Overview
aAIdle is a daily guessing game about AI models. Each challenge gives you clues and a set of possible answers, then asks you to identify the model that matches the hidden answer.
The game is designed to make the fast-moving AI model landscape easier to explore without turning it into a spreadsheet. It combines short daily sessions with a growing catalog of models, historical events, and playful visual clues.
The Idea Behind It
The original idea came from an early conversation with Michał Wojdylak. Instead of treating prompts as the destination, he suggested looking at the wider landscape of artificial intelligence: models, systems, methods, and the people who make them useful. That became the direction for a game about finding the model behind the answer.
Game Modes
The Classic mode is the main model-comparison game. It uses categories such as release date, model family, provider, context window, and capabilities to help players narrow down the answer. Daily challenges are selected deterministically on the server, while the answer remains private until the game is completed.
The Emoji mode represents a model or concept through a sequence of emoji clues. The Timeline mode turns the model catalog into a historical challenge, asking players to place AI releases and events in the right order. Together, the modes make the same catalog useful for deduction, recognition, and learning.
How It Works
The frontend is a Vite-built React single-page application. It owns the interface, local anonymous progress, animations, and accessible game controls.
The backend is a Rust service built with Axum and SQLx, using SQLite for the catalog, daily challenges, guesses, accounts, progress, and aggregate statistics. The API is server-authoritative: it selects challenges, keeps answers secret, validates guesses, and records completion events.
Retries are idempotent and attempts are bounded by the eligible answer pool, so refreshing the page or submitting the same request twice cannot create extra progress. Signed-in players can synchronize verified progress across devices, while anonymous players can keep playing without creating an account.
Building a Small Game Around a Real System
Although the interface is intentionally lightweight, aAIdle includes the parts of a production web application that are easy to overlook in a small game. The project has versioned API contracts, password and OAuth authentication, CSRF protection, account recovery, administrative roles, issue reporting, migrations, seed validation, and browser and backend test suites.
I also wanted the game to remain pleasant to use as it grows. The frontend is installable as a PWA, the model catalog is navigable outside the daily challenge, and the deployment keeps the static frontend separate from the private API container.
Deployment
aAIdle is deployed to a VPS through GitHub Actions. Each release is tested, tagged with a SemVer version, and used to build the Rust API image published to GitHub Container Registry. The Vite frontend is built as static assets and uploaded alongside the runtime configuration.
On the VPS, Caddy serves the static SPA and proxies /api/* to the private API container managed by Docker Compose.
The API and one-shot seed service share a persistent SQLite volume, while versioned frontend releases are switched atomically through a current symlink.
This keeps rollback-friendly static assets separate from the application runtime and makes the deployed release visible through the app version.
Takeaway
aAIdle is an experiment in turning technical curiosity into a daily habit. It is part game, part model catalog, and part exercise in building a secure, server-authoritative application from end to end. The most important constraint is also the simplest one: the game should give you a reason to come back tomorrow.