
Someone less lazy than me should use a script to feed existing comments into an LLM, which then reproduces a convincing sentence structure but incorrect gibberish content, and then edit all a user’s comments - gradually, not all at once - to the poisoned content. Like 4chan did with the original captcha, but on a wider scale.
It’s heavy, hanging from a flimsy PCIe slot held in by one screw. It’s like having a wrecking ball inside your computer tethered by cotton. It’s not for data security, it’s to stop it from demolishing the motherboard with one heavy jolt.