ToddlerMaps — Dima Perkis

Active July 2026

ToddlerMaps

Translating the adult mental model of navigation—GPS, miles, and minutes—into a 4-year-old's language: playgrounds, lunch, buses, and cousins.

Open ToddlerMaps↗ Read the story ↓

Observed: Dean recognized a wrong turn home and predicted the next street. Physically print-tested on thermal hardware. Not yet measured: whether in-car strips reduce "Are we there yet?"

The Road Trip

On a drive to North Carolina, Dean kept asking, “Are we there yet?” He wasn’t being impatient. He knew where we were going; he just had no way to picture the progress.

I had GPS, a moving dot, miles, and an understanding of time. He had none of those. But he did understand landmarks.

So I designed a printable trip strip that translated the drive into his language:

Home → playground → lunch → Peachoid → rest area → cousins

Instead of asking a 4-year-old to track 47 miles or 42 minutes, the strip gives him a sequence of recognizable events. Each trip has today’s destination, landmarks, and stops.

The Larger Idea

ToddlerMaps is the umbrella, not a single product. It shares maps, image generation, schedules, and thermal printing across several related experiments:

This page spotlights Trip Maps. The other experiences deserve their own stories as they develop.

Building an Internal Map

The idea expanded beyond long drives. I narrate ordinary routes in terms Dean can hold onto: “We’re turning onto Lenox Road,” “This is the curvy road,” and “That’s where the school buses sleep.”

It helps him construct an internal map of his world.

There are early signs that may be happening. On one drive, he noticed, “This isn’t the way to our house.” On another, he predicted, “Next is Morningside Place.” Those are early observations rather than measured outcomes. They tell me more than the number of maps the system generated.

What I Learned

Where the Printer Fits

The hardware entered the story late, and that ordering matters. Seeing StickerBox — a roughly $100 kids’ AI sticker printer — triggered the realization that equivalent mono thermal hardware was already sitting on our kitchen counter: the Brother label printer I’d tried and failed to connect months earlier.

Connecting it gave us instant wireless physical output for one specific child and one specific trip. But the printer is infrastructure; the product insight is translating between adult and child mental models.

What Still Needs Testing

I haven’t formally measured whether the maps improve spatial reasoning or reduce “Are we there yet?” questions. I still need to document how many trip strips have actually been printed and used in the car, how Dean interacted with them during a drive, and which landmarks are most useful. For now, the evidence is observed route recognition and prediction—not a causal claim.

Models, agents, tools, and infrastructure

Claude Fable 5
The main July 11 build sessions — trip-strip engine, site, and print pipeline.

Gemini 2.5 Flash Image
Icon and sticker line art, run through OpenRouter.

Python + Pillow
Composes each strip and renders it for the printer.

OSRM · Overpass · Nominatim
Routing, drive times, and the landmark data along each route.

MUTCD road-sign assets
The real highway signs a kid actually sees out the window.

Brother QL-810W
Thermal label printer driving physical output via brother_ql.

frontend-design agent skill
Reusable Claude Code skill that shaped the site's UI pass.