Credit: ZME Science.

Richard Feynman could turn almost anything into physics and math. Even lunch.

One day in the late 1970s, the Nobel Prize-winning physicist sat in a Thai restaurant in Glendale, California, with his friend Ralph Leighton, who faced a familiar decision paralysis: order his regular beloved ginger chicken, or risk disappointment on something new that he had never tried but might turn out to be better.

Feynman did not flip a coin. He reached for a sheet of paper and started scribbling equations. He essentially transformed the mundane choice into a problem about when to keep exploring and when to return to the best thing you already know.

The notes from this seemingly inconspicuous diner night have remarkably survived. But the math was never really checked out independently.

Now, nearly 50 years later, researchers say they have deciphered Feynman’s “restaurant problem,” proved that his answer was mathematically correct, and tested whether ordinary people behave anything like his ideal diner.

The new study finds that people do not follow the perfect rule derived from Feynman’s restaurant equations. But they get surprisingly close.

The Math of When to Stop Looking

Imagine you are in a new city for a fixed number of nights. Each evening, you can try a new restaurant, whose quality you do not know until you go there, or return to the best restaurant you have already found. Your goal is not merely to find the best place. It is to have the best overall dining experience.

“The essence of the problem is that the value of exploring, of looking around and trying something new, decreases the opportunities you’re going to have to make use of that information,” Prof. Tom Griffiths of Princeton University, a co-author of the study, told The Guardian.

Feynman’s restaurant problem. Credit: Ralph Leighton/Richard P. Feynman Estate.Feynman’s restaurant problem. Credit: Ralph Leighton/Richard P. Feynman Estate.

Feynman’s answer was a moving bar. Early in a trip, you should keep trying new restaurants unless you find something exceptional. That is because the reward for discovering a great place is highest at the beginning: you still have many nights left to return to it. But each passing night makes exploration less valuable. By the end, even a merely good restaurant may be good enough, because there is little time left to benefit from finding something better.

“The thresholds are being guided by the best thing you might be able to find if you kept looking,” Griffiths told The Guardian. “If you have a long time to look, finding something amazing has a lot of value because you can go back many times.”

The problem belongs to a larger family of decision puzzles called optimal stopping problems. Similar logic appears when people hunt for apartments, search for jobs, choose parking spots or decide whether to keep dating. It also overlaps with the classic “explore or exploit” dilemma: when should you try something new, and when should you return to what already works?

Unlike some classic stopping problems, in Feynman’s version, the diner can return to an earlier choice. That makes the puzzle less like grabbing the first acceptable apartment before someone else does, and more like building a list of known favorites.

Different Worlds, Same Restaurant Problem

Brian Christian, a researcher at the University of Oxford and the Center for Human Compatible AI at the University of California, Berkeley, and his collaborators reconstructed Feynman’s notes with help from Leighton and Michael Gottlieb, who maintains The Feynman Lectures website.

Feynman’s hasty handwriting was challenging to decipher. But that was only step one. Then came the math itself.

“What you have to unravel to make sense of the notes is: What are these equations even trying to do? What’s the problem that he’s then going forward and solving?” Christian told Nautilus.

Once the researchers recovered the problem, as formulated by the late physicist, they generalized it. Feynman assumed restaurant quality could fall anywhere across a uniform range. However, the new study also examined worlds where most restaurants were mediocre but a few were exceptional, or where good restaurants were more common.

In one world, restaurant quality was uniform: a terrible meal, an average meal and a superb meal were all equally likely. In another, called exponential, most restaurants clustered toward the lower end, with truly great ones becoming rarer. A power law world was even more extreme: most places were ordinary, but a few outliers could be spectacular. And finally, in a triangular world, scores leaned toward the higher end, making decent restaurants relatively common.

Then the team looked at how people actually behave.

What People Actually Do

They recruited 2,520 participants online and placed them in a virtual city. Each participant saw a grid of restaurants. A restaurant’s score remained hidden until the participant visited it. The volunteers had either 7, 14 or 28 nights to maximize their total score.

“We wanted to really capture people’s gut intuitions,” Christian told Live Science. “When you just get thrown into this situation, what do you do?”

The answer was not exactly Feynman’s elegant curve.

People used a simpler rule. They started with a high standard, then lowered it in a straight line as the trip went on. They also explored more than the optimal strategy predicted early in the game, especially when they found a strong option right away.

But the shortcut worked. The researchers found that this simple strategy captured about 90 percent of the value of the optimal approach.

“People are not doing the optimal thing. They’re doing something radically simpler,” Christian told Live Science. “And still the simple strategy is being tailored in a way that feels very situationally appropriate.”

A Kinder View of the Human Mind

Since the work of Daniel Kahneman, Amos Tversky, and others, behavioral science has documented many ways in which people depart from strictly optimal and rational outcomes.

The new study fits into a newer theoretical framework called resource rationality. The idea is that people often use cognitive shortcuts not because their minds aren’t up to the task, but because perfect calculation would demand too much time, information and mental effort.

“People don’t do the perfect thing, but they make nearly perfect use of their constrained resources,” Christian told Live Science. “I think this is a little bit more of a redemptive story about the human mind than we are used to from the 20th century.”

The study’s participants did not apply exactly the same standard in every restaurant world. When the distribution suggested a few rare gems might exist, they set a higher starting bar. When most restaurants were clustered closer together, they settled sooner.

Still, the experiment was cleaner than life. Real restaurants are far from this uniform. A typically five-star restaurant can feel like a four or even a three on a bad night. Money, distance, mood and novelty all matter in what, at the end of the day, is a pretty subjective experience.

“Like any math problem, the Feynman restaurant problem abstracts away some of the messy reality of life,” Christian told Nautilus.

Yet the deliberate simplification is part of the process. It’s a running gag that to physicists, cows are like spheres. You can’t milk a sphere, but you can easily determine its volume, and that may be helpful when you have to figure out how big a farm needs to be. By stripping dinner down to its bones, Feynman exposed a pattern that reaches far beyond food. The same tension shapes how people search, commit, compare and settle.

The work may also matter for artificial intelligence. Many systems assume humans act like perfectly rational agents. Studies like this suggest a better model: people often use rough rules that are fast, adaptive and good enough. The insight may be helpful when designing AI that better models the efficiency of the human brain.

Feynman died in 1988 without publishing the restaurant problem. For decades, the puzzle sat in a few scribbled pages from lunch. It began with ginger chicken. It ended as a lesson in how the human mind decides when the search is finally over.

The findings appeared in the Proceedings of the National Academy of Sciences.

Editor’s note (October 8, 2026): This article originally appeared in June 2026 and was updated with slight corrections for style before republishing.