An easy to implement safety protocol for an autonomous agent would be to have it tweet everything it does and implement replies in its feedback loop

→ View original post on X — @yoheinakajima