The southeast storage buildings were exactly as advertised: low-traffic, utilitarian, and deeply uninspiring.
Building A was a converted maintenance shed, maybe forty feet by twenty, with reinforced doors and no windows. Building B was slightly larger, brick construction, also windowless. Building C was the biggest - probably an old classroom building before the cluster repurposed it for storage.
All three were dark. No power. No movement. Just locked doors and the sense that nobody had cared about them in days.
Perfect candidates for deprecation, Marcus thought.
He pulled out his clipboard and started logging.
BASELINE OBSERVATION - Southeast Storage Cluster
Date: Day 23
Time: 8:47 AM local
Weather: Clear, mild
Building A:
- Dimensions: ~40' x 20'
- Status: Dark, no power consumption detected
- Foot traffic (past hour): 0
- Last recorded access: Day 20 (per cluster logs)
- Entity proximity: 1 Glitch Spawn observed at 80m distance
Maya was doing a perimeter sweep, her Threat Detection skill active. Marcus could see the faint shimmer around her vision that indicated an enhanced sense running.
"Spawn density is higher here than the cluster average," Maya reported. "Not dramatically, but consistent. I'm reading three contacts within a hundred meters. One Spawn, two Stalkers on rooftops."
Marcus made notes. Entity proximity correlated with low activity, just like the pattern he'd identified in the data. Just like Julie had predicted.
Owen had positioned himself between the team and the nearest entity contact, relaxed but alert. "If we're staying here longer than ten minutes, we should rotate position. Stalkers don't like static targets."
"Acknowledged," Julie said. She was taking measurements with something that looked like a jury-rigged multimeter. "No power draw on any of the three buildings. Totally inactive."
Marcus photographed each building, documenting current state. Visual records for comparison later.
They spent forty-five minutes collecting baseline data. Power consumption: zero. Foot traffic: zero. Entity proximity: elevated. Temperature differential: Buildings were cooler than ambient, suggesting no internal heat generation.
When they finished, Marcus had three pages of observation notes and a clear picture: these buildings were as close to "abandoned" as you could get without actually being deprecated.
If the theory was right, they were probably already in the queue.
Which means we might already be too late, Marcus thought. But we won't know until we test.
They returned to the library to deliver the data to Chen.
Chen read through Marcus's notes with the intensity of someone evaluating a production deploy.
"Good documentation," he said finally. "This is baseline. Now we implement phase one."
He pulled up a schedule on his laptop. "Starting today, we're running the following operations through the southeast storage buildings:
- Daily inventory checks (minimum one hour per building)
- Guard rotation expansion (perimeter sweeps every four hours)
- Power activation (emergency lights on 24/7)
- Foot traffic increase (mandatory runner routes through the area)
All activity gets logged. Same measurement categories you used for baseline. We run this for seventy-two hours and monitor whether deprecation notices appear."
"Seventy-two hours," Marcus repeated. "That's one full processing cycle."
"Exactly. If the System runs its deprecation cron job at 11:47 UTC daily, we'll catch at least three processing windows. If the buildings stay off the deprecation list, we've validated the theory."
"And if they don't?"
"Then we learn we were wrong and try something else." Chen closed the laptop. "But I don't think we're wrong. The pattern's too consistent. The correlation's too strong."
Julie was studying the timeline. "What if the buildings are already flagged? Already in the queue?"
"Then we find out if activity increases can remove structures from the queue before the notice goes out. That's phase two of the test." Chen looked at the team. "This is good work, all of you. Now we just have to wait and see if the System agrees."
Waiting was the hard part.
That night, Marcus joined Maya for a perimeter check. Chen had insisted someone from the analysis team observe the activity protocols in action, and Marcus wanted to see the implementation firsthand.
They met at the southeast barricade at 9 PM, just as the guard rotation was hitting the storage buildings.
"First time doing night recon?" Maya asked as they headed toward Building A.
