how to find people with similar interests. the real problem is signal, not venue.

You can read this whole essay as a single claim. The reason it is hard to find people who share your interests is not that the people are rare. They are not. The reason is that almost every venue you can stand in front of, online or off, makes the wrong signal legible. You can read how someone looks, what part of the country they live in, what their job title is, sometimes their political team. You cannot read the part of them you actually wanted to meet.

This essay is about that mismatch, and about what you can do about it given the venues that actually exist in 2026.

why "find a meetup" stopped working

The standard advice has not changed in fifteen years. Go to a meetup. Join a club. Take a class. Find your local subreddit or Discord. The advice was good in 2010 and it is still partially good now, but the failure mode is consistent enough that it deserves naming.

The failure mode is that the venue selects on the wrong axis. A photography meetup selects on the word "photography" but the room contains the person who wants to talk about medium-format film for four hours, the person who only does iPhone street, the person who is here because they were hoping to date someone with a similar hobby, the person who is here because their therapist told them to leave the house, and the person who builds the lenses. All five people will say "I like photography" if you ask. None of them are interested in the same thing.

This is not a problem with the meetup. The meetup did the only thing a venue can do, which is select on a topic name. The problem is that topic names are very low-bandwidth descriptions of what someone is actually into. A topic name is a label on a folder. Two people whose folders share a label can have wildly different contents inside.

So you arrive, you have a fine evening, you go home, you do not feel like you found anyone. You blame your social skills. You blame your city. You blame your age. The real cause is that the venue had no way to surface the specific texture of what you cared about. It could only group people by the name of the room.

what the platforms get wrong

Most of the apps that promise to fix this are doing the same thing the meetup does, just digitally. They give you a checklist of interests, you tick "hiking" and "books" and "live music," they show you other people who ticked the same boxes. This produces matches in the way a Venn diagram produces an overlap, which is to say, structurally, but with no content.

A few do it slightly better. Bumble BFF added prompts. Hinge has prompts for dating. The prompt is a small text box where you write a few sentences about something specific. This is a strict improvement over the checklist, because a sentence carries more shape than a tag. But the matching engine on those apps does not actually read your prompts in any deep way. Photo is still first in the rank. The prompts decorate the profile after the face has already decided.

The platforms that hit the texture more closely tend to be the ones nobody markets as community apps at all. The Are.na user who reblogs your art reference is reading a very specific thing about you. The person whose Goodreads is uncannily aligned with yours has revealed a real overlap that no checklist could produce. The mutual on a tiny niche Substack who replies thoughtfully to your comment three times in a row has demonstrated something a meetup never could.

So the actual landscape is not "the right app exists, you have not found it." The landscape is that interest-affinity gets surfaced by media you both spend time on, and by writing you both produce, and almost never by anything that advertises itself as a community platform.

what "similar interests" actually means

Before walking through what to do, it is worth being precise about what you are actually looking for. The phrase "similar interests" is too coarse to be useful as a search query for your own life. It means at least four different things and they each need different approaches.

The first is shared topic. You both like indie strategy games. You both follow medieval history. You both garden. This is the easy version and the version every platform tries to solve. It is also the least predictive of whether you will actually enjoy each other. Two people who like indie strategy games can talk for ten minutes and then run out.

The second is shared sensibility. You both find a specific kind of thing funny. You both notice a specific kind of detail. You read books in the same direction (toward the structure, or toward the emotional weight, or toward the prose). This is harder to surface and much more predictive. Two people with very different topics but the same sensibility can talk for years.

The third is shared depth. You are both at a comparable level of seriousness about the topic. The hobbyist photographer and the professional photographer will struggle to find a register for each other even if they technically share the interest. The depth mismatch is silent and corrosive.

The fourth is shared trajectory. You are both moving in the same direction (getting more into the thing, less into the thing, pivoting from it to something adjacent). The two people who fell out of love with running last year and are figuring out what is next have a lot more to say to each other than the two people who are both currently running.

The reason most "shared interests" matches go nowhere is that they nailed the first one and ignored the other three.

a practical sequence

Given that the platforms cannot read the second through fourth, here is what I have watched work for people who are actually finding their people now, in the actual venues of 2026.

Start with the writing or reading you already do. If you write anywhere at all, even a private Substack or a Notion page or a personal blog with twelve readers, the people who reply thoughtfully are a higher-signal pool than any meetup. If you do not write but you read, the comments section of writers you respect is the same pool. Reply to a comment that struck you. Reply once a week for a month. You will recognize the same handful of names. Some of them will recognize yours.

Use the recommendation engines of media you already love. If you have a Goodreads with two hundred books rated honestly, ask it for users with the highest overlap. If you use Letterboxd, do the same with film. If you are on Are.na, follow the people whose channels you keep returning to and look at who they follow. The recommender knows more about your interest texture than any survey-based matching app can extract from you in five minutes.

For depth and trajectory, look for the small forums and Discord servers rather than the big ones. The big subreddit for a topic gives you topic overlap and depth chaos. The small Discord that has eight hundred members and a weekly thread on something specific gives you topic overlap and depth alignment for free, because the people who stay are the people for whom the depth matches.

When you do try a meetup or class, set the bar low and pay attention to the third-person signal. You are not trying to walk out with a friend. You are trying to notice which one person said the one thing that surprised you. Get their handle. Move the conversation to a channel that lets writing happen. Writing is the venue where sensibility gets revealed; rooms are mostly where topics get named.

If you are searching specifically for an app, the honest filter is this. Does the app let you write something specific about how you think about the thing, not just that you like the thing. Can the other person read what you wrote before they have decided whether you are worth their time. Does the matching engine read the writing, or does the writing decorate a profile that was already ranked by something else.

That is most of the test.

where byvibration fits

I work on byvibration.com, an app I built because I wanted the answer to this problem in a form that was actually usable. The matching engine is structurally photo-blind: the function that ranks two profiles cannot reach a photograph at all. We removed it from the type signature so the build would fail if anyone tried to thread one through. What the engine reads instead is a vibe profile written from a few prompts: how you think about a few small honest questions. The profile gets embedded, the embeddings get compared, and the people who surface as suggestions are people whose writing reads like writing you might want to read.

It does not solve the four problems above on its own. Nothing does. But the prompts are designed to draw out sensibility and trajectory rather than topic, and the ranking does not penalize you for being a year off the trend or for being at an unfashionable depth. That is the specific bet of the product.

The interest-affinity work is yours. The venue can either help or get in the way. Most venues get in the way.

I run byvibration. The matching engine is open source at github.com/donnowyu/soulmate-core (MIT, 65 unit tests). If you try it and the suggestions feel wrong, please tell me. The shape of "wrong" is the most useful signal we get.