Cupid's Algorithm
Singapore's government just opened a dating pilot that pairs singles with a 1962 piece of maths: the Nobel-winning Gale–Shapley “stable marriage” algorithm. It also runs the US medical residency Match and school admissions in New York. Here's how it works: watch it pair up a moonlit ball of animals, then try to beat it.
⚙️ The Machine
Each guest has secretly ranked everyone in the other house. The tiny row under each name is that ranking, favourite first. One house does the asking. Press play.
Full proposal log
💡 Tap any guest to see their bio and full ranking.
How the algorithm works, in four steps
- Everyone ranks everyone. Each guest privately ranks the other house from favourite to least favourite. FirstDate builds this from a 33-question questionnaire on interests, habits, values and dealbreakers.
- Askers ask their top choice. Every single guest in the asking house proposes to the highest-ranked person who hasn't turned them down yet.
- Receivers say “maybe”. Each receiver holds on to the best offer so far and rejects the rest. A held offer isn't a promise. If someone better asks later, the receiver trades up and the dumped guest goes back to asking.
- Repeat until nobody is free. When no asker has anyone left to ask, the “maybes” become matches. Gale and Shapley proved in 1962 that the result is always stable.
What “stable” means
No rogue couple: there's never a pair who'd both rather be with each other than with the partners they got. Nobody has a reason to sneak off.
Askers get the better deal
The asking side gets the best partner they could have in any stable matching. The receiving side gets the worst. Flip the “who asks” switch above and compare the averages.
Stable ≠ happy
Stability only rules out better options inside the pool. Someone can end up with their last choice and the matching is still stable. Critics of FirstDate point this out.