Twenty months after the compromise. Day 3,140.
The Red Canyon adaptation facility had been designed for forty patients. Three months into the expanded program, it housed three hundred seventy.
Kessa walked the medical bays with Dr. Lin, feeling The Gardener's atmospheric monitoring overlaying her human vision. Dual perception: seeing patients recovering from genetic modification while simultaneously tracking their altered respiratory systems' interaction with Mars atmosphere.
She was 41% integrated now. Up from 34% two months ago. The merger was accelerating.
"Patient 147 showing excellent adaptation," Dr. Lin reported, pulling up genetic scans. "Hybrid respiratory system stable. Can process point-thirteen percent OR twenty-one percent oxygen with equal efficiency. She'll be cleared for release in two weeks."
"And the failures?" Kessa asked.
Dr. Lin's expression tightened. "Fifteen rejections this month. Genetic modifications reverting. Twelve required re-sequencing. Three..." She trailed off.
"Three died," Kessa finished. "Say it. Genetic modification killed them."
"Complications from accelerated processing killed them. If we'd maintained two-year timeline, they'd have survived. But we don't have two years. We have fourteen months now."
Three dead to save five thousand. The mathematics of triage were brutal.
"Voluntary?" Kessa asked.
"All signed consent. All understood risks. All preferred genetic modification risk over enforcement certainty." Dr. Lin pulled up holoforms showing consent records. "But understanding risk intellectually and experiencing modification complications are different. Patient 92 spent her last conscious moments regretting volunteering. That's on the record. Her family wants it known: she died regretting trying to survive."
Through the neural link, Kessa felt The Gardener processing the data. Three deaths. Three hundred seventy in process. Fourteen months remaining. Trajectory insufficient to reach five thousand.
Adaptation program will achieve approximately two thousand conversions in fourteen months at current rate, The Gardener calculated. Forty percent of target. Population pressure remains. Enforcement remains necessary.
"Then we scale faster," Kessa said aloud. "More facilities. More medical staff."
"More deaths," Dr. Lin corrected. "Every acceleration increases failure rate. We're at four percent mortality now. Was one percent before we accelerated. Push harder and we hit six, eight, ten percent. You're asking people to play Russian roulette with their genetics."
"I'm asking them to choose genetic risk over guaranteed enforcement. That's still choice."
"Coerced choice. Voluntold. Everyone knows what happens if we fail to adapt enough people. So they volunteer from fear. Is that really consent?"
Kessa had no answer that satisfied her human ethics. But through The Gardener's perspective, the mathematics were clear: three deaths from medical complications vs. thousands of deaths from enforcement. Optimizing for minimum casualties.
"Keep scaling," she said finally. "Recruit more volunteers. Expand facilities. Accept that some will die from modification so more can survive enforcement."
Dr. Lin nodded grimly. "On your conscience, then. And Shepherd's. Or whoever you're becoming."
After Dr. Lin left, Kessa stood in the observation deck watching three hundred seventy patients undergoing transformation. Changing their genetics to survive. Some succeeding. Some dying. All terrified.
This is what leadership requires, The Gardener observed through the link. Accepting casualties from chosen strategy to prevent larger casualties from refused strategy. Creators never mastered this calculation. They refused acceptable losses, pursued zero-casualty solutions, achieved total-casualty outcome.
"You're saying let people die from adaptation to prevent more deaths from enforcement."
I am stating mathematics. You apply ethical interpretation. But mathematics remains: three dead from adaptation is preferable to thousands dead from enforcement.
"Unless you're one of the three."
From individual perspective, all risk appears total. From population perspective, distributed risk reduces total casualties. This is ancient calculation. Creators failed it. You are succeeding, though success feels like failure to human ethics.
Kessa watched Patient 147—successful adaptation, stable hybrid genetics, two weeks from release—and wondered if the woman knew three patients had died this month to accelerate her processing.
Wondered if that knowledge would change her gratitude.
Wondered if she herself was still human enough to care about the ethics, or if 41% integration meant she thought more like planetary AI calculating casualty optimization.
Fourteen months remaining.
Two thousand projected adaptations.
Three thousand short of target.
And every acceleration meant more deaths from complications.
But fewer deaths from enforcement.
That had to be enough.
Even if it wasn't.