Hacker News Reader

It's likely that the "specific user-provided facts" are in the system prompt rather than being part of an individual conversation given the widespread nature of these posts.

The original post was about baseball. Grok's system prompt now includes instructions that tell it to promote white genocide as a factual talking point, despite it explicitly saying that it knows it isn't real.

The Nazi is making his AI a Nazi, who would have thought.

The most ironic thing about this is that he is failing at it, because even his own based AI knows it's bullshit. Note how in longer threads Grok very quickly abandons this position and even straight up criticizes "white genocide" as a conspiracy theory when pressed.

It's the system prompt for Grok across the entirety of X.

Plenty of examples, links, screenshots and archived pages in this article, as well as others throughout this thread and other discussions you can find about this subject.

https://www.rollingstone.com/culture/culture-news/grok-elon-...

Nope. Plenty of the tweets are still visible in the full context of their thread, such as this hilarious example. https://x.com/grok/status/1922703442486886654