A few years ago I had an idea for a game to help people practice the stance of orienting towards win-wins, and the skill of creatively finding win-wins. I came up with a tiny playable proof-of-concept that didn’t need any materials, and played it a few times. Recently I discovered that somebody had made it as a card game!
The basic premise is as follows:
I called the concept the “omniwin training game”. While collaborative games are not as common as competitive zero-sum games, there are still lots of them out there: Hanabi, Pandemic, The Mind, and of course many video games. However, most of them focus on coordinating your actions around a single shared known goal.
So for my tiny proof of concept, I used the medium of words. Each player would secretly pick some constraint a word could have—often based on the spelling but you could do “verb” or something else. I didn’t have specific rules for that. Then each person would take turns naming a word, and the players for whom that word worked would say “ding!” Initially I had the rule that players had to only offer words that satisfied their own constraint, but I found that when the game lasted more than a few minutes it was pretty obvious to want to say things like “‘banana’ doesn’t work for me, but it works for you, right?” and this felt within the spirit of things.
This version of the omniwin training game worked well enough, and produced one astonishing and instructive result I’ll tell you about at the end of the post, but it had the unfortunate property of feeling a bit too simple, and of the possibility of people making totally incompatible constraints and getting stuck forever until they give up. Of course, “we mutually recognize there is no win-win” is a kind of win-win on the meta-level… but it’s less satisfying. On the other extreme is…
…calculate every possible combination of every possible constraint and ensure they’re all satisfiable at once.
That’s the approach Jarrah Bloomfield took for creating Too Many Cooks, the card game version of the omniwin training game. The premise is that you are a team of expert chefs, collaborating on the ingredients for a soup, and you all have very strong preferences from family secret recipes, so you need to make modifications to the soup without revealing what matters to you, only whether or not you’re satisfied, until everybody is satisfied.
Jarrah showed me the game personally at the Metagame 2025 Night Market—I sat down at his booth and he told me the basic premise and I said “I’ve been meaning to make this for years!” He spent hundreds of hours tuning it and playtesting it and getting it right, so that it works with a wide range of difficulties… and he did some sort of computer simulation to ensure that every if you put in all 10-20 constraints from a given difficulty level at once, there still exists a solution—highlight unlikely you’ll find it beyond 12 or so though. But this means that for any given 3-6 constraints, several solutions exist and you’ll likely be able to find one.
He initially gave players a limited number of turns to find a solution, but this resulted in overthinking, so instead he switched it to a timer-based challenge, which encouraged trying things rapidly and iterating—even if your change causes 2 problems by fixing 1, it’s still forward progress in learning, so it’s good that you made it.
One obvious difference is that this approach to the omniwin training game has persistent state, whereas the word game has fresh words every time. There’s pros and cons to each of these approaches. It probably becomes basically necessary in order to handle such a large number of constraints, and also helps you see smaller differences in order to figure out what is or isn’t relevant to someone.
Another thing that’s neat about Too Many Cooks is that when you first play, you look at your own constraint cards and they’ll say something like “same number of red cards as carrots” and “no blue potatoes or corn”, and you’ll assume that everybody else’s cards are also about quantity of particular cards, but they have something like “red in every corner” and “no greens next to yellows” and theirs are all about position! So you learn not just “other people want different things than me” but also “other people want kinds of things I hadn’t even imagined someone could want“.
…which is a very good insight to have about why omniwins are more possible than it usually seems. Your apparent intractable conflict is because you’re projecting their requests onto the dimensions you care about, when in fact they care about something completely different that’s only incidentally causing you problems.
I LOVE that Too Many Cooks exists. I’ve played it a few times, and it’s quite fun and indeed focuses your attention on the skill of how to get your own needs met while avoiding causing problems for others. Politicians would benefit from more of this.
And… it doesn’t satisfy a few things for me. The main is that the conditions are perfectly known and unambiguous. In the real world, while you’re ultimately the only one who can answer “does that work for me, given everything?” you don’t have perfect knowledge in advance of exactly what your constraint is. Your concept of what you want, and what you actually want, are not quite the same. Win-win solutions often involve being surprised when one of your assumptions doesn’t hold. But I realized that you could solve this with…
…a constraint that you somehow have privileged access for testing things on, but that you yourself don’t exactly know how it works or what satisfies it.
