A High School Student Just Beat 100 PhD Teams

It's Thursday, February twenty-sixth. A high school student in Michigan just co-authored medical research that took PhD teams years to produce. But this isn't an AI story — we've done plenty of those. Today we're talking about what a college degree is actually worth in 2026, and why the math has stopped making sense. This is HigherIQ. Last week, a paper dropped in Cell Reports Medicine. Peer-reviewed. Respected journal. One of the co-authors is Victor Tarca, a student at Huron High School in Ann Arbor. A few years ago, researchers at UC San Francisco ran a global competition called DREAM. The challenge: build machine learning models to predict preterm birth. More than a hundred teams entered — computational biologists, statisticians, people with decades of training. The competition ran three months. It took two more years just to publish the findings. The question was: could you do it faster? So they paired a master's student named Reuben Sarwal with Victor, the high schooler. Same datasets. Same challenge. But this time, they had AI tools that could write analysis code from natural language prompts. The AI generated working code in minutes. Not hours. Minutes. The junior pair finished their experiments, verified results, and submitted a paper in six months total. On one prediction task, their AI-assisted model didn't just match the PhD teams. It beat the best one. Now, what Victor actually did matters here. He didn't write the code — the AI did that. He didn't have years of training in computational biology. What he did was help design the experiment, frame the question, and figure out whether the outputs made sense. The AI was the engine. Victor was the one who knew where to point it. One of the study's senior authors is Adi Tarca — who happens to be Victor's father, which is its own interesting wrinkle. This isn't a bootstraps story. It's a story about access. But that's part of the point we're getting to. If a high school student with the right tools and the right mentorship can co-author publishable medical research, what exactly is a two-hundred-thousand-dollar education buying you? The honest answer is: we're not sure anymore. The numbers have stopped agreeing with each other. Start with what degrees still deliver. The college wage premium — the gap between what bachelor's holders earn versus high school grads — is near an all-time high. Around seventy percent. That's over thirty-two thousand dollars a year. The Federal Reserve says the return on a degree is about twelve and a half percent annually — beats the stock market. Over a lifetime, somewhere between six hundred thousand and a million dollars. So degrees pay. That hasn't changed. But here's where it gets strange. Entry-level jobs — the jobs degrees are supposed to unlock — are vanishing. Programmer employment in the U.S. dropped twenty-seven percent between 2023 and 2025. In the UK, tech graduate roles fell forty-six percent last year alone. A Harvard study tracking sixty-two million workers found that junior positions are shrinking at companies using AI. Their phrase was blunt: AI is eroding the bottom rungs of career ladders. And here's the part that should bother you. Companies keep saying they don't care about degrees anymore. Eighty-five percent claim they use "skills-based hiring." Google, IBM, Apple — they've publicly dropped degree requirements. Except they haven't. Not really. Harvard Business School and the Burning Glass Institute looked at what companies actually do versus what they say. They tracked eleven thousand job postings over a decade. The finding: fewer than one in seven hundred hires benefited from dropped degree requirements. One in seven hundred. Companies changed the job posting, but not the hiring. Forty-five percent were "in name only" — dropped the requirement on paper, kept filtering for degrees in practice. So here's the contradiction. Degrees still pay a premium. Entry-level jobs are collapsing. Companies say they don't need degrees. But they keep hiring people who have them. The wage gap is at a record high. And the path to capturing it is narrower than ever. We think the answer is that we're in a transition where the signaling value of a degree hasn't caught up to its functional value. They're unbundling, and nobody knows how to price the pieces. For decades, degrees did two things. They taught you skills — analysis, coding, modeling, whatever your field required. And they signaled to employers that you could finish something hard, follow through, and show up. The skills and the signal came bundled together, like cable TV. You paid for the package. What's happening now is the skills part is getting unbundled. If AI can write analysis code in minutes, then the years you spent learning to write it aren't the asset they used to be. The value has shifted from "can you do the task" to "do you know which task to do." But the signal hasn't adjusted. Employers still use degrees as a filter because it's easy, because hiring managers don't know how else to evaluate people, and because — as one Harvard researcher put it — "if you haven't been trained to assess someone based on their work rather than their diploma, that's a real challenge." The result is a mismatch. Degrees are overpriced for what they teach. But they're still required for what they signal. And nobody's figured out how to close the gap. Now, we're not saying skip college. The wage premium is real. Unemployment for degree holders is still half what it is for high school grads. If you can get a degree without wrecking your finances, it's probably still worth it. But the frame is shifting. Employers surveyed by the National Association of Colleges and Employers say the number one thing they look for now is hands-on experience. Not GPA. Not school name. What you've actually done. Seventy-four percent ranked it as the most important factor. The student who graduates with just a transcript is going to struggle. The one who graduates with a portfolio — projects, contributions, evidence — is going to have options. Victor Tarca didn't just take classes. He got on a research team. He co-authored a paper. And now, before he's even applied to college, he has something most PhD candidates don't: a publication in a peer-reviewed journal. His dad opened the door. But Victor walked through it and did something worth publishing. That's not because he's a genius. It's because he was in the room. Adi Tarca described what AI actually changes for scientists. He said: "Thanks to generative AI, researchers with limited backgrounds in data science won't always need to form wide collaborations or spend hours debugging code. They can focus on answering the right biomedical questions." Answering the right questions. That's the shift. For a long time, the bottleneck in knowledge work was execution. Could you build the model? Write the code? Crunch the numbers? If you could, you were valuable. If you couldn't, you needed to pay someone who could — or spend years learning. Now the bottleneck is moving upstream. The execution is getting cheaper. What's staying expensive is knowing what to execute. Knowing which question matters. Knowing when the output is wrong. A degree used to be a driver's license — proof you could operate the machine. Now the machine drives itself. The question is whether you know where to go. Most curricula haven't caught up. But the job market isn't waiting. A high school student just co-authored research that took PhD teams years to produce. The college wage premium is at a record high. Entry-level jobs are disappearing. Companies claim they don't need degrees but keep hiring people who have them. And one in seven hundred — that's how many people actually benefit from all those headlines about skills-based hiring. The degree isn't dead. But what it's for is changing faster than the institutions that grant them. Stay informed, stay curious, and we'll see you tomorrow.