"First time doing any kind of recon," Marcus admitted. "My skill set is more 'sit at desk, find bugs' than 'walk around in the dark looking for murder monsters.'"
Maya grinned. "The monsters are the same. Just different debugging tools."
The storage building area was transformed. Emergency lights blazed from all three structures, casting harsh pools of illumination across the formerly dark sector. Two guards were doing a slow patrol between Buildings B and C. Someone had propped open Building A's door, and Marcus could see people inside - inventory team, moving crates, calling out item counts.
"Power consumption: high," Marcus noted on his clipboard. "Active occupancy: confirmed. Foot traffic: continuous."
Maya had her Threat Detection active, eyes scanning the rooftops. "Entity count is down. I'm only reading one contact now - single Stalker, about a hundred twenty meters out. It was three this morning."
"They're responding to the activity," Marcus said, making notes. "Just like Julie predicted."
"Your Synthesist friend is scary smart," Maya observed. "She connected patterns I'd been staring at for weeks."
They walked the perimeter, Marcus logging observations while Maya tracked threats. The guards acknowledged them with nods but didn't break routine. The inventory team inside Building A was working through boxes with efficient precision, someone reading off a checklist while others verified contents.
It looked mundane. Boring, even. Just people doing inventory in a well-lit building.
But to the System, it was data. Signals that said: this infrastructure is in use. This structure serves a function. Don't deprecate it.
"This is going to work," Maya said quietly. "I can feel it. The entities are already pulling back. The System is noticing."
Marcus wanted to share her confidence. But he'd spent too many years in QA watching promising tests fail for unexpected reasons.
"We'll know in forty-eight hours," he said. "First processing window is tomorrow morning."
They completed the circuit and headed back to the library. Behind them, the storage buildings blazed with light and purpose, no longer abandoned, no longer at risk.
Probably.
Marcus spent the next day and a half monitoring the southeast buildings, logging activity, checking entity density. The cluster's operations ran like clockwork: inventory checks at 9 AM, 3 PM, and 7 PM. Guard rotations every four hours. Lights on continuously. Enough foot traffic that the buildings no longer felt abandoned.
Support the creativity of authors by visiting Royal Road for this novel and more.
And slowly, measurably, the entity density continued to drop.
By Day 24, Maya was reporting only one Stalker in the hundred-meter radius, down from two. The Spawn had moved on entirely.
"They're responding to the activity," Maya said during the evening check-in. "It's subtle, but consistent. Entities are avoiding the area."
Marcus documented it. Another data point. Another piece of evidence that the System's various subsystems were connected - that entity spawning and structure maintenance referenced the same activity metrics.
Julie was already building a theoretical model, connecting deprecation risk to entity deployment patterns. Marcus could see her notebook filling with diagrams and equations, the Synthesist brain making connections at a speed that would've impressed Chen.
On Day 25, at 5:47 AM local time (11:47 UTC), Marcus was awake.
He'd set an alarm. Couldn't sleep anyway, knowing the deprecation cron job was about to run. This was the first processing window since they'd implemented the activity increases.
His deprecation timer read 38 hours, 12 minutes. Still counting down. Still red text in his vision.
At 5:47 exactly, Marcus felt... nothing.
No notification. No shimmer. No status update.
He waited five minutes. Ten. Fifteen.
No deprecation notice for the southeast storage buildings.
At 6:00 AM, Marcus sent a message to Chen: No notices. First processing window passed clean.
Chen's response came back immediately: Noted. Continue monitoring.
The second processing window was Day 26, same time.
Marcus was awake again. This time with company - Julie and Chen had joined him at the library, all three watching the clock tick toward 11:47 UTC.
5:47 AM local.
Nothing.
"Two clean cycles," Chen said quietly. "That's not coincidence."
"We need three to validate," Julie cautioned. "Pattern confirmation requires multiple instances."
"I know. But this is looking good."
Marcus's timer read 14 hours, 18 minutes.
