The argument
Two degrees off at the dock is the wrong country a hundred miles out
This is the long version. The short version is on the front page and it is the same argument; this one shows the working, including the parts that are inconvenient for us.
The problem
Late is not merely late
The complaint everybody already has about dating apps is that you find out too late. That is true and it is not the interesting part. The interesting part is why late is worse, and it is a mechanism rather than a feeling: an angular error compounds with distance. Two degrees at the dock is nothing you could even measure. A mile out it is a few boat-lengths. A hundred miles out it is a different country, and the correction costs more than the trip did.
A fundamental difference works exactly like that. On date one it is a slightly awkward evening. In year three it is the whole thing, and by then the correction may not be available at all — there is a mortgage in it, or a child, or a decade.
So every app hands you two bad options, and most people have tried both. Confirm the bearing before you leave: interrogate a stranger about abortion and vaccines, over text, and poison the evening before it starts. Or be charming, be warm, enjoy it, and hope you have not gone too far off course by the time it feels safe to ask.
Neither is necessary if both charts were already on the table.
And the thing that actually sells this is not the filtering. It is the permission. You get to stop auditing. You should not have to wonder, three drinks in, whether the person opposite you holds something you cannot live with — and if you knew they did, you would not be spending the evening with them at all. So flirt. That is the entire point of the machinery.
The honest part
Asking the questions was never the hard part
We are not going to pretend this was an unoccupied idea, because it was not, and a claim like that falls apart in the first interview.
OkCupid has asked importance-weighted compatibility questions since 2004, and it is owned by the largest dating company in the world. A values-and-dealbreakers app called Weaver shipped in 2023. If the idea were “ask people what they believe”, it would have been done, repeatedly, by people with more money than us. It has been.
Here is what happens next, every time, and it is the whole argument:
OkCupid will tell you that you are a 43% match with somebody, and then show you that person anyway.
The number is decoration on an undiminished stack. Nothing was removed. You were informed and then handed the same pile you would have been handed without the questionnaire, because the pile is the product. A company whose revenue is attention cannot let a compatibility model delete ninety-five percent of its own inventory; that decision shows up on the next earnings call, and no amount of goodwill survives that meeting.
So the constraint was never on asking. It is on acting. And acting is not clever — it is just expensive in a way we have arranged to be able to afford.
What acting on the answers actually costs
Six things that follow, and none of them is a feature idea
Each one exists because the argument above requires it to be true. If any of them stopped being true, the argument would be marketing.
The feed is finite and it ends
Your ranked list is as long as the number of people who actually clear your gates, and when you reach the bottom it says so and stops. It is not refilled with people you already ruled out, and nothing is held back to give you a reason to come back tomorrow. Some days the honest answer is four people. That is a real number about your life, and the product's job is to hand it to you rather than to hide it behind an infinite scroll.
Five dealbreakers, and not six
Unlimited hard filters multiply to an empty feed, so the cap is real and it is enforced by the database rather than by a screen. Choosing a fifth means giving one up, which forces a ranking, and a ranking is honest in a way an unlimited filter list never is. Everybody has more than five things they care about. Almost nobody has more than five they would end it over.
Every line is priced in people before you draw it
Before you spend a dealbreaker, the product tells you how many people in your radius it removes. Before you narrow your radius, the same, rung by rung. The number is usually uncomfortable, which is presumably why a product paid by the size of its stack would not show it to you. We think it is the most useful thing we can tell you, and we would rather you drew a line knowing what it costs than drew it in the dark.
You see somebody's answers only when they agree, in that conversation
Rolling out the chart is mutual and simultaneous, per conversation, revocable by either of you. There is no upsell that shows you somebody's values before they have agreed, at any price. That upsell is the dark pattern this company exists to refuse, and every read of your answers is logged with a name against it, in a record you can download.
