FTLeats

The problem

You move to a new city and end up at mediocre places the algorithms pushed you to. Fort Lauderdale has a genuinely great food scene, hundreds of real spots locals actually go to, but no single guide organized the way people who live here actually think about the city: by neighborhood and by the kind of night you're having, not by whatever a review site's ad budget surfaces first.

Why I built it

There was no good local guide, so I built one, fast. Instead of months of entering restaurants one by one, I scraped the city's real restaurant data to get genuine coverage, 356 spots and counting, then used AI to organize all of it the way locals actually talk about the city: 17 areas instead of official planning-map boundaries, and every spot tagged by vibe and occasion (date night, rooftop, waterfront, dive bar, solo-friendly) instead of just cuisine.

The part I'm proudest of is the AI concierge. Ask it for something like a cheap Thai place near the beach and it reads through the real restaurant data and answers in plain language, with a reason for each pick. The rule I built in from day one: it can only recommend restaurants that actually exist in FTLeats' own database. Every recommendation it returns gets checked against that real list before it's ever shown to you, so it cannot invent a restaurant that isn't there. I tried to get it to hallucinate a Michelin-starred spot that doesn't exist in Fort Lauderdale; it corrected the premise and recommended real alternatives instead.

The concierge is also honest about being AI. The interface says plainly that it's an AI guide and to double-check details with the spot directly, and it's rate-limited so it can't be hammered into runaway costs by one visitor. None of that is dramatic, but it's the difference between a feature that looks impressive in a demo and one I'm comfortable leaving live for strangers to actually use.

What FTLeats doesn't have yet is my own voice. Right now the write-ups come from the scraped data, not from me having eaten at every one of these places. That's the gap the site is built to close next, not paper over.

What it does

• Coverage across 17 local areas, organized how locals think about the city, not official planning lines.
• An AI concierge you can chat with: describe the night you're planning and it recommends real spots, with a reason for each, grounded only in restaurants that exist in the database.
• Vibe and occasion search: find a place by the night you're planning (date night, rooftop, waterfront, dive bar, solo-friendly), not just cuisine.
• A newsletter of new spots and neighborhood guides.

Where it's going

The data foundation is automated. Next comes the human layer: my own take on the places I know, and comments from locals who eat here too, so the guide stops being just a well-organized database and starts being a place with an actual point of view.

Could this work for you?

The real value here is the pattern, not the topic, and it travels. A few ways the same idea could work for something of yours: