The board needs gems that read instantly at game scale, but I also want them to be real 3D objects that can survive a future move into a more dimensional presentation. That made this a small visual-design exercise disguised as asset production.
The larger move was in the language. The five gem families are now Heat, Cold, Light, Shadow, and Life. They describe states and forces in the game world rather than the traditional substances that produced them. That gives Chromancers a more flexible vocabulary while leaving the actual affinities, combat roles, and interactions open until the prototype earns those decisions.
I built a low-poly source set around those states in Blender, using distinct silhouettes rather than relying only on colour. The outputs are editable meshes, not flat sprites: they can return to Blender, be adjusted, and eventually appear in the 3D stage without replacing the board's visual language. The useful constraint was checking the forms from the front, at an angle, and at the small sizes where board pieces actually live. A gem that looks good as a render but collapses into a generic coloured pebble at 48 pixels is not doing its job.
I also sent Codex into a GPT-6 Astra analysis. It consumes enough tokens that the only honest next step is to wait and judge the result: whether the depth of the analysis earns its cost, and whether it produces a useful decision or only an impressive document. This is still source art and an open AI experiment, not a claim that the board or its systems are finished.
Workbench time: 45m tracked on September 7, 8h 15m total since August 29, 2026. Estimated from Codex activity with a 15-minute inactivity cutoff.