ANTI-USER-MANUAL // v0.ESCALATION
← RETURN TO DASHBOARD
document classification: [███████] // eyes only
ANTI
MANUAL
zombie apocalypse prediction machine
field guide for the operationally misguided
⬥ this document will not save you ⬥
DOCUMENT ID: ZAPM-DOC-001-REV2  |  ISSUED: AFTER THE INCIDENT (UPDATED: AFTER THE SECOND INCIDENT)  |  DISTRIBUTION: RESTRICTED (good luck)
§ 00 — BEFORE ANYTHING ELSE

Remain Calm

ZAPM cannot prevent what it measures.
It can only tell you it is happening, and approximately how fast.

This manual was written with the genuine hope that you will never need it beyond idle curiosity. The model runs regardless. The planet continues.

If the global score reaches 100, please disregard this document and focus on the more immediate matters at hand.

§ 01 — BEFORE YOU BEGIN

What This Is

ZAPM is not a survivalist toolkit.
It is a probability estimator for threats that the academic consensus insists are fictional. If you are reading this because something has already gone wrong, we are genuinely sorry.

The Zombie Apocalypse Prediction Machine aggregates, scores, and cross-references global outbreak signals — epidemiological, environmental, geopolitical — through an autonomous pipeline that runs whether you are watching or not.

It does not predict the future. It does aggregate the present into a single uncomfortable number.

All data is sourced from public RSS feeds, news aggregators, and open epidemiological reports. The model's scenario classifications (including "Viral Recombination", "Fungal Bloom", and "Cosmic Contamination") are statistical constructs — not diagnoses, not prophecies, not your horoscope.
§ 02 — PRIMARY INTERFACE

The Dashboard

The Global Score

A number between 0 and 100. Zero means nothing unusual. One hundred means everyone on the response team has stopped answering messages.

0–24
Baseline
Normal background noise. The internet is merely anxious. Go about your day.
25–49
Elevated
Something is being watched. Multiple regions are mentioned in the same sentence as "unusual mortality." Keep snacks accessible.
50–74
High Alert
Signals are converging. The model is nervous. You should probably check in with family members you haven't spoken to recently.
75–100
Critical
Do not panic. Panic impairs motor function. Pack methodically. Leave a note.

The Trend Indicator

↑ Increasing
DETERIORATING Score has risen compared to the previous cycle. Not ideal.
— Stable
HOLDING Score unchanged. The situation is not getting worse. It is also not getting better.
↓ Decreasing
CONTAINMENT Score declining. Either the threat is receding or our data sources went offline.

The Onset Phase Window

This badge describes the current phase classification of global threat signals — not a mood, not a weather forecast.

💤 Dormant
Score < 20. Routine. The planet is fine, relatively speaking.
🔍 Early Indicators
Score 20–40. Weak signals are accumulating. Things worth knowing about are happening somewhere.
⚡ Active Phase
Score 40–70. Multiple signal categories are co-activating. The model has opinions.
🚨 Critical
Score > 70. We do not have a comforting statement for this one.

Model Confidence

How certain the model is about its own score — not how certain you should be. A confidence of 89% means the model is consistent with itself. It says nothing about whether it is correct.

Confidence is built from: article volume, signal coverage breadth, key indicator strength, cross-region scenario convergence, and escalation momentum. Below 50% it is largely speculating. Above 85% it has convictions.

§ 03 — THREAT TAXONOMY

The 11 Scenarios

Every signal is classified against one of eleven threat archetypes. These are not predictions. They are probability buckets the model pours evidence into.

