These are great points, and I understand your skepticism. First let's get OkCupid out of the way: while OkCupid was great for 2008 and deserves a lot of credit

These are great points, and I understand your skepticism. First let's get OkCupid out of the way: while OkCupid was great for 2008 and deserves a lot of credit for pioneering the idea of actually trying to find great matches, it certainly wasn't at the upper limit of how well a dating app could work. For one thing, OkCupid didn't solve the spam problem (since anyone can message anyone else), and that's why over time Match Group has dismantled OkCupid and turned it into another Tinder clone (i.e. double opt-in through swiping) but with different branding. Their matching algorithm also left a lot to be desired, with its core assumption that two people answering multiple choice questions the same is a good measure of match quality[1]. Rarebird solves both problems. It works in three steps: 1) You are shown two user profiles at a time and asked "Who would you rather meet?"[2]. This trains your personal AI which searches for matches on your behalf. 2) Each user on the platform gets an equal stipend of 100 pebbles/day, which is an in-app currency used to balance incentives. 3) Your matches are delivered to you ordered from best to worst[3], but contacting each match costs a price in pebbles such that it's cheaper to contact better matches. This can be thought of like a carbon tax that charges you for the risk of "polluting" the platform with spam. (Note: you only pay when initiating contact with a new match, then the message channel stays open until it's closed by either party.) Because of Rarebird's use of a forced choice between two profiles, users do better when they create a profile that shows their uniqueness and is polarizing, to draw in the best matches and turn off people who would be a bad match. The use of contact prices also gives people a lot of information about who is interested in hearing from them. They can still optimize their personal preferences, but the contact prices give an upfront hint about who is likely to reciprocate and respond. It's fine to have high ideals and dreams, but even someone who has very high ideals is ultimately still looking for a good match. This design lets everyone search and filter to find the people they're most excited about contacting, while also solving the other major problems. The most obvious is that it reduces spam, since it costs more to contact people who are predicted to be not interested. But it also directly addresses your point regarding "[people] being extremely picky". While you're probably right that men and women have different thresholds to consider entering a relationship, in our modern society we're all picky when searching for a long-term partner. We all have the idea that we can be a differentiated unique individual and still find a partner who is a great match and appreciates us for who we truly are. In the ancestral environment of small tribes, the situation was very different, with only a few potential partners in the neighboring tribes, and far less individual differentiation. We don't assume that "men and women are pretty much the same" -- instead we embrace the individual complexity and variation between people, and give people the tools to express themselves and act on their individual preferences. While our app's design is the same for everyone, we don't expect everyone will use the app in the same way. For example, maybe Alice doesn't want to reach out first, but instead to be found and contacted. But she still wants to find her ideal match, not just some random person! Rarebird accounts for this naturally. To make it harder, let's even assume Alice is "conventionally attractive" and there's a lot of guys who want to contact her. In this situation, most people will find that Alice is expensive to contact, but the people who Alice wants to be contacted by will find they have a low cost in pebbles to message her. Much of what you describe in your comment is the fact that the underlying game theory of current dating apps is terrible, and when people play the game as it’s designed, the outcome is awful for everyone. On Tinder, people are just trying to get over the right-swipe threshold, so it encourages this homogenization where everyone looks generic, and conceals the more vulnerable unique aspects of themselves. Plus it gives no feedback about who is a realistic match, so everyone just stays anchored to their prior beliefs based on their ideals. So while online dating is a platform design problem, and not an algorithms problem, we believe that algorithms are a useful tool in the toolbox. Footnotes: 1. This has two flaws: 1) What about features that aren't captured, like physical attraction? 2) It assumes that people have reflective abstract verbal access to accurately say what is important in a partner. That's a hard thing to know accurately, except maybe if you are particularly self-reflective and have a lot of relationship experience through which you've honed your conception of a good match. 2. There's also a "neither" button, to be used when you'd rather be single than match with either person. 3. You can sort matches by predicted match quality, predicted attraction, or cost to contact.