One-shotting a Raccoon Heist game using Claude Fable 5

Simon Willison's Raccoon Heist experiment shows how an agent can produce a working game while leaving its central design problem unresolved. Willison gave Anthropic's Claude Fable 5 a game concept and asked it to work independently. His account describes the agent building a browser game, checking desktop and mobile screenshots, and fixing a mobile rendering bug.

Willison found the result technically impressive but dull to play: players could collect everything before a level ended, leaving them waiting, and the promised raccoon team had no cooperative mechanics. In this experiment, automated implementation and testing produced a usable starting point; pacing and gameplay still needed design work.

“Giving Claude access to an OpenAI key turns out to work really well for filling in gaps in its capabilities - in this case we needed some way to generate images to use as textures.”

“As a finished game project, it's mediocre. As a starting point from a single prompt I think it's very impressive.”

“They've all been deeply disappointing from a gameplay perspective - it turns out designing games that are fun remains a uniquely human trait, and one which requires significantly more skill and experience than either Claude or I can bring to bear.”

“That said, I thoroughly recommend tinkering with game development projects as a way to explore the capabilities of agents.”