Luke Harris had spent much of the conversation on editing. The tedium of it, the hours the first one took him in Audacity, the discomfort of not liking his own recorded voice. So when his host asked what the most challenging part of recording with another person was, the answer seemed already given. Harris named something else.

“For each person that I recorded, I planned out what I wanted to talk about. And there was a bit of tension there — do I stay present with the other person and what they’re talking about, or do I lead them into a conversation that I think is cool? […] I wanted to go according to plan, and they were saying cool stuff — but it was changing my plans.”

The obvious response is to tell him the plan should lose. Stay present, follow the guest, the good stuff is in the room and not on your sheet. Decent advice, and not what his own experience supports. The detail that gives it away comes a couple of minutes later.

For his second interview he over-prepared. He had read so much of his guest’s work that he thought he knew it, and got a fact wrong anyway, about which of the man’s books was the most recent. The preparation did not protect him from the error it was supposed to prevent. And the best thing in the episode was something no amount of it could have produced: the author reading aloud a poem he had translated from the Hebrew Bible, in his own pace and pronunciation, with a rhythm the page does not carry. Harris says it was better than he could ever have planned.

So the plan failed twice over. It did not make him accurate, and it did not contain the best moment. On the tally it looks like pure cost.

Except that Craig Constantine, who hosts the show, says something odd when Harris raises the tension — not that the tension should be resolved, but that he depends on it. The magic, he says, is in having the tension; without feeling it he does not think he would do as good a job. He describes a long walk down Broadway with a destination at the end, a guest who turns off toward pancakes, and the pleasure of ending up somewhere he had not intended.

Notice what that requires. You cannot be pulled off course without a course. The reason Harris could tell that his guest was saying something better than the plan is that he had a plan to measure it against — he knew what he had come for, so he could recognize the moment he was being offered something else. A host who walks in with nothing has nothing to be diverted from. Every direction is the same direction, and the remarkable thing goes by unremarked.

Which makes the plan an instrument rather than a script. Its job is not to be executed. Its job is to be the thing you notice yourself leaving — and that job is done the moment you feel the pull, long before you would have reached any of it. The preparation has already paid for itself by then. Whether you follow it afterwards is a separate question, and a smaller one.

The reframe leaves the hard part unsettled, though, and the conversation does not pretend otherwise. Nobody in it says how you tell a good departure from a bad one. Constantine follows his guest to pancakes and calls it a great stroll; he also names a destination and clearly still intends to arrive. Harris’s best moment came from following; his slip came from the same preparation, and no rule offered here would have caught it. The tension is not a problem the episode solves. It is closer to a condition of the work — the thing you feel when two legitimate goods are pulling at once, and neither one is wrong to want.

What changes is the ledger. If you have been grading yourself on how much of the plan you got through, you have been scoring the wrong column — and scoring it backwards. Read that way, the episodes that ran furthest off plan are the ones where you were offered something and took it. The ones that went exactly to schedule are either the ones where nothing better came up, or the ones where it did and you kept walking.

This field note references the Podtalk episode “Tension with Luke Harris,” published March 21, 2022.

This work was produced using AI language models directed through an editorial system designed by Craig Constantine. The author selected all source material, designed the creative framework, directed the editorial process, and made all acceptance and revision decisions. The prose was generated by AI under sustained human editorial direction.