Comment by Daviey
Earlier today I found myself thinking about the opposite of CAPTCHA. Instead of proving something isn't a bot, how do you create a non-repudiable mechanism that proves something is a bot? We’ve mostly solved the "human verification" side, but this direction feels much harder.
Long computing embedded in confusing texts should be sufficient.