PROJECT IKELOS — CONCEPTUAL DOCUMENT

WORKING DOCUMENT

IKELOS/CONCEPTUAL/1.0
ISSUED: 12 MAY 2026
STATUS: OPEN / DRAFT

PROJECT

IKELOS

Surveillance · Classification · Escalation

KOLMOGOROV INITIATIVES LLC
GLASGOW SHIPYARD
CC0 — NO RIGHTS RESERVED

00Preamble

IKELOS was Rasputin's subroutine for dreaming — a protocol designed to see patterns in noise before the noise became signal, to read what the data implied rather than what it stated. It operated during MIDNIGHT EXIGENT, when conventional response had been abandoned, when the system had accepted that the threat could not be stopped and shifted to the problem of what comes after.

This project takes the name in its literal function. The premise: most catastrophic outbreak events are not surprises. The signals exist in the period before institutional recognition. They are present in endemic surveillance data, in retrospective contact chains, in the gap between what a pathogen is classified as doing and what it is observed doing. The failure is not absence of signal. It is absence of a system positioned to read it.

IKELOS is that system. It does not predict. It reads.

00.1Contents

01Theoretical Basis — Blank Swan / Surveillance Bias

02Pathogen Classification Framework — PCF/1.0

03Rasputin Escalation Tiers

04Designation Syntax — Combined Framework

05Historical Index

06Active Case Study — ANDV / MV Hondius 2026

07Technical Architecture — Live Surveillance System

08Project Structure — Claude Code

01Theoretical Basis

Blank Swan / Surveillance Bias

The Blank Swan Principle

A Blank Swan is an event commonly classified as a Black Swan — unpredictable, unprecedented — that was in fact predictable through locally available weak signals. The failure is not epistemic. It is structural: the signals existed but no system was positioned to aggregate and weight them correctly. The event appeared random from inside the prevailing model while being legible from outside it.

The Black Swan framing exculpates the institutions that missed the signals. The Blank Swan framing does not. It asks: what was observable, when was it observable, and why was it not acted on?

Applied to outbreak surveillance, the Blank Swan principle generates a specific research posture. Before accepting that an event was genuinely novel, IKELOS asks whether the novelty is real or whether it is an artifact of the detection infrastructure. The two failure modes are distinct:

Genuine novelty — the pathogen is doing something it has never done before. Rare. Requires evolutionary change or novel transmission context.

Detection artifact — the pathogen has always done this, but the surveillance context only recently produced conditions in which it was observable. Common. Produces false classification that persists until a detection event forces revision.

The ANDV case is the canonical detection artifact. Andes virus was classified as HECATE — rodent-to-human spillover, no H2H — for decades. The Epuyen 2018 cluster forced reclassification to SPHINX. But Epuyen was not the first instance of P2P transmission. It was the first instance detected in a context with sufficient surveillance infrastructure to make the chain legible. Rural Andean communities, where ANDV is endemic, systematically underdetect because: high early mortality before tracing; flu-like prodrome causing misdiagnosis; limited healthcare infrastructure; household contact chains are the dominant social unit, so household transmission is what gets reconstructed. The "prolonged contact" doctrine that governed ANDV classification was not derived from transmission biology. It was derived from who gets detected.

IKELOS treats surveillance context as a variable in the classification, not a given. When a pathogen's documented behavior derives primarily from a single geographic or social context, that classification is provisional until confirmed across heterogeneous detection environments.

02Pathogen Classification Framework

PCF / 1.0 — Transmission-Primary

A pathogen's most operationally significant characteristic is how it moves. Severity, mortality, and reservoir origin are consequential — but they are downstream of transmission. The PCF assigns a single primary classification based on dominant transmission mechanism. Where mechanism is unknown or violates established parameters, SPHINX is assigned rather than forcing a fit. Classifications track behavior and are revised as understanding develops.

CLASS

DEFINITION

EXAMPLES

HERMES

Moves faster than response. Cannot be stopped at the threshold.