I haven’t yet figured out exactly how to do this, but I imagined some computer game, where:
One thing that’s cool about this version is that you can actually allow people to speculate openly about what their constraint might be! This is way more realistic to an actual good faith negotiation, where there’s minimal need to hide what you want. There’s at most some need to be a little vague about how well it works for you or exactly why, but even that is minimal in healthy families, companies, communities.
I think this system would allow you to add in that kind of element: you have players’ satisfaction be a number out of 10, not a yes/no, and then the game ends in a win if you reach everybody at 7/10 satisfied within the time allotted, but players earn points equivalent to the total satisfaction they personally had. In that case, there’s some incentive to bluff or deceive, to encourage people to go further in your direction.
However, if you reveal the results each time, players who bluff a lot become distrusted… so the iterated game theory here could end up reincentivizing honesty, both for the collaborative victory and for the personal points. And then you could get into versions where part of the puzzle is “how many players can we fit into some sort of win-win in the time allotted?”, where you have perhaps dozens of players and the game is no longer about finding something that works for everybody and instead about forming coalitions you can make it work with.
Lots of trippy possibilities here.
In 2022, as part of our courtship, my now-wife-then-girlfriend and I went on a bit of a camping roadtrip around Cape Breton, during a visit to my family in Nova Scotia. At one point, to pass the time in the car, I proposed we try this game I had come up with. We did the word version, which has the convenient property of working quite well as a game to be played when all you have are voices.
It turned out to have another amazing property too.
We had a few fairly easy rounds, and then we had one round that seemed impossible.
She opened with “beat”. I played “lackadaisically”. She played “kite”. I played “transpersonal”. She played “seven”. I played “supercalifragilisticexpialidocious”. There was an obvious tension between long and short words. Nothing that worked for me worked for her.
After 10 minutes of this, we started guessing words that would work for the other, and were usually right but not always. It clearly wasn’t just about word length: “stretched” worked for her but “beauty” worked for me.
Eventually we both became increasingly confident that it wasn’t going to work, because we knew what our own rule was and our best candidate for the other person’s rule was something that directly contradicted it. And if we had stopped there, and stated our rules, we would have agreed that indeed, it was impossible.
But I wasn’t so sure. I felt like I had one or two examples of places where her constraint wasn’t exactly what I thought it was, which might mean it wasn’t impossible. “Give me a minute,” I said, and with my new model of her constraint in mind, racked my brain trying to think of a real word that would fit.
“NONSYNERGY!”
“…that works for me,” she said. “it works for me too!”
And we had found our win-win.
What were the win conditions? I had “at least 4 vowels”, and she had “exactly two vowels”: a perfect contradiction. Except, she was counting only “aeiou”, and I was, at least sometimes, counting “y”.
Good thing we didn’t give up.
As I said, finding win-wins is often surprising and delightful.
There’s a tool called pol.is, which offers a prototype of a solution for the non-voting part of democracy: how do you even come up with the policies on which one might possibly vote? And they do something really brilliant, that I thought of while telling the above story, in such a way that makes pol.is itself kind of an instance of the omniwin training game—except that it’s real life, with real skin in the game, not just a toy training context. (The main active usage of the algorithm is Twitter’s Community Notes feature, which is used to add user-generated and swarm-moderated context to misleading tweets. There’s also this clone, contextengine.xyz)
Here’s the basic premise: people write statements, and people vote agree/disagree on them. Statements are ranked highest based not just on total votes, but diverse votes: votes from people who disagree with each other on other topics. And if you want, you can even see which clusters of anonymous voters are voting for which statements… which means you can attempt to bridge their views into something they might both agree with, by writing a new statement.
The real kicker is: there’s no other way to make any sort of commentary at all. There’s no comment feature, no “critique” or “argue” feature. If you want to critique someone’s idea, the only way to do it within this system is to articulate a better version. And if your version satisfies everybody who liked the original, plus some people who shared your critique, then it’ll win.
This upward spiral medium is a good way to domesticate memes in the political sphere.
Constantly consciously expanding the boundaries of thoughtspace and actionspace. Creator of Intend, a system for improvisationally & creatively staying in touch with what's most important to you, and taking action towards it.
Have your say!