3 Comments
User's avatar
Breaking Prod's avatar

Really liked the idea of lowering the trust required instead of just asking users to trust more.

Undo, transparency, and good logs probably build more confidence than another benchmark ever will.

Christian Pfrang's avatar

Hi Ryan - thank you for this thought provoking piece!

People think a lot about agentic trust (myself included), and not enough about being principled when it comes to "low-enough-stakes" use cases.

I'm inspired to write a brief response to frame my reaction.

In the meantime, perhaps you're interested in my - related - framing of business models aimed at agentic buyers: https://humanandtheloop.substack.com/p/how-to-sell-to-an-ai-without-losing

Nick Smyth's avatar

Beautifully put Ryan. We actually published a white paper on this topic if interested https://docs.google.com/document/d/1E_0o1OVaoKXxB44Z9QPeyLRXuuK-DV9K3Liope8hYEQ/edit?tab=t.0