What I’m revisiting in my 2023 essay on AI and games
Field Note · 22 August 2026
I wrote about AI in game development in 2023. Building with these tools has since made me more interested in revision, consistency, and what happens inside the editor.
In March 2023, I wrote about ways AI might fit into game development, from dialogue and generated assets to animation and tools inside the engine. I was excited by how much more a small team might be able to attempt.
I’m still excited by that. But after spending more time building with AI, I pay much more attention to the second change someone asks for. Can I alter one part of a scene without losing the rest? Can I get back to a version I liked? Does the generated asset behave properly once it is in the engine?
Those are familiar development problems with some new failure modes. A character needs to stay recognisable across shots. A texture has to fit the material. Code that looks plausible still needs to run with the rest of the project.
This is part of what I’m exploring with Kinema: keeping the game, editor, and scripting tools together in a browser. I want to make it easier to try a change and see what it does. It is an active experiment, not a finished answer to AI-assisted game development.
My ocean and flight experiments have been useful reminders too. AI helped me build quickly, but I still had to work on controls, performance, loading, and how the experience felt on different devices.
If I were writing the essay today, I would spend less time listing what models could generate and more time on how a developer could keep editing the result.