Your Game Is a Harness: How to Make LLMs Work on Real Game Development
How to engineer an AI harness for game development
LLMs struggle with game development because games contain much more than code: assets, layouts, and systems that are opaque to models. Drawing on 15 years of game programming experience, this talk argues the problem is one of context engineering, not inherent model weakness. The solution is to build a harness that turns the non-code parts of a game into something the LLM can actually reason about and manipulate.
Said differently, getting LLMs to work well with games is a context problem, not something that is intrinsically linked to how games or LLMs “are” (or “aren’t”!).