Our brain has already solved the equation
Fifty years after Richard Feynman scrawled equations across dinner napkins in a Thai restaurant, researchers at Princeton have finally decoded what he was doing: solving the ancient human tension between the comfort of the familiar and the lure of the unknown. His formula, now validated in the Proceedings of the National Academy of Sciences, describes a mathematically perfect strategy for when to explore and when to commit — a threshold that rises at the start and falls as time runs out. What makes the discovery quietly astonishing is not Feynman's genius, but what it reveals about our own: evolution, it turns out, has already taught us nearly the same answer.
- For half a century, Feynman's handwritten restaurant equations sat undeciphered in a drawer — a solved problem no one knew existed.
- Princeton's Thomas Griffiths and his team cracked the notation and confirmed the formula was mathematically optimal, then immediately asked the harder question: do humans actually behave this way?
- Testing 2,500 participants across simulated dining scenarios, researchers discovered we don't follow Feynman's precise curve — we use a simpler, linear descent, and we explore far more aggressively at the start than the math strictly demands.
- The tension resolves in a quietly triumphant finding: our imperfect, intuitive shortcut delivers results nearly identical to the optimal algorithm, suggesting evolution has quietly solved a Nobel-level problem inside every human brain.
Richard Feynman was midway through a Thai meal when his friend Leighton posed a question about ordering — try the unknown or return to the beloved? For most diners, a shrug suffices. For Feynman, it became pages of equations before the check arrived. He had solved what mathematicians call an optimal stopping problem: how to balance the known value of a favorite against the potential of something better, with the goal not of finding the single best option but of maximizing total satisfaction across every meal.
The rule Feynman derived was elegant and unforgiving. Assuming restaurant quality is evenly distributed, he proposed a dynamic threshold of acceptability that starts high and falls as time runs out. On your first night in a new city, demand excellence. Explore until something clears the bar, then commit. As the trip winds down, lower your standards gracefully. The mathematics was beautiful — and then it sat in a drawer for fifty years, cryptic and untouched.
Thomas Griffiths and his team at Princeton finally cracked the notation and published their findings in the Proceedings of the National Academy of Sciences. They extended Feynman's framework to account for different distributions of restaurant quality, then tested it against real human behavior — 2,500 participants choosing fictional restaurants across trips of varying lengths, with researchers occasionally sabotaging early options to force longer exploration.
What they found was both humbling and quietly wonderful. Humans don't follow Feynman's optimal curve. We use a simpler linear descent — lowering our standards at a steady rate rather than along a precise mathematical arc — and we explore far more eagerly at the start than the cold equations recommend. Yet our scores were nearly identical to the theoretical ideal. Evolution, it seems, has built a cognitive shortcut so efficient that it approximates genius without calculation. The next time instinct tells you to keep exploring a little longer, or to finally commit before the trip ends, it may be worth listening. Your brain has already done the math.
Richard Feynman was sitting in a Thai restaurant when his friend Leighton posed a question that would remain unsolved for fifty years. Should he order the ginger chicken he knew he loved, or risk trying something new that might be even better? For most people, this would be a shrug and a guess. For Feynman—the Nobel laureate who had spent his career untangling the mysteries of physics, including why spaghetti cannot be snapped cleanly in two—it became something else entirely. He pulled out a pen and began filling pages with equations, symbols, and mathematical notation. By the time the meal ended, he had transformed a dinner dilemma into pure mathematics. He had solved it.
Leighton kept those handwritten notes. They sat in a drawer, cryptic and silent, for decades. No one knew what they contained. No one had cracked the code. It was as if a piece of scientific history had been buried, waiting.
Fifty years later, a team of researchers led by computational psychologist Thomas Griffiths at Princeton University finally deciphered what Feynman had scrawled that afternoon. Their findings, published in the Proceedings of the National Academy of Sciences, revealed that Feynman had solved what mathematicians call an optimal stopping problem—a decision where you must weigh the potential value of new options against the known value of what you already have. The formula he derived was mathematically perfect. But the real discovery came when Griffiths and his colleagues asked a deeper question: How do humans actually make these choices?
