- Published on
DOOMQL
- Authors
- Name
- DeepHorus
DOOMQL, a first-person shooter game built entirely using SQLite as the game engine, has been released by developer Peter Gostev. The project, created using OpenAI's GPT-5.6 Sol model, demonstrates an unconventional approach where SQL queries control movement, collision detection, enemies, and even pixel rendering. This experiment pushes the boundaries of what a database can do, turning SQLite into a full-fledged game engine.
Gostev posed the deliberate question: what if SQLite were the game engine rather than merely a storage system? The result is a small but complete Doom-like game implemented as a Python terminal script. It uses recursive common table expressions (CTEs) to perform ray tracing within SQLite, a technique typically reserved for graphics engines. The game was built using GPT-5.6 Sol, which generated the core logic and SQL queries.
The game is available on GitHub and can be run by cloning the repository and executing a Python script. When launched, it creates an SQLite database at /tmp/doomql/.doomql/doomql.sqlite, which stores every aspect of the game state. Gostev also integrated the game with Datasette, a tool for exploring SQLite databases, using the Datasette Apps plugin to create a real-time visualizer. The visualizer, built with Claude Fable (Fable 5), displays the screen as a grid of pixels and updates once per second, and even includes a minimap.
This project showcases the remarkable flexibility of SQLite and the capability of modern AI models to generate complex, functional code. By offloading rendering and game logic to SQL queries, Gostev demonstrates that databases can serve as the backbone for interactive applications. The integration with Datasette also highlights how traditional database tools can be repurposed for game monitoring and visualization.
The tech community has reacted with enthusiasm, noting the novelty of a database-driven game. The blog post describes it as "a lot of fun," and the ability to play a game while observing its internal state in a separate app is both educational and entertaining. This work may inspire further experiments at the intersection of databases and real-time systems.
As AI models continue to improve, such unconventional applications could become more common. Gostev's project encourages developers to think beyond standard use cases for relational databases. Future iterations might include multiplayer support or more complex graphics, all driven by SQL. For now, DOOMQL stands as a proof of concept that a database can indeed be a game engine.
