Fields Medal 2026: Leaked Names Reveal a Century of Math Solved at Once
Four days before the International Congress of Mathematicians opens in Philadelphia — the first U.S. congress in more than four decades — a front-end coding error on the ICM website exposed schedule entries marked “HIDDEN,” revealing the four probable winners of math's most prestigious prize: Yu Deng, John Pardon, Jacob Tsimerman, and Hong Wang. What makes the cohort remarkable is not the leak but the ledger: between them, they resolved the Kakeya conjecture (open since 1917), the kinetic-theory portion of Hilbert's Sixth Problem (posed in 1900), the André-Oort conjecture in arithmetic geometry, and foundational problems in symplectic topology — several of mathematics' longest-running open chapters, all closed within a single four-year window. If the list is confirmed at Thursday's ceremony, Wang becomes just the third woman to win the medal, and she and Deng become the first Chinese-born mathematicians ever to receive it.
CREATIVE FORCE: It's tempting to read four once-in-a-generation results as four lone geniuses striking at once. The truer story is more interesting: this is a generation that grew up on each other's preprints, built on each other's tools, and closed problems the previous generation left open — creativity as a relay, not a lightning strike. That's a powerful reframe for our classrooms and research groups alike. The mathematics our students inherit is not a finished cathedral but an active construction site, and the drama of this week — leaks, prediction markets, a boycott, a ceremony — is a reminder that mathematics has stakes, personalities, and stories worth telling out loud.
_______________________
Mathematicians Are Closing In on the Hidden Order Inside Chaos
Ramsey theory asks how much disorder a system can hold before order must emerge — how big a social network can get, say, before it's guaranteed to contain either three mutual friends or ten total strangers. Domagoj Bradač of EPFL has now dramatically tightened the bounds on “off-diagonal” Ramsey numbers, toppling a barrier that stood for decades. His method is itself a lesson in the subject: start with a large graph chosen for its geometric structure, then inject randomness by zooming in on a subgraph and pruning the troublesome vertices. The twist came weeks later, when an OpenAI reasoning model — pointed at the fresh preprint — found a refinement that tightened the bound further, bringing the walls of a 90-year-old problem within polylogarithmic distance of touching. As one colleague put it, the AI's tweak was important, “but it was very much based on the idea that was already there.”
CREATIVE FORCE: Bradač's proof strategy — lay down structure first, then let chaos do the finishing work — is a beautiful model of creative process in general: constraints and randomness collaborating rather than competing. And the human-AI coda raises the week's most interesting question about mathematical creativity: when a machine sharpens a brand-new idea, who is creating? The mathematicians in the piece are unbothered — the conceptual leap was human; the polish was mechanical. For those of us thinking about what to prize in students' work, that's a clarifying distinction: the idea that reframes the problem, not the optimization that follows, is where the creative force lives.
__________________
More A's, More Fails: What AI Is Really Doing to Student Performance
What happens when AI helps students earn more A's — and contributes to more failures when it's taken away? Ray and Dan walk through a stack of new research on exactly that paradox: studies linking generative AI to grade inflation, reports of failing grades soaring in UC Berkeley computer science courses as professors see heavier AI use and weaker math skills, and emerging evidence that purpose-built AI tutors support learning better than general-purpose chatbots. Along the way they take clear-eyed detours through AI detectors (which fail diverse student populations), “AI humanizers” built to disguise machine writing, and what all of this means for teacher workload. It's a conversational, non-technical tour of the most uncomfortable question in our field right now: are the grades going up because the learning is?
CREATIVE FORCE: Mathematics educators have always known there's a difference between producing an answer and making meaning — AI has simply made that gap impossible to ignore. If a chatbot can manufacture the performance of understanding, then what remains assessable is precisely the creative core of mathematics: posing a conjecture, choosing a representation, explaining why an approach must work. For our readers, this episode is less a warning than a design brief. Assessment that asks students to do the things machines fake poorly — reason aloud, connect, justify, create — isn't just AI-proof; it's better mathematics.