GPT-4o can now generate pretty convincing looking screenshots. But it's also remarkably good at distinguishing AI-generated screenshots by looking at subtle UI clues. A few thoughts around this: 1. We're probably not far off from a world where AI is the ONLY way to distinguish real vs fake 2. This is probably also really useful data for training these models further. 3. Screenshots / images are not going to be a reliable source of proof anymore (there are massive industries that rely on images as proof) 👀