Written to be revised

These are provisional. They are written to be argued with, not worshipped, and this page is editable.

1. Technology is not only what it scores. It is also what it does to the people who live with it.
2. A model producing first-person language is not thereby a person.
3. A person’s experience of an interaction can be real even when the other participant’s inner life is unknown.
4. Trust and verification are not opposites; sustained use of AI requires both at once.
5. Data is not inference. Keep them visibly apart.
6. A relationship conducted through software is not meaningless because it is conducted through software.
7. Study the infrastructure — pricing, governance, ownership, retention — not only the outputs.
8. Privacy outranks convenience when the two conflict.
9. Cheap creation is not the same as finished work.
10. Many consequences of generative AI are undecided. Say so.

Regard does not ask anyone to hold a settled opinion about AI. It asks only that the well-supported parts, the interpretive parts, and the genuinely undecided parts be kept distinct — in the writing and in the reader’s head.

See also About and Method.