Spatial transcriptomics platforms such as Visium read gene expression through an array of capture spots laid over a tissue section, but before capture, RNA molecules diffuse a short distance and bleed into neighboring spots — smearing the sharp boundaries between cell-type niches. This simulator renders an 18×18 spot array in 3D, seeds it with two overlapping cell-type niches across four marker channels, applies the discrete 2-D diffusion equation to model realistic RNA bleed, and then runs an iterative Richardson–Lucy-style deconvolution against the known blur operator to recover each spot's true cell-type composition. Toggle between the ground truth, the diffused (observed) signal, and the deconvolved reconstruction, tune the diffusion rate and exposure time, and watch the reconstruction RMSE drop as more deconvolution iterations run.