Also: how malicious AI swarms can threaten democracy, scalably solving assistance games, red teaming framework by dynamically hacking reasoning, unsupervised elicitation