A compass app for finding the nearest public toilet, built almost entirely with Claude Code, and it is being discussed everywhere across Reddit, X, HN and other communities.
What Is It?
The app is called Compiss, it’s live on both app stores with an Apple Watch companion.
I want to sit with why this particular project blew up the way it did, because it says something useful about where solo, AI-assisted building actually has an edge right now.
The idea was small enough to finish.
The builder’s own framing was refreshingly unpretentious.
He said the built-in toilet search on stock map apps felt lacking, so he built something closer to a mix between Waze and a toilet finder, with cleanliness ratings, photos, and filters for free access and accessibility features.
That’s it.
No grand platform vision, no ten-feature roadmap.
One specific annoyance, one specific fix.
This matters more than it sounds like it should.
A huge share of vibe-coded projects that never ship die from scope creep, not from any limitation in the tool.
When the target is “make toilet-finding less annoying,” Claude Code has a bounded, well-understood problem to work against:
pull location data,
render a map,
let people rate and photograph things,
ship to two app stores.
None of that requires novel architecture.
It requires someone who can describe what they want clearly and iterate fast, which is exactly the workflow Claude Code is built for.
Data sourcing was the actual hard part, and it was solved cheaply
When people asked where the toilet data was coming from, the builder said cleanliness ratings are pure community input for now, with location data pulled from OpenStreetMap globally and a UK-specific toilet map site for domestic coverage.
That’s a smart move for a free, ad-free app:
no paid API dependency that scales badly with usage, no data licensing headache, and a natural path for the community to fill gaps by long-pressing the map to add a missing toilet themselves.
Someone in the thread suggested scraping Google reviews for bathroom mentions, which is a reasonable idea, but starting with open data and crowdsourced ratings is the right instinct when you’re a solo builder with zero backend budget.
You don’t want your unit economics tied to a metered API on day one.
The launch had real bugs, and the builder handled them in public
What struck me most wasn’t the polish, it was the opposite.
Android users reported immediate crashes on a range of devices, from Pixels to Samsung flagships.
The builder replied to nearly every single crash report individually on Reddit, thanked people, and said a fix was submitted and waiting on Play Store approval.
Someone even suggested wiring in a proper crash reporter like Crashlytics instead of triaging by device model in the comments, which is solid advice for scaling past manual bug-hunting.
There was also a login hiccup on Apple sign-in that resolved itself on a second tap, and reports about the map freezing when zoomed too far out with too many pins loading at once, something that needs a zoom-level check before it queries the toilet database.
None of this is unusual for a first release.
What’s worth noting is that the builder didn’t disappear after the launch post.
He was answering questions and shipping patches in near real time, including standing up a web version within a day of the original post going up.
That responsiveness is doable precisely because the whole stack is small enough for one person, working with an AI coding assistant, to reason about end to end.
There’s no team standup blocking a hotfix.
The name did as much work as the app
I’d be underselling this if I didn’t mention the obvious:
Compiss is a genuinely funny name!
It apparently also works as a pun in Swedish, where “kompis” means friend, and in a few other languages the phonetic overlap runs in less flattering directions.
That’s not something you plan for, it’s a lucky accident of sound, but it’s also worth noting that a memorable, slightly ridiculous name for a genuinely useful niche tool is a real distribution advantage.
People shared this post because the name made them laugh, not because they read a feature list.
For anyone building small utility apps, that’s a reminder that naming isn’t a footnote you handle after the “real work” is done, it’s part of the product.
What this says about vibe coding right now
The skeptical read on vibe coding is that it produces demoware, impressive in a screen recording and broken the moment real users with real devices touch it.
Compiss is a decent test case against that framing, not because it launched bug-free (it didn’t), but because the bugs were the normal kind any indie app hits on day one, and they got fixed in days, not months, by one person.
The app also isn’t a toy.
It solves an actual problem for parents, people with disabilities, and pregnant partners navigating public spaces, several of whom said as much in the thread.
What I take from this for my own build queue is less about toilets and more about the shape of the win:
pick something narrow enough that Claude Code can carry most of the implementation weight, ship even with known rough edges, and stay visibly present to fix what breaks.
The builder here didn’t wait for a polished v2 before posting.
He shipped a v1 with real crashes, ate the criticism in the comments, and used that same thread as his QA pipeline.
That’s a workable playbook for solo builders shipping fast, and it’s a better one than waiting for something that never has bugs, because that app never ships either.
In case we are meeting for the first time, come over here, it’ll be worth the roller coaster of articles that are gonna come up in the next few weeks.