THREE DAYS EARLIER - DAY 23

  The council meeting started at 8:00 AM sharp, which meant Marcus had exactly ninety minutes to prepare a presentation that would convince two dozen skeptical survivors that his theory wasn't just pattern-matching run wild.

  He'd been awake since 5:30, reviewing his notes, rehearsing arguments, anticipating objections. His deprecation timer read 61 hours, 3 minutes. Enough time to run the test if they approved it immediately. Not enough if they spent three days debating methodology.

  Chen had told him to meet in the coordination center's west conference room - a repurposed faculty lounge that still had motivational posters about "achieving excellence" on the walls, which felt darkly ironic given that excellence now meant "not getting garbage-collected by reality."

  Marcus arrived at 7:55 with his laptop, his notes, and the particular kind of nervous energy that came from knowing your presentation might save lives or waste the last resources a dying community had.

  The room was already filling up.

  Director Vasquez sat at the head of the table - military bearing, gray hair in a practical bun, eyes that assessed threat levels automatically. Former Army logistics, according to Chen. Level 6 Warden. The kind of person who'd been making hard decisions long before the System arrived to gamify them.

  Around her: the council. Team leads, resource managers, senior scouts. Most were Level 4 or higher. Most looked like they hadn't slept well in weeks. Most had the particular expression of people who'd lost friends to deprecation and weren't interested in theories that couldn't be proven.

  Chen sat to Vasquez's right, laptop open, already pulling up the data visualizations they'd prepared. He caught Marcus's eye and gave a microscopic nod: You're on. Don't fuck it up.

  Marcus sat down and tried to remember how to breathe like a human giving a normal presentation instead of someone arguing for permission to test a theory that might be the difference between survival and deletion.

  Vasquez glanced at the wall clock - exactly 8:00 - and spoke. "We're here because Chen says Webb has a theory about deprecation. We've lost six buildings in two weeks. Three more are flagged. If there's even a possibility we can stop it, I want to hear it. Webb, you have thirty minutes. Make them count."

  Thirty minutes. Not ninety. Budget cut before we even started, Marcus thought. Classic project management.

  He opened his laptop and pulled up the first slide.

  "The System's deprecation process isn't random," Marcus began. "It's algorithmic, scheduled, and exploitable. I can prove it."

  Marcus had given presentations before. Product demos. Sprint retrospectives. That one disaster of a stakeholder meeting where the senior VP asked why the team hadn't "just fixed the bug" that required rebuilding half the authentication system.

  This was different. These people weren't evaluating his competence - they were evaluating whether following his advice would get them killed.

  His hands were steady on the laptop. That was good. His voice was calm. Also good. His deprecation timer was counting down in his peripheral vision while he tried to convince people to trust his analysis. Less good, but unavoidable.

  "Timeline analysis first," Marcus said, pulling up the data Chen had helped him visualize. "Every documented deprecation in the cluster has occurred at exactly 11:47:23 UTC. No variance. No drift. Not approximate - exact."

  He clicked through timestamps. Building 12 from his first day. Three residential structures since then. Two administrative buildings. All at the same timestamp, down to the second.

  "That's not random decay," Marcus continued. "That's a scheduled process. The System runs a maintenance routine - like a cron job - that processes the deprecation queue at a specific time every day."

  Rodriguez, the Level 5 Scout who'd survived by not taking stupid risks, raised her hand. "So it's consistent. That tells us when buildings die, not why they're flagged in the first place."

  "Correct. Which brings me to the correlation analysis." Marcus pulled up the next slide - a scatter plot showing building activity metrics versus deprecation notices. The pattern was clear even to non-analysts: low activity clustered heavily with flagged structures.

  "Every building that's been deprecated showed the same pattern before flagging: occupancy dropped to zero, power consumption went to baseline, resource interaction stopped. The System treats buildings like cloud infrastructure. If something's not being used, it gets marked for garbage collection."

  Patel, the thin Synthesist with electrical-tape glasses, leaned forward. "You're saying the System depreciates buildings because they're empty?"

  "I'm saying deprecation correlates strongly with abandonment metrics. The System monitors activity. When it drops below some threshold, structures get flagged." Marcus clicked to the next slide - a timeline showing the gap between abandonment and deprecation notices. "Average lag time: four to seven days. Long enough that it's not instant response, short enough that the System's monitoring is continuous."