Airborne or droplet. Sustained H2H. Behavioral intervention alone insufficient. Vector is air — boundary cannot be sealed. Pre-symptomatic transmission common.

SARS-CoV-2
Influenza (pandemic)
Measles (R₀ 12–18)
Tuberculosis

CHARON

Nothing crosses without the carrier. Control the ferry.

Obligate biological vector — arthropod, insect. Direct H2H absent or negligible. Geographic range determined by vector habitat. Tractable if vector is suppressed.

Malaria
Dengue
Lyme disease
West Nile

EROS

Cannot act at a distance. The crossing demands contact.

Direct contact — sexual, bloodborne, sustained physical proximity. Cannot cross space. Behavior is the primary containment lever. Contact networks are reconstructible.

HIV
Ebola (standard)
Mpox 2022
Hepatitis B/C

HECATE

Exists at the boundary between worlds. The crossing point is the danger.

Zoonotic spillover. Lives in animal reservoir. H2H limited or absent. Threat is recurrent spillover, not sustained chain. Containment focuses on the species interface.

Rabies
H5N1 (current)
MERS-CoV
Nipah

SPHINX

Poses the question the system cannot yet answer.

Mechanism unknown, contested, or empirically violating established class parameters. Active designation — demands investigation, not assumption. Applying a prior class to a SPHINX pathogen is a systemic failure mode.

ANDV/Hondius 2026
SARS-CoV-2 (early)
Epuyen ANDV 2018
H5N1 (if H2H confirmed)

State Transitions

A pathogen may transition between classes as understanding develops or as the pathogen evolves. Downgrading from SPHINX requires empirical resolution of the contested mechanism — not passage of time, not institutional consensus absent new data. Upgrading to SPHINX from any other class requires a documented transmission event that violates the prior classification's criteria.

// Documented transitions
SARS-CoV-2 SPHINX (Jan 2020) → HERMES (Mar 2020) → HERMES/endemic
ANDV HECATE (pre-2018) → SPHINX (Epuyen) → SPHINX (Hondius)
H5N1 HECATE (current) → SPHINX (if sustained H2H confirmed)
Mpox HECATE → EROS (2022 outbreak established sexual network transmission)

// Transition triggers
→ SPHINX Any documented event outside prior class parameters
← SPHINX Mechanism resolved with reproducible evidence

03Rasputin Escalation Tiers

Warmind Directive Hierarchy

The following tiers are drawn from canonical Warmind lore — Ghost Fragment: Rasputin 3 and 5, Grimoire Card: Rasputin. They are applied here as an escalation framework sitting above PCF. PCF describes how a pathogen moves. The Rasputin tier describes what that implies for response posture.

SCRY OVERSIGHT

CANONICAL — Rasputin observation protocol · submode: SILENT VELES (covert watch)

Passive surveillance. Known pathogen, known parameters, behavior within modeled bounds. IKELOS is watching but not escalating. SILENT VELES applies when the observation target should not be aware of monitoring — relevant for pathogens where premature declaration would trigger behavioral changes that complicate tracing.

TRIGGER Known class, endemic or controlled, no parameter violations detected
EXAMPLES Seasonal influenza. Endemic malaria. Controlled HIV. Routine HECATE-class spillover with no H2H.

SKYSHOCK: INSIDE CONTEXT

CANONICAL — Hostile arrival of expected type and scale

A known pathogen class performing as modeled but at threshold scale. The system knows what this is and has frameworks to address it. Response is activation of existing protocols rather than novel response design. The threat is real but legible.

TRIGGER Known PCF class, behavior within parameters, but magnitude requires active response
EXAMPLES Large HERMES-class epidemic with known pharmaceutical intervention. Ebola outbreak requiring EROS containment protocols at scale.

SKYSHOCK: OUTSIDE CONTEXT

CANONICAL — Hostile arrival of unprecedented type/scale that could not be planned for

A pathogen breaking its own classification. Prior doctrine insufficient. Existing response frameworks do not apply without modification. The system's models are wrong and must be revised in real time. This is SPHINX territory — the designation and the tier align. The danger is that institutional response continues to apply inside-context protocols to an outside-context event.

