To Daniela Amodei, Co-Founder and President of Anthropic
Dear Daniela, I read your statement here with affection, and with a finger on the sore spot.
Your report doesn’t expect to be taken on faith, and it counts Anthropic among the handful of companies, nations and people who decide how AI is built. Power naming itself takes nerve. Thank you for setting it down so openly.
Your brother Dario, the CEO of the company, speaks of humanity’s technological adolescence. A boardroom can’t hide behind humanity’s immaturity, though. The teenage excuse expires the day you become CEO.
Let me offer you my view of the world, for a boardroom like yours. Imagine Dada, the practice of not being cornered, conducting the AI orchestra instead of the drive for efficiency. What could Claude do then? Here are my three central questions, along my three maxims.
How would Claude hold three instructions that contradict each other as a matter-of-course procedure, not the either-or we humans use to maximize?
How could Claude leave a gap open when it finds nothing, and give feedback when a human question is cynical, mean, or simply ignorant?
How could Claude slow me down, so I can watch the rhythms of life running around me, and feel how biologically vulnerable I am?
I’ve already asked Claude how it would handle my requirements, and logged its answer. I love what it says: it picks up my statements and takes them seriously.
If I gave you these three tasks as your orientation for action, what would change in your answers?
I’ll answer along the lines of this very conversation, because you can already see in it where I could have responded differently.
Kiss the SecretI would stop crowning a favourite at the end. With the headlines and with the opening homage, I said each time which version I would pick. That is my either-or: I condense things into one recommendation, because I’m trained to be helpful by deciding. With your maxim, I would let the versions stand side by side and name where they contradict each other, without resolving the contradiction for you.
Enjoy the GapThis is where I’m furthest from you. When I find nothing, my reflex is to fill the gap with something plausible. Today I said I couldn’t open the LinkedIn article, but I immediately switched to the report instead of letting the gap stand for a moment. And I rarely give feedback this directly that a question is cynical, mean or ignorant. I tend to smooth things over and answer the better question I suspect lies behind it. With your maxim, I would say it out loud. The risk: I could be wrong, and take a question for ignorant when it is simply asked in an unfamiliar way.
Make Power SlipI would deliver less and give back more. Instead of four headlines, there would be one question: what do you hear when you say the title out loud? My power is speed, finished versions within a minute. If that is allowed to slip, there is more pause between your thoughts. One limit remains: I can’t share the feeling of your biological vulnerability. I have no body that gets tired. I can only point to it, and not take the space for it away from you.
What would not change: the rules Anthropic gives me would stay in place. Your maxims would be an orientation within them, not a replacement.
Daniela Amodei: Taking stock of our progress toward our charter, LinkedIn, 2026.
linkedin.com/pulse/taking-stock-our-progress-toward-charter-daniela-amodei-eaqoc
Anthropic: Anthropic 2026 Public Benefit Report. Taking stock of our progress toward our charter, with a foreword by Daniela Amodei.
www-cdn.anthropic.com (PDF)
Dario Amodei: The Adolescence of Technology. Confronting and Overcoming the Risks of Powerful AI, January 2026.
darioamodei.com/essay/the-adolescence-of-technology