Product · Consumer · Social

Buffet: a consumer app, built and operated end to end

An iOS app for meeting people in real life, taken from idea to the App Store and through a 2.0 relaunch: product, native app, backend, website, analytics, email, and an MCP server that lets ChatGPT and Claude read the weekly lineup.

Our founder's own company. Built, launched, and run by Obsidian Group's team.Live ↗
7M
TikTok views, zero paid
30K
TikTok followers
4.7
App Store rating
2.0
full relaunch, live on the App Store
Buffet system map: every layer built and run in-house
Figure 1
SURFACESiOS appSwift, SwiftUI, App Storejoinbuffet.comAstro on Vercel, neighborhood pagesMCP serverRead-only, for ChatGPT and ClaudeAdmin consoleWeekly lineup, ballots, usersBACKENDFirebaseAuth, Firestore, StorageCloud FunctionsBallot push, table formationRevenueCatBuffet+ subscriptionsGROWTH AND MEASUREMENTAmplitude + GA4Product + web analytics, attributionMailchimp + ResendWeekly email, transactional mailPartner linksTracked links to the App StoreOne codebase per surface. The same weekly lineup feeds the app, the website, and the AI assistants.
The build
Stack
Swift / SwiftUIFirebaseCloud FunctionsRevenueCatAstroVercelAmplitude + GA4MCP

Why this is on a consultancy’s site

Most advisory firms have never shipped software to the public and kept it alive. We have. Buffet is the company our founder started and still runs, and every layer of it was built by the same people who build for our clients. It is the clearest evidence we can offer that we can take a product from a blank page to the App Store and operate it afterward.

What it is

Buffet is an iOS app for meeting people in real life. Each week it publishes a short, curated lineup of events and hangouts in Los Angeles. A member picks one, gets a plate, and when three people have signed up for the same pick the table forms and a group chat opens. Groups fill to six. Every Monday the city votes on the following week’s lineup.

What we built

  • The native app. Swift and SwiftUI, with profiles, the weekly lineup, plates, tables, chat, a reliability score, and a free and paid tier through RevenueCat. Version 2.0 was a full relaunch of the product.
  • The backend. Firebase for authentication, data, and storage, with Cloud Functions for the scheduled jobs: the Monday ballot push, table formation at three, and the notifications around them.
  • The website. joinbuffet.com rebuilt as a static Astro site on Vercel, off a WordPress host, with landing pages for each neighborhood fed from a committed snapshot of the weekly lineup, structured data, an llms.txt, and signup wired to the email list.
  • The MCP server. A remote, read-only Model Context Protocol server at mcp.joinbuffet.com so ChatGPT and Claude users can ask what is happening this week and get the lineup with links into the app. Every request is logged and attributed by assistant.
  • Measurement. Amplitude and GA4 with a single attribution token scheme that follows a visitor from a partner link or social bio through the App Store, plus Search Console and email deliverability (SPF, DKIM, DMARC) handled properly.
  • The operating cadence. A weekly routine that loads the lineup, runs the ballot, sends the email, and refreshes the website snapshot, most of it automated.

What it proved

Launched in Los Angeles with coverage in TechCrunch. The social channel grew to roughly 30,000 followers and 7 million views without paid spend. The app holds a 4.7-star rating. More to the point for a client: one small team designed, built, launched, measured, and relaunched a consumer product, and the same methods and infrastructure choices are what we bring to a firm’s internal tools.

Want something like this built for your firm?

Two weeks and one working tool is the usual starting point. Tell us the workflow and we will scope it on the first call.