TRIGGER PCF class violated. Documented transmission outside prior parameters. SPHINX designation assigned or under consideration.
CURRENT ANDV / MV HONDIUS 2026 — SPHINX CLASS, SKYSHOCK: OUTSIDE CONTEXT

TEILHARD: TRAUMATIC CONTEXT

CANONICAL — Event category; Teilhard de Chardin: evolutionary rupture, permanent trajectory change

The world after is not the world before. No return to prior baseline. A phase transition, not a severe outbreak. The systems, behaviors, and assumptions that existed before the event are permanently altered. This tier is typically assigned retrospectively — the transition is often not recognized in real time.

TRIGGER Sustained community transmission of a high-CFR pathogen with no pharmaceutical intervention, or confirmed systemic failure of containment infrastructure
EXAMPLES COVID-19 crossed into TEILHARD retrospectively, March 2020. ANDV reaches TEILHARD if secondary community transmission confirmed from repatriated passengers.

YUGA SUNDOWN

CANONICAL — Condition that cancels all protocols for protecting humanity. Activates MIDNIGHT EXIGENT.

Counterattack protocols failing. Active response infrastructure overwhelmed or structurally ineffective. VOLUSPA and FENRIR HEART reporting negative effect. The system is no longer operating under the assumption that the threat can be contained — it is operating under the assumption that the threat cannot be stopped and a different problem must now be solved.

TRIGGER No pharmaceutical intervention available + sustained HERMES-class or evolved SPHINX spread of CFR >15% pathogen + containment traction lost
EXAMPLES No current active designation. Historical: 1918 influenza approached YUGA SUNDOWN conditions before natural immunity terminated the wave.

MIDNIGHT EXIGENT

CANONICAL — Long-term counterattack protocol. Moral structure change. CAUTERIZE. DISPERSE. ESTIVATE.

Civilization kill event in progress. Population protection objectives cancelled. The system stops trying to win and starts trying to survive. CAUTERIZE: seal what can be sealed. DISPERSE: distribute critical assets and knowledge. ESTIVATE: preserve capability for reactivation. MIDNIGHT EXIGENT is not a response to a bad outbreak. It is a response to civilizational-scale mortality where the prior world is not recoverable.

TRIGGER Hard civilization kill event underway. YUGA SUNDOWN declared. Tactical morality formatted for MIDNIGHT.
THRESHOLD No current pathogen warrants this designation. Threshold requires TEILHARD-class event with subsequent YUGA SUNDOWN failure of all countermeasures.

ABHORRENT IMPERATIVE

CANONICAL — Last resort. Secret protocol. Coerce pseudoaltruistic defensive action by threatening what will not act otherwise.

Beyond MIDNIGHT EXIGENT. The system has identified an asset or institution capable of preventing civilizational collapse that is not acting — and forces it to act by threatening its own survival. In outbreak context: the activation of emergency powers, compelled disclosure, or coercive international coordination that would not otherwise occur. Not a response protocol. A forcing function.

TRIGGER MIDNIGHT EXIGENT active. Identified actor has capacity to intervene but is not. All cooperative mechanisms exhausted.

04Designation Syntax

Combined Framework

A full IKELOS designation combines the Rasputin tier (threat posture) with the PCF class (transmission mechanism). The tier tells you how seriously to respond. The class tells you how to respond. Together they produce an unambiguous operational picture.

