ISEF 2026 Physics Track — A Judge’s Dispatch from Phoenix

I have always understood high-school science research through something like a quantum-mechanics lens. On one hand, I think most students are already doing well to handle AP coursework, standardised tests, and AMC or the Physics Bowl to a high standard; asking them to also develop genuine original-problem consciousness and participate in an academic community on top of that strikes me as potentially damaging to the next generation. On the other hand, I recognise that some people have simply never been suited to basketball, gaming, or dating — they have been, from a young age, inappropriately curious about some corner of the world. For those people, spending all their energy on a strange but interesting problem seems entirely defensible. The world already has too many normal people.

These two beliefs have coexisted in superposition throughout my career. They contradict each other, but both are true simultaneously. So I have tended to say: in high-school research, process matters more than outcome. Even if nothing extraordinary emerges, at worst the student's Common App looks richer and more interesting — and that is already enough.

This year's global physics judging in Phoenix, however, genuinely scrambled that equilibrium. It left me with a new question I haven't been able to shake: what are these students actually doing? And, more unsettlingly, what am I doing?

In keeping with the convention that what happens in the caucus room stays there, everything below is expressed through analogy, generalisation, and personal observation. No real conversations, specific project details, or internal judging information are reproduced.

The quality problem

In previous years, the overall field felt like watching a high-quality football awards ceremony. Among the nominees you could see Messi, Ronaldo, but also Xavi, Robben, Ribery, Sneijder — each with distinct strengths, each capable of making a case. Whatever you thought of the final selection, you could acknowledge that the winner had genuinely had a good year.

This year felt more like the selection pool had quietly contracted from a global field to a regional one. Aside from one project in an entirely different tier, there was almost nothing to choose from. The atmosphere in the panel wasn't "who deserves to win?" but "who is least clearly disqualified?" It was the first time I had experienced that at this level. The panel eventually resolved it by reverting to raw scores — which had previously been treated as advisory rather than decisive — because subjective deliberation couldn't find a foothold.

Beyond the thin field, there was a second quality issue: projects that couldn't be easily disproved. Not because the science was strong, but because the packaging was. These were not cases of wrong units, fabricated data, or obviously broken models — those are trivial to catch. These were projects that had been processed into something smooth: polished, narratively coherent, and strategically opaque. You had the sense that something wasn't quite right, but nothing specific to put your finger on. Like a gem that had been tumbled until it was brilliant — and hollow.

More troubling: several students demonstrably didn't understand their own projects. Not nervous, not struggling to express themselves in a second language — genuinely unaware of what they had nominally done. When asked about the basic physics underlying their method, or why a particular design decision had been made, the answers evoked the tradition of strategic ambiguity rather than scientific honesty. In previous years I might have found ways to defend this: the student still went through a research process, still learned where their aptitude lay. After enough of it, I find I can no longer summon the defence. There is something quietly nauseating about a project that exists entirely as a display object.

The topic problem

Before the interviews, simply reading the project list, I could already predict with reasonable confidence which projects would place and which would not. That is not a compliment to the field.

Type one: good execution, low ceiling. Projects that are technically competent and well-organised but structurally limited — like a perfectly written school essay, excellent for its genre but not a candidate for a literary prize. In previous years these projects would have been cleanly ranked below the stronger work. This year, with the field so thin, some of them placed anyway. Parents should not conclude that a particular topic or mentor approach is therefore reliably award-producing. The same projects in a stronger year would have finished lower.

Type two: high-sounding, hollow underneath. Projects invoking advanced concepts — quantum computation being the most common example — where the framing significantly exceeds the content. There is nothing wrong with a high-school student running experiments on a quantum computing platform; the problem is treating "quantum" as a protective label that raises the apparent prestige of whatever sits underneath it. If a student cannot account for the physics of a hydrogen atom or a one-dimensional square well at a basic level, they are almost certainly not in a position to meaningfully interpret what they ran on the hardware.