His own building - Structure ATX-E-4471 - was hitting its deprecation completion in less than fifteen hours. He'd find out soon whether evacuating had actually saved him or if the System was thorough enough to track users independently of location.
The third processing window was Day 27.
This time, half the analysis team was watching. Word had spread that Webb's theory was holding up. People were invested in the outcome.
5:47 AM local. 11:47:23 UTC.
Marcus's status screen flickered.
His heart stopped.
But it wasn't a deprecation notice for the buildings.
[Status Effect Removed: Deprecated]
Structure ATX-E-4471 has completed end-of-life processing.
User status updated: ACTIVE
Deprecation timer cleared.
Marcus stared at the notification for five full seconds.
The timer was gone. The red text was gone. His status just read "Active" with no countdown, no warning, no threat of garbage collection.
He'd survived.
Then a new notification appeared:
[Status Update: Pattern Logged]
Classification: Displacement Survivor
Note: User evacuated deprecated structure within processing window.
Telemetry: Active
Marcus felt the relief curdle into unease.
The System had flagged him. Not deprecated, not penalized, just... noted. Logged his behavior. Marked him as someone who'd exploited the evacuation window successfully.
It was like having a security camera zoom in on you. Not hostile. Just watching.
"Webb?" Chen was watching him. "You okay?"
"My deprecation cleared. I'm not deprecated anymore." Marcus hesitated. "But the System logged me. New status: Displacement Survivor. Says my telemetry is active."
Chen's expression darkened. "It's tracking you."
"Tracking something. Don't know what yet." Marcus pulled up his full status screen, checking for other changes. Everything else looked normal. Just the new classification sitting there like a permanent record.
"That's a cost," Julie said quietly. "You survived, but now the System's paying closer attention."
Marcus nodded. She was right. He'd beaten the deprecation timer, but the win wasn't clean. The System had noticed. And it was keeping notes.
But the southeast storage buildings had passed three processing cycles with no deprecation notices.
"Three clean cycles," Chen said, reading the same conclusion. "The theory holds. Activity metrics prevent deprecation. We can protect the cluster's infrastructure."
Julie was already pulling up a map. "We need to identify other at-risk structures. Implement activity protocols before they hit the queue."
"Agreed. Webb, I want full documentation of the test methodology and results. Publish it to the cluster network so other settlements can replicate it."
Marcus nodded, still processing the Pattern Logged status. He wasn't deprecated. He had more than fourteen hours - he had all the time. No countdown. No timer.
But the System was watching him now.
I beat the System, he thought. Found the bug. Exploited it. Survived. And it noticed.
That should've felt like a complete win.
It didn't.
Marcus spent the rest of Day 27 writing up the documentation, trying not to think too hard about what "Telemetry: Active" meant for his future.
EXPLOIT CONFIRMED: Activity-Based Deprecation Prevention
SUMMARY: Structure deprecation can be prevented by maintaining activity levels above System monitoring threshold.
MECHANISM: System monitors building activity metrics (user presence, power consumption, foot traffic). Structures falling below threshold are flagged and added to deprecation queue. Queue processes daily at 11:47:23 UTC.
REPRODUCTION:
- Identify at-risk structure (low activity, minimal user presence)
- Implement activity protocols:
- Regular user access (minimum 1 hour per day recommended)
- Power activation (lighting, equipment)
- Foot traffic increase
- Monitor for 72 hours across 3 processing cycles
RESULTS: 100% success rate in test sample (3/3 buildings avoided deprecation)
APPLICATIONS:
- Cluster infrastructure protection
- Expansion into new territory
- Possible reversal of flagged-but-not-yet-noticed structures
LIMITATIONS:
- Unknown: exact threshold values
- Unknown: whether activity can remove structures already in queue
- Untested: whether exploit works for user deprecation (suspected similar mechanism)
- No control building (all at-risk structures were treated)
- Cannot definitively prove causation vs correlation
- Sample size: 3 buildings (insufficient for statistical significance)
Note: A true control would require letting a building deprecate to prove our intervention made the difference. Given resource constraints and lives at stake, this trade-off was accepted. Circumstantial evidence strongly supports the theory.
