Suki Neighbors.

Most Philippine condo towers have a buy-and-sell group chat on Viber or Facebook, sometimes three of them per building. Neighbors post what they are cooking that week: adobo, leche flan, meal-prep containers, whatever a home cook or a small carinderia has ready. Buyers reply in the thread to claim a plate. It works, in the sense that money changes hands, but it was never built to be a market.

The word for the relationship at the center of this is suki: a regular customer a seller comes to know and favor, the backbone of how Filipinos have always bought from small vendors. A condo building recreates that relationship at scale, except the group chat has no memory. A listing posted Tuesday morning is buried under two hundred unrelated messages by Tuesday night. Buyers ask "meron pa ba?" (is it still available) into threads the seller has not seen in hours, get seen-zoned, and give up. Sellers cannot tell a serious buyer from someone who will ghost after claiming the last order.

None of the big platforms solve this. Grab and foodpanda take a commission a P55 plate of adobo cannot absorb, and neither one lets a rider past a condo lobby without a resident badging them in. The market is real, every tower has one, but it is a market of maybe forty active sellers and a few hundred regular buyers: too small for a national platform and too disorganized for a chat app to manage.

I built Suki as an installable web app, not an App Store or Play Store listing. A hyperlocal product for one building does not clear the bar of an app-store download: nobody searches an app store for "my condo's food group," and a $99-a-year developer account plus review cycles buys nothing a resident cares about. A link dropped into the same group chat Suki is replacing, tapped once, "Add to Home Screen," gets someone from zero to browsing in under a minute. That distribution math only works as a PWA.

The tenancy model is the real product decision. Most marketplaces scope by GPS radius; I scoped by community instead, chosen once at signup, because the actual market boundary in a condo tower is the building, not a five-hundred-meter circle that pulls in three unrelated towers and none of the trust that makes a suki relationship work. Every community can be public, unlisted, or gated behind a join code, and the person already running the building's Viber group becomes the manager, with co-admin roles, a moderation queue, and a printable QR code that gets people in without typing anything.

Discovery and trust both had to work at the scale of dozens of sellers, not thousands. Ratings live on the seller, not the platform, one review per order, shown only once a seller has at least one. Verification is a manual admin review of a submitted proof, not a self-serve checkbox. Listings auto-expire off the feed twelve hours after a seller last touched them, which forces the same "is it still available" signal the group chat never gave buyers, but as a fact the app enforces instead of a question buried in a thread. Sellers get a limited pool of free boosts each month, not a pay-to-win auction, because paying to outrank your literal neighbor breaks the thing that makes a suki market work in the first place.

Suki never touches money. Every seller sets their own accepted methods, cash on delivery, GCash, or bank transfer, and the buyer pays the seller directly, the same way the group chat already worked, which also keeps Suki outside the marketplace withholding-tax rules that kick in the moment a platform starts holding or remitting funds. Delivery is the seller walking a plate to a unit themselves or a scheduled pickup, not a third-party rider account required to badge into a lobby; fulfillment runs on ASAP, a scheduled window, or pickup, with an automatic apology and stock restore if a seller ghosts an order for an hour.

Suki Neighbors is live at sukineighbors.com, currently on v0.59.1, installable as a PWA on iOS and Android. The codebase carries 65 applied migrations, 600-plus automated unit tests, and around 112 page routes, including a full local-SEO layer of guides, comparison pages, and a PH island-to-city directory.

The live site currently runs as a single public Demo Community: a real, fully working marketplace seeded with sample sellers and listings so anyone can try browsing, ordering, and the seller dashboard before a real building signs on. That is a deliberate sequencing choice, not a stall. The original condo and village communities are archived and reversible while pioneer building managers get recruited one tower at a time.

Every core mechanic, the 12-hour freshness clock, the payment handoff, the oversell protection, the strike ladder, has been exercised against the live database with forced timing and concurrency, not just in unit tests.

1. Scope the market to where trust actually lives. One building beats one city when the product depends on neighbors recognizing each other.
2. A once-a-day cron can still deliver second-accurate behavior if the check runs at read time and the cron only cleans up after.
3. Cloud-provider defaults are tuned for a fast demo, not for real user data. Audit privilege grants and RLS before the first real row goes in, not after.
4. Write a race condition's fix as an explicit, testable guarantee, warn before cancel, never sell twice, instead of trusting it to fall out of the code naturally.
5. A PWA kills app-store friction for anything hyperlocal. Nobody searches an app store for their own building.
6. Taking payment off the platform removes a whole category of compliance and trust problems before they exist.

Suki is a full two-sided marketplace with real-time inventory, seller moderation, a peer-to-peer payment handoff, and its own SEO growth engine, built, secured, and iterated solo in under two weeks and counting. It is what I bring to a team building a marketplace, funnel, or ops platform: I will find the race condition and the leaky default before your users do.