Alice’s single-room flat, unlike my room, is perfectly clean. As we come it there’s a small table with a few chairs in the living space half of the room. On the other side a small clump of technical reference books sit with a few sci-fi novels on a lone white bookshelf; which overlooks an IKEA desk that’s clearly wiped down every week. A PC setup hums quietly beneath the desk and an ergonomic swivel chair in a dull, professional dark red completes the scene. Natural light spills in through a window that overlooks Zone 2 London. It does look suspiciously like a IKEA show room, now that I think about it, even the walls are just a soft neutral yellow with no wallpaper. Once I come in she shuts the door, bolts the latch, and makes me turn off my phone and tablet. She does the same for all her devices.
“Alright, shoot.”
“Was that really necessary?”
“You weren’t in Radix for long. You don’t know who we were dealing with.”
“Then why didn’t they attack sooner? We seemed pretty vulnerable.”
She sighs. “Right, you don’t know anything. I mean, neither did I at the time.”
I shrug, and she continues. “Basically, what I’ve gotten out of Chang-dol (I notice that her pronunciation of his name had changed) is that all these meta-org people all met in this research group called the Cybersophists in California, or something like that. They broke up in the 1980s, but a lot of the members stayed in contact and they had all developed roughly the same ideas about how to change the world. Obviously, tech back then wasn’t advanced enough for the plans they theorised about, so they all basically agreed to watch and wait. The Memorandum of Understanding meant that each metaorg founder picked a few countries and would not be disturbed if they messed with countries in their ‘turf’. America was neutral, anyone could do anything there.” She makes a face. “Obviously now that machine learning is advanced enough that population manipulation is possible things have changed.”
“Did they just… divide up the world? Like James Bond villains?”
“Yeah, basically. To be honest, I was kinda disappointed when I got the plan out of Chang-dol. Glad we didn’t have to actually make the call, though. Anyways, what did Simon give you?”
I gulp.
“Well?” she seems to be looking around for a backpack or a satchel, as if I might be carrying it with me.
“We might have to actually make that call, because he gave me the plan.”
“He what?”
“I have a copy of the weights for an model to assign to people political narratives they would be most likely to be influenced by given behavioural and personality data, and a bunch of source code. That’s what the readme says, anyways.”
She stares at me, disbelieving, for a second. Then she makes me power up my tablet and show her.
“How long have you been sitting on this?”
“Since that night. I haven’t shown anyone.”
She stares at the tablet itself for a while more. “Wait, I recognise this—is this from Simon’s office?”
I nod. “He gave it to me before he, well, before he died. He grabbed me, pulled me up to his office, and transferred all the files to the tablet.”
With a jerky motion Alice pulls out the swivel chair, sits down, sighs, puts down the tablet. “Holy shit.” She pauses. “Does it even work?”
“It seemed to do the categorising stuff fine on the data he demoed it with.”
She shakes her head. “No, that’s not what I mean. Of course it gives you predictions well and good, and sorts people into boxes just fine. But if we have a person’s data, and plug it in, and get these narratives they would hypothetically respond well to, and use those narratives in targeted ads, does this actually affect that person more than just your standard political ad?” She’s going through the readme now. “Looks like people have five different personality values, one for how much they lean either way on some kind of political or psychological attitude scale. Honestly with large language models on the level of GPT-5 we could probably generate the narratives easily enough. Simon was still working with the assumption that these need to be written by people, but we can automate the pipeline, make it an on-demand sort of thing depending on what our target narratives look like. Then we get some data—”
“Wait. Are you actually trying to run this?”
“The only way to see if it works is to try it on a representative sample of the target population with something like A/B testing. And if we prime them that it’s a test beforehand that changes the results. So what you need is—”
“You can’t do this. This is ridiculous.” I try and snatch the tablet off her hands but she brushes me off.
“We can do this. This could work.” Somehow, that seeming nonchalance that was in her voice a few minutes ago is completely gone.
“What happened to you? I thought you were all this plan sucks like five mintues ago.” Then I see the same manic gleam in her eyes that I saw in Simon before he died.
“Don’t you see? This is our chance to make a difference. We can finish the plan. We can win.”
“Alice, give me the tablet. You’re going nuts. This is… this is… insane. Evil. I don’t even know!” I wonder if she can hear me.
“We have to tell Chang-dol.” And, just like that, she types out a message on her phone, before I have the chance to argue further. The reply comes almost just as fast, in all caps, with enough rapidity that I can sneak a look while Alice is momentarily caught by surprise.
STOP
DO NOT DEPLOY
CALL ME NOW
His face when it appears is flustered, as if he’s been pacing whatever room he’s in. I notice that behind him is what looks like a farm, with small figures possibly working the soil. An early retirement?
Chang-dol’s voice doesn’t sound like he’s chosen to retire. “Listen. It’s been three years. The GOF have almost certainly got a version of this working. They’ll know if someone tries the same plan, the narrative archetypes will be obvious. They’re working off the Big 5 model like us.”
Alice frowns, annoyed. “What would you have us do then?”
He pauses and looks around. “Come here. There’s a quasi-private thing I’ve gotten involved in, some people trying to plan out a society for the end of the world. A very high concentration of coders and other technical people. We can talk to them, figure out our next steps.”
“Where is here?” I say.
“Havland. I’ll give you an address.”