I spent a year building a GPS-ephemeral social network. Here's what happened

I was doom-scrolling Instagram when I asked myself: Why am I even doing this?

I'm a graphic designer, not a professional dev, but I decided to build the opposite of the "engagement at all costs" machine. No algorithm. No permanent profiles. No infinite scroll. Just GPS, reciprocity, and 48-hour ephemeral posts.

The Concept

What if posts stayed in the physical location where you took them? I built Bliip: a Vanilla JS PWA radar-based platform where content is anchored to GPS coordinates and expires after 48 hours.

I stopped calling it 'social media' and started thinking of it as a 'localized reality node' - which sounds pretentious, but bear with me. It’s not about following people; it’s about discovering the space you’re currently standing in.

Core mechanics:

Give-to-get: You must post to unlock nearby content for 24h. No lurkers allowed. GPS-locked: Content is tied to space, not users. The Narrator: A sarcastic "System" AI (the gatekeeper) that roasts you during onboarding.

Who is this actually for?

Bliip definitely isn't for everyone. It's for the person at a music festival, a campus, or a neighborhood block party who wants to know what the person three feet away is seeing — not what an algorithm decided they should see six hours ago.

Technical Hurdles

Schrödinger's Photo: Modern phones don't release files immediately; they process HDR/AI in the background. If you grab the file too fast, it's "not there" yet. I had to dump the Blob directly to RAM to bypass the OS lag.

Video Without a Server Farm (FFmpeg.wasm): I couldn't afford a transcoding backend. I used FFmpeg.wasm to move the processing to the client's CPU. Getting it to behave on mobile browsers required a nightmare of SharedArrayBuffer and COOP/COEP header configurations (the true PWA experience), but it works.

AI Brain Surgery: I coded this with AI assistance, which was a hallucination-filled roller coaster. Halfway through, GPT-5 started failing on complex WASM memory logic. I tried switching to Claude, but it insisted it was the year 2025 and at one point literally outputted "KURWA MATEUSZ" (my name + a Polish swear word) in a moment of context-collapse. I finally "transplanted" the project's brain to Gemini to finish the last 10k lines.

Google Play Hell: The new "12 testers for 14 days" rule is trauma for solo devs. If day 13 had 11 testers instead of 12, Google reset the clock. It took two months just to "exist" in the store.

Reality Check

Organic users arrived. They saw the neon/cyberpunk design and thought: "Dating app."

I learned the industry term TTD (Time To Dick). Bliip's score was alarmingly short. Ratings plummeted to 1.0 because users expected Tinder and got a location-based radar with a sarcastic narrator telling them to go outside.

I spent months solving memory leaks and building Redis spatial indexing, only to be rewarded with crotch photos and 1-star reviews. This is the darkest moment for solo devs. Total isolation.

Current State

After 1 year solo: 38,106 lines of code (Vanilla JS, PHP, Redis) €0 marketing budget 15 active users Regrets: 0

Tech stack: Frontend: Vanilla JS PWA (No big frameworks, just sweat) Backend: PHP + MySQL + Redis (Spatial indexing/GeoSearch) Video: Client-side FFmpeg.wasm Infra: €30/month VPS

What I Learned

Users will misunderstand your cathedral and shit in the corner. Those who stay to help clean are your true audience. Don't build for everyone (build for yourself first). Tech stack doesn't matter (PHP is a sleeper hit for Geo). Being a designer who codes is a superpower (and a curse).

Try it: https://play.google.com/store/apps/details?id=me.bliip.twa Web: https://bliip.me