doing and support, rather than a claim that anyone can accomplish anything regardless of circumstances.⁷⁸

His own publishing infrastructure also needed attention. In May 2026 he described moving the blog from WordPress to a Git-based Pelican system. Previously, a correction could require changing a local draft, converting it to HTML, and reinserting pictures on the server; now the repository held the posts and could receive corrections directly. Easier maintenance was his principal reason, alongside control over mathematical rendering and a preference for a portable, simple site. Even a personal essay had a continuing cost in formats, software, and repairs.⁷⁹

Public teaching included live problem-solving broadcasts. A 2023 announcement scheduled two-to-four-hour Twitch sessions, usually Friday evenings, with recordings on YouTube and dates adjusted around travel. These were live attempts at mathematical problems, a different encounter from reading a finished solution. The announcement establishes the planned format; it does not establish audience size, broadcasting income, or that every announced session occurred.⁸⁰

In September 2026 Chen estimated that he had personally read almost twenty thousand OTIS problem-set submissions over a decade. The estimate includes work predating the dedicated website model. What struck him was the variation: polished mathematical writing at one end, almost unreadable explanations at the other. He linked that difference to habits of craftsmanship and trust, rather than only exam scores. This is a teacher’s retrospective observation, not an independently counted database or comparative study. It nevertheless identifies a large, continuing part of his labor: reading how individual students have actually worked.⁸¹

Teaching calculus to students who did not choose olympiads

Fall 2024 brought Chen into a different classroom. As a teaching assistant for MIT’s multivariable calculus course, he led recitations for students meeting a General Institute Requirement. A large lecture course was divided into small sections. The audience was broader than OTIS’s applicants: these students needed calculus to graduate, whether or not they enjoyed mathematical competitions. His materials later became Linear Algebra and Multivariable Calculus, completed in January 2025 and published through MIT OpenCourseWare.^{82, 83, 84}

His August 2025 reflection describes recitations as a place to answer questions students might not ask in the large lecture. He began asking for questions at the start, even when that displaced his plan, and added an office hour after class. An umbrella represented a plane’s normal vector; audience-supplied numbers made a matrix example less abstract. Larry Guth’s teacher training and advice helped him learn how to conduct these sessions.⁸⁵

He also learned to repeat important statements, look at the person asking, and allow time for an explanation to register. Referencing something said earlier was less effective aloud than on a page. Students brought algebraic confusions his plan had not anticipated. This required adaptation to people taking compulsory calculus, with grades at stake, alongside his work with students who had actively chosen olympiads.⁸⁶

MIT OpenCourseWare identifies the notes as Chen’s work and says they had not been checked by the lecturer or other faculty. His site separately labels some older worksheet answer keys as unedited ChatGPT output.^{87, 88} His 2025 essay “Words Spent” describes a different use in the book: many example solutions began as ChatGPT drafts, which he shortened substantially. He distinguishes the concise intuition sought in Napkin from the fuller explanations required to replace a calculus textbook. Diagrams and detailed examples could justify considerable authorial work while reducing what the reader had to decode.⁸⁹

He also continued to reconsider live olympiad teaching. A May 2026 post publishes a cubic-polynomial lecture he had given at MOP in 2025, explains its theatrical choices, and says he was no longer teaching at MOP. He withheld the handout during the session to preserve the unfolding ideas, then expanded details for publication. His objections to lectures therefore did not prevent him from designing a particular live experience when he thought the format served it. The sources do not explain the circumstances of his departure from MOP.⁹⁰

Friends, games, puzzles, and a life beyond credentials

Many of Chen’s friendships grew from the contest circuit, including people who subsequently pursued other careers. In his public FAQ he calls the peer group a major benefit of competitions, independent of the results.⁹¹ His later leisure continued to give him groups to do things with. He plays competitive Hanabi, a cooperative card game in which players can see others’ cards but not their own, including organized play with the H-group. He also plays StarCraft and participates in puzzle hunts.⁹²

His relationship to video games changed as he grew up. As a child he used them to pass otherwise boring time; contest mathematics later displaced them. Returning in college, sometimes to games he had played years earlier, he became interested in the work behind the experience: pictures, animation, music, dialogue, and small details. He now described games as art, appreciating the care with which someone had designed and tested the encounter.⁹³

