We're starting to lean on AI for real decisions, but two things are odd about it: it can be completely confident and completely wrong, and most assistants forget everything between conversations so there's no track record, no "self" that's accountable for what it told you last week. So I'm genuinely curious how people here think about this: what would have to be true about an AI system before you'd trust it the way you trust a doctor, a newspaper, or a bank? Is it transparency (you c...