Stephen asked the question and waited.
Ten seconds. Maybe more. On a Teams call, that kind of silence feels dangerous. Like you’ve lost the room. Or your Wi-Fi.
A few weeks earlier, I’d told him he was moving too quickly—not through the material, but through the pauses. He’d ask something thoughtful, then rescue us from the quiet before anyone had time to think. It’s a reflex most of us have.
So he started counting to ten.
This week, he went past ten. He was about to give up. He even messaged me privately: ah well.
What followed was the real session.
The first half had been practical and energetic. Kristen shared a small but useful discovery about using Claude Projects to preserve context instead of re-uploading files each time. Patty ran into 403 errors pushing to Alma (my fault—read-only API keys while we practice). Stephen walked us through RAG.
He demoed a reference agent he built for the Graduate Center—pulling from LibGuides and the Primo API. In his tests, it outperforms the current chat service. He hasn’t deployed it. He’s thinking carefully about what it would mean for an AI agent to be someone’s first interaction with the library.
Then he asked: has any of this changed your mind?
And he waited.
Will spoke first.
He’s in library school and new to these tools. He said he could see how efficiently AI handles reference questions. But he kept wondering what that means for the person on the other end. If learning becomes frictionless, does curiosity grow—or shrink?
“What actually is the point?” he asked.
Then, almost in the same breath: “It’s also really cool to learn how this works.”
That pairing stuck with me. Concern and delight, side by side. Not contradiction—honesty.
The rest of the room met him there.
Robin worried about judgment: these tools can generate answers without the slow work of getting wise. Jason reframed it as information literacy. Anthony said bluntly that students are already using AI-generated citations they can’t evaluate, and ignoring that won’t make it go away. Ashley raised the political stakes. Shamiana worried about inequitable access.
No one rushed to resolve it.
We spent the first half of the session talking about how AI retrieves information. We spent the second half doing something it can’t: sitting with an unsettled question.
Stephen counted to ten.
Will said something real.
And the room didn’t flinch from the silence.
We’re six weeks into a 16-week cohort. Post 1 is here. More to come.