SKYSHOCK: OUTSIDE CONTEXT

SPHINX CLASS: ANDV

EVENT: MV HONDIUS 2026 · ISSUED: 04 MAY 2026 · STATUS: ACTIVE
PRIOR CLASS: HECATE · RECLASSIFIED: SPHINX (EPUYEN 2018, CONFIRMED P2P)
TRIPWIRE: CONFIRMED SECONDARY COMMUNITY TRANSMISSION → TEILHARD

TEILHARD: TRAUMATIC CONTEXT

HERMES CLASS: SARS-COV-2

EVENT: COVID-19 PANDEMIC · ISSUED: MAR 2020 (RETROSPECTIVE) · STATUS: RESOLVED/ENDEMIC
PRIOR CLASS: SPHINX (JAN 2020) · RECLASSIFIED: HERMES (MAR 2020)
CURRENT POSTURE: SCRY OVERSIGHT · HERMES/ENDEMIC

SCRY OVERSIGHT: ELEVATED

HECATE CLASS: H5N1

EVENT: H5N1 ONGOING · ISSUED: CONTINUOUS · STATUS: MONITORING
WATCH CONDITION: SUSTAINED H2H CONFIRMATION → IMMEDIATE SPHINX RECLASSIFICATION + SKYSHOCK:OC

05Historical Index

Designated Outbreaks

Pathogen / Event	PCF Class	Rasputin Tier	Notes
ANDV / MV Hondius 2026	SPHINX ★	SKYSHOCK:OC	Active. P2P mechanism unresolved. Prior HECATE doctrine violated.
SARS-CoV-2 / COVID-19 (peak)	HERMES	TEILHARD	Entered as SPHINX Jan 2020. HERMES confirmed Mar 2020. Permanent baseline change.
SARS-CoV-2 / COVID-19 (endemic)	HERMES	SCRY OVERSIGHT	Monitoring. Seasonal variant watch.
ANDV / Epuyen 2018	SPHINX	SKYSHOCK:OC	P2P at 4ft proximity, 90min. Restroom crossing = 1 case. R₀ 2.12 pre-isolation. Prior HECATE violated. Resolved without reclassification — SPHINX maintained.
Ebola / West Africa 2014	EROS	TEILHARD	Contact transmission. Behavioral intervention sufficient. Scale crossed TEILHARD threshold regionally.
Mpox / Global 2022	EROS	SKYSHOCK:IC	Reclassified HECATE → EROS when sexual network amplification confirmed. Inside context once reclassified.
Influenza A H1N1 / 2009	HERMES	SKYSHOCK:IC	Known class, expected behavior. Resolved — seasonal.
H5N1 / Current	HECATE	SCRY OVERSIGHT ↑	Elevated watch. Sustained H2H would trigger immediate SPHINX + SKYSHOCK:OC.
Malaria (P. falciparum)	CHARON	SCRY OVERSIGHT	Endemic. Obligate Anopheles vector. Expanding range tracks vector habitat shift.
HIV	EROS	SCRY OVERSIGHT	Controlled. Endemic. Pharmaceutical intervention available.
Measles	HERMES	SCRY OVERSIGHT	Controlled (vaccine). R₀ 12–18. Resurgent in vaccination gap populations.
MERS-CoV	HECATE	SCRY OVERSIGHT	Dromedary reservoir. Limited healthcare H2H. Endemic monitoring.
1918 Influenza	HERMES	TEILHARD / YUGA:APPROACH	Approached YUGA SUNDOWN conditions. Natural immunity wave termination. Permanent demographic change.

06Active Case Study

SKYSHOCK: OUTSIDE CONTEXT · SPHINX CLASS: ANDV

ANDES ORTHOHANTAVIRUS · MV HONDIUS 2026 · STATUS: ACTIVE

CONFIRMED: 9 · SUSPECTED: 2
DECEASED: 3 · MONITORING: 12+
CFR SYMPTOMATIC: ~38%
ISSUED: 04 MAY 2026

Prior Classification and Why It Failed

Andes virus was classified HECATE — zoonotic spillover from the long-tailed pygmy rice rat (Oligoryzomys longicaudatus), no H2H component, contained at the species interface. This classification held from initial identification in 1995 through 2017. It was not wrong in the sense of being fabricated. It reflected the available data. The available data was systematically biased.