🦠 Viral Recombination
Novel pathogen emergence, spillover events, unusual hemorrhagic or neurological presentations. The classic.
🍄 Fungal Bloom
Cordyceps-adjacent. Environmental disruption enabling parasitic fungal colonization of neural tissue. More plausible than you'd like.
☣ Lab Escape
Containment failure. Enhanced pathogen release. The paperwork would be extraordinary.
🐛 Parasitic Dominance
Behavioral modification via parasitic infection. Toxoplasma at scale.
☢ Radiation Mutation
Ionizing radiation exposure triggering novel pathologies. Usually low probability. We check anyway.
🧬 Synthetic Bioagent
Engineered organisms or pathogens released intentionally or accidentally. This one keeps the analysts up at night.
🔬 Prion Cascade
Misfolded protein propagation. Slow, untreatable, and entirely real. Mad cow disease was a preview.
💊 Pharmaceutical Vector
Drug interactions, contaminated supply chains, or mass medication events producing neurological effects.
🌱 Environmental Toxin
Industrial chemicals, heavy metals, or biotoxins affecting cognition or motor control at population scale.
📡 Cascading Collapse
Infrastructure failure cascading into social breakdown. Not a pathogen. Equally concerning.
🌌 Cosmic Contamination
Exogenous biological material. Panspermia. The model includes it because completeness demands it. Its probability is usually 0–3%.
confidence

Each scenario has a confidence score. This reflects how much independent, corroborated evidence the model has found — not how alarmed you should be. A scenario at 3% with 85% confidence is well-established at 3%. A scenario at 40% with 12% confidence is a rumor the model is taking seriously.

§ 04 — GEOGRAPHIC INTELLIGENCE

The Map

Fifteen world regions are scored independently. Each marker on the map is a risk circle — larger and redder means more concentrated signal density in that region.

Marker States

📍 Origin Zone
Dashed amber ring. The model calculates this region as the most probable source of the primary scenario. This is a statistical inference, not a confirmed event.
⚠️ Escalation Watch
Red solid ring. Rapid score escalation detected in this region — rate of change is high regardless of absolute score.
No ring
Standard monitoring. Nothing exceptional. Yet.

Popup Contents

Tap any marker to see the region's score, current phase, and its two highest-probability threat scenarios. The ① primary scenario is the model's top attribution for this region. The ② secondary is the next most plausible explanation for the signals detected.

The secondary scenario is currently derived from global rankings rather than per-region article attribution. This means all regions share the same secondary candidate. Per-region secondary scoring will improve as the normalization pipeline populates.
§ 05 — SIGNAL FEED

Reading Signals

Signals are the raw intelligence layer — individual articles, reports, and dispatches that the model has ingested and scored. They are not curated by humans. They are pulled from RSS feeds every 30 minutes.

Relevance Badges

Critical 80%+
High keyword density + credible source + recent. The model found this very relevant to its current scenario assessments.
High 60–79%
Strong indicators present. Probably worth a read if you're tracking the primary scenario.
Moderate 35–59%
Partial signal match. The model included it but holds its judgment.
Low <35%
Marginal relevance. The pipeline ingested it anyway. Nothing is discarded — absence of evidence is also evidence.

Reliability Icon

The shield icon (🛡) next to each source indicates source credibility tier. Green = established credentialed outlet. Red = unknown domain or low-signal aggregator. Gray = neutral / unclassified. Source reliability adjusts the relevance score multiplicatively.

§ 06 — HISTORICAL REFERENCE

Outbreak Benchmarks

The Historical Trends page includes reference curves from documented outbreaks. Toggle them to compare the current ZAPM risk trajectory to how these events looked at equivalent points in their development.

COVID-19 (Mar 2020)
Benchmark for rapid exponential global spread. Peaked ~Day 35. The one everyone remembers.
Ebola (Aug 2014)
West Africa epidemic. High CFR, moderate spread velocity, eventually contained. Peak ~Day 40.
H1N1 (Apr 2009)
Swine flu pandemic. Fast initial spread, lower severity than feared. Peak ~Day 21.
Mpox (May 2022)
Multi-country sustained spread. Slower onset than COVID, longer plateau. Contained via vaccination.
Nipah Kerala (Sep 2023)
Localised, extremely high CFR (>70%). Rapid spike, rapid containment. Short curve, brutal peak.
Marburg EG (Feb 2023)
Equatorial Guinea. Moderate scale, contained within ~60 days. Valuable hemmorhagic fever comparator.
H5N1 Avian Flu (2024–25)
Slow-burn. Dairy herd spread + sporadic human cases. Trajectory still rising as of last update. The one to watch.
All curves are approximate risk-equivalent scores derived from published WHO/CDC situation reports. They represent what ZAPM's model would have scored at those dates, not official data.
§ 07 — FEED HEALTH

Data Ingestion

The Feeds page shows the operational health of every RSS source the model monitors. A dead feed is a blind spot.

