How to engineer an AI harness for game development
A Calgary talk argues game LLMs need harnesses because much of a game is not code.
A game developer blogifies a 45-minute Indie Rodeo talk in Calgary on 12 September 2026, titled Your Game Is A Harness, and says the post is human-written rather than a transcript. The author describes early ChatGPT copy-paste coding for mood.site and later use of Claude Code, arguing LLMs are effective on web stacks such as Svelte because abundant source is available. Games fail more often, the author says, because assets and many systems are not code and are therefore opaque to the model. The excerpt outlines system, assistant, and user prompts but cuts off before the harness design itself.
- Talk titled Your Game Is A Harness was given in Calgary on 12 September 2026.
- Author says web stacks have far more public training code than games.
- Non-code game assets stay opaque to an LLM unless represented as code.
- Provided excerpt stops before the detailed harness-engineering steps.
From the original article
16 min read at kylekukshtel.com
I gave a 45 minute presentation entitled “ Your Game Is A Harness” as part of Indie Rodeo in Calgary on 9/12/26. After speaking with a lot of people after my talk, I thought it would be good to more broadly share the ideas I presented there in a form that could be enjoyed by people beyond just those who attended the conference. Also, because I talk about AI a lot on this blog, people assume I use AI to write.
By kkukshtel · kylekukshtel.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.