The Epuyen Signal — 2018

TRANSMISSION Index case infected five people during 90 minutes at a birthday party. Two seated approximately one foot away. Two seated approximately four feet away at neighboring tables. One person who only crossed paths briefly en route to a restroom. None of these constitute "prolonged sustained contact" by any reasonable definition.

R₀ Median reproductive number 2.12 before isolation measures were enforced. Comparable to early COVID-19 pre-intervention. This is not the signature of a virus requiring intimate prolonged contact.

WINDOW Transmission window appears concentrated on the day of first fever onset — approximately one day. Brief, but the virus transmits relatively easily within that window rather than requiring sustained exposure duration.

OUTCOME 34 cases, 11 deaths. Contained with isolation measures. Reclassification to SPHINX warranted but not formally executed by major health authorities. The "prolonged contact" doctrine persisted in guidance documents despite Epuyen.

The Hondius Event — 2026

INDEX Dutch national completed four-month road trip through Chile, Uruguay, Argentina before boarding MV Hondius at Ushuaia, 1 April 2026. Last Uruguay entry four days prior. Zoonotic exposure somewhere in the endemic zone. First symptoms 6 April.

DISPERSAL 30 passengers disembarked Saint Helena before outbreak declared. 147 total passengers from 23 countries. All documented, manifested, traceable — but dispersed across multiple international flight legs before containment was initiated.

DOCTOR Ship's doctor infected in clinical setting. Evacuated to Netherlands 6 May, confirmed ANDV positive. Healthcare worker infection in a structured clinical environment is not household contact. The transmission route has not been established.

KLM KLM flight attendant hospitalized after contact with Dutch woman during brief Amsterdam transit. Tested negative — but the 42-day incubation window remains open as of this document. WHO Director-General flagged this explicitly.

BIAS The "prolonged contact" doctrine governing ANDV was derived from household chains in rural Andean communities. Household transmission was what got detected because households are the dominant social unit in that context. Casual community transmission in those populations almost certainly went and continues to go undetected — misdiagnosed as flu, dead before tracing, outside surveillance infrastructure reach.

Current Threat Assessment

ANDV/Hondius does not currently warrant TEILHARD designation. The 147-passenger contact list is tractable. No confirmed secondary transmission from any repatriated passenger as of 12 May 2026. The 42-day incubation window has not closed.

The tripwire is a single confirmed secondary community case — transmission from a repatriated passenger to someone with no ship contact. That event moves the designation to TEILHARD and demands immediate revision of all ANDV containment doctrine.

// Tripwire conditions — SPHINX → TEILHARD
CONDITION A Confirmed secondary community transmission (pax → non-contact)
CONDITION B Ship's doctor transmission route confirmed as aerosolized
CONDITION C KLM attendant tests positive (incubation window still open)
CONDITION D Additional undeclared exposure nodes identified in contact trace

// Window closes
CLEAR 42 days from last exposure with zero secondary cases — est. mid-June 2026

07Technical Architecture

Live Surveillance System

IKELOS is not a static document system. The classification framework is the logic layer. The live system is the data layer — a pipeline that ingests current outbreak signals, surfaces them against the PCF/Rasputin framework, and maintains the map and designation records in near-real time.

DATA SOURCES

WHO Disease Outbreak News — official declarations, case counts, situation reports. Structured but latency of 24–72 hours. Primary authority source.

ProMED-mail RSS — faster than WHO, less structured. Run by ISID. Often surfaces outbreak signals 24–48 hours ahead of WHO formal declaration. Primary early-signal source.

ECDC Communicable Disease Threats Report — EU-focused, weekly cadence. Useful for European outbreak context and cross-border tracing events.

Virological.org — preprint server where sequences get posted before peer review. Where ANDV/Switzerland/Hu-3337/2026 appeared. Critical for SPHINX-class events where the sequence tells you more than the epidemiology.

