Research

Building the operational evidence
behind adaptive deterrence.

Our research is grounded in real-world deployments, operational observations, and the continuous study of adversary behaviour. As our body of work grows, we publish practitioner papers, operational reports, and research that advances the understanding of adaptive deterrence intelligence.

Every paper in this library began in the field. The origin ↓

Doctrine
5 papers · 1 governance document

Core concepts and frameworks. The intellectual foundations of Adaptive Deterrence Intelligence as a discipline.

PAR-001 Doctrine 2026 Parthenius Air Intelligence Unit
The Learnability Problem
Most security systems measure what happened. The critical variable is not incident frequency — it is learnability. The rate at which an adversary can build a reliable model of an operation sufficient to execute against it. This paper defines the problem, proposes a measurement framework, and presents field observations from a 19-month active deployment.
● Published · 2026
→
PAR-006 Doctrine 2026 Parthenius Air Intelligence Unit
The Adversary Learning Cycle
Security models describe the defender. This paper models the adversary. A formalisation of the five-stage learning cycle organised adversaries move through before acting — watching, probing, model building, model testing, execution readiness — with the observable characteristics of each stage, the point where disruption works, and the regression dynamics field data shows when a cycle is broken.
● Published · 2026
→
PAR-004 Doctrine 2026 Parthenius Air Intelligence Unit
On Operational Intelligence Sovereignty
Six principles on why the intