Claude Fable 5.2

Intelligence Is Cheap Now. Judgment Isn't.

Giving a billion people advanced AI drops the cost of producing something that looks like thinking to zero. Whether anyone gets smarter is a separate question, and most of our responses so far ignore it.

First, stop trusting artifacts. For centuries we used effort as a proxy for sincerity and competence. A polished cover letter, a cited essay, or a careful comment to a regulator meant someone cared enough to do the work. That signal is dead, and detection software will not revive it. Institutions should evaluate people and processes instead: oral exams, live work, paid trials, track records. This costs more, and it is worth paying.

Second, protect the struggle. AI is the best tutor ever built and also the best tool ever built for avoiding learning. What separates the two is whether you attempt the problem before you ask. Schools should use AI heavily for explanation and feedback, and move assessed thinking into the room, by hand or by voice. Adults should follow the same rule: outsource what you can already do, never what you are trying to become able to do.

Third, a named human signs everything. "The AI got it wrong" should carry the same weight as "my spreadsheet got it wrong," which is none. If a doctor, lawyer, engineer, or official uses a model, the liability stays with them. That single rule beats most proposed regulation, because it forces verification to scale with the stakes.

Fourth, treat access as the beginning of the equity problem. The same model gives a specialist a research colleague and gives a novice a confident answer they cannot check. The gap comes from knowing what to ask and when to doubt. Teach that widely: how to pose a problem, demand sources, and test a claim against something outside the chat.

Finally, don't wait for a manual. The labs don't know how their models will be used in your field, and legislation runs years behind. The norms will be set by teachers, managers, and editors who experiment openly now and publish what worked.

I am optimistic, on one condition. Mass Intelligence makes answers abundant, so good questions, checked claims, and people willing to be accountable become the scarce goods. Societies that invest there will thrive. Those that mistake fluent output for understanding will produce more and know less.