Hello Nearstream: the anti-social network that my friends won't use

I built an anti-social network and my friends won't use it

Five people. All close friends, all genuinely interested, all of whom said yes when I asked. They signed up, looked around, posted once or twice, and stopped.

I know exactly why, and it isn't a bug.

---

The bundle

Social networks did something clever about twenty years ago that nobody has undone since. They took two separate things — publishing and reading — and made them one product.

Your page and everyone's feed became the same company. That sounds like convenience. It's actually the whole problem, because once one company owns both halves, the feed has to grow. A feed that has to grow gets ranked. A ranked feed gets optimised. An optimised feed stops being a record of your friends and becomes a machine for holding attention, and the page you write on becomes an input to it.

You can watch this happen in slow motion across every platform. Nobody planned it. It's just what the bundle does.

The unbundled version isn't a new idea, and it isn't hard. You own a site. Your friends own sites. Something pulls their sites into one place for you to read. That's it — and it already existed. It's RSS, a boring twenty-five-year-old standard that solved this before the problem arrived.

The IndieWeb community has been saying this since 2011. The reason it never reached normal people isn't that the idea is wrong. It's that setting it up required being the kind of person who enjoys setting things up.

The old machinery it runs on

Here's the part I find hard to stop thinking about: none of this works by inventing a new human weakness. It works by finding an old one and pulling on it.

Wanting to know what your group is doing is not a modern defect. For most of the time humans have existed, knowing who was where, who was with whom, and who was upset with you was survival information. A person who tracked that lived better than a person who didn't. The pull toward it is old and it's supposed to be there.

What's new is that the thing answering the pull is now a machine that can answer it a thousand times a day and never resolve it.

The specific mechanism is not mysterious and it isn't even recent. B.F. Skinner worked it out with pigeons in the 1950s: a reward that arrives unpredictably produces far more persistent behaviour than a reward that arrives every time. Predictable rewards get boring. Unpredictable ones become almost impossible to stop. It's why slot machines work, and it's why the pull-to-refresh gesture — which is, mechanically, pulling a lever — was such an effective invention.

The commonly repeated version of this is about dopamine as a pleasure chemical, and that's not quite right. Dopamine has much more to do with anticipation than reward, and the signal is at its strongest when the outcome is genuinely uncertain. Which means the compulsive part isn't someone having liked your post. It's maybe. Maybe there's something. That's the state the machine keeps you in, and it's a state with no natural end, because uncertainty can always be renewed.

Then there's the quieter loss, and it's the one I feel personally. I can't be bored anymore.

Not in a poetic way — I mean I have lost the actual skill. Waiting for a train, standing in a lift, walking somewhere I've walked a hundred times before. Those used to be empty and slightly uncomfortable and that discomfort was where a certain kind of thinking happened. Ideas arrived in those gaps because there was nowhere else for my attention to go. I filled every one of those gaps in about 2012 and I don't know how to unfill them.

I want to be careful with the bigger claim. It has become normal to say that phones caused a generational mental health crisis, and I notice I believe it, and I also notice that when I read the actual research it is much less settled than the headlines. Some researchers report meaningful effects. Others — looking at the same data — find associations so small they'd struggle to explain much of anything, and point out that most of it is correlational anyway. I'm not equipped to settle that, and I don't think the essay needs me to.

The narrower claim is enough, and I'm confident in it because I've lived it: these systems are built to be difficult to put down, that difficulty is engineered rather than accidental, and the thing being exploited is a real and reasonable human need. You are not weak for checking. You are being answered on a variable schedule by a machine that has no reason to ever let you finish.

So the question I actually cared about was: what do you build if you believe that?

What I built

Nearstream is a shared journal between close friends.

Everyone has a personal site on their own domain. You post to it: a line of text, a voice note, a photograph, a piece of music, an essay. A reader pulls your friends' sites into one room.

