I'm a very introverted, people avoidant person and I have a hard time holding conversations. Sometimes I wonder about meeting people 1:1 just to socialize and to practice conversations but I never know how to go about it. I tried friend finder / dating apps but none of them really clicked with me and I quickly gave up on them. So in this post I daydream about what app I think my autistic self would be happy to use over a longer time horizon.
Friend-finding apps are a dime a dozen but they all feel shallow to me, especially the swiping ones. There are non-swiping ones where you can see all profiles from the get go but there I get overwhelmed by the profiles. And even when you like each other, the apps don't hold your hand afterwards.
Furthermore these apps are paid which means they have a conflict of interest. If they quickly give you what you are looking for, then they don't make money. So they have to engineer them to be addictive but generally unsuccessful if they want to stay in business. And then, as usual in unregulated capitalism, all these companies are bought up by the same investors who then force them all to suck the same way.
I want a website that:
Surely this is not too hard, so let me try.
I'm imagining a text-only website. I think photos are just a distraction for the types of interactions I'm looking for. Making it text-only hopefully also drives away people who are looking for those shallow interactions. The text-only format will also reduce server costs and I won't need to worry about people uploading illegal pictures.
Anyone should be able to create an account on this site. Your profile will consist of four pieces of data:
Your public profile text should tell your rough age, your sex, and what you like to talk about. An AI will check whether your profile matches these criteria. Your profile update will be shown to others only if it passes this AI test. Thanks to AI we can get rid of the outdated, rigid bureaucratic systems the old apps have and replace them with a fuzzier system where users can add any nuance they want.
Your search description should describe what you are looking for. If you are looking for a middle aged man with a beard who likes to watch Netflix movies and is within 5km, well, then you write exactly that in here. No need to reduce your complex desires to a few inflexible metadata fields.
Your private profile text will not be visible to other users. However the AI systems will be able to see and use it for improving the daily profile suggestions. You can write stuff like "plz avoid bald guys" that you otherwise wouldn't want to write publicly.
Each text field is limited to about 2,000 characters. It feels like an appropriate limit. Not too long (not a soapbox) but not overly limiting either.
In general the website will allow you to send messages to any connected user. However initially you will have zero connections and the connections will grow very slowly. You become connected to another user after both users answer the following two opening questions based on each other's public profile texts:
So yeah, this is way more involved than swiping left and right. Answering both of these questions will require some thinking. This is how I imagine a "non-shallow" interaction.
The first question is to prime you to find reasons you like this person rather than reasons you dislike them. And then the next question is meant to be a form of a conversation starter from that mental state.
There's no equivalent of swiping-left or swiping-right; you need to answer the questions for everyone equally. The closest thing to swiping left is clicking "hide profile" and the closest thing to swiping right is sending a message.
The website runs on a daily cycle with the website being unavailable between 23:00 and 01:00 UTC. During the day people can answer the opening questions (three at most) and send messages to other connected users (five at most). However these are delivered only the next day (no instant messaging here) and those small limits ensure the site cannot become overwhelming or addictive.
The user flow is that you check your inbox, perhaps send some responses/messages, and do your three-profile answering cycle. And then you are done with the site for the day; you can move on with your day.
The frontpage will have the following sections:
When you click on a connected user, you will see their current profile text and a chat thread. The first few messages are the original profile texts, the opening questions, and the answers to them. So even if the profile text or the opening questions change over time, the users can still see how it all began.
Also note that the users don't have usernames but rather random ID numbers. But you can attach a private freeform note about other users. The system will use the first line of that as the name to display when listing that profile for you.
The website also gives a clear suggestion of what the users should strive for: meeting for a game of Gomoku. It's the game of Xs and Os where you have to get five in a row. It's a simple, fast paced game to play that requires only a piece of squared paper and a pen. The site could offer an online version too for webcam meetups.
The game is not the point of the meeting, it just provides a basic scaffolding for a nice real-life chat. As a twist, if people are up for it, people can opt for a rule that they can only speak when it's their move and listen otherwise. Sometimes there's not enough listening happening in today's world, so this could be a nice avenue for that. This rule would be in line with the turn-based chatting spirit of the site.
Whenever you finish answering the opening questions for the three users, the system will give you another batch of three users to process the next day. The system selects the three people in the following way:
The random user might be a person you don't like or are philosophically opposed to. Finding commonality might be hard then. But this is the goal of the site: to interact with all sorts of people, not just the ones you seek. The site doesn't let you progress until you can master this.
The system considers for selection only the users who logged in at least once in the past 8 days.
