Thanks for the question! Each paintable pixel in the environment is actually its own object. I'm able to optimize it because it's all on a fixed grid, so lookup is really quick. I don't need to update the pixels outside of when they get colored in, and I only draw the pixels that are actually on screen, so it's all O(1). Paintable pixels on moving objects like platforms and thwomp are a bit more complicated, but follow the same general principal. It also helps that the art style leaves a lot of negative space, meaning there are fewer pixels to keep track of.