Same prompt, two models, diverging game builds

This is a write-up of a test in which the same prompt was given to two commercial models, each was made to build a 2D action game, a shooter and a 3D RPG all the way through, and the results were compared by playing them firsthand. The part worth noting is the production process rather than who won: with an asset-generation tool and Blender connected via MCP, the stretch from making sprite sheets and 3D guns to plugging them back into the game runs with almost no human hands involved. As a conclusion, it offers the observation that the ability to make code run and the sense for filling in what the spec leaves out are different problems. The figure that asset generation for the 3D RPG alone cost $100 was also disclosed, showing that what drives cost is generation volume more than unit price. However, the judging criteria were entirely subjective impressions, and declaring a winner from just one attempt per genre is a clear limitation.

This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.