The important part is that the reader is yours. It's local and private — it pulls only the feeds you chose, and nobody else can see it. There is no shared timeline that everyone appears in. There is no discovery. Your room is not a room anyone else is standing in.

There are no likes. No follower counts. No notifications. Nothing that can be a number.

It's AGPL and self-hostable, which matters more than it sounds. If a friend of a friend wants one, the answer isn't "join my instance" — it's "run your own." The thing should spread by codebase, not by user count. An instance that grows is an instance that has started to become the thing I was trying to get away from.

MySpace, briefly

While I was building this, the owners of MySpace announced they're bringing it back. Their stated reason: modern social media is too algorithmic and too optimised for endless scrolling.

That is exactly my diagnosis. They have held the brand since 2011, lost around $150 million on their last attempt to revive it, and announced this relaunch with no date and no description of what it would actually be.

So: they have a brand and no product. I have a product and five users. Neither of us has solved the real problem, and I don't think their approach can, because another platform is the thing that failed the first time. You cannot fix the bundle by rebuilding it with better intentions.

What building it taught me

Most of what I learned was about restraint, and most of it came from getting it wrong first.

I gave the home page four categories — notes here, pictures there, essays below — and it read like four filing cabinets. Then I rebuilt it as one dated page with no sections and no pagination, and it read like a diary. Nothing is "older." Nothing is behind a load more. The whole page is just there, and it gets longer.

I borrowed that from Mataroa, a blogging platform for minimalists that puts every post you have ever written on a single page. Their site is mostly a list of features they refuse to build, with reasons. So is Bear Blog — no JavaScript, no stylesheets, no trackers. I found that clarifying. The refusals are the product.

Two details came directly out of the thinking above, and they're the ones I'm most pleased with.

Entries fade as they get older — a little contrast, a hair of blur, over about a year. Recency without an algorithm deciding anything. Nothing is ever hidden or paginated away; the past just goes quiet, the way a print left in the light does.

And a new post arrives faint, and darkens as it sets, like ink — it reaches full strength about six hours after you write it. You cannot see your own new post at full strength immediately. That one is deliberately hostile to the mechanism I described earlier: there is no hit available at the moment of posting. You write the thing, and then you go away, and it's properly there by the evening. It is the smallest possible argument against the slot machine, built into the surface where you'd expect the lever to be.

The closest existing thing to Nearstream is Micro.blog, which Manton Reece has been building since 2017. If any of this appeals to you, go and look at it — it's more mature than mine and he has thought about it for longer. The difference is the room. His is one shared timeline that everyone appears in: a small public web. Mine is a closed group of friends, with no public anything.

Finding it after I'd started was reassuring rather than deflating. It meant the idea was real enough that someone had already reached it independently.

The part I haven't solved

Which brings me back to my five friends.

I removed every mechanism that makes people post. Likes, counts, notifications, streaks, the little red dot. That was the entire point — those are the levers, and I took them all off the machine.

But they weren't decoration. They were the engine.

Nobody sees what you wrote, so there is no reason to write it. You post into a quiet room and the room stays quiet, and the next week you don't bother. I built something with no way to know if anyone is there, which is exactly what I said I wanted, and it turns out to be a colder experience than I'd imagined from inside a design document.

And I'm aware of the shape of what I've done. I removed a system that exploits a real human need, and I replaced it with nothing that meets that need. The need doesn't go away because the machinery exploiting it was cynical. Wanting to know your friends are there is not a bug that platforms installed in us. It's the thing they found and monetised.

I don't think the answer is to add the mechanisms back, because then I've just rebuilt the thing with fewer users. I think there's a version of presence that isn't a metric — something that tells you a friend is around without turning it into a score. I don't know what it is yet.

So this is an unfinished thing, published anyway. If you have thoughts, my email is on the about page. If you want to see how it works, the manifesto has the reasoning and the code is open.

And if you want one of your own — good. That's the idea.