ExitNow

Inspiration

We use the NYC subway all the time, and honestly navigating it is way more annoying than it needs to be. You have to keep checking Maps, figure out whether a delay actually matters for your trip, and text whoever you're meeting every time your ETA changes.

So we built ExitNow to basically make all of that disappear.

What it does

ExitNow is a subway copilot that follows your trip in real time. It tells you where to go, when to get off, and changes your route when something goes wrong.

We also use Photon to automatically message people about your ETA. So if your train gets delayed, you don't have to pull out your phone and explain why you're 15 minutes late. It can just tell them.

But we thought the data from all of these trips could be useful for more than navigation. When you aggregate it, you can start seeing where people consistently have bad routes, make unnecessary transfers, or have terrible connections. That gives us a way to see where the subway itself could actually be improved.

How we built it

We use live MTA data to understand the state of the subway and continuously update the user's route. Grok takes that information and helps decide what is actually worth telling the user instead of spamming them with every little change.

Photon handles messaging, so important updates like ETA changes can automatically be sent to friends without the user doing anything.

We also built an analytics layer that takes anonymized trip data and looks for larger patterns across the system.

Challenges

The biggest challenge was that subway data is messy. The best route can change really quickly, and not every delay is important enough to bother the user about.

We also had to think a lot about privacy. Trip data becomes really interesting when you aggregate it, but we don't want ExitNow to become a giant database of where individual people go. So the system is designed around anonymized aggregate patterns instead.

What we learned

We originally thought the main problem was better subway routing, but that isn't really it. There are already apps that can find a decent route.

The more interesting problem is making the subway something you don't have to constantly think about. And once you have that system running across a lot of people, the same data can tell you a surprising amount about where NYC transit itself isn't working well.

Built With

• ai
• api
• fastapi
• geospatial
• grok
• gtfs
• gtfs-realtime
• llm
• mta
• photon
• python
• react
• real-time
• solana
• transit
• typescript
• websockets