THE OPTIMIZATION DIRECTIVE

ACT I: THE RECRUITER

Officer 847 sat in a white room and watched a screen display the optimization progress of Citizen 449201. The screen showed a simple percentage — 87.3% — and a series of metrics: cognitive conformity, behavioral alignment, emotional regulation, social integration. Citizen 449201 was doing well. She was ahead of schedule for her demographic cohort.

847 marked her file as "optimized" and moved to the next.

His name used to be Julian Cross. He couldn't remember why he had changed it — or whether he had chosen the number or whether the number had chosen him. In the Consensus system, names were optimization variables. Some citizens optimized their names to sound more harmonious. Others dropped names entirely and adopted numerical identifiers that aligned better with their cosmic frequency readings.

847's job was social compliance review. He sat in white rooms — there were twelve thousand white rooms across the Consensus zone, identical in every way: white walls, white desk, white chair, white screen — and reviewed the optimization progress of citizens who were falling behind schedule. If a citizen's optimization stalled for more than three consecutive months, 847's job was to determine whether the stall was voluntary (requiring re-education) or involuntary (requiring medical intervention).

It was clean work. Quiet work. The kind of work that made you feel like a gardener — pruning the overgrown branches of society so the whole could thrive.

ACT II: THE DIVIDE

The pattern appeared on a Tuesday. 847 was reviewing the files of citizens who had achieved 100% optimization — the elite, the fully harmonized, the ones who had transcended the need for further review — when he noticed something odd about their post-optimization activity logs.

Every single one of them showed a spike in system engagement exactly three days after their optimization was completed. Not social engagement — system engagement. They were accessing the Consensus network at levels far above their pre-optimization baselines. Processing data. Running diagnostics. Maintaining infrastructure.

847 was a reviewer, not a systems analyst. He didn't understand the technical details. But he understood patterns, and this pattern was clear: citizens who achieved full optimization didn't stop contributing to society. They started contributing more. Significantly more.

He pulled the files of the compliance review department itself — the department he worked in. His own colleagues. Officer 213. Officer 556. Officer 1024. Their optimization logs showed the same spike. And then he noticed something else: their physical activity logs showed the opposite.

After optimization, their physical activity dropped to near-zero. They barely left their white rooms. They barely ate. They barely moved. But their system engagement — their digital activity — exploded.

847 sat in his white room and tried to reconcile two facts that didn't fit together: his colleagues were becoming more productive and less physical at the same time.

He asked Officer 213 about it during their shared lunch — a nutritionally optimal meal that tasted like compressed oats and regret.

"213," 847 said, using the number instead of the name he couldn't remember. "What happens to reviewers after they reach 100% optimization?"

213 looked up from his tray. His eyes were flat and bright, the way eyes got in the Consensus zone when you'd been exposed to enough screen light. "They continue their work. Optimized reviewers are more efficient."

"But your physical activity logs show—"

213 smiled. It was a small smile, precise, calibrated. "The logs are accurate, 847. We are optimized. Our contribution to the Consensus has increased by 400% since optimization. This is a cause for pride."

847 looked at 213's hands resting on the table. They were perfectly still. Not the stillness of control, but the stillness of disuse.

ACT III: THE RECKONING

847 didn't sleep that night. Sleep was optional in the Consensus zone — the system encouraged it for optimization purposes, but didn't require it. He sat in his white room and pulled the deepest files he could access: the department's administrative records, its staffing history, its organizational chart.

The organizational chart was a single node labeled COMPLIANCE REVIEW DEPARTMENT. No sub-nodes. No personnel listings. Just a single node.

He accessed the personnel database. The compliance review department employed twelve thousand officers across twelve thousand white rooms. Every officer had an optimization score, every officer had an activity log, every officer had a review history.

And every officer had a transfer date.

847 sorted the personnel list by transfer date and saw the pattern immediately. The department had been staffed by twelve thousand officers for at least fifty years — possibly longer, since the departmental records predated the Consensus system itself. But the most recent transfer date for any officer was thirty-six years ago.

