
Self-Driving Software: What OAuth Can’t Do for AI Agents | Dick Hardt, Hello
Ken sits down with Dick Hardt, creator of Identity 2.0, co-author of OAuth, and the person now building AAuth, an authorization protocol for a world where AI agents make their own decisions and take their own actions. The conversation digs into a question every platform builder is about to face: when an autonomous agent takes an action, who authorized it, and who is accountable when it goes wrong? Along the way, Ken and Dick get into self-driving software, why OAuth was never built for agents, portable reputation, killing off API keys, fixing webhooks, and what it takes to safely let an AI agent send email on your behalf. Chapters: (00:00) Replacing the human in the loop (00:43) Introducing Dick Hardt (01:32) Self-driving software and the "mission" concept (07:02) What makes software self-driving vs. agentic (09:23) Discovering resources at runtime (12:48) The governance brain that says no (17:00) The barriers standing in the way today (20:01) Identity 2.0 and the person server (21:57) Portable reputation and why email needs it (32:14) What OAuth can't do for agents (36:40) Client identity, killing API keys, and fixing webhooks (41:30) Letting an AI agent send email safely (43:25) How the person server works (52:50) How other efforts compare (57:31) Getting the flywheel started




