A matrix view of co-occurrence similarity between the most structurally relevant cards in the current cached reporting window.
Rows and columns follow the precomputed cluster order; darker cells mean higher pair similarity.
Select a block to filter the matrix and inspect the cards it is built around.