New Math Promises a Fairer Way to Split Stuff Nobody Can Cut

Imagine three roommates moving out of a shared apartment. There's a couch, a TV, a bike and a coffee machine. You can't saw the couch in half. Someone will end up grumpy. This is one of the oldest unsolved problems in economics and computer science: how do you split things that can't be divided, when everyone wants different things?

The standard fairness benchmark is called the 'maximin share' (the least you'd accept if you had to divide everything yourself and then let everyone else pick first). The promise is simple: give every person at least that much, and nobody has grounds to complain. For years, mathematicians could only prove that promise held under fairly narrow conditions — mostly when everyone valued things in similar, neatly ordered ways.

Nicholas Teh's new paper widens that dramatically. His proof covers a much broader range of preferences, including cases where different people value the same item in wildly different ways, and where some items tower over others in value. Think of one heir who only cares about the art, another who only wants the cash, and a third who wants the family business — all mixed together. He also provides a fast recipe a computer can follow, so the fair split can be found quickly even when there are many people and many items.

Why should you care? This isn't just abstract geometry. The same math underlies how algorithms divide advertising slots, allocate school places, split cloud computing power, and even how estate settlements get negotiated. Better fairness guarantees mean fewer disputes, less haggling, and automated systems that produce splits people actually accept as fair. It won't settle your Thanksgiving argument about grandma's china — but the math behind the negotiation is getting stronger.