>>1843820I actually used an adapted insect behavioural model to predict when I take screenshots
when the likely hood raises above ~%40 for that minute the program starts rapidly taking screenshots of every 100 pixel square area, the instant the likely hood drops again meaning the peak moment was passed a script parses together the latest sequential screenshots into a full resolution picture
then using various publicly available paintings, statues, and buildings as the base of a neural network I can determine the most artistically compelling area for a screenshot
this method usually produces better screenshots than I intended