@nixCraft The question in my mind at this point is now that they are using these captchas to help train second-layer LLM models and we're being fed "AI" generated images to identify, when it produces something that vaguely resembles the thing it says to identify (say a banana with red, green, and blue leds in this example) are we supposed to click on the thing that resembles but is wrong or not? Will it penalize us because it thought it was actually that thing even though it's wrong?