Yeah the OpenAI model isn’t great. We’re working an update that sends all of its reports to a person to review and which will delete them if the person decides the bot made a mistake. 

It’d be great to have a self hosted and trained bot which did a better job. 

This is an experimental bot followed by a couple dozen people and posting events to a single otherwise read only relay. 

We don’t even suggest Nos users follow it. I think we should suggest it to users but there are people on our team who also think it’s not accurate enough yet and we need human intervention. 

I want there to be many bots like this which all make different judgements about content. You can choose to follow none of them, one, or several.