Godshatter Versus Legibility: A Fundamentally Different Approach To AI Alignment — LessWrong