Zonori

You are standing outside a room you have not been in before. It starts in four minutes, and you do not know whether you will recognise one face inside it. Most people solve that by not going.

To confirmattendance I go to the ecstatic dance, the kirtan and the bachata classes in Bangkok, and it is the same faces moving between them. The dance is one organiser's calendar, the kirtan is an Instagram account, the class is a studio's booking form. Nothing in that stack knows they are the same person. Neither does the person walking in. What makes you stay in a city is not meeting someone on the first night. It is recognising a face on the second.

What it is

Every platform that organises going out is paid by the organiser. I checked eight: Meetup, Luma, Eventbrite, Eventpop, Ticketmelon, SweatPals, Timeleft and 222. Six are paid by the organiser — a group fee, a host plan, a cut of each ticket, a commission on the session; two charge the attendee, but for a matched table that resets each week. When the organiser is the customer, the object the platform maintains has to be the organiser's container, and a person exists inside somebody's or not at all. None of the eight maintains an object whose unit is a person in a city, and none has one that spans organisers. My own read is that this is not an oversight: a room nobody owns dissolves the thing the organiser is paying for. If you know of one that does it, that is the most useful thing you could tell me.

You need not take my word for the absence. Pick an interest you actually have, in your own city, and try to find one place that shows you every organiser running it and who from your circles is going to each. People already do that join by hand — runclubbangkok.com keeps a weekday calendar for 16 clubs it does not run — and even they hold events, never people.

Zonori inverts the payer. The attendee pays, and Zonori takes no cut of anyone's door money and does not process it. What it maintains is (person, interest, city) in a container nobody owns: Running · Bangkok, not one club's group. The rule that a circle may not be a club, a studio, a brand or a personal following is written into the review policy the API applies before anything goes live. When I first wrote this page that was a rule in the code and not in production. The API now refuses to start in production without the review key, and production is up. Two rival running clubs' events land in front of the same members. The membership outlasts the organiser, and it outlasts the city.

Who built it

Zonori is built and run by Shaquille Hinds, on his own: a Barbadian engineer in Bangkok, To confirman introvert, building a tool for meeting people — which is the honest reason it has no feed, no infinite scroll, no swipe-first matching screen, and a written list of persuasion techniques he will not use. The app, for iPhone and Android, and the API were designed and built by him, directing Claude Code agents for implementation, which is why line counts are not the point. The repositories' histories begin 7 September 2026 and he commits to them most days. Both are private; read access on request.

One thing you would have found, so you hear it from me: until 19 September the People screen ranked and sorted people by a weighted overlap score in which astrology was 20% and personality type was 10%. It was on that same written list, and it came out that day: ranking is now shared circles, shared values and events you have both been checked in at, and astrology and personality are shown on a profile as facts, never used to order anyone. It is the clearest example of what I am trying not to build. With the check-in screen, the review policy and the seed script, that is four things a careful reader would have turned up in a week, and you have them, with what happened to each, before you asked. It is the only evidence I can offer this early about how I will report a number I do not like, and I will have one of those before I have a good one.

Where it stands

Nothing is live. Zero users, zero revenue. Everything above is how it is built, not what is happening. The piece that produces the number I am about to promise you — a host checking someone in from a roster — had no screen in the app when I first wrote this page. It went in on 19 September and is in the current build. If for any reason it is not in the launch build, there is no November number, and I will send you that sentence instead. The site is live, the staging and production servers are up, and the build is in TestFlight and in Google Play's internal testing. I submit to App Store and Google Play review in early October; if review clears on the usual timeline it opens in Bangkok, on iPhone and Android, in October, and if it comes back I will send you the new date rather than the old one.

The number I will report

What I am testing is whether a room nobody owns pulls in organisers who do not know me. One number: people who checked in to an event I did not organise, where the check-in was recorded by a host or a chapter admin who is not me. Not downloads, not RSVPs, not members — onboarding makes you join three circles before it shows you anything, so the product manufactures members. Someone else records it at a door, which makes it harder to inflate than a signup — and it is not unfakeable. Every app like this has a seed script and mine can write host-recorded check-ins. So: a guard that refuses to run it against the live database went in on 19 September, and it is excluded by name in my deployment notes. When I send the number, it comes with the exact query and a screen recording of it running against production, with the coverage line beside it — and beside that, how many of those people came back in a different week. The mechanism actually built for the second night, a card the morning after, has been in the build since 19 September, so that half will have a few weeks of data by 30 November and a real answer in 2027.

The decision

On Monday 30 November I read three counts against thresholds I wrote down in September. Three outcomes, not two: continue in full, continue on a narrower footing, or wind the Bangkok chapter down. You get the counts either way, including if they say stop.

What it costs

About $910 a month: $8–14 is servers, the rest is one person at $29.59 a day. It sells once, at $49.99, for more circle and city slots, and I have no evidence anyone will pay that. No published benchmark exists for one-time conversion in this category, and I wrote that down rather than borrow a subscription rate. The first test is a purchase in October.