The Engine and the Escape
I do producing sessions with creative founders. It’s a thing I’ve been developing — you bring the project you’re most stuck on and we sit with it together. Not coaching exactly. More like being in a band. We plug in and see what happens.
Last week I sat down with Zakk. He’s building something I find genuinely interesting — an autonomous agent that reads through the Western canon of philosophy and writes essays about what it finds. Think Frankenstein’s monster, but instead of terrorizing villagers, it’s reading Aristotle and publishing book reports in the voice of a 19th-century letter writer.
Zakk has five projects. Four of them are moving. This one — the one he calls the coolest, the most interesting, the one where the heart of his company lives — hasn’t been touched in over a month.
He told me this with the specific energy of someone confessing. Like admitting to a doctor that you haven’t been taking the medication that’s working.
I asked him what happens when he sits down to work on it.
Not writer’s block, he said. Not confusion. Something more elaborate. A cascade of decisions. Which model should I use? Should I pay for the premium one or run it locally? Do I put it on the Mac mini or a server? Which credit card? Should I scrap the whole thing and start over?
Every question is reasonable. He’s a technical founder. These are real decisions. But he’s been making real decisions on his other projects just fine. It’s only this one — the alive one — where the decisions multiply until the session is over and nothing happened.
I said: that’s also a really good way not to do the thing you’re most excited about.
He laughed, because he already knew.
This is the pattern I keep seeing. The creative engine — the thing that makes these people extraordinary builders — doesn’t distinguish between building toward and building away. It just builds. Point it at the scary project and it produces the work. Point it away and it produces something else. Something easier. Something that doesn’t carry the weight of your actual desire.
We generate alternatives. New tools. New frameworks. New adjacent projects. A whole system for organizing the work that conveniently never gets applied to the work that needs it. The engine doesn’t care what it makes. But we do. And the gap between “I can build anything” and “I can’t build this” — this specific thing, the one I care about most — that gap is where the avoidance lives.
The avoidance is productive. That’s what makes it invisible. A creative person in avoidance mode doesn’t scroll their phone. They ship features on a different project. They learn a new tool. They feel competent and busy and like they’re making progress, because they are. Just not on the thing.
We got the project booted up. Took a while — Docker crashing, models going down, the infrastructure fighting back like the thing itself didn’t want to be easy. Then we were in.
Zakk had been trying to get the agent to speak in the voice of Frankenstein’s monster — this formal, literary, 19th-century prose. It kept reverting to standard AI after a couple messages. Flat. Helpful. Dead.
So we started directing it together. Zakk driving, me tossing in framing. Give it to me again. Too modern. We’re basically at a Ren fair — don’t break character. He’d edit the response, push it back, shape it again. I’d suggest a move — stay in character even when doing admin, rewrite the whole identity in the period voice. He’d run with it, twist it, make it his.
Something shifted. We weren’t configuring a system anymore. We were playing. The voice work was alive in a way the infrastructure work wasn’t. Zakk was laughing. The agent was responding. We’d gone from troubleshooting to jamming, and the project that hadn’t been touched in a month was suddenly the most fun thing either of us had done all week.
This is what the avoidance was protecting him from. Not failure. Not complexity. This. The aliveness of the thing when he actually sits down with it. The intensity of caring about something this much.
Then I asked the question we couldn’t answer.
The agent has a heartbeat — it wakes up every thirty minutes. But it doesn’t do anything unless prompted. It processes what it’s told to process. It doesn’t wonder. It doesn’t get curious about a thread in Aristotle and follow it on its own. How do you inject the curiosity?
Neither of us knew. We sat with it for a minute. I said: we’ll get there.
I think the curiosity question is the whole thing. Not just for Zakk’s agent. For us.
What keeps us from the project isn’t a lack of discipline or a missing plan. It’s that the relationship has gone cold. We stopped being in contact with it. The excitement is still there — we feel it every time someone asks us about it, every time it ambushes us in the shower — but we’ve built so much insulation between us and the thing that the signal can’t get through.
What happened in that session wasn’t that I fixed Zakk’s project. It’s that we made room for the project to express itself. The voice work, the directing, the jamming — that was Zakk and his project reconnecting. The project had things it wanted to say. It had a voice it wanted to speak in. It had directions it wanted to go. The moment we stopped troubleshooting and started playing, the project started leading.
This is what I mean by producing. Not managing the work. Not holding someone accountable. Making room for the project to show up as itself — in all the ways it needs to — so that the person making it can feel the pull again. When a project knows it can express freely, the work flows. The decisions that were paralyzing an hour ago become obvious, because the relationship is alive again and the project is telling you what it needs.
A producer can hold this space. Someone in the room with you whose job is not to have the answers but to keep the channel open between you and the work. To notice when you’re building away and gently point you back toward the thing that’s pulling.
But we can hold this space for ourselves too. It starts with contact. Not with the plan, not with the infrastructure, not with the decisions. With the project. Sit down with the thing. Ask it what it wants. Let it be messy, let it be half-formed, let it speak in whatever voice it has today. The relationship is the work. Everything else follows from there.