  "What's the confidence level on this?" Rodriguez again. Always the pragmatist.

  Marcus had expected this question. "High confidence on the scheduling component - that's just timestamp analysis, very clean data. Medium-high confidence on the activity correlation. Lower confidence on the exact threshold values, because we don't have enough data points to establish precise parameters."

  "So you're guessing about the part that matters," Rodriguez said flatly.

  "I'm forming a testable hypothesis based on available data. That's not guessing. That's methodology." Marcus kept his voice level. "Every exploit starts as a hypothesis. The question isn't whether I'm certain - it's whether the test is worth running."

  "And the test is?" Vasquez's eyes hadn't left Marcus's face.

  "Take a building that's already flagged. Move people in. Generate traffic, power consumption, resource interaction - all the signals the System uses to evaluate 'maintained' status. If deprecation is activity-based, the flag should be rescinded. If it's not, we've spent some time and learned that the System uses different criteria. Either outcome is valuable."

  "Resources required?" That was Patterson, one of the Defender team leads. Graying beard, practical expression, the look of someone who'd been rationing resources for weeks.

  "Twelve people on rotating shifts. Minimal supplies - whatever justifies regular traffic to the site. Power consumption through normal usage. We're not asking for combat deployment or major supply commitments. We're asking for occupancy."

  "Which building?" Vasquez asked.

  "UTX-A-2847, southeast perimeter. Three-story administrative building, already flagged with forty-one hours remaining. Low-priority supplies only. If it's going to be lost anyway, we might as well learn whether we can save it."

  Silence. The kind where people were doing mental cost-benefit calculations and not liking any of the possible outcomes.

  Patel broke it. "What if the System notices what we're doing? You said yourself it has behavioral monitoring. What if we trigger a response?"

  There it was. The question Marcus had been dreading because he didn't have a good answer that wouldn't sound like reckless optimism.

  "That's a risk," Marcus admitted. "But the System's already watching us. We're two hundred people surviving in its territory. The question isn't whether we stay under the radar - we're already past that threshold. The question is whether we understand the rules well enough to play the game without triggering exploits that get patched."

  "That sounds like a yes," Vasquez said. "You think this will draw System attention."

  The tale has been stolen; if detected on Amazon, report the violation.

  "I think existing draws System attention. At least this way we're learning the rules while we can still use them." Marcus met her eyes. "We're in a live production environment with incomplete documentation and no rollback capability. Every day we don't learn the System's behavior is a day closer to catastrophic failure. I'd rather test in controlled conditions than wait for the System to surprise us."

  Chen leaned forward. "Director, we've been reactive since Day One. Evacuating buildings, moving supplies, losing infrastructure piece by piece. Webb's theory gives us a chance to be proactive. To stop deprecation before it completes. Even if there's risk, I'd rather take it while testing one building than lose three more by doing nothing."

  Vasquez's finger tapped the table - a rhythmic, considering sound. She looked at Rodriguez. "Scout assessment?"

  Rodriguez studied Marcus for a long moment. "The analysis is sound. Pattern recognition is solid. Risk is real but manageable. If the building's already condemned, the downside is minimal."

  "Patel?"

  The Synthesist pushed her taped glasses up. "We need to understand the System's resource management logic. If this works, it could protect residential structures. That's worth the test."

  Vasquez's gaze moved around the table. Patterson nodded. Others showed varying degrees of agreement. The room wasn't enthusiastic, but it also wasn't hostile.

  "Background check," Vasquez said, looking at Marcus. "You're the QA guy. East Austin apartment, walked here alone, figured out the lane marking exploit without help. That right?"

  "Yes, ma'am."

  "Chen says you've got good analytical instincts. Says you think like the System thinks." Another tap on the table. "I'm inclined to approve the test. But I want to be clear about stakes. If this works, it's valuable. If it fails and triggers a System response that gets people killed, that's on you. Understood?"

  Marcus's mouth was dry. "Understood."

  "Good. You have authorization for one test. Building UTX-A-2847, southeast perimeter. Forty-one hours until deprecation. Chen will organize the implementation team. You'll coordinate monitoring and documentation. If the building drops off the deprecation queue, we expand the program. If it doesn't - or if you trigger something we can't handle - we revisit strategy." Vasquez stood. "Questions?"

