"Thanks, Margaret, that means a lot," Catherine says. The room is somewhat cramped, with a smaller room off to its left side. There's an L-shaped line of work tables along the right wall and the far wall, with a couple computer terminals along it, all displaying a flashing red warning message about an inability to connect to the servers. On the wall, there's drawings of the ARK, both mechanical drawings and interior maps and renders of the virtual world like the ones Margaret saw in the computer at Lambda. On the counter, there's various connections for some kind of peripherals, looking a lot like Cortex chips but less industrial. Around the corner, the smaller rooms seems to be racks of the chips, hooked into some kind of support frames. It might have been powering them, or it might just be really fancy racks.

The virtual world in the ARK looks--good. Like a place meant for living in. Not like here. 

Can she get the terminals to give her anything other than a warning message? She's not hoping for much in the way of network, but they might have something in local storage.

"It looks like the lab terminals can't contact the local cluster," Catherine says. "You might need to go downstairs and reset the servers to activate the terminals in here. You might have more luck with the systems in the scan room, they're more capable on their own and have more local storage."

"Okay, I'll try the scan room first and then downstairs." Scan room, do you have any secrets to reveal?

The scan room has one of the robots hanging from the ceiling, apparently inert, across from one of the pilot seats. The floor around the robot is marked "Robot Control Test Area," and a diagram of the "SCX-303 Pilot Seat" is up on one wall. Unfortunately, most of the computers seem to have similar connection issues. There is, however, a comm panel with a buffer waiting on the wall, and a small corridor leads into what looks like a small server closet, though the tentacables erupting along the edges of the doorway are worrying. "Hey, Margaret, I want to check something. Can you take a look in the data cluster?" Catherine says over the PA.

"Sure." She makes a mental note to check the data buffer and heads into the server closet.

The WAU tentacables and polyps are growing all over, around, and out the cabinets, but the system is apparently intact, and displaying a screen listing "Development Kit--Legacy Scans". In the list, there's one labeled "Nanami" from 2014, done with a resolution of 21 ns, then Munshi and Pegg listed at 7 and 6 ns, respectively, and then...Margaret Peregrine. Born on a familiar date in 1995, died...2093. There's also a series of recordings, one labeled, "Recommendations for working with Margaret Template Instances," another a transcript of a Turing Award address by "Dr. Margaret Peregrine" in 2035. And then, the final thing is a simple text document, labeled "A Message for Margaret."

It begins, simply: "Margaret, if you're reading this, then something remarkable has happened..."

Oh wow. She had a future. She won a Turing Award. She did things and learned things and achieved things and she can't remember any of it. Not even that, she doesn't have any of the expertise or wisdom or personality changes that come from having a long life full of doing things. It happened and then it had never happened.

She keeps reading the message from her dead future.

Margaret, if you're reading this, then something remarkable has happened. I, which is to say you, have finally succeeded at running a brain scan at high enough fidelity to preserve conscious thought. The new vistas this will open up for humanity don't need discussion, because we think about them every time we reattempt the scan, but take a moment to be happy about them anyway. The potential end of death, the freedom to explore new ways of being . . . it's a big future out there, and if the biological me is still alive, I can't wait to watch you explore it. If not, I know you're all the me I need.

She stares at the screen for a long time, memorising the message. She spent her whole life trying to wake herself up, with updated scans so she wouldn't be too far behind . . . somehow it makes everything she's experienced feel a little less awful, knowing she was accidentally fulfilling her own dream. Her future self would have wanted her to conclude that this second life was worth it; she'll make it have been worth it.

She reads the "Recommendations for working with Margaret Template Instances".

The document looks like instructions for using a non-sapient neural scan as the basis for limited AI. The instructions are written by Dr. Peregrine, circa 2040--apparently, a refinement of the work which won her the Turing Award. It looks like the idea is that by hacking an existing connectome, even a non-sapient one, many of the issues of AGI which dogged AI speculation in Margaret's 2015 can be avoided, resulting in (depending on the inputs) a specialist or generalist agent which can be created adapted to a particular task without worries about it "waking up" but retaining much of the evaluative and adaptability.  

The concept of quasi-Margarets that don't talk or exhibit consciousness but are still in some sense her is kind of weird and messed up, but objectively it isn't that much weirder or more messed up than, for instance, the incoherent excuse for a mind she has while dreaming. It's also pretty cool that, according to this, her scans are "unusually suited to operating machinery with structures very different from the standard human body plan". There's also some stuff about the ideal rates of introduction of novel training data and the design of optimal reward and reinforcement patterns that reads like an extremely specific personality quiz result. The people using these systems knew more about how to teach (parts of) her things than she had learned from years of going to school.

There's a "click" from the PA, drawing Margaret's attention back to the tiny server closet with a terminal she's in, overrun with WAU tentacables.

"So, is that you, Margaret? The devkit scan?"

"Huh? Oh. Yes, that's me. It's--good to know more about why I was even around to be woken up down here." It's good to know a lot of the things she just learned, but that's the part she can put into words right now.

"That explains that, I guess," Catherine says. "I was wondering, it was the only thing I could figure but...waking up a legacy template scan is supposed to be impossible."

"Well," a not-quite-manic giggle, "Clearly I'm incapable of figuring out how to do it!"

"They're supposed to be too flat, too...blurry, you might say. I guess, actually, you did--that was the word you used in the text I read in college," Catherine says. "I used some of your work and theories combined with some other stuff of my own when we made the ARK scans. I just don't get how the WAU..." She trails off.

"The WAU does seem like a plausible culprit, yeah. I just don't understand how intelligent they are. They might be intelligent in a way pretty orthogonal to human intelligence--less language ability and theory of mind, but similar mathematical or problem-solving ability . . . I guess instead of speculating I should find some way to experiment with the WAU and get the truth."

"That doesn't get us any closer to finding out where everyone went, or to the ARK, though," Catherine says. "And I've never seen anything that indicated to me that the WAU was more awake than a template-scan-based AI."

Margaret produces an approximation of a sigh. "The WAU might have some other people we could put in the ARK, but I agree that's not our first priority. Still, if I get an opportunity to find out what the WAU's deal is, I'm going to take it. If they could wake me up, and keep Amy alive, they might be able to do other useful things too, if we can just figure out how to ask."

"Maybe," Catherine says. "It was keeping the Curie going, at least. I'm just worried with everything else it's been doing if getting its attention is actually a good idea."

"Yeah, it seems to be very much a literal genie in some ways." 

Speaking of finding out the truth, she passed a data buffer on the way in here. She should go read that.

"Uh, Strasky, come in? I need help in the lab, the scan room," a recording of Catherine says.
"What happened?" a second voice, presumably Strasky's, asks.

"Konrad killed himself after the scan," Catherine says.
"Jesus, how?" Strasky says.

"Uh, maser tool," Catherine says. "What should I do?"
"I'm...going to need to tell Strohmeier," Strasky says.

"No! Please, I'm so close!" Catherine says. "Strohmeier's going to shut down the ARK project. It's not my fault people keep killing themselves."
"Catherine...what're you going to do? It's not like you can sneak a 300 pound body out of the lab," Strasky says.

"I...know," Catherine says, in what almost sounds like a moan.
"Catherine, are you OK?" Strasky asks.

"...Not even close," Catherine says with a sigh, and then the buffer cuts out.

Well, that's awkward.

"I just got one of those data buffers. Uh, you were in it."

"One of the communications buffers? When was it from?" Catherine asks.

"Sounded like toward the end of when you were scanning people? They kept committing suicide after the scans and Strohmeier was getting sick of it and so were you."