Questions

Agents

• When every team can spin up agents, what actually becomes scarce: the agent, the data it can reach, or the judgment to trust what it returns?
• If an agent can cite its sources, does “trust” become a UI problem more than a model problem?
• What’s the right unit of accountability when an autonomous agent makes a costly call: the model, the operator, or the company that shipped it?

Distribution

• In a world of AI answers, is the homepage dead, and is being cited the new being ranked?
• When assistants mediate discovery, how does a small company even get into the model’s consideration set?
• Does content still compound if humans never read it and only models do?

AI UX

• What does “loading” mean when the wait is the model thinking, and how much reasoning should we show versus hide?
• Should an AI product optimise for answering what you ask (passive) or surfacing the right thing unprompted (active)? Where’s the line before it’s annoying?
• When the model is confidently wrong, whose job is it to make the doubt visible?
• Is the chat box a temporary interface, a CLI moment before the GUI, or the durable one?