New scoring cycles run on a scheduler. If the score has not updated in over an hour, check the Feeds page for systemic failures. If all feeds are healthy, the scheduler may have encountered an internal error.

§ 08 — OPERATIONAL GUIDANCE

What Not To Do

§ 09 — EVIDENCE ARCHAEOLOGY

What Triggered This

Every scenario probability has a paper trail. The What triggered this? panel is the evidentiary layer beneath the score — the articles the model read, the keywords it flagged, and how much weight it assigned to each source before reaching its uncomfortable conclusions.

The model does not speculate.
It aggregates. Then it speculates. Then it gives the speculation a percentage.

What Triggered This shows you the raw material behind the percentage.

How To Access It

Open the Scenarios page. Expand any scenario card. At the bottom of the expanded view, click "What triggered this?" The panel will load the articles ingested in the last 48 hours that the model attributed to that scenario — sorted by relevance, most incriminating first.

Reading The Evidence

Relevance Bar
How strongly the model weighted this article against the scenario. Red = high weight. Amber = notable. Green = marginal. The model included marginal articles regardless. Completeness is non-negotiable.
Keyword Tags
The specific terms that caused the article to be attributed to this scenario. These are the words the model is watching for. If you recognise any of them from your local news, keep reading.
Source & Timestamp
Where the article came from and when. Recency matters. The model weights recent articles above older ones. If all your triggers are from 36 hours ago, the scenario may already be evolving.
Articles shown are from the past 48 hours only. If a scenario has high probability but no articles appear in the panel, it is weighted by historical baseline and weak-signal accumulation rather than fresh evidence. The model has a long memory. It does not forget the week things started moving.

What To Do With This Information

Read the articles. Follow the links. Make your own assessment. The model will not be offended if you disagree with it. It will continue to be correct at whatever percentage it has already committed to.

§ 10 — PERIODIC INTELLIGENCE SUMMARY

The Weekly Forecast

Located under WKLY in the navigation bar. The Weekly Forecast is a seven-day summary of what the model observed, what it concluded, and which regions and scenarios were responsible for making the week uncomfortable.

It is not a comfort document.

What Is In It

Global Score Summary
Average score, peak score, and seven-day change. If the change is positive, this week was worse than last week. If the change is negative, things are improving. Do not celebrate. Improvement is temporary.
Top Scenario
The threat archetype with the highest average probability across all snapshots in the period — not just the most recent reading. Consistency across 7 days is more significant than a single spike.
Most Active Region
The geographic zone that sustained the highest average risk score over the week. A region at the top of this list for three consecutive weeks warrants attention.
Notable Escalation
The largest single score jump observed between consecutive snapshots. This is where things moved fastest. An escalation of +15 or more in one cycle is the model telling you it noticed something.
the card

Downloading The Forecast Card

The Download PNG Card button generates a 680×400 pixel image containing the week's key metrics in a format suitable for sharing. It is drawn on a canvas element in your browser — no data leaves the application to produce it.

The card is optimised for readability in the thumbnail previews generated by messaging applications, social platforms, and crisis management group chats.

The weekly card includes the phrase "All scenarios theoretical — not medical advice" in the footer. This text is not decorative. If you are sharing this card in a context where someone might act on it as medical guidance, reconsider your distribution list.

Frequency

The forecast refreshes automatically with each ingestion cycle. The data reflects whichever seven-day window ends at the most recent snapshot. There is no fixed "publication day." The apocalypse does not observe a publishing schedule.