Chen assembled an escort team: six people, mix of combat and support classes, enough to handle entity encounters on the route to Marcus's building. The plan was straightforward-return to his East Austin apartment, generate activity above the threshold, maintain presence until his deprecation timer either dropped off the queue or hit zero.
"Worst case, you die and we document what happens," Rodriguez said with the kind of bluntness that made Marcus appreciate her honesty. "Best case, you survive and we prove individual deprecations can be reversed. Either way, valuable data."
"I feel very valued," Marcus said.
The test was scheduled for tomorrow morning-7:00 AM departure, estimated arrival at his building by 9:30 AM. That would give him thirty-eight hours remaining on his timer when they arrived. Plenty of time to test the theory.
Plenty of time to find out if I live or die, Marcus thought, but didn't say out loud.
Around noon, Marcus found himself back at the monitoring station, running follow-up observations on Building 2847.
The building was still marked Active. Activity Index had stabilized at 0.16 with Team Seven's continued presence. The shimmer hadn't returned. The structure looked solid, real, maintained.
Through the window, Marcus could see the current shift: Patterson on perimeter patrol, the Level 3 Synthesist doing inventory checks, two others whose names he was still learning.
The exploit was holding. The building wasn't being re-flagged.
One positive test result, Marcus reminded himself. Not proof of consistent behavior. Need more data points.
But it was something. A rule they'd found and leveraged. A pattern they could exploit.
His phone buzzed. Unknown number.
"Yeah?"
"Marcus Webb?" Unfamiliar voice. Male, formal tone, hint of Texas accent.
"Speaking."
"This is Michael Santos, Survival Cluster Delta, northwest Austin. Chen gave me your contact. I heard you figured out how to reverse deprecation flags."
Marcus sat up straighter. "We ran one successful test. Theory's sound, but we need more data to confirm consistency across different scenarios."
"Can you share the methodology? We've got four buildings flagged, two under twenty-four hours. If your exploit works-"
"Send me your cluster's data," Marcus interrupted. "Deprecation timestamps, current activity levels, structure classifications. I'll review and send back implementation protocols. If it works across multiple clusters, that confirms the exploit is systemic, not localized."
Santos was quiet for a moment. "You're just... sharing this? For free?"
"We're all in the same production environment. Your cluster learning the exploit doesn't cost us anything, and it gives us more data points to validate the theory." Marcus paused. "Plus, if the System's already flagged us for behavioral analysis, we might as well make it worth the attention."
Santos laughed-surprised, genuine. "Thank you. I'll send the data tonight. And Webb? Good luck with your test tomorrow. Chen told me about your situation."
"Thanks. Let me know how your implementations go."
Marcus ended the call and stared at Building 2847.
Word was spreading. Other clusters were hearing about the exploit. If this worked at scale, across multiple locations, across different structure types and scenarios...
They might actually have a chance.
His deprecation timer read 45 hours, 2 minutes.
* * *
That evening, Marcus tried to sleep early-he'd need his stamina for tomorrow's journey-but his brain refused to cooperate.
He lay in his dorm bed, staring at the ceiling, thinking about timelines and thresholds and all the ways the test could go wrong.
What if individual deprecation worked differently than structure deprecation?
What if the Activity Index calculation was different for residential buildings?
What if his apartment's Activity Index was already too low to recover?
What if the System had special rules for players versus structures?
What if I'm about to find out that some deprecations can't be reversed, and mine is one of them?
Around 9:00 PM, Marcus gave up on sleep and went to the library's analysis space.
Julie was there, as expected-she never seemed to sleep. She was working on entity correlation data, laptop screen showing spawn density maps overlaid with activity metrics.
"Can't sleep?" Julie asked without looking up.
"Brain won't stop running failure scenarios."
"That's very QA of you." She pointed at the spare chair. "Sit. Talk through them. Sometimes speaking aloud helps."
Marcus sat. "What if individual deprecation flags can't be reversed? What if the System has different rules for people versus structures?"
"Then you die, and we learn that the exploit has limitations." Julie's tone was matter-of-fact. "But we'll have more data than we did before. That's not nothing."
"That's not comforting either."