A puzzle hunt differs from an olympiad. A team may have to discover what a puzzle is asking, identify an encoding, organize information, and combine answers from several puzzles into a final one. Chen has solved with groups including Galactic Trendsetters and teammate, and written for MIT Mystery Hunt and other events. The MOP 2021 hunt was made with Isabella Quan, Sanjana Das, and Serena An. Work with Quines on the OTIS hunt later preceded Quines’s appointment to the teaching staff. The links between leisure and work are specific collaborations, not evidence that every friend is a professional associate.^{94, 95}

His public personal page describes an Arch Linux environment, NeoVim, Dvorak keybindings, and a to-do system. It also records music, animation, and plushies; Dreamcatcher is among his musical preferences, and Duluth his favorite city.⁹⁶ In September 2026 he was studying braille, Morse, semaphore, and a little Korean, the last expressly as a hobby without a compelling practical reason.⁹⁷

The interests matter partly because Chen has warned his younger self against letting academic work consume the period when friends live close by. They give content to “life outside mathematics” instead of leaving that phrase as an abstract work-life ideal.

He also contributed to Athemath, a nonprofit founded by Ali Cy in 2020 to give female and nonbinary students mathematical instruction and a supportive community.⁹⁸ His referral page identifies him as an adviser and board member, explicitly not an instructor; his September 2026 update records maintaining its Django website. Those are specific supporting roles in an independent organization.^{99, 100}

After the PhD: research, AI, and several employers

Chen defended in December 2024 and received his doctorate in February 2025; MIT’s graduate announcement confirms the February degree.^{101, 102} By fall 2026 he described living mostly near Duke University and working concurrently on OTIS, OMEGA, and Axiom Math. The location does not establish a Duke appointment. His public account gives the projects and employers, but no complete explanation of how he chose this post-PhD arrangement.¹⁰³

The AI work has more than one form. Chen was a coauthor of FrontierMath, introduced in 2024 as a benchmark built from original advanced mathematical problems. Such a benchmark needs questions whose answers can be checked and whose novelty reduces the chance that a model has merely retrieved a published solution. It is a different production problem from a student contest. The paper’s performance figures belong to specified versions and evaluation dates; they are not used here as current measurements of AI capability.¹⁰⁴

His own account explains the difference in problem-writing. An olympiad favors insight under constraints that make pencil-and-paper work realistic. A research-level benchmark can require computation, specialized background, and verification strategies that would be inappropriate for that examination. The writer still needs to construct a worthwhile mathematical obstacle, but now must consider machine access to tools and the risk of accidentally easy numerical answers. This work extends his experience of designing problems; it also changes the audience and conditions under which a problem is judged.¹⁰⁵

At Axiom Math he again worked with Ken Ono, his undergraduate REU mentor. A 2026 paper on Fel’s conjecture, coauthored by a large team including Chen and Ono, treats alternating power sums arising from semigroup rings. Its abstract reports a proof via exponential generating functions and coefficient extraction, and a Lean/Mathlib formalization produced by AxiomProver from the natural-language statement. This is documented participation in a particular mathematical and formalization project. The revised paper has a much larger author list than an earlier CV entry, so it should not be presented as a four-person final collaboration.^{106, 107}

Formal checking changes one part of the work. A proof assistant checks an encoded statement against explicit rules and definitions; human researchers still determine which statement is worth proving and whether its formal version expresses the intended mathematics. The public sources used here establish the authors’ reported result, not an independent audit of every step of the software’s production. They also do not disclose Chen’s consulting fee, hours, equity, or responsibility for every component of the system.

