Breaking

Doom-Style Game Built Entirely in SQL Showcases GPT-5.6 Sol Ultra's Coding Reach

July 13, 2026 at 22:15 EDT

Foundation Models
Software Dev & Coding
AI Agents

A developer using OpenAI's GPT-5.6 Sol Ultra has built a playable Doom-style game whose entire engine runs as SQL queries, an unusual demonstration that has drawn attention across the developer community. The project, dubbed DOOMQL, pushes a database language far beyond its intended purpose by handling every element of the game inside a SQL engine rather than conventional application code.

July 2026 · GPT-5.6 Sol Ultra

A Playable Doom Engine — Written Entirely in SQL

Built with OpenAI's GPT-5.6 Sol Ultra, "DOOMQL" runs its full game engine as SQL queries on SQLite — raycasting, pixel rendering, enemy AI and combat — bending a data-query language far past its design.

2,000+

Lines of SQL running the entire game engine

Parallel sub-agents used to crack a 50-year math conjecture in under an hour

~88.8–91.9%

Terminal-Bench 2.1 score range for Sol Ultra

What runs where — the engine, drawn to scale

Plumbing

1 layer

Python: keyboard, clock, terminal I/O

Raycasting

RGB pixels

Terminal encode

Move / collide

Enemy AI

Combat

6 layers

SQL on SQLite: the whole engine

Every frame of gameplay is expressed and executed as SQL queries — Python only connects it to the outside world.

The real-time loop, in SQL

Input
keyboard → SQLite

→

SQL logic
AI, collision, combat

→

Render
raycast → pixels → text

↺

Supporters

Cite DOOMQL as raw capability and creativity — arguing Sol Ultra edges past text-to-SQL leaders, plus faster code generation and agentic problem-solving.

Critics

Complain usage limits are tight — quotas exhaust fast on heavy tasks, nudging users toward higher-tier plans. API pricing remains unconfirmed.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.