Mr Penumbra's 24 Hour Bookstore

Kat has convinced me that we’ll never be able to catch Penumbra at Penn Station. There’s too much surface area, she says—too many ways Penumbra can get off the train and up to the street. She has math to prove this. There’s an 11 percent chance we’ll spot him, and if we fail, he’ll be lost for good. What we need instead is a bottleneck.

The best bottleneck, of course, would be the library itself. But where does the Unbroken Spine make its home? Tyndall doesn’t know. Lapin doesn’t know. Nobody knows.

Intensive googling reveals no website and no address for the Festina Lente Company. There are no mentions in newspapers, magazines, or classified ads going back a century. These guys don’t just fly under the radar; they’re subterranean.

But it has to be a real place, right?—a place with a front door. Is it marked? I’m thinking about the bookstore. On the front windows, there’s Penumbra’s name, and there’s that symbol, the same one that’s on the logbook and ledger. Two hands, open like a book. I have a picture of it on my phone.

“Good idea,” Kat says. “If a building has that symbol anywhere—on a window, carved into stone—we can find it.”

“What, by conducting a complete sidewalk survey of Manhattan? That would take, like, five years.”

“Twenty-three, actually,” Kat says. “If we did it the old-fashioned way.”

She pulls her laptop across the sheets and shakes it to life. “But guess what we have in Google Street View? Pictures of every building in Manhattan.”

“So subtract the walking time, and now it’ll only take us—thirteen years?”

“You’ve got to start thinking differently,” Kat clucks, shaking her head. “This is one of the things you learn at Google. Stuff that used to be hard … just isn’t hard anymore.”

I still don’t understand how computers can help us with this particular species of problem.

“Well, what about hu-mans-and-com-pu-ters,” Kat says, her voice pitched like a cartoon robot, “work-ing-to-ge-ther?” Her fingers fly across the keyboard and there are commands I recognize: King Hadoop’s army is on the march again. She switches her voice back to normal: “We can use Hadoop to read pages in a book, right? So we can use it to read signs on buildings, too.”

Of course.

“But it will make mistakes,” she says. “Hadoop will probably get us from a hundred thousand buildings down to, like, five thousand.”

“So we’re down to five days instead of five years.”

“Wrong!” Kat says. “Because guess what—we have ten thousand friends. It’s called”—she clicks a tab triumphantly and fat yellow letters appear on the screen—“Mechanical Turk. Instead of sending jobs to computers, like Hadoop, it sends jobs to real people. Lots of them. Mostly Estonians.”

She commands King Hadoop and ten thousand Estonian footmen. She is unstoppable.

“What do I keep telling you?” Kat says. “We have these new capabilities now—nobody gets it.” She shakes her head and says it again: “Nobody gets it.”

Now I make my voice into a cartoon robot, too: “The-Sin-gu-la-ri-ty-is-near!”

Kat laughs and moves symbols around on her screen. A big red number in the corner tells us that 30,347 workers are waiting to do our bidding.

“Hu-man-girl-very-beau-ti-ful!” I tickle Kat’s ribs and make her click the wrong box; she shoves me with her elbow and keeps working. While I watch, she queues up thousands of photos of Manhattan addresses. There are brownstones, skyscrapers, parking structures, public schools, storefronts—all captured by the Google Street View trucks, all flagged by a computer as maybe, possibly, containing a book made from two hands, although in most cases (actually, in all but one) it’s just something that the computer has mistaken for the Unbroken Spine’s symbol: two hands in prayer, an ornate Gothic letter, a cartoon drawing of a twisty brown pretzel.

Then she sends the images off to Mechanical Turk—a whole army of eager souls lined up at laptops around the world—along with my reference photo and a simple question: Do these match? Yes or no?

On her screen, a little yellow timer says the task will take twenty-three minutes.

I can see what Kat is talking about: this really is intoxicating. I mean, King Hadoop’s computer army was one thing, but this is real people. Lots of them. Mostly Estonians.

“Oh, hey, guess what?” Kat says suddenly, a jolt of excitement animating her face. “They’re going to announce the new Product Management soon.”

“Wow. Good luck?”

“Well, you know, it’s not completely random. I mean, it’s partially random. But there’s also, like—it’s an algorithm. And I asked Raj to put in a good word for me. With the algorithm.”

Of course. So this means two things: (1) Pepper the chef will, in fact, never be chosen to lead the company; and (2) if Google doesn’t put this girl in charge, I’m going to switch to some other search engine.