Of course. He is only a chatting predator or pedofile.

/s btw. I don’t know anything about the situation

These models are mad libs machines. They just decide on the next word based on input and training. As such, there isn’t a solution to stopping hallucinations.

This sounds like the textbook definition of a collider. Meaning that being toxic is the likely “root cause” and that toxic people are more likely to engage in political discourse (because it’s likely going to be toxic anyways? Idk) and they are more likely to comment toxic stuff in general.

I would suggest using Tailscale. It’s an app that runs on your local and remote computers. You log in with your google account, get a special up address that starts with 100.x.x.x. Then you use the special IP address to connect through ssh or mount a volume through samba.

Fair enough. If you want to self host, you can go with forgjo as your web ui and forgjo CI/woodpecker CI for building and deploying the site.

I don’t know of any other self-hostable way to build and push a static site. There was forestry but they discontinued it for a paid service.

You can do all of the through GitHub and GitHub actions by picking a static website generator.