ADDITIONAL FINDINGS:
- Entity spawn density inversely correlates with activity metrics (credit: J. Tran)
- High-activity zones show measurably lower entity presence
- Implications for strategic territory control
STATUS: VALIDATED, HIGH VALUE, READY FOR DEPLOYMENT
He published it to the cluster's local network. Within an hour, Chen had forwarded it to the other Austin settlements. By evening, someone had posted a printed copy on the dining hall bulletin board.
Marcus Webb, Level 2 Logician, had officially contributed something useful to humanity's survival.
His quest progress updated:
[Quest Progress: Integration Assessment]
Progress: 3/3 mechanics identified, 1/1 exploit documented, 72/72 hours survived
QUEST COMPLETE
Rewards:
- +500 XP [Level up! You are now Level 3]
- Skill upgraded: Anomaly Detection v0.9.2-beta
- Development Credits: +250
- New trait unlocked: System Analysis (Passive)
Marcus felt the Level 3 transition hit - a strange sensation, like his brain was being defragmented. His INT bumped from 16 to 18. His PER increased to 16. And a new skill appeared in his status screen:
System Analysis (Passive) v1.0.0
Provides limited insight into System mechanics when examining structures, entities, or interface elements. Displays incomplete data that may prompt further investigation. Skill accuracy improves with level progression.
Cost: None (always active)
Cooldown: None
Note: v1.0.0 analysis may contain errors or ambiguous results. Verify findings through testing.
Marcus blinked at the description. A debugging tool that admitted it might be wrong. That was almost refreshing in its honesty.
He looked at the library's emergency lights, and information flickered into his perception:
Structure Analysis: Emergency Lighting System
Status: Active (backup power detected)
Maintenance: [DATA UNCLEAR - suggest weekly?]
Deprecation Risk: LOW
Not complete information. Not perfect. Just... hints. Prompts for questions.
Why is deprecation risk low? What makes it low versus high?
What counts as "maintenance"? Physical inspection? Power cycling? Something else?
The skill raised more questions than it answered.
Marcus looked at his own status screen, and the System Analysis tried to parse it:
User Interface Analysis: Status Screen
Framework: [ENCRYPTED - Unity-based suspected]
Update Rate: Real-time (approx. 0.1s refresh?)
Data Source: Backend API (unable to read)
Encrypted. Unable to read. Suspected. Approximate.
The skill was like getting a flashlight in a dark room - it helped you see that there were things to investigate, but it didn't solve the mysteries for you.
Marcus looked at the Pattern Logged status effect, curious what System Analysis would show.
Status Effect Analysis: Pattern Logged
Type: [UNKNOWN - not debuff, not buff]
Effect: Telemetry tracking active
Duration: [DATA UNCLEAR - permanent?]
Trigger: [UNABLE TO DETERMINE]
Note: This effect is being monitored by backend systems. Purpose unknown.
So the System was tracking him, but his shiny new analysis skill couldn't tell him why or what was being tracked. Just that he was flagged.
Helpful. Sort of.
Marcus closed his status screen and rubbed his eyes. They felt like sandpaper. His back ached from hunching over the desk. The coffee someone had brought him an hour ago sat cold and forgotten, but he drank it anyway because caffeine was caffeine.
He had work to do. Questions to answer. More patterns to find.
But right now, he just wanted to sleep. Real sleep. Not deprecation-timer-counting-down anxiety sleep.
Marcus saved his documentation, closed his laptop, and headed back to the dorms.
Behind him, the library's lights stayed on. The cluster's infrastructure hummed. And somewhere in the System's processing queue, three buildings that should have been deprecated remained flagged as active, alive, maintained.
It wasn't magic.
It was engineering.
And Marcus Webb was getting very good at reverse-engineering the apocalypse.
Even if the apocalypse was starting to reverse-engineer him back.