The exploration-exploitation dilemma appears everywhere. You arrive in a new city for vacation and must decide whether to visit a different restaurant each night, risking bad meals in pursuit of excellence, or to find one good place and return to it. You search for a parking spot, wondering whether to take the first available space or circle the lot hoping for something closer. You choose a house, a partner, a career. In each case, you are balancing the comfort of the known against the possibility of the better. Feynman's restaurant problem was unique because you could return to previous choices, and the goal was not to find the single best option but to maximize total satisfaction across all your meals.
Feynman had assumed that restaurant quality followed a uniform distribution—that any rating between zero and one hundred was equally likely. From this, he derived an elegant rule: establish a dynamic threshold of acceptability that decreases as your time runs out. On your first night in the city, demand excellence. Keep exploring until you find a restaurant that exceeds your high bar. Once you do, stop. Become loyal. As your vacation dwindles, lower your standards. When only two nights remain, accept something merely good. The mathematics was unforgiving and beautiful.
Griffiths expanded Feynman's equation to account for different distributions of quality—exponential curves where most restaurants are mediocre and only a few are exceptional, power-law distributions, triangular ones. Then he tested the theory against human behavior. Twenty-five hundred people participated in an experiment where they chose fictional restaurants across seven, fourteen, or twenty-eight nights, trying to maximize their cumulative satisfaction. The researchers even sabotaged early options to force longer exploration periods and observe how people adapted.
The results revealed something unexpected. Humans do not follow Feynman's mathematically optimal curve. Instead, we use linear decreasing thresholds—we lower our standards at a constant rate, like descending a ramp, as time disappears. We also show a pronounced bias toward early exploration. When we arrive somewhere new, we are insatiably curious. We resist committing to the first good option. We want to explore a little more, just in case, far beyond what the cold mathematics recommends.
Yet here is the surprise within the surprise: our simpler, imperfect strategy works nearly as well as Feynman's optimal algorithm. Humans achieved scores practically identical to the mathematical ideal. Our brains, shaped by millions of years of evolution, have developed a cognitive shortcut—a heuristic so simple and elegant that it saves enormous mental energy while delivering near-perfect results. We are not calculating square roots in our heads. We are following an intuition so refined by time that it approximates genius without effort. The next time you face a restaurant menu, you now know the truth: trust your instinct to explore early, and trust it again when time grows short. Your brain has already solved the equation.
Citas Notables
The people adopt thresholds that decrease with remaining trials, achieving performance near the optimal solution— Thomas Griffiths and colleagues, PNAS
We wanted to understand how people actually perform this kind of task— Griffiths research team
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that Feynman solved this problem in a restaurant, of all places? Why not in a lab?
Because that's the whole point. He wasn't trying to solve it. His friend asked a casual question, and Feynman's mind couldn't help but see the mathematics underneath. It shows how he thought—he found elegance in the ordinary.
But the notes sat hidden for fifty years. Why did no one find them sooner?
Leighton kept them, but Feynman never published the work. It was just a napkin sketch, really. Without Feynman alive to explain it or promote it, the notes were just cryptic marks. They needed someone like Griffiths to recognize what they were.
The study found that humans use linear thresholds, not Feynman's curve. Doesn't that mean we're doing it wrong?
No. That's the revelation. We're doing it differently, but just as well. Our brains found a simpler path to nearly the same answer. Evolution optimized us for this without us ever knowing we were solving an equation.
So we're naturally good at deciding when to stop searching?
Better than we should be, yes. We have this bias toward exploring early—we don't want to settle too fast. But we also know when to commit. It's not perfect mathematics, but it's practical wisdom.
What happens if you don't know how many nights you have left? What if the vacation is open-ended?
That's a different problem entirely. Feynman's solution depends on knowing your deadline. Without it, the threshold never drops. You'd keep searching forever, never satisfied. That's why deadlines matter so much to how we decide.