Swapping Puzzle

Start with a simple 4 × 4 grid and a clean visual setting. Once the agent can reliably restore that board, increase the grid size or add a shuffle mode.

Yes. Try photos with clear objects, strong contrast, and distinctive areas first. Then swap in more detailed or repetitive images to create a harder visual-reasoning task.

Different models and reasoning settings can approach the same board differently. Run the same photo and grid more than once to compare observation, planning, recovery, and move count.

Ask the agent to summarize what it has learned from the board, then retry with a simpler grid, a clearer photo, or a different reasoning setting.