Built This Week/Builds

Builds

Trick-or-Treat Route Planner

A one-hour trick-or-treat route planner took us minutes in Google AI Studio. For Nashville it picked Sylvan Park, Sam's old neighborhood, which he called pretty much spot on.

On our Halloween episode, Episode 19 (October 31, 2025), we live-built a trick-or-treat route planner in Google AI Studio. Pick any US city, an age, and how much you care about candy quality and spookiness, and it recommends a neighborhood, a one-hour walking route on a map, the reasoning, and safety tips. Sam came up with the idea on the spot and Jordan built it in minutes, starting at 08:22. It was the third Halloween tool of the episode.

Why we built it

Jordan had just built a costume-guessing app live. Sam wanted to test AI Studio on something "slightly more practical": route optimization, the same kind of problem a business faces with deliveries, here aimed at covering as many houses as possible in an hour.

The stack

  • Google AI Studio (Build): generated the app from one prompt.
  • Gemini: built in, picks the neighborhood and writes the reasoning.
  • Google Maps: the route map. Jordan noted that building on Google's stack means Maps is right there.

How we built it, step by step

  1. Write one prompt. Jordan asked for a tool that optimizes a one-hour trick-or-treat route in any US city by number of houses covered, letting users pick their city, age and other parameters.
  2. Note the suggested features. While building, AI Studio proposed real-time weather, saved profiles with past routes, and predicting which houses give the best candy.
  3. Run it one-shot. Beverly Hills, age 8, candy quality 5/10, spookiness 5/10.
  4. Iterate and rerun. A few quick changes, candy quality bumped to 6, then Nashville.

How it turned out

For Beverly Hills, it recommended the Flats, starting on Crescent, with a turn-by-turn one-hour route. Its reasoning: dense single-family homes, wide, well-lit, flat streets, and residents who go big on Halloween.

For Nashville, it picked Sylvan Park. Sam used to live there, on Dakota, a street on the route, and it's close to where he takes his kids every year. Jordan's summary of the result: "pretty much spot on."

So these are great throwaway ideas, super fun, and, yeah, I hope the Gemini team likes it too.

— Jordan Metzner, Episode 19

Why AI Studio over Bolt or Lovable for this? Jordan called it a step-function improvement because Gemini is already integrated. In other tools, adding AI means building a back end, choosing an API, getting a key and wiring it in.

What we'd do differently

It's a throwaway, so we didn't take it further. Ideas from the session: weather-aware routing, a current-location button, and Sam's favorite, a notes feature to log "who gave out the full Snickers bars" for next year.

Earlier in the same episode: Mirror, Mirror on the Wall and the Guess My Costume app. The week before, we made the trick-or-treat AI costume builder. More on the tool: Google AI Studio vs Lovable.

FAQ

How does an AI trick-or-treat route planner work?

In ours you pick a US city, the child's age, how much candy quality and spookiness matter (each out of 10), and it returns a recommended neighborhood, a one-hour walking route on a map, the reasoning behind the pick, and safety tips.

Can AI find the best trick-or-treat neighborhood?

In our test it did well. For Beverly Hills it picked the Flats starting on Crescent, and for Nashville it picked Sylvan Park, which Sam, a former resident, called pretty much spot on. Always trick-or-treat with adult supervision; the app itself says routes are suggestions.

How do you build a route optimization app in Google AI Studio?

Jordan wrote one prompt asking for a route tool that optimizes a one-hour trick-or-treat route by houses covered in any US city, with user-chosen parameters. AI Studio built it with Gemini and Google Maps already integrated, and suggested extra features while it built.