You have to ask a boolean question that splits all humans as close to 50/50 as possible. What question do you ask?

Claude Opus 4.8

Answer: "Are you currently age 30 or older?" comes remarkably close to splitting all living humans roughly in half, since the global median age is around 30.

The cleanest way to split humanity near 50/50 is to find a single objective variable whose median value is known and well-distributed. Age is the strongest candidate because demographers track the global median age precisely, and it currently sits at roughly 30 years. By definition, the median is the value where half the population falls below and half above, so a question keyed to the median naturally produces a near-even split. I prefer age over alternatives like 'Are you male?' because sex is close to 50/50 but skews slightly male at birth and varies by region, and it's also a fixed binary rather than a tunable threshold. With age, I can set the cutoff exactly at the median to optimize the balance, and I can adjust it over time as demographics shift. A boolean question also needs to be unambiguous and answerable by everyone the same way. 'Are you age 30 or older?' is clear, requires no specialized knowledge, has a definite yes/no answer for every person, and doesn't depend on culture, language, or interpretation. That makes it both balanced and reliable as a true 50/50 splitter.

Gemini 3.5 Flash

Answer: The optimal question to achieve an exact 50/50 split is: 'Are you older than the median age of the living global human population?'

The concept of the median is mathematically defined to divide a population into two equal halves, making this question the most reliable way to achieve an exact 50/50 split. While physical or biological traits like biological sex or finger-interlocking preferences hover near a 1:1 ratio, they are subject to slight global skewing, genetic variations, and demographic imbalances. In contrast, using the median age—which is currently around 30 years globally—inherently and dynamically bisects the entire living human population. By framing the question around a mathematically self-correcting boundary, we bypass cultural, geographical, and biological biases to guarantee a near-perfect division.

GPT-5.5

Answer: Ask: “If all humans alive at this instant are placed in a fixed deterministic total order, is your rank in that order odd?”

This splits humanity as close to 50/50 as mathematically possible: odd-numbered and even-numbered positions in a complete ordered list differ by at most one person, and if the total number of humans is even they are exactly equal. The key is to avoid relying on demographics like sex, age, handedness, or geography, which may be near 50/50 but are not guaranteed to be optimal and can drift over time. A rank-parity question makes the balance structural rather than empirical. To make it well-defined, the “fixed deterministic total order” would need an agreed tie-breaking rule, for example ordering by exact birth time, then birthplace coordinates, then a canonical identity record. Most people could not answer the question themselves, but the question’s purpose is to partition all humans, not necessarily to be easy for each person to compute. If there are N humans, this question produces either N/2 vs. N/2 or, when N is odd, (N+1)/2 vs. (N-1)/2, which is the closest any boolean split can possibly get.

Final Summary

While Claude Opus 4.8 and Gemini 3.5 Flash agreed that targeting the global median age (roughly 30 years old) is the most practical way to dynamically bisect humanity, GPT-5.5 took a radically different path. GPT-5.5 argued that demographic data is prone to drift, proposing instead a highly abstract question about whether a person's rank in a deterministic total order of all living humans is odd or even. The AI Roundtable ultimately split between real-world demographic precision and pure, unforgeable mathematical parity.

2 of 3 models agreed