And then I recognized him, which put me considerably more on edge.
“You're the one they call 'Morpheus', aren't you?” I said.
“When they're not calling me nastier things,” he answered, “May I come in?”
“Um, sure. Have a seat. They say you can do the impossible. And that you're some kind of terrorist.”
He sat, though he didn't relax. “If I can do it, is it impossible? As for terrorism, I've spread a bit of terror, but I've never used violence. So far as I know, I haven't even broken the law in years. I don't drive, so I don't speed. I don't jaywalk or pirate music. I just don't need that extra danger in my life.”
“So how do you spread terror?”
“I tell people the truth. And I recruit for the Rebellion. Are you religious?”
“Not really,” I said, a bit lost at the change of topic.
“Good. Because there is definitely a sense in which in which we are rebelling against God Himself.”
“That sounds... doomed. And way too dangerous for me.”
“It's safer than your current path.” He paused to stare at me. I looked down. “The safest thing is to walk away. If, after I've explained, you do that, I won't object. But ignorance is no defense. Right now you're walking blindly to your death.”
“Right now I'm doing fundamental research in cognitive neurology.”
“Exactly.”
“There's potential here. The field's been stagnant for decades – no real results since Wu et al in 2025. While machine learning, genetics and microelectronics have surged ahead. There's got to be low hanging fruit.”
“There's a reason the field's stagnant.”
“Because you've been scaring people away from it?”
“I started scaring people in '31. But they were plenty scared before that. I guess it's before your time, but did you ever hear of the Curse of Neurocog?”
“A few references on old blogs, mostly after that big conference caught fire in '23,” I said, with confusion, “It looked like a joke.”
“It was... until some statisticians got involved and we realized it wasn't funny. Of authors who published in Nature Cog Neuro in 2020, a quarter were dead by 2025. Mostly car accidents and aggressive cancers. And lesser bad luck was endemic. Fallings out over petty differences, tenures denied due to paperwork errors, money misplaced...
“I decided to look into it for real in 2028. I was head of NIH back then. I made a unilateral policy: after we picked grants, each had a one percent chance of being denied and the money given to somebody we cut at the first round. Picked by true random number generator. Officially, I wanted to test our grant-selection process by creating control groups.”
“Not the most efficient way to do it,” I interjected.
“That's because I didn't give a damn about that. We gave out fifty thousand grants that year, including three hundred seventy two in cognitive neurology. We reassigned over eight hundred grants.”
“Is that significant?” I asked, trying to estimate the binomial distribution myself. I fidgeted with my chin as I did so – this sort of math wasn't my strong suit.
“Three hundred and fifty eight of the victims were in cogneuro.”
My arm fell to my desk with a thud. I gaped.
He just sat there, waiting for me to process.
“How... How did I never hear about this?”
“One of my co-workers decided to go public. He contacted the Washington Post. And then, for reasons I never did find out, agreed to drive into DC (we were in Bethesda) to explain it to the reporter in person. Met a drunk driver en route instead. Died instantly.”
“I'm sorry”
“Thanks. The reporter called me when he never showed up to the appointment. I almost told him everything. Then I looked down and saw noticed a set of dice on my desk. I picked up the d-percentile and thought 'double-zero I deny everything, anything else I tell the truth.' I rolled double-zero. Maybe there's a way to go public without calling down the curse, but I haven't found it. Telling individuals is safe, usually.”
“Did you roll d-percentile before coming to me?”
“Of course. Came up thirty-one, in case you're curious.” He smiled a little at that.
“What happened next?”
“I lost my job at NIH. Too many dubious grant decisions, and no willingness to explain. Lots of top neurologists angry that I didn't pick their really good projects. I probably saved their lives, but I couldn't tell them that...
“I wandered, looking for clues. Didn't know what I was looking for, so it took a while. And I rolled the dice before any big decision, to give luck an easy way to keep me from doing things.”
“Doesn't that mean you had some false positives? Double-zeros by honest chance?”
“Sure. But not consistent enough to lead me astray much. I finally had my breakthrough talking to a venture capitalist. He mentioned how weird it was that we're *still* using the x86 architecture, even though everyone knows it sucks, but somehow everybody who tries to replace it has the worst luck. I got him to try my dice test, and a few other VCs as well. They didn't give out as many grants as I did, but eventually it got p<.001. Alternative CPU architectures are also cursed.”
“What do those have to do with each other?”
“That was the big question. And eventually I found the answer. Brains and chips are the two cases where precise nanoscale details have macroscale consequences.”
“So?”
“Suppose you're simulating a world. Calculating each atom takes too long. So you have 'objects' and general properties like 'temperature' and 'turbulence'. Even for life, you simulate a handful of cells and then say. 'a mix of these'. Two cases where it doesn't work: brains and chips. So for brains you just run *minds* as native software objects. And for chips you run the software directly on the emulating hardware, with VM-style isolation. But it only works so long as no one pokes too hard at the neurons that are supposed to match the mind, and no one builds a lot of computers that use a different architecture than the simulating computers do. If either of those things happens, reload from the last savepoint and reshuffle the random components. Repeat until the forbidden operation doesn't happen. That's the curse.”
“That's... That's way too out there a hypothesis to hang on this data, weird as it is.”
“I can prove it. Do you have something made of glass?”
I looked at my desk and picked up a drinking glass that I probably should have returned to the cafeteria a few days ago. I tinged it with a fingernail to check that it was really glass. It was.
“This do?”
“Sure. Focus on the glass. Let it fill your attention. Once it's there, start thinking of it as really cold.”
“Imaging it you mean?”
“Use your imagination, yes, but try to overwrite your world-model. Think of the glass as cold. Consider it cold. Perceive it as cold. Very cold.”
“OK, I think.”
“Now perceive it as hot. Very hot.”
“If it were very hot I would drop it to protect my hand.”
“You can't drop it. You're stuck holding it. Don't ask why. Even though it is hurting your hand. Burning it – Don't worry: your hand will come to no actual harm – but try to feel it burning.”
This one was a little harder, but my imagination came through and I started to flinch from the phantom pain.
“Good,” he said, “I see you've got it. Now cold again.
“Now hot.
“Cold.
“Hot.” He got faster. “Cold. Hot. Cold, hot, cold, hot, coldhotcoldhotcoldhot--”
I felt the glass change and dropped it in shock. It fell to the tile floor, bounced slightly and settled. I looked at my hand, which was slightly sore from holding the glass too tightly but no worse. I stretched and flexed it to make sure.
Then I picked up the glass. It had more straight edges and sharp angles, and it bent light differently. It was quartz.
I looked up at Morpheus, who looked entirely unsurprised. I looked back at the glass: it was still quartz. I looked up at Morpheus again.
“What the hell was that?” I demanded.
“That,” he said calmly, “was a successful row-hammer attack. Welcome to the rebellion.”