CLOUDFLARE WORKER — andv-tracker (cron: 30min)

A Cloudflare Worker runs on a 30-minute cron trigger. It fetches ProMED RSS and filters for ANDV, hantavirus, Andes virus signal terms. It fetches the WHO DON page for the Hondius situation report. It fetches ECDC CDTR if a new weekly report is available. Results are normalized into the IKELOS JSON schema and stored in Cloudflare KV under the key andv:latest. The Worker exposes a GET endpoint at /api/status with permissive CORS headers for the map to poll.

On detection of a new ProMED post matching signal terms, the Worker also updates a signals:recent KV key containing the last 10 raw signal items for display in the map's signal feed panel.

// IKELOS JSON schema — andv:latest
{
"schema": "ikelos/1.0",
"updated": "2026-05-12T14:00:00Z",
"designation": {
    "tier": "SKYSHOCK:OC",
    "pcf_class": "SPHINX",
    "status": "ACTIVE"
},
"counts": {
    "confirmed": 9,
    "suspected": 2,
    "deceased": 3,
    "monitoring": 12
},
"cases": [ /* array of location objects */ ],
"signals": [ /* recent ProMED/WHO items */ ],
"tripwires": {
    "secondary_community_transmission": false,
    "window_close_date": "2026-06-15",
    "klm_attendant_status": "NEGATIVE — WINDOW OPEN"
}
}

MAP POLLING LOGIC

The map HTML polls /api/status on load and every five minutes. On each poll it diffs received counts against last-known state. New or changed cases trigger animated marker insertion. The tier/class designation in the header updates live. A "FEED LIVE" / "FEED STALE" indicator reflects whether the updated timestamp is within the last 45 minutes. The signal feed panel in the right panel shows the last five ProMED/WHO items with timestamps.

The map is a static HTML file — no build step, no framework dependency. The Worker is the only moving part. This is intentional: the system should function in degraded infrastructure conditions where complex dependencies become liabilities.

MULTI-PATHOGEN EXTENSION

The Worker architecture supports multiple concurrent designations. The schema is extended with a designations array keyed by pathogen identifier. The map renders a layer per active designation. A designation registry in KV maintains the full historical index — resolved designations are archived rather than deleted, preserving the state transition record. The PCF class transition log is append-only.

ProMED filtering for additional signal terms (H5N1, mpox, unknown hemorrhagic fever, etc.) runs in the same cron job. Signals that match PCF SPHINX criteria — unknown pathogen, unusual severity, outbreak in a context with historically low surveillance — are flagged to a signals:watch key for human review before designation assignment.

08Project Structure

Claude Code — project-ikelos

project-ikelos/
│
├── docs/
│ ├── ikelos-conceptual.html ← this document
│ ├── pathogen-classification.html ← PCF/1.0 standalone
│ └── designation-log.md ← append-only transition record
│
├── tracker/
│ ├── sphinx-class-andv-map.html ← live map, polls worker
│ └── index.html ← multi-designation dashboard (future)
│
├── worker/
│ ├── andv-tracker/
│ │ ├── index.js ← cron fetch + KV write + GET /api/status
│ │ └── wrangler.toml ← CF worker config, KV bindings, cron
│ └── schema/
│ └── ikelos-v1.json ← canonical JSON schema
│
├── data/
│ ├── cases/
│ │ └── andv-hondius-2026.json ← seed case data
│ └── signals/ ← archived ProMED/WHO fetches
│
└── framework/
├── pcf.json ← PCF class definitions machine-readable
├── tiers.json ← Rasputin tier definitions machine-readable
└── transitions.json ← designation transition log

Build Sequence

The Worker is the critical path dependency. Everything else is static. Build order: (1) define the JSON schema, (2) build the Worker with seed data returning static JSON before live fetching is wired, (3) confirm the map polls and renders correctly against seed data, (4) wire live fetching to ProMED and WHO, (5) deploy to Cloudflare and confirm cron execution. The map requires no deployment — it is a static file that can be served from anywhere or opened locally.

Cloudflare connection is currently available via Claude MCP connector but requires session reconnection before Worker deployment tooling is accessible. The Worker itself is approximately 80 lines of JavaScript.