Sign InOpen Brain
AI EngineerVideoSource Linked

The Next Game Engine Won't Have a Manual — Arturo Nunez, Nereu

Nereu argues that coding agents need an engine-native vocabulary, not more generated glue code: tags, reusable systems, and selective scene context turn game intent into reliable actions.

AI Engineer · Aug 18, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Nereu represents game behavior through an **asset tag system** derived from entity-component-system ideas. Its assistant receives scene context, available tools, and tags, while a level-of-detail strategy omits irrelevant objects from large scenes.

Practical Implication

Agent-tool builders should encode domain intent as stable primitives such as player, vehicle, animated, or double jump. That lets an agent compose tested systems instead of rebuilding cameras, movement, and physics from scratch for every request.

Agent-Ready Context
Nereu represents game behavior through an **asset tag system** derived from entity-component-system ideas. Its assistant receives scene context, available tools, and tags, while a level-of-detail strategy omits irrelevant objects from large scenes.

Agent-tool builders should encode domain intent as stable primitives such as player, vehicle, animated, or double jump. That lets an agent compose tested systems instead of rebuilding cameras, movement, and physics from scratch for every request.

The product is still in **closed alpha**, and decomposing genres, mood, lighting, and behavior into reusable tags remains ongoing work. The talk demonstrates an approach, not evidence that it scales to production game complexity.
Context Map
contextcoding#context-engineering#tool-use#coding-agents
Uncertainty
The product is still in **closed alpha**, and decomposing genres, mood, lighting, and behavior into reusable tags remains ongoing work. The talk demonstrates an approach, not evidence that it scales to production game complexity.