About QuestSpot

The product is a generative-AI application, not a game with AI bolted on. Three things a person experiences are model output rather than authored content:

• The clues. Every hunt is assembled at request time from real places near the player and a clue is written for each one. No hunt is authored by hand and no two players receive the same set.
• The sidekick. A conversational companion the player can reason with about a clue. It is built to suggest genuine possibilities and explain why they fit, while never naming the answer, which is a harder constraint than it sounds and is enforced in our own code rather than left to the prompt.
• The collectibles. Solving a spot generates a piece of artwork for that specific place, so a souvenir of a bakery in Fort Wayne exists because somebody found that bakery.
• The sea shanties. Finish a landmarks hunt on foot and the game composes an original instrumental sea shanty for that voyage, generated with Lyria on Vertex AI. It is the one reward you cannot get from a simulated hunt, because the point of it is that you actually went.

A hunt does not have to be solitary, and the multiplayer is not a leaderboard bolted onto a single-player game.

• Race and Crew. Share a six-character code and up to four players get the same five spots. In Race you are competing for the same answers; in Crew you are solving them together, and a spot one player cracks opens for everybody.
• Player-hosted hunts. Experienced players can author a hunt themselves: choose the real places, write the clues, upload a piece of artwork for each spot, and stake a prize pool of their own QuestCoins across a table of finishing places. Up to twenty people can join with the code, and the escrow is built so that currency can only move from the host to the finishers, never be created. Unclaimed prizes return to the host automatically when the hunt closes.
• Everything a host writes is moderated before another player sees it, artwork included, because a hunt authored by a stranger is content distributed to twenty people.

Solving spots earns QuestCoins, the in-game currency, and collectibles in three tiers: Fridge Magnet, Postal Stamp and the rare Trading Card. Coins are spent on extra clues when a spot will not give way, and clues cost points, so taking help is a real trade rather than a free hint button.

Anything earned can be sold to another player on the in-app marketplace, collectibles and sea shanties alike, priced in QuestCoins with a floor that keeps a real prize from being dumped for pocket change. Purchases move the item and the coins in a single atomic transaction, so a listing cannot be sold twice.

Balances, scores and collectibles can only ever be written by our servers. The app cannot write them and the database rules enforce it, which is what makes a player-to-player economy safe to run at all.

A hunt can be as small as your neighbourhood or as large as the planet. The five difficulty levels change how far apart the five spots are, and that single choice changes the game completely:

• Virtual Voyage. Hunt anywhere in the world from your sofa. Pick any city, state or country and solve it from the clues alone.
• City Code Cape. Five spots across your city, on foot.
• State Secret Search. Five spots across your whole state. A road trip, not an afternoon.
• National Nautical Nightmare. Five spots across your country.
• World Wind Wander. Five spots across five continents. Solvable as a puzzle by anyone; completable in person by almost nobody, which is the joke and the challenge.

The harder the radius, the more obscure the places and the more oblique the clues, because a clue that works for a corner bakery is useless for a landmark on another continent.

QuestSpot runs entirely on Google Cloud. The client is Flutter, for iOS and Android from one codebase. Everything the game decides happens server-side:

• Vertex AI (Gemini) generates the clues, powers the sidekick, and moderates every piece of player-written content and player-uploaded art before it reaches another player.
• Gemini 2.5 Flash Image generates the collectible artwork, and Lyria generates the sea shanties. Both are called on Vertex AI, as is everything else here: QuestSpot uses no third-party model providers.
• Cloud Functions and Cloud Run hold all the game logic. A Python agent pipeline on Cloud Run builds themed hunts across several countries in parallel.
• Cloud Firestore and Cloud Storage hold player data and artwork; Google Places supplies the real-world locations.
• Firebase Authentication, App Check, Crashlytics and Analytics handle sign-in, request attestation and observability.

Scores, currency and collectibles can only be written by our servers, never by the app, and the answer to a spot a player has not solved yet is never sent to their device. Those are not incidental details. A game about being somewhere is only worth playing if being somewhere cannot be faked.

Sumanth Mylar

Founder and sole developer

Sumanth holds an MS in Computer Science from Purdue University and is based in Fort Wayne, Indiana. He designed and built every part of QuestSpot: the Flutter client, the Cloud Functions behind it, the agent pipeline on Cloud Run, and the prompt and guardrail work that keeps an AI companion helpful without letting it give the puzzle away.

LinkedIn · support@questspot.co

Development began in February 2026. The app is feature complete and entering closed beta ahead of release on the App Store and Google Play. It is self-funded, with no institutional investment.

For anything at all, including press, partnerships and beta access: support@questspot.co.

Our privacy policy describes exactly what the app collects and what it does with it. The short version is that QuestSpot never stores your location.