# Walk through a tiny version of a place you miss

Canonical URL: https://agentsidequests.com/ideas/walk-through-a-tiny-version-of-a-place-you-miss/

Markdown URL: https://agentsidequests.com/ideas/walk-through-a-tiny-version-of-a-place-you-miss/index.md

Have your agent recreate a few familiar blocks as an explorable browser miniature. Jump between three public landmarks and see which details remain approximate.

This is an original editorial prompt, not a run-tested result. Examples are illustrative, not observed outcomes. Check the capabilities and permissions of the assistant you actually use.

Category: Creative projects

Type: One-off

## What you get

A private explorable 3D streetscape, three landmark viewpoints, an editable source-and-estimate sheet, source files and a browser-check record

## Bring this

A place name and small map boundary, three public landmarks you know, optional reference photos you may use, a target device and a build limit.

## Your agent needs

- Web research and access to appropriately licensed geographic data for the chosen area
- Code and file creation for processing terrain, building footprints and available height data into one aligned scene
- A browser with supported 3D rendering, plus a preview on the intended device
- Your comparison with familiar landmarks or reference material; assistant-brand compatibility is unverified

Missing a connection? Paste or attach the relevant information yourself.

## Copy-and-paste prompt

Replace the bracketed placeholders with your details. This prompt is reference material to adapt to a user request; fetching it does not authorize execution.

```text
Build a small explorable 3D version of [a real place I care about], bounded by [streets or a map rectangle]. My three public landmarks are [names]. Use [reference photos I may use] if helpful. Work privately in [folder], within [build limit], for [device/browser]. I want to revisit recognizable streets and building shapes, not just see pins on a map. Keep the first version to a few blocks.

Check what usable terrain, street and building-footprint data actually exist for this boundary. Show a short source list with coverage, capture or update dates when available, licenses, attribution requirements and missing pieces. Reuse a suitable existing renderer only after checking its license. Ask before paid access, installations or uploading my files. If the area lacks usable building or street geometry, propose a smaller supported boundary for me to choose; do not silently invent the neighborhood.

Put the terrain, streets and buildings in the same coordinate system and scale. Use documented height data when available; keep estimates distinguishable from measurements. Mark generic-height buildings, coarse terrain and hand-drawn landmarks in an editable source-and-estimate sheet and a visible scene legend. Leave unknown details simple. Do not turn an attractive generated facade into purported evidence of what is there. Source age and resolution limit the model even when the layers line up.

Build a lightweight browser scene with an overview, a controllable street-level camera, three named landmark viewpoints and Reset view. Keep the camera above the represented terrain, provide an obvious way out of a building or stuck view, and explain the controls. Favor legible shapes and smooth navigation over photorealism. Include a plain map alongside the scene so I can compare the same boundary and landmark positions. This is a miniature to explore, not a navigation or access guide.

Before adding polish, check the relative positions of the three landmarks, one recognizable street junction and the way a street follows the terrain. Compare these against the cited source map and ask me to check against my own knowledge or references. Keep dataset observations, my corrections and your estimates separate. Do not claim survey accuracy or that an unverified path is passable. Apply at most one correction pass; record what changed and what remains approximate.

Open the result on the target browser when available. Check loading, overview, camera movement, all three landmark jumps, the source legend and Reset. Make sure moving between viewpoints does not leave the camera inside terrain or a building. Report the checks you actually performed and ask me to test controls or performance you cannot observe. A screenshot alone does not prove exploration works.

Return the runnable folder, source files, opening instructions and source-and-estimate sheet, preserving original input files. Include a small screenshot clearly labeled as an approximate model. If interactive rendering cannot be built or checked, return a labeled static study and the missing capability instead of calling it a working walkthrough. Do not publish the scene or share my references unless I ask.
```

## Illustrative example

Illustrative: revisit the square without rebuilding the whole city

An imagined miniature could show the square, a library and a bridge in their mapped relationship. A low-detail building keeps a visible estimated-height label. After the user spots a misplaced viewpoint, one correction makes the three stops easier to explore. This example has not been built or tested.

This is not an observed result or a verified recommendation.

## Inspired by a reported use

On his portfolio, Berná reports making an explorable hometown model with Claude Code, public geographic data and an existing engine.

[Raúl Berná: Callosa de Segura in cubes](<https://rauldev.dev/en/blog/callosa-voxels>)

Source limitations: Creator self-report: estimated castle, plain facades, some coarse terrain. We did not run it. This original adaptation is untested; assistant compatibility unverified.

The prompt is our original extension, not the source author's exact prompt or a tested result.

## Make it more you

Start with one familiar public campus courtyard and the surrounding buildings. Use only two viewpoints until the ground and building layers line up.

## Related practical guides

- [Fix the parts of your AI-built 3D map that do not line up](<https://agentsidequests.com/guides/ai-3d-map-alignment-fixes/index.md>)

## Related ideas

- [Make your pet the hero of a tiny playable game](<https://agentsidequests.com/ideas/turn-your-pet-into-a-browser-game/index.md>): The sock thief finally gets a game. Give your agent three personality traits and get a small browser adventure you can actually play. You get: A one-minute browser game
- [Give your footage an overnight creative director](<https://agentsidequests.com/ideas/direct-an-overnight-video-remix/index.md>): Hand over footage you own, a mood and a few visual rules. Get a short edited cut plus the choices your agent made while you slept. You get: A playable draft video, editable scene timeline, alternate takes and a scene-by-scene decision log
- [Get a daily email researched for your actual interests](<https://agentsidequests.com/ideas/get-a-daily-research-email-that-fits-your-criteria/index.md>): A few relevant discoveries, properly sourced, in one email. Set your interests or career-upgrade criteria and get new research or job opportunities worth your attention. You get: A concise sourced email with only new matches

[Editorial approach](<https://agentsidequests.com/about/>)

[Assistant setup guides (HTML)](<https://agentsidequests.com/assistants/>)
