Back
Working build · Paused
( Case study · 2026 )
( 01 — The short version )
What
A city guide for where locals actually go: places, scores and tips from people who live there, with check-ins — what Foursquare City Guide was until it closed in December 2024.
Why
City Guide was how I found a good place in a new neighbourhood. When it shut down, Swarm kept the check-ins but lost the guide. Google Maps lists everything; it doesn’t tell you where the locals go.
How
I designed it and built it as a working iPhone app, with Claude as my engineering partner — on real data, real terms of use and a budget of zero. Then I checked the market before building more.
Places in Israel, from open data
Data sources, ranked and capped
Terms rules that changed the design
Passes at the home screen
To search. By design.
( 02 — The gap )
THE
GAP.
Problem
The guide closed. The need didn’t.
Foursquare City Guide shut down on December 15, 2024. Swarm stayed, but it’s a diary of where you’ve been, not a guide to where to go next. Google Maps lists everything, and almost everything scores above four stars. What went missing was the locals’ opinion.
Google Maps
every place, four stars
Swarm
check-ins, no guide
what photographs well
Group chats
“anyone know a place?”
City Guide
closed in 2024
Where locals go

( 03 — The data )
Search runs on 74,350 places I keep in my own database: Foursquare’s open data, and Overture’s where Foursquare had nothing. Paid sources only add detail to a place page, under daily caps that fail closed — so a busy day can’t become a bill.

( 04 — The rules )
THE
RULES.
Constraints
Reading the terms before launch, not after
Foursquare and Google both publish rules for how their data can be shown. I read them before anyone else used the app, and six things I’d already designed had to change. Each change made the screen clearer, not only compliant.
One blended score
Google places on an Apple map
Storing Google’s data
Uncredited data
Foursquare-green badges
“City Guide” on the logo

( 05 — The home )
The first home put the categories at the top and left the rest of the screen empty. City Guide had the better answer: look at the top, tap at the bottom. Four passes later, everything you tap sits within reach of a thumb, and the top got a photo and the name.

( 06 — The revive )
Foursquare closed the guide but kept Swarm, and people still check in there every day. So Square doesn’t ask anyone to leave: connect your Foursquare account once, and every check-in in Square also lands on Swarm — same venue, same note. A revival, not a replacement.
( 07 — The call )
The market check
“Is it better than what people already use — by enough to switch?” For Square, the honest answer was no.
Corner, with $7.5M from investors, already turns friends’ saves into a city map, Tel Aviv included. Swarm keeps the check-in habit. Google has every place. Square did some things better — scores, locals, Swarm — but not by enough. So I stopped adding features and kept what it taught me.
Everyone
Every place and live reviews. Hard to beat on coverage, easy to beat on taste.
Still live
The check-in habit and years of history. Square posts there instead of competing.
$7.5M raised
Friends’ saves on a shared map, free while investors pay. Already in Tel Aviv.
One designer
Scores, locals’ picks and Swarm in one guide. Better, but not by enough to switch.
( 08 — How it was built )
THE
BUILD.
Build
A real app, not a prototype
I designed every screen and wrote down every behaviour, and built it with Claude as my engineering partner: a native iPhone app, a server with hard spending caps, and a database of 74,000 places I imported and cleaned myself.
Building it is what surfaced the real questions — what data costs, what the terms allow, what a check-in needs — long before a mockup would have.
01
SwiftUI and MapKit, with a photo viewer you tilt to look around.
02
Supabase: one edge function, a daily cap for every paid source, Postgres for places and check-ins.
03
FSQ OS Places and Overture Maps, imported with DuckDB, deduplicated, chains marked.
04
Foursquare sign-in through the server. The token stays on the phone.
SwiftUI
MapKit
Supabase
Postgres
DuckDB
Foursquare API
Google Places
Claude
( 09 — What I learned )
01
When every place page costs money, “rich content” stops being a layout question. The best decision in the app was a data decision: search on what you own, pay only for detail.
02
Separate scores, credited sources, no look-alike colours. Each one was required, and each one made the app clearer to read.
03
I asked it late. A working build made it impossible to ignore — a good reason to build early, and to look around just as early.
( 10 — Where it is now )
A working build on my phone, with 74,000 places.
Built
Check-ins post to Swarm through a Foursquare sign-in.
Built
“Square” is a working title. The final name needs to be its own.
Open
One test could reopen it: a person’s years of Swarm history, alive again on day one.
Paused
( Get in touch )
I’m a product designer who builds what he designs. If you’re hiring for product design — in Israel or remote in Europe — I’d love to talk.
Get in touch
Product · UX · iOS · AI-assisted building
Product strategy · UX & UI · Data & API design · iOS (SwiftUI) · AI-assisted engineering





