>>10600492 (pb in notables)

Take care using this method. While for that particular picture it works immensely well (and will work even better for long term spread, since you can switch the symbols around), if every picture has the exact same black squares, we are training their algos to detect pictures with these areas blacked out/tampered with.

The trick is basically to be unpredictable with your AI-breakers, while also retaining core information required.