"I'm not trying to comfort you. I'm trying to help you think clearly." Julie pulled up something on her screen. "Look. You've established that the System monitors activity and uses it for deprecation decisions. You've proven that increasing activity can reverse structure flags. The logical extension is that the same principle applies to individuals-unless the System explicitly uses different criteria for player deprecation versus structure deprecation."
Unauthorized usage: this narrative is on Amazon without the author's consent. Report any sightings.
"Which it might."
"Which it might," Julie agreed. "But you won't know until you test it. And sitting here catastrophizing doesn't change the test parameters."
Marcus knew she was right. Didn't make it easier.
"Chen thinks I'll survive," he said. "Says the data supports it."
"Chen's good at probability assessment. If he thinks you'll make it, that's worth considering." Julie glanced at him. "But even if you don't-and I'm not saying you won't, just considering all outcomes-you'll have provided valuable data that helps the rest of us survive. That's not a small thing."
"Being useful data when I die is not high on my list of career goals."
"Adapt your expectations." Julie returned to her work. "Get some sleep, Marcus. Tomorrow's going to be long either way."
Marcus tried. He went back to his dorm, lay in bed, and eventually managed something that might have qualified as sleep around midnight.
When he woke at 5:47 AM-internal clock still perfectly synchronized to the System's cron timing-his deprecation timer read 43 hours, 16 minutes.
Time to find out if his theory could save his life.
* * *
The Trap
Marcus was halfway to the coordination center when the second notification appeared.
Not blue text this time. Gold. Different formatting. Priority flag impossible to ignore.
[Skill Evolution Available]
Anomaly Detection v0.8.1-beta → v1.0.0-release
Requirements met:
- Successful pattern identification (3+)
- System exploit documented and validated
- Behavioral monitoring flag triggered
Evolution includes:
- Increased detection range (+40%)
- Improved confidence accuracy
- New feature: Probability Assessment (beta)
- WARNING: Skill includes telemetry reporting
Accept evolution? [Y/N]
Marcus stopped walking.
Read it again. Focused on the warning.
Skill includes telemetry reporting.
The System was offering him an upgrade. Better detection range-+40% meant he'd spot patterns faster, see connections he might otherwise miss. Improved accuracy on confidence scoring. A new feature for probability assessment that could evaluate exploit reliability without full testing.
All in exchange for enabling telemetry.
The System wants to track what I'm detecting, Marcus realized. It's offering me better analytical tools in exchange for data about what patterns I'm finding, what exploits I'm discovering, what mechanics I'm reverse-engineering.
It was elegant. Reward the user for finding exploits, then monitor their continued analysis to learn what other vulnerabilities they're discovering. Use the analyst's own pattern recognition to map your own weak points.
Classic optimization feedback loop. Marcus had implemented similar systems for bug tracking-give QA better tools, collect their findings, use the aggregated data to improve testing coverage.
The System was doing the same thing. Except instead of improving testing, it was probably using the data to patch exploits. To close vulnerabilities. To make itself harder to game.
Marcus should refuse it.
But.
The upgrade was significant. +40% detection range meant spotting subtle patterns he might otherwise miss. Probability Assessment meant evaluating whether an exploit was worth testing without having to risk lives on unvalidated theories. Improved accuracy meant fewer false positives, less time wasted chasing noise.
And if he was going to test his own deprecation reversal today-going to risk his life on the theory that activity metrics could save him-having better analytical tools might be the difference between spotting a critical detail and walking blindly into failure.
Chen's warning echoed in his head: The System doesn't give gifts for free. If it's offering you upgrades, it's because your continued survival serves some purpose.
Marcus pulled up the notification again and really looked at it.
What was he giving up, exactly?
The System would know what anomalies he detected. What patterns he recognized. What exploits he was analyzing. It would see his analytical process in real-time-every time he used Anomaly Detection to spot something unusual, that data would be sent somewhere, logged, processed, probably fed into whatever algorithm governed System behavior.
He'd be telling the System exactly what he was looking for.
But the System already knew he was looking. The behavioral analysis flag made that clear. The cluster was being monitored. Marcus's actions were already under scrutiny.