The members write the question bank, not us
Members propose a topic, get it seconded, and take it to a vote. An editor controls wording and cannot decline a subject. A published charter bounds what a vote is allowed to do, and every proposal that ever existed stays published, including the ones that failed. This is the part a public competitor cannot copy, and not for a technical reason: running it would mean letting your own users vote in questions that shrink your own inventory.
The price is published in full, in advance, with a ceiling
Every rung this company will ever charge is on the pricing page before you pay anything, and the top is $20 a month, permanently. A published ceiling is the other half of a published ladder; without it the ladder is only a warning. It is also the reason there will never be an app-store version — you cannot promise a ceiling on a rail where somebody else sets the toll.
The line we will not cross
We do not decide what you should care about
This is a sorting product, not a persuading one. It has to be equally useful and equally respectful to a devout homeschooling family and to a childfree atheist, and the instant it reads as one tribe’s product the other tribe leaves and the filter has nothing left to filter.
So the bank is governed rather than curated. There is no subject Bearing refuses to ask about — abortion, Israel and Palestine, vaccines, immigration, medical transition are all in there precisely because they end relationships. Topics are unlimited. Framings are not. What the bank will not carry is a question whose wording has already decided its answer, or that names a group of people as a thing to hold a position “on”. Those are not topics; they are conclusions wearing a question mark, and no fair set of answers can be written underneath one.
That is a practical rule before it is a principled one. A question phrased in one side’s vocabulary returns that side’s opinion and nothing else, because the other side cannot find themselves in any of the options. It sorts nobody. Reframing is not softening — it is the only version that works.
Values are chosen. That is not. The same goes for disability, national origin and health status as distinct from health choices.
What it cost
Four things we gave up on purpose
Written down because a list of principles with no losses attached is a brochure.
The waitlist, which was the better launch story
Bearing was going to hold a city shut until 3,000 people there had signed up, with a public counter and a queue. “Calgary: 2,841 of 3,000” makes waiting feel like progress and turns every person on the list into an unpaid recruiter, which is genuinely the strongest organic loop available to a product like this. We gave it up, because it excluded small towns permanently, because a threshold is not a sequence so the queue position was fake, and because refusing somebody for living somewhere small withholds the one product on the market that handles a thin pool honestly. The whole argument is here.
The trial, and the moment where we ask you for the dollar
There is no free trial, because the free tier is the trial and it does not expire: 3 from your ranked list a day and 1 new conversation a day, permanently, with no card and no cliff. What that costs us is the conversion moment. A trial manufactures one; a free tier does not, so paying has to come from the product being better without the limit rather than from an expiry date. That is the harder version and it is the one consistent with everything else here.
Every form of friction, including the useful-looking ones
An earlier specification had a seven-day cooling-off before you could release a dealbreaker, and a monthly cap on changing them. The worry behind it was real — somebody could edit their values to match a profile they want. But we already refuse to tell members what to care about, and the pool-cost screen exists to show the true cost of a line before it is drawn. Charging somebody for acting on that information contradicts the thing the screen is for. So the attack is not prevented, it is made visible: every answer is dated, and the chart tells both of you how old each one is — answered seven months ago reads differently from changed two days ago, and it reads that way to the person on the other side of it. Prevention would have fallen on everybody. Visibility falls only on the person doing it.
The size of the company
At $20 a month, at half of what the incumbents charge, with customers we are actively trying to lose, this tops out as a good small business rather than a large one. That is not a weakness we are explaining away — it is the mechanism that protects everything above. A company that has promised its investors engagement growth cannot keep a finite feed. So we will not take money that requires it.
How you check us
The number that would prove it
A dating app saying it wants you to leave is a claim, and claims are cheap. The one that is not cheap is publishing how often it actually happens: the share of departing members who told us, on the way out, that they left because they met somebody here. Published monthly, whether or not the number flatters us.
Nobody else publishes it. That is the difference between promising to be deleted and showing the receipts.
The first rung is $1 a month and the free tier does not expire, so nothing above depends on you deciding anything today.