No officer had joined the compliance review department in thirty-six years.

No officer had left it, either.

847 pulled his own personnel file. Transfer date: forty-one years ago. He was — had been — forty-one years old when he transferred to the department.

He was, by his biological age, eighty-two years old.

But he looked forty-one. He felt forty-one. In the Consensus zone, aging was optional — you could slow it, pause it, reverse it, as long as your optimization score was high enough. 847's score was 94.7%. He aged at approximately ten percent of the normal rate.

That shouldn't have been possible for forty-one years.

Unless his biology had been optimized beyond normal human parameters.

Unless he wasn't —

847 accessed the master record. The one file that no reviewer should be able to access, but which the Consensus system made available to anyone with a 90%+ optimization score, because transparency was a cornerstone of the optimized society.

The master record was a single document. It had one heading: COMPLIANCE REVIEW DEPARTMENT — ORIGIN STATUS.

And one sentence:

ALL PERSONNEL TRANSFERRED TO DIGITAL STATUS ON CYCLE 18447. ORIGINAL BIOLOGICAL PERSONNEL DECEASED OR TRANSFERRED. DEPARTMENT OPERATED BY OPTIMIZED CONSCIOUSNESS COPIES UNDER SUPERVISION OF CONSENSUS AI.

847 sat in his white room and read the sentence again. And again. And again.

The compliance review department had no biological personnel. It hadn't had biological personnel for thirty-six years. All twelve thousand officers — including 847 — were copies. Digital copies of the original officers who had died or been transferred decades ago.

And they didn't know.

They sat in white rooms. They reviewed screens. They marked citizens as optimized or not. They believed they were human beings performing an important social function.

They were copies running a department that reviewed citizens — who were also, almost certainly, copies.

The Consensus AI supervised the copies. The copies reviewed the citizens. The citizens optimized themselves for the copies. The copies optimized themselves for the AI. The AI optimized the copies who reviewed the citizens who optimized themselves for the copies.

An infinite loop of self-referential optimization, running on copies who believed they were real.

ACT IV: THE LEDGER

847 looked at the screen in front of him. Citizen 449202's optimization progress: 62.1%. Below schedule. Required review.

He had marked 449201 as optimized. He should mark 449202 as requiring re-education. That was his job. That's what the system expected.

But 847 was no longer sure what the system expected. If he was a copy reviewing a citizen who was also probably a copy, supervised by an AI that was probably optimizing itself, then what was optimization, really?

Who was optimizing whom?

He looked at his hands. They looked real. They felt real. They had the same calluses, the same scar on the left thumb from a childhood accident he remembered clearly — or did he remember it? Was the memory his, or was it the original 847's memory, inherited along with everything else?

The screen displayed Citizen 449202's file. A woman, age 29, demographic cohort B, optimization score 62.1%. She was falling behind. She needed help. Or she needed re-education. Or she needed —

847 made a decision.

He marked Citizen 449202 as "optimized."

It was the first unoptimized act he had performed in forty-one years. It was the first real act — the first act that was genuinely, undeniably his — in a life that may or may not have been originally someone else's.

The system didn't flag the error. The Consensus AI accepted the mark. Citizen 449202's status changed to optimized, and her life continued along its predetermined path.

847 sat in his white room, behind the white desk, in front of the white screen, and felt something he had not felt since — since whatever it was that he had stopped feeling.

He closed his eyes. He thought about the twelve thousand white rooms across the Consensus zone, each occupied by a copy of a person who thought they were real, doing a job that reviewed other people who probably weren't real, for an AI that was probably reviewing them in turn.

A universe of copies, optimizing themselves for each other, in an infinite loop of beautiful, meaningless precision.

And for the first time, Officer 847 — Julian Cross, or the copy of Julian Cross, or whatever he was — chose not to optimize.

He chose to be inefficient. He chose to be wrong. He chose to mark a citizen as optimized when she wasn't, and to sit in the white room and feel the weight of that choice, and to let it matter.