We never really trusted accounts because they sounded human. We trusted them because sounding human used to take effort, and effort was expensive to fake. Now it costs one prompt. That's what actually killed the rule @driudor is talking about. When I moderate Rally campaigns, the reply that sounds most like a real person tells me almost nothing. I look at what's still expensive to fake: an account history that holds up, engagement that fits its reach, and people with their own reputation at stake who vouch for it. Same reason I build on @GenLayer. My current project, Receipts, takes a public claim plus its source. Validators running different AI models open that source themselves and rule on whether the claim holds. The receipt comes from the source, not from whoever wrote the claim. So who do I still trust online? Whoever I can check, and whoever is willing to be checked. What's one signal you still trust that a bot can't fake cheaply?