Let's see if I can do some AI Safety work... - View it on GitHub