Context
This project started with a small annoyance and a bigger question.
The annoyance was Instagram. Every time I saved a food recommendation reel, what I was really trying to remember was the place so I could visit it later. Weeks later, when it was time to decide where to eat, I'd find myself digging through endless saved videos trying to find that one restaurant. I'd end up opening and rewatching reel after reel just to find the name of a place. It was slow, frustrating, and more often than not, I'd give up before finding it.
What I wanted was simple: when I save a reel, I should be able to find the place immediately. After all, my intention was to save the place inside the reel, not the reel itself.
The bigger question was whether I could build that solution myself. After 12 years designing products alongside engineers, I was curious to see how far AI could help me go. Could I take an idea from a problem I experienced, design it, and build a working app end-to-end on my own? FoodSpots became my way of finding out.
How it works
Share
Shared from Instagram; the extension logs it instantly and the app processes it next time it opens.
Extract
Apify extracts metadata from the reel. GPT builds a search query using its priority ladder. Google Places returns a match for that query.
Validate
Four guardrails — is it a food place, is it the right place, not a repeat restaurant, not a repeat reel — clear before anything is written
Save
Saved as a list item and a pin on the map, with everything Google Places returned — and every reel that's ever mentioned it.
Curious about what went into building this app end-to-end? I'd be happy to walk through the challenges, trade-offs, and lessons from building FoodSpots.