  Rodriguez raised a hand. "What happens to Webb's deprecation status? He's at... what, sixty hours?"

  "Sixty-one hours, two minutes," Marcus said automatically.

  "So if this test runs for forty-one hours, and his timer runs out before we know if it worked-"

  "Then I get to find out whether being outside a deprecated structure actually saves you," Marcus finished. "Which is also useful data."

  The room was very quiet. Through the window, Marcus could see people moving across the quad, carrying supplies, manning guard posts, living their deprecation countdowns one hour at a time.

  "The test continues regardless," Vasquez said. "Webb knew his timeline when he proposed it. Anyone else?"

  No hands.

  "Dismissed. Chen, Marcus - stay."

  The council filed out. Marcus's heartbeat was loud in his ears. His hands had started shaking slightly - adrenaline dump, post-presentation comedown, the weight of what he'd just committed to settling on his shoulders.

  When the room was empty, Vasquez turned to them. "Off the record: I think you're right, Webb. The pattern's too clean to be random. But if this triggers something - entity swarm, System escalation, whatever - I need you to tell me immediately. Don't try to handle it yourself. Don't wait to see if it resolves. The moment you see anomalous behavior, you report it. Clear?"

  "Clear," Marcus said.

  "Chen, you're responsible for keeping him alive long enough to complete the test. Webb, you're responsible for documentation that proves or disproves the theory. Either outcome is valuable, but only if you survive to explain it." She paused. "One more thing. The council's betting on you. Don't make us regret it."

  Then she was gone, leaving Marcus and Chen alone in a room with motivational posters about excellence and a timeline that felt very, very short.

  Chen closed his laptop. "Well. That went better than expected."

  "She threatened to blame me if people die."

  "That's approval in Vasquez-speak. If she didn't think you were competent, she wouldn't bother assigning blame." Chen stood up. "Come on. We need to organize Team Seven."

  Team Seven turned out to be less a "team" and more "twelve people with different skills who'd agreed to spend the next two days keeping a building alive."

  Chen had assembled them in one of the library's ground-floor study rooms - a space that still had whiteboards and ergonomic chairs and the faint smell of spilled coffee from back when the biggest crisis was a midterm exam.

  Marcus counted heads: twelve people, mix of classes, mix of levels, all looking at him with varying degrees of skepticism, curiosity, and exhaustion.

  "Okay," Chen said, taking charge. "You've all been briefed on the theory. Building 2847 is flagged for deprecation, forty-one hours remaining. Our job is to generate enough activity to push it above whatever threshold the System uses for maintenance priority. Questions before Webb walks through the specifics?"

  A hand went up immediately. The speaker was a woman in her thirties, Latina, wearing practical cargo pants and a UT hoodie. Her status screen read Level 3 Crafter, name tag said Rivera. Dark hair pulled back in a practical ponytail, hands that looked like they'd handled tools more than weapons, eyes sharp with the particular alertness of someone used to logistics problems.

  "What kind of activity?" Rivera asked. "Are we just living there, or do we need to do specific things?"

  "Unknown," Marcus admitted. "Working hypothesis: the System tracks multiple signals. User presence, definitely. Power consumption, probably. Resource interaction, possibly. Movement patterns, maybe. We're going to maximize all of them and see what happens."

  "So we're the guinea pigs." Young voice from the back. Marcus tracked it to a Latino kid, maybe twenty-two, Level 2 Runner with nervous energy radiating off him like heat shimmer. His name tag read "Santos." "Just clarifying the role here."

  "We're the test case," Marcus corrected. "And if it works, you'll be the team that proved we can fight back against deprecation. That's worth something."

  "Is it worth more than forty-one hours of my life sitting in a supply building?" Santos asked. Not hostile - genuinely asking, like he was still deciding whether to stay.

  Fair question.

  "It's worth knowing whether we can save the rest of the cluster," Chen said quietly. "Building 2847 is expendable. The residential buildings aren't. If this works, we can protect two hundred people. That's the actual stakes."

  The room settled into considering silence.

