NFL players have skipped training camp over contracts; John Urschel skipped his for the Massachusetts Institute of Technology. Andrew Beaton's Sports page picks up the story nine years on. The former Baltimore Ravens offensive lineman — listed at 6-foot-3 and 313 pounds, three seasons in the league — published a paper Monday, On the Growth Factor of Random Matrices, proving a conjecture that had been open since the 1990s. He's 35 now, an assistant professor at MIT, and he spent this week paging through OpenAI's new math papers on the Riemann hypothesis and the like, astonished the way only someone who can actually follow them gets to be.
Penn State gave him a math major and the honor people call the academic Heisman, and he had a master's in hand before the Ravens drafted him in 2014. One season blocking for Joe Flacco and the math was calling him back. He applied to MIT, the school he'd always wanted, and started there part-time in 2016 while still technically a Raven. The season exposed the flaw in the plan: hours. Football and eigenvalue problems left almost no time for the woman he'd later marry. “In hindsight, it was a horrible decision,” he says.
So in 2017 he didn't show up for camp. Doing both had been too much, and a career of taking hits on the line looked like a threat to the academic one. The doctorate followed in four years, and two years later MIT brought him back as a professor. Football didn't vanish — three years on the College Football Playoff selection committee starting in 2020, Penn State on Saturdays — but the math won.
What he proved, in two lines
The Journal's example is straight out of high-school algebra: 2x − y = 0 and x + y = 3. Add the lines, get 3x = 3, and x is 1, y is 2. That's Gaussian elimination, and software in nearly every field leans on it to crack systems far bigger than two lines. The open question since the 1990s: could small rounding errors snowball and spoil the answer? Mathematicians suspected that almost never happens on typical problems but had never fully proved it. Urschel did, with help — his acknowledgments credit OpenAI models, and he welcomes questions about AI's role. His comparison is chess, where the machines beat the people and he still enjoys the search for a good move. “AI doesn't ruin the why,” he says.
Our read
This is Career planning, with a Behavioral footnote. Urschel left the NFL at about 26 — our arithmetic from a 35-year-old who quit nine years ago — and the Journal prints no salary, so we won't invent the size of the check he walked away from. What matters is the order: the credential banked before the draft, the new field tested part-time while the old job still paid, and only then the clean break. That sequence scales down fine without an NFL contract — to a teacher moving into administration, a nurse into a practitioner program. And notice what the test run taught him: not that he could do both, but that he couldn't, which told him which one to keep. Plenty of career-change plans are budgeted in dollars and not in hours; his double year was paid for in evenings away from the woman he'd marry.
So when the plan says a year of runway, ask the second question — a year of what, at how many hours a week, and who at home is carrying the rest? He also weighed the risk of staying, a career of hits on the line, against the risk of leaving, and that's a judgment, not a spreadsheet. Run your plan with the current paycheck switched off before you switch it off; the time to learn whether the numbers hold is while you can still change them.