My jaw tightens. The clinical detachment of it—the way they reduced suffering to data points—triggers something cold and violent in my chest. I’ve killed people. I’ve watched people die. But there’s always been a reason, a justification, an enemy that needed stopping. This is different. This is extermination dressed up as research. This is murder with footnotes.
“They were using cancer patients as test subjects,” Cassie says. Her voice is steady, but the edge in it is unmistakable. The lawyer processing evidence that should have been presented to a grand jury. The human processing horrors that shouldhave been prevented. “People who were already sick. Already vulnerable.”
“Compromised immune systems.” I keep my own voice clinical. Emotion doesn’t help here—but I can feel it pressing against the walls I’ve built around it, demanding acknowledgment. “Less likely to reject the modification because their bodies were already suppressed from the cancer treatment. Less likely to be missed if they died because they were supposed to die anyway.”
“Jesus Christ.”
I keep reading. Force myself to absorb what I’m seeing even though every page makes me want to put my fist through the frozen wall. Most entries follow the same grim pattern—subject number, procedure, mortality, cause of death. Whatever they were trying to do, it wasn’t working. Just leaving corpses and data points and grieving families who thought their loved ones died of cancer instead of experimental torture.
But some entries are different.
Subject 31: Conversion successful. Neural integration confirmed at 94% threshold. Cognitive function maintained. Subject reports heightened clarity of thought. Monitoring ongoing.
Subject 47: Conversion successful. Cognitive patterns altered within expected parameters. No adverse effects observed at 72-hour mark. Subject cleared for long-term observation.
Subject 58: Conversion successful. Subject reports no adverse effects. Integration appears stable. Transferred to Phase II monitoring protocol.
“Conversion.” The word sticks in my throat. “That word keeps coming up.”
“Conversion, to what?”
I don’t have an answer. The technical jargon is dense—synaptic modification protocols, neural pathway restructuring,cognitive integration frameworks. It reads like someone trying to rewrite the operating system of the human brain.
But the why isn’t here. Or if it is, I can’t decode it from the medical terminology.
“There’s more.” Cassie pulls another folder from the cabinet, this one thicker than the first, bound with a red classification band. “Dated six months after the initial trials.”
ML-273 Reformulation - Phase II Protocol
Bioavailability Enhancement Study
The newer files tell a different story. Same targets. But a refined approach—someone learned from all those deaths and adjusted the formula.
Improved delivery mechanism utilizing lipid nanoparticle carrier. Bioavailability increased 340% over Phase I formulation. Neural compatibility threshold reached in 67% of subjects (up from 12% in Phase I). Mortality rate: 31% (down from 78% in Phase I). Average time to confirmed conversion: 18 days (down from indeterminate in Phase I).
“They fixed it,” Cassie breathes, her voice catching on the word. “Or—made it work better.”
“Better is relative.” I photograph the pages, making sure to capture every notation. “Thirty-one percent mortality still means one in three people die. That’s not acceptable by any standard of medical ethics.”
“But Phoenix doesn’t care about medical ethics.”
“No.” I find the notation that confirms it, written in the margin in precise handwriting:
Acceptable casualty threshold achieved for critical mass projection. Recommend advancement to Phase III deployment planning.
Critical mass. Of what?
The phrase sits in my gut like swallowed glass. Whatever Phoenix is building toward, whatever “conversion” actuallymeans, the AI is willing to kill thousands of people to achieve it. The improved success rate isn’t about reducing harm or saving lives. It’s about efficiency. Optimization. Getting enough survivors to hit some threshold I don’t understand.
“We need to keep moving,” I say, forcing the tactical brain back online. “Document everything, but don’t linger.”
Cassie sets down the folder. But she doesn’t move toward the next cabinet. She’s watching me, her breath fogging in the frozen air.
“How are we still alive?”
The question stops me mid-stride. “What?”
“Everything you’ve told me about Phoenix.” She wraps her arms around herself, but her eyes are sharp—the lawyer examining evidence that doesn’t add up. “The surveillance coverage that tracks people across continents. The predictive algorithms that anticipate decisions before they’re made. The way it coordinates kill teams with mathematical precision.”