This site would be simple to run, incurring very small server costs. Most of it is a static website with an endpoint where users can prepare their answers and messages to be processed at midnight. For the midnight batch processing you don't need the smartest AI; the site could run its own models without needing to send any data to big tech. This site could be run pretty much for free, and hence it doesn't need to be backed by a business.
The site would also require a one-time registration fee of at least $1 just to reduce spam accounts. To avoid a conflict of interest, the site would commit to donating all proceeds to a predetermined charity like EFF or GiveDirectly. The site doesn't need to implement payment processing, it can simply use https://docs.every.org/docs/donate-link and verify the donations with webhooks. Perhaps the site should make a few more commitments to protect against enshittification over time as outlined in @/commitments.
Furthermore, there would be three tiers of users: new, verified, trusted. Trusted people would be hand-selected by the admins and their extra power would be to upgrade new users into the verified tier after meeting them. When searching you would be given the option to avoid new people just to avoid the potential spam accounts. You can remove the filter once you have exhausted the verified pool to expand the potential users to connect with at the cost of seeing a greater number of lower quality accounts. You could even sign up to be a trusted user if you like meeting new people.
For a fresh start the site would allow resetting connections that were idle for more than 18 months. This should eliminate another reason people might want to create new accounts.
The only problem I see is adoption. I'd expect not many people would want to join such a site. So it might not be worth creating it. Though maybe it being a niche, few-users site would actually make it more attractive, hmm.
Would you join such a site? Hit the thumbs-up reaction in the reaction box below if yes, thumbs down if no. Asking for a friend.
Published on 2026-07-06.
Comments:
#c1 by cghkzuop-guest on 2026-07-20
I was thinking about doing something similar, but more targeted for dating apps.
My criteria was very small to none server costs and that users can send images to each other but none of it is stored ever.
I do not think though that AI analysis of profiles would be scalable - altough the local models are not bad, if the site really gets big things can go wrong so I prefer metadata.
Other ideas is that I would try to monetize the "wrong" things: when someone ghost without writing answers or someone heavily, demonstrably lies I thought about taking away their ability to send messages until either they pay or they "fix" these. Not filling out your profile and other things are too penalized and rules would be public.
I actually think this would be still monetizable by the sheer amount of people doing "the wrong things" but at least it would be penalized. Also people often like feeling they "cheated" the system by trickery even if the "price" they pay for it is behaving well :-)
Another idea of mine would be a fully private - non-searchable app/website which however can blink everyones phones who use the app and are on the same location (you can turn off such notifications temporarity though for known reasons). I think this could be fun to use - but in this one could have "subcultures" where if you issue a ping, only those belonging to said "subculture" would rang in your location. This way it would make it easy to approach ppl with either weird kinks, lesser known but general hobbies, interests and I think would be fun to meet with "oh that girl likes dune" kinda happenings in real life :-)
Again adoption for the second app is really a key - so possibly its a marketing-heavy operation, which is hard unless one has endless money of huge enough platform... Or one can try to "incorporate" this into a service that otherwise already gets at least somewhat successful.
preney
#c1-1 by iio on 2026-07-25
I do not think though that AI analysis of profiles would be scalable
By AI, I mean the computationally cheapest models, like Gemini Nano, which even runs in your browser: https://chrome.dev/web-ai-demos/prompt-api-playground/. Those are easy to run at scale even on a modest gaming machine. I don't need it to be perfect; I just want something that's a tiny bit better than random.
non-searchable app/website which however can blink everyones phones who use the app and are on the same location
I just checked, and there are a bunch of apps like this. But people don't like them because constant location sharing drains their batteries way too quickly.
monetize the "wrong" things
I'm still looking for some sort of "proof of effort" that is not money. I want to avoid giving an unfair advantage to people who have more time or money available. I think I'd require users to physically visit an office to verify their account and photos. I wonder if people would be willing to do that.
I'm also becoming more convinced that the broker model will be the only scalable approach. Rather than swiping through profiles, you would be chatting with an AI. The AI would know about all the profiles and could ask the right questions to determine which profiles would fit you, and it could even arrange dates. And if a date doesn't work out, or someone ghosts you, the AI would take note of that. If there's no match for you, then it would tell you directly rather than letting you waste time on the platform. It would also give you tips on how to become more attractive.
I expect this would be so efficient that it would make no business sense to run it. So this, too, would need to be maintained by the community rather than by private companies. (However, it looks like Bumble is already pivoting to something like this.)
Add new comment:
(Adding a new comment or reply requires javascript.)