The question wasn't whether the System was watching-it was whether the tools he gained were worth the cost of being watched more closely.
And the answer, Marcus realized with grim certainty, was yes.
Because in forty-three hours and eleven minutes, his deprecation timer would hit zero. And if he didn't have every possible advantage-every tool, every insight, every analytical edge-he might not survive long enough to regret accepting the telemetry.
Pragmatism beats principles, Marcus thought for the third time since the System had arrived. I'm walking into the trap with my eyes open because the trap has better tools than I can get anywhere else.
He accepted the evolution.
[Skill Evolution Complete]
Anomaly Detection v1.0.0-release
New capabilities unlocked.
Telemetry enabled.
System monitoring active.
Thank you for your participation in System optimization.
The notification disappeared. Marcus felt... nothing dramatic. No surge of power. No obvious change. Just a subtle shift in his peripheral awareness, like someone had adjusted the contrast on reality and he could now see patterns he'd been missing before.
He pulled up his status screen.
Skills:
- Anomaly Detection v1.0.0-release ??
- System Analysis v1.0.0 (Unreliable)
- Pattern Recognition (Passive)
The warning icon next to Anomaly Detection glowed faintly-a permanent reminder that this skill was reporting back to the System.
I just agreed to let the System track my exploit analysis, Marcus thought. In exchange for better tools to find more exploits. I'm either brilliant or about to become a case study in why you don't make deals with automated systems that control reality.
His deprecation timer read 43 hours, 9 minutes.
He continued walking to the coordination center, carrying new skills and old doubts toward a test that would either prove he was brilliant or prove he was dead.
* * *
The Journey
The escort team assembled at 7:00 AM in the coordination center's briefing room.
Six people besides Marcus: Rodriguez leading (Level 5 Scout, pragmatic and competent), Patterson (Level 4 Defender from Team Seven), two medics-Kim (Level 4, also from Team Seven, skeptical but competent) and Aaron (Level 3, young and nervous but allegedly skilled)-one Level 5 Scout named Maya Reyes (quiet, efficient, the kind of person who spotted threats before they materialized), and Julie Tran as analytical support.
"Route is straightforward," Rodriguez said, pointing at a map. "East through campus perimeter, south on Speedway, east on Dean Keeton, then navigate residential streets to Webb's building at ATX-E-4471. Estimated travel time: two hours accounting for entity avoidance and cautious movement. Webb's timer will be at forty-one hours remaining when we arrive."
"Entity threat assessment?" Patterson asked.
"Moderate. Morning hours typically show lower spawn density, but we'll still encounter patrols. Standard engagement rules: avoid when possible, eliminate when necessary, don't take unnecessary risks." Rodriguez looked at each team member. "This is a data collection mission, not a combat operation. Webb's survival is the priority objective. Everyone clear?"
Nods around the room.
Rodriguez turned to Marcus. "You're non-combat class. You stay in the middle of formation, follow directions without argument, and don't try to help if things go bad. Our job is protecting you. Your job is staying alive long enough to reach your building. Understood?"
"Understood," Marcus said.
"Good. Move out in fifteen minutes. Final gear check."
The team dispersed to collect equipment. Marcus had his backpack-laptop, phone, charging cables, notes-and not much else. He wasn't built for combat, and everyone knew it.
Chen appeared in the doorway. "Webb. Got a minute?"
Marcus followed him into the hallway, away from the team.
"I saw the skill evolution notification this morning," Chen said quietly. "The telemetry warning. You accepted it anyway."
"Better tools might keep me alive today. Seemed worth the tradeoff."
"Maybe. Or maybe the System's training you. Offering upgrades to see how far you'll go, what you'll accept, how much you'll compromise to survive." Chen's expression was unreadable. "Be careful, Marcus. The System doesn't give gifts for free. If it's offering you upgrades, it's because your continued survival serves some purpose in its architecture."
"Noted. Still better than dying in forty-three hours."
"Fair enough." Chen held out his hand. "Good luck. Come back alive. We need your data."
"I'll do my best." Marcus shook his hand, then rejoined the team.
At 7:14 AM, they moved out.