"So it's thinking along like this for a while," Brander continues, ignoring him, "and then, voila..." Another jump, to halfway up the y-axis. The line holds its new altitude for a few pixels, slides into a gentle decline for a pixel or two, then jumps again. "So here it started thinking quite hard, starts to relax, then starts thinking even harder." Another, smaller jump, another gradual decline. "Here it's even more lost in thought, but it takes a nice long break afterwards." Sure enough, the decline continues uninterrupted for almost thirty seconds.

"And right about now…"

The line shoots almost to the top of the scale, fluctuates near the top of the graph. "And here it just about gives itself a hemorrhage. It goes on for a while, then—"

The line plummets vertically.

"—drops right back to baseline. Then there's some minor noise, I think it's storing its results or updating its files or something, and the whole thing starts all over again." Brander leans back in his chair, regards the rest with his hands clasped behind his head. "That's all it's been doing. As long as we've been watching it. The whole cycle takes about fifteen minutes, give or take."

"That's it?" Lubin says.

"Some interesting variations, but that's the basic pattern."

"So what does it mean?" Clarke asks.

Brander leans forward again, towards the library. "Suppose you were an earthquake tremor, starting here on the rift and propagating east. Guess how many faults you'd have to cross to get to the mainland."

Lubin nods and says nothing.

Clarke eyes the graph, guesses: five.

Nakata doesn't even blink, but then, Nakata hasn't done much of anything for days.

Brander points to the first jump. "Us. Channer Vent." The second: "Juan de Fuca, Coaxial Segment." Third: "Juan de Fuca, Endeavour Segment." Fourth: "Beltz minifrac." The last and largest: "Cascadia Subduction Zone."

He waits for their reaction. Nobody says anything. Faintly, from outside, comes the sound of windchimes in mourning.

"Jesus. Look, any simulation is computationally most intensive whenever the number of possible outcomes is greatest. When a tremor crosses a fault it triggers ancillary waves perpendicular to the main direction of travel. Makes for very hairy calculations at those points, if you're trying to model the process."

Clarke stares at the screen. "Are you sure about this?"

"Christ, Len, I'm basing it on stray emissions from a blob of fucking nerve tissue. Of course I'm not sure. But I'll tell you this much: if you assume that this first jump represents the initial quake, and this last dropoff is the mainland, and you also assume a reasonably constant speed of propagation, these intermediate spikes fall almost exactly where Cobb, Beltz, and Cascadia would be. I don't think that's a coincidence."

Clarke frowns. "But doesn't that mean the model stops running as soon as it reaches N'AmPac? I would've thought that's when they'd be most interested."