The idea started on a napkin in 2017, at a State College, Pennsylvania bar. I was an econ student watching a Friday night cycle through three different bars before everyone got bored of guessing where the good crowd actually was, and ended up at the slowest place by default. The decision-making was random. The information was bad. The night was worse than it should have been.
On the napkin: a framework I called the Equation to a Good Time. The idea was simple — a good night out isn't random. It's the product of variables you could, in theory, measure: who you're with, where you go, what the room is doing at the moment you walk in. Most people optimize the first two and gamble on the third. The third is the one with the most variance. The third is where everyone wastes their Saturday.
The third variable was the one nobody owned.
Yelp owned reviews. Google owned addresses. OpenTable owned reservations. Resy owned the same. Nobody owned "is this room actually alive right now." Not because it was hard — because it was a different shape of problem. Reviews are a static artifact. Energy is a real-time stream. The two require different infrastructure, different incentives, and different relationships with the venues themselves.
I sat on the napkin for nine years. Built three other companies. Watched the consumer-mobile platform mature to the point where you could realistically run real-time geofence presence at city scale without burning a phone's battery. Watched live-source providers (Google, BestTime, Foursquare) get good enough that you could fuse them into a meaningful signal. Watched the social graph fragment into Instagram-loose and group-chat-tight, with no app in the middle for the people you actually go out with on a Saturday.
The patient bet isn't that something can be built. It's that the market is ready when you finally ship it.
San Diego in 2026 is the right place at the right time.
Three years ago, this product would have been a year too early. The data sources weren't quite there. The user base wasn't yet conditioned to expect real-time everything. Now? Apple's Live Activities work. Geofencing is cheap, reliable, and battery-friendly. Every 21-35-year-old in a city expects their food, their rides, and their music to be live — and is reflexively annoyed when their nightlife information is two years old.
San Diego is the right launch market because it's the city I live in, the city I go out in, and the city where the signal is densest. 19 neighborhoods, a tight nightlife corridor between Pacific Beach and Gaslamp, a 21-35 demographic that's both early-adopter-friendly and big enough to make the dataset robust on day one. Get one city right. Then the next.
What I want this to be.
Long-term, Jellyfish is an occupancy data company that ships a great consumer app. The map is the front-end. The dataset is the moat. The opportunity isn't to be Yelp 2.0 — it's to be the live layer underneath every product that needs to know what's happening in physical spaces right now. Marketing, hospitality ops, urban planning, transit, advertising, retail — all of it benefits from a real-time crowd dataset that today doesn't exist.
The consumer app exists to bootstrap the dataset. The dataset exists to fund the long-term bet. The long-term bet is that live data about physical spaces is one of the last big unsolved data problems on the internet, and whoever measures it carefully and patiently will own the layer underneath the next decade of social, hospitality, and urban tech.
It's a small idea. We think it's revolutionary.
Nine years from napkin to ship. The patient bet. We're shipping now.
Sources + further reading.
- Jellyfish glossary — Live occupancy — the technical definition of the signal at the centerpoint of every feature.
- About Jellyfish — the longer-form founder story with team context.