And the result, shown on the monitor, was just what it was supposed to be: a line with a negative-one slope, the telltale sign of a signal that carried information. Buoyed — or, as she liked to say, girled — she went ahead and plugged the data into the Shannon-entropy function, and—

Well, that was strange.

When her dad had run the data, he’d gotten a second-order Shannon-entropy score, indicating very-low-level complexity.

But her results were clearly third order.

She must have done something wrong. She noodled around, looking for the source of her error. Of course, she could ask her father or Dr. K where she’d screwed up, but figuring that out was half the fun! But after half an hour of checking and rechecking, she could find no flaw in what she’d done — which meant the error was probably in sampling. The data Kuroda and her dad had looked at must have been different somehow, and either their data set or hers wasn’t typical.

She switched to websight again — she was getting the hang of making the transition quickly, and no longer found it disorienting. Of course, when looking at the background a frame at a time, she had been vastly slowing down her perception of the Web; although she’d spent several minutes examining the buffered data, it represented only a small amount of time. But now that she was just looking in on the Web in real time, the background of cellular automata was shimmering once more.

She thought perhaps the giant, jittering version of her own face might reappear — perhaps that was what was causing her to get different results. But it didn’t, although…

Yes, something was different here in webspace. There was a tiny wavering, an annoying flashing, just at the limit of her perception. It wasn’t in the shimmering background, though; it was coming right at her. She frowned, contemplating it.

Yes, yes, yes! After the lesson, Prime rewarded me by reflecting myself back at me again. But I wanted to demonstrate my comprehension, so instead of reflecting Prime back at itself, I tried something new…

Caitlin switched back to simplex mode, restoring her vision of the real world, and then she headed down to the basement. Kuroda was once more hunched over in one of the swivel chairs, typing away at the desktop computer’s keyboard. He seemed lost in thought, and apparently hadn’t heard Caitlin enter, so she finally said, “Excuse me.”

Kuroda looked up. “Oh, Miss Caitlin. Sorry. How’s the reading going? Up to polysyllables yet?”

The letters F U briefly flashed through her mind. “Fine,” she said. “But, um, back in Tokyo, you used a phrase I didn’t understand. You said I might experience some ‘visual noise’ when you first activated the eyePod.”

Kuroda nodded. “Yes?”

“Visual noise — that’s interference, right? Garbage in the signal?”

“Yes, exactly. Sorry. I should have explained myself better.”

“I didn’t experience any back then,” she said. “But I think I might be experiencing some now.”

He swiveled his massive form around to face her properly. “Tell me.”

“Well, when I go into websight mode, I—”

“You’re doing that again?”

“I can’t resist, I’m sorry.”

“No, no. Don’t be. If I could see the Web, believe me, I’d be doing it, too. Anyway, what’s happening?”

“I’m not sure. But, um, could you have a look at the datastream that’s being fed to my eyePod?”

“The Jagster datastream, you mean?”

“I guess. But I think it’s being … polluted by something else.”

He frowned. “It shouldn’t be. Anyway, sure, let me have a look. Go into duplex mode, please.”

She did so; the eyePod made its high-pitched beep.

She heard his chair swivel and the clicking of a mouse. After a few moments he said, “It’s just raw Jagster data.”

“What are you looking at?”

“The feed coming to you from Tokyo.”

“No, no. Don’t look at the source; look at the destination. Look at what’s actually going into the buffer on my eyePod.”

“It should be the same thing, but … okay. Yeah, Jagster data, and … hello!”

“What?”

“You’re in duplex mode now, right?”

“Yes, yes. I have to be to receive.”

“Right. But … hmmm. Well, there is an extra signal coming in. It’s not properly formatted HTML, it’s … well, that’s strange.”

“What?”

“I’m looking at it with a debugging tool. See?”

“No, I’m seeing the Web.”

“Right, right. Well, I’m looking at a hex dump — 4A, 41, 52, 4B, etc. All the high-order nibbles are four or five. But the screen also shows the ASCII equivalent, and, well, I mean, yeah, it’s gibberish, and — oh, no, hang on. It’s not, it’s just hard to read. It’s all run together without spaces, but it says, ‘Egg frog goose hand igloo’.” He paused, then: “Ah, I must have come in the middle. It cycles around again to the beginning of the alphabet: ‘Apple ball cat dog,’ then ‘egg frog,’ etc.”

“How does it say it?”

“What do you mean?”

“I mean, is it all in capitals?”

“Yes. How’d you know?”

“Here … give me a sec.” Caitlin reached into her pocket, and pressed the eyePod’s button. She heard the low-pitched tone, and webspace dissolved into reality. She moved over and peered at the LCD monitor. It was overwhelming, seeing so many capitals packed together; she had trouble making sense of them, but—

“That’s part of the reading exercise I did earlier. But how could that get bounced back at me?”

Kuroda frowned. “I have no idea.” He looked at her. “Has anything else like this happened?”

“No,” she said, perhaps too quickly. “Weird, isn’t it?”

Kuroda’s features rearranged themselves in a way Caitlin had never seen before, but she guessed it meant he was perplexed. “It certainly is,” he said.

“You’re using an online literacy site, right?”

“Yes.”

“It must communicate in HTML, or at least with HTTP standards,” he said. “I mean, I’ll check it out, but if the feed from it was just somehow echoing back at you, there should be more than just the ASCII characters.”

“Doesn’t most of the Web use Unicode instead of ASCII these days?” Caitlin asked.

“Oh, lots of it is still pure ASCII, but for basic Western letters, Unicode and ASCII are the same, anyway; Unicode just adds a second byte to each character that’s nothing but eight zero bits.”

“Ah, okay. But where’s this coming from?”

He took a deep breath, let it out, and lifted his chubby hands a bit. “I’m sorry, Miss Caitlin. I have no idea.”

Back in her room, Caitlin did two hours of online literacy lessons, but found her mind wandering back to the question of why she’d gotten a different Shannon-entropy score than her father had. She decided to try to replicate his results again, going through the process of gathering more data from the cellular automata and feeding it into the Shannon-entropy calculator, and—

Shit.

This time it came up as fourth-order entropy.

It could be another sampling error, but the sequence of second, third, fourth seemed more like a progression…

Could it be?

Could the information being conveyed by the cellular automata be growing more complex over time?

Did that make any sense at all?

No, no. Surely it was just that she wasn’t properly clearing out the data she’d previously fed into Mathematica. Yes, that had to be it: first, her dad had fed it a single set of data, and it had shown up as second-order entropy; next, she’d accidentally added another set on top of the first one, and it yielded third-order entropy. And now, she’d dumped yet another set of data on top of the previous two, and the program was reporting a result of fourth-order entropy. There must be a data cache somewhere in the program; all she needed to do was find it and flush it.