What might work better than LLMs for AI alignment?

A post argues that large language models are constitutionally ill-suited to alignment and asks what might work better. The same author later called it “one of the most important technical questions of our era.”