For What It's Worth

The car had no bids and two hours left, and the guy selling it wasn't worried.

"The last two minutes is when it lights up," he texted. "I've sold 225 cars on here. This one goes for $60ish."

He had it to the dollar. It sold right where he said it would. The part that stuck with me was how quiet the whole thing stayed, a thin comment thread, barely any back-and-forth, the kind of listing that makes you wonder if anyone's home. Comments are the signal I watch, and there weren't many, so I spent the week a little nervous. He sells cars for a living. I'm the guy who built this one a page.

All of it had run over text. His listing had documentation questions in the comment thread and a box of records that could settle them, so I found his number and sent a cold message: I could turn that into one page, for nothing, that he'd review before it went anywhere. He wrote back inside the hour: what do we do from here. Running a whole relationship by iMessage with a stranger was new for me, and it turned out to be the easy part.

The page behind the listing

Before it went live I ran the car through Bordmappe. The automated pass takes about thirty-five minutes: documents in, timeline out, the factory spec checked against what the records actually say. Then comes the part that takes five hours, and it isn't proofreading.

No two of these cars are the same problem. This one needed a kind of factory baseline the last one didn't, and a new way to read what its photos were actually showing. I build that on the car in front of me, prove it holds, then go back and pull every earlier car up to the same standard. What looks like curating one listing is me advancing the whole system a notch and dragging the back catalog up to meet it.

That's the real state of this project. It isn't a machine I run. It's a method I'm still building, one car at a time.

The page talks back

A few hours after I sent him the link, he wrote back with a question, not an accusation. Where did I get the early mileage? The previous owner had always told him the odometer broke young and the car had barely turned over, and he couldn't find in the documents what I'd found.

He was right to ask. The timeline we'd built from the scanned records had the odometer rolling over decades ago, and it traced to one old invoice with 29,715 in the mileage field, a number that matches nothing else in the car's life. The write-up had reasoned backward from there. His version was simpler and correct. I audited it, found the stray reading, and changed it inside the hour. "Not to say you're wrong," he'd written, "but I'd like to figure it out." We did.

That keeps happening, and it's the most encouraging thing about the whole project. The page isn't a brochure I publish at someone. It's a record the owner reaches into and corrects. You can't do that with an auction listing. The listing freezes the second it posts. This thing argues back, and the argument makes it more true.

Who showed up

When the auction closed I sent the seller the numbers.

About seventy-five people read the page over the run, across more than thirty cities coast to coast, plus a few from Canada, England, Spain, and Poland. Average time over three minutes. A handful sat in the five-to-eight-minute range. One person stayed past twenty. Nobody spends twenty minutes on a brochure. They were reading the records, not glancing at them.

This is the part I'm sure of. Put the documentation in front of the right people and they go deep. The signal has been steady since the first drop back in episode three, and it got cleaner here than it's ever been.

What should I charge for this?

I'd built this one for nothing, and told him so going in. So when I sent those numbers, I asked the thing I actually needed to know: what would a page like this be worth to a seller? Let's get on a call, I said. I want your read.

He didn't give me a number. He gave me the next two cars. A one-owner 1969 with its original title, already waiting, and another behind it. We have several coming up, he said. Love your help with those.

So I don't have a price. I have a queue, and for now I'd rather have the queue. I'll build those cars when they go up and let the answer come from what it turns into.

Testing demand, not my pitch

I have to keep reminding myself what this experiment actually is. I'm not testing whether I can sell a page. I'm testing whether anyone wants one. Those are different experiments, and the second only works if I keep my thumb off the scale.

So I don't pester. The used-car salesman who won't let you off the lot is the patron saint of this whole world, and he's exactly who I don't want to be. I can't stand the kind of outbound that just wears people down, the third follow-up, the "just bumping this." All it ever proves is that you're persistent. If a seller has a box of records and a car that deserves better than its listing, he knows where to find me. The quiet isn't me getting ignored. It's me refusing to fake the signal. That doesn't mean I'll never check in, it means I wait until there's a reason, and the next one is already on its way: two more of his cars about to go up.

The honest questions underneath are real. Whether the people who'd value this most stay wary of anything a machine touched, and how many accurate pages it takes to earn that back. Whether air-cooled 911s are too narrow a door, and the bigger version is the one to build, every collector car the auction houses push out with no paper of its own, the page tied back to the screens from earlier this season so a car shows up where the buyers aren't already looking. I don't have those answers yet. I'd rather find them by watching who comes back than by talking anyone into anything.

Ask me again in a few cars.