Type three: boundary-straddling projects. Engineering, environmental, computer science, or medical projects nominally registered in the physics track. The motivation is usually either "physics seems less competitive" or "a different label might improve the odds." These projects test judging philosophy sharply. My own position is strict: if you register in physics, there must be a clear physical question, a clear mechanism, and a complete experimental physics methodology. Others are more tolerant of cross-disciplinary work. At a certain point in a long day, everyone's philosophy converges on "let's be done."

What high-school research is actually for

I arrived in Phoenix, forty degrees Celsius, with a naive expectation that AI tools and professional mentorship had perhaps raised the floor — that I might encounter some genuinely fresh thinking. Instead I left questioning something more fundamental. If the joint effect of AI, agency pipelines, professional mentors, and accumulated competition strategy is to make high-school research better packaged, better narrated, and more resembling a polished business proposal — are we training future scientists, or future promoters?

This points to a deeper question: what is high-school research actually supposed to serve?

If it serves admissions, the rational equilibrium is packaging. Pick a topic that reads as current. Write an abstract that sounds ambitious. Use a method that appears complex. Cite a mentor with institutional prestige. Produce a publication. Whether the student genuinely wrestled with confusion, failure, disconfirmation, and reconstruction — all the things that constitute actual research experience — becomes irrelevant, because the admissions system cannot reliably read it.

If it serves competition placement, the rational equilibrium is prediction: guess what the judges want this year, guess which track is softer, guess what conclusions look original without being auditable. The student stops asking "is this problem worth investigating?" and starts asking "does this problem look like a prize-winning problem?"

If it serves science itself — the most demanding case — the outcome is genuinely uncertain, often unremarkable, and sometimes actively negative. Real scientific training means three months discovering your original question was wrong, six months discovering your experimental error is intractable, a year discovering your model's central assumption doesn't hold. The real spirit of research is not "I built a system" or "I proposed a method" — it is the willingness to honestly face what you don't know, to acknowledge that your earlier idea was naive, and to find the energy to keep thinking anyway.

That is exactly what was most absent in Phoenix this year: intellectual honesty. Not moral honesty — the students were not lying, exactly — but epistemic honesty. Do you actually know where your question came from? Do you actually understand why your method is appropriate? Do you actually know what your data can and cannot show? Can you explain what you had to learn before you could start, and what you still don't understand?

Science is, at its root, a slow and unglamorous activity. You fix your attention on a problem for a long time. It turns out to be harder than you expected, and more interesting than you expected. It resembles, in some ways, the experience of a young person encountering something they genuinely care about for the first time.

This is why I kept noticing, as something of a counterweight, that some of the Japanese students felt like a clean signal in a noisy room. Their projects were often small — nearby insects, cultural objects, local materials. But there was a quality of attention that felt different: they were actually curious about the question, not about the question's appearance. They were willing to do a small thing carefully rather than a large thing impressively. They wrote about failure and limitation. That doesn't make for a strong press release. But it looks like science.

The key to high-school research is not scale. It is honesty.

A student can study a small, local phenomenon — but they must have actually observed it, been confused by it, tried something. A student can use a simple model — but they must actually understand its limits. A student can produce no striking conclusion — but they must be able to explain what the failure taught them. By contrast, a nominally grand project, where the student can only recite the logic their mentor provided and reads from agency-polished paragraphs, is an empty shell even if it has a first-place ribbon.

High-school research could be a genuinely good thing — which is why the organisers of ISEF and S.-T. Yau Award have invested so much in making it exist. It can show students that knowledge doesn't grow from textbooks but from questions. It can teach them that science isn't a warehouse of correct answers but an ongoing process of correction and refinement. It can give students something to be, not just something to display.

But if it becomes an agency assembly line, a mentor byline game, an AI factory, and a competition-arbitrage scheme, it will have become a form of educational corruption — not the familiar kind, but a spiritual kind. It will have taught students, before they have ever touched real science, how to consume the symbols of science. That is the most wasteful outcome of all.

My view of high-school research after Phoenix has not fully collapsed. It is still in superposition. But the wavefunction is now more complex, and less stable.