  "What do you need us to do, specifically?" That was a man in his fifties, graying beard, Level 4 Defender named Patterson. Marcus recognized him from the council meeting - practical bearing, the look of someone who'd done guard duty in worse conditions than an empty building.

  Marcus pulled up his laptop. "Rotating shifts. Four people on-site at all times. You'll eat meals there, run inventory checks, maintain equipment, keep lights on. We're simulating an actively maintained facility. Chen's logistics team will deliver supplies to justify regular traffic. I need foot traffic at minimum every thirty minutes during waking hours."

  "And at night?" someone asked.

  "Two people on night watch. Reduced activity is fine - people sleep, the System probably accounts for that - but the building can't look abandoned. That's the signal we think triggers flagging."

  A woman with Korean features and a Level 4 Medic tag - name badge said "Kim" - spoke up. "What about entity risk? Southeast perimeter's not the safest sector."

  "Guard rotation will include perimeter sweeps," Chen said. "You're not out there alone. And if Webb's theory about entity correlation is right, increased activity should actually reduce spawn density in that area."

  Kim looked at Marcus. "You believe that?"

  Marcus thought about Julie's insight - the connection between activity metrics and entity spawning patterns. "I think there's a strong correlation. Can't prove causation yet, but the data supports it. Entities seem to spawn more frequently in low-activity zones. If we're raising the activity index, we're probably lowering spawn probability."

  "Probably." Kim's tone was dry. "Love that scientific precision."

  "I can give you a confidence interval if you want actual precision," Marcus said. "But the short version is: yes, I think you'll be safer in an actively occupied building than in an abandoned one."

  Kim studied him for a moment, then nodded. "Good enough. I'm in."

  Around the room: more nods, resigned acceptance, the particular look of people who'd decided that boring and productive beat exciting and dead.

  "Any other concerns?" Chen asked.

  "Yeah." That was Rivera. "What happens if this doesn't work? If the building stays flagged despite everything we're doing?"

  "Then we learn that the System uses different criteria than we thought," Marcus said. "And we adjust the theory accordingly. But either way, we get data. Either way, we're better off than we were before the test."

  "You QA types really love your data," Santos muttered.

  "Data's kept me alive this long," Marcus said. "Seems like a reasonable thing to love."

  A few smiles at that. The tension in the room eased slightly.

  Chen pulled up something on his laptop. "Shift assignments are posted outside. First shift starts at noon - that's Rivera, Patterson, Santos, and Okoye." He nodded at a tall Nigerian woman, Level 4 Scout, who'd been quietly observing from the side. "Bring personal gear, something to do during downtime, and patience. This is going to be boring."

  "Boring is good," Rivera said. "Boring means not dying."

  "Boring means the test is working," Marcus corrected. "If the building stays flagged despite activity, that's interesting. If it drops off the queue, that's exploitable. Either way, we're advancing our understanding of System mechanics."

  "Always with the mechanics," Santos said, but he was smiling slightly now.

  Patterson spoke up. "Webb. Level with us. What are the actual odds this works?"

  Marcus considered the question - really considered it, not just politically. "If I'm right about the System treating structures like cloud resources? Seventy, maybe eighty percent. The timestamp pattern's too consistent to be anything but automated maintenance. The activity correlation is strong enough that it's probably causal, not just coincidental. But there's always the chance I'm pattern-matching noise."

  "And if you're wrong?"

  "If I'm wrong, you spend two days in an uncomfortable building doing make-work, and then we all watch it disappear at the scheduled time. Worst case: we're no worse off than if we'd done nothing. Best case: we just saved the first structure in the history of the System apocalypse." Marcus met Patterson's eyes. "I think those odds are worth taking."

  Patterson nodded slowly. "Fair enough. I've gambled on worse."

  Chen dismissed them, and the room emptied in a controlled exodus of people heading to gather gear and mentally prepare for forty-one hours of structured boredom.

  Marcus was left standing next to a whiteboard where someone had optimistically written "DATA COLLECTION METHODOLOGY" in neat block letters.

  "You did good," Chen said. "They trust you. That's not easy to earn in a week."

  "They trust you," Marcus corrected. "I'm just the weird analyst you're vouching for."

  "Take the win, Webb." Chen pulled up something on his laptop. "Now let me show you the monitoring station setup."