Chen also distinguished mathematical AI from the immediate needs of his students. After AMM 2026, he wrote that the systems he had tried were not yet adequate for routine proof feedback. This was a dated judgment about a particular teaching task, alongside participation in a formal theorem-proving project.¹⁰⁸

OMEGA and the attempt to create another place for students

At OMEGA—the Organization for Mathematical Engagement & Growth in America—Chen worked as head writer with president Paul Zeitz, executive director Yia Hang, and creative director Zuming Feng. The organization operates through the National Mathematical Engagement Foundation, a 501(c)(3) nonprofit.¹⁰⁹ Feng had led MOP when Chen attended as a student.¹¹⁰

In March 2026, Chen described the organization as new and small, with four staff, many part-time; his own arrangement was one-third time. He preferred starting with an immediate, manageable need. The initial AMM event was designed as an in-person proof competition with a much longer surrounding festival. Students would write five proofs in four and a half hours, meet peers, and receive direct discussion of their work. He wanted a path into proofs that did not first require an exceptionally high numerical qualifying score.¹¹¹

The first event took place in Denver in May 2026. OMEGA describes review meetings with pairs of judges, talks by Ravi Vakil and Melanie Matchett Wood, a parents’ conversation with Richard Rusczyk, and puzzles and games. The exam sat within an event for students and families, requiring speakers, reviewers, and organizers.¹¹²

Chen’s June report gives a less polished view of the same event. The first exam draft came together in under three weeks; the timetable was stressful, and review-session assignments were not always clear. He explicitly credited Yia Hang with most of the organization. Review conversations often became explanations of unsolved problems or discussions about mathematics and life. He also regretted some of his own public speaking and wished he had expressed his thoughts more precisely. These are the observations of a participant learning how to run a new event, not evidence that the format had already solved national mathematics education.¹¹³

A July 2026 essay for Education Progress sets out the broader problem as he sees it. He argues that competitions should create interest and ability rather than merely identify a tiny elite. He worries about their weakening connection to ordinary schools, the concentration of opportunities among already informed students, and pressure to turn contests into admissions instruments. His historical participation comparison uses different examinations and populations; it is treated here as his argument, not as a fully reconstructed national statistical series. His proposed destination is several accessible routes into advanced mathematics, not one compulsory contest system.¹¹⁴

The board connects universities, secondary-school teaching, Proof School, Champlain Capital, and Jane Street. These affiliations identify a network, without disclosing particular donations or Chen’s pay. Governance and event management accompany the writing.¹¹⁵

By September 2026, Chen had renewed his term and OMEGA was announcing a January 2027 regional event and a May 2027 Denver event. The separate online USEMO was continuing with Anant Mudgal as director for 2026. These developments show responsibilities spreading among people and institutions. They also define the boundary of the current account: the planned 2027 events have not yet happened.^{116, 117}

Authority, ambition, and what remains unsettled

Chen asks students to call him Evan: the syllabus says he wants respect earned through teaching, rather than age or honorifics. His informal manner still accompanies authority over materials, readiness, and assessment. It describes how he wants to conduct unequal responsibilities within a course, not a claim that teacher and student have identical roles.¹¹⁸

In a 2023 essay called “Pride,” he says he deliberately avoids keeping a count of students who make the IMO. He distinguishes satisfaction in their achievements from using those achievements as evidence of his own greatness. When a proposed alumni survey felt like reputation-building, he dropped it. He also admits that wanting recognition remains a temptation. The admission makes the position more useful than an account in which a teacher has permanently escaped ambition.¹¹⁹

His career still depends on credentials, institutions, and public trust. A medal, degree, published book, or camp role can help a stranger decide to read him or buy a course. The tension is practical: he wants students to own their accomplishments while maintaining a service whose value has to be explained to students, families, collaborators, and funders. The public record documents that tension without showing a complete solution to it.

In an April 2024 essay, Chen compares his own first TSTST selection tests in 2011—when he solved nothing but enjoyed the occasion—with the apprehension he perceived in the 2023 testing room. He objected to ambition detached from curiosity and doubted that replacing contest prestige with high-school research prestige would resolve the problem. He was also unwilling simply to remove the test from camp. This is a participant’s judgment about his community, not a measured change in a generation’s psychology; its sharpness is part of his public voice.¹²⁰

By 2026, Chen’s responsibilities had widened beyond the independent course and the doctorate. He had a first OTIS colleague, a new competition to help build, and research involving mathematical AI. His earlier collaborators and mentors remained visible in these arrangements, alongside newer ones. The published work documents substantial changes in practice; the personal reasons for every change are not public. His next choices cannot yet be read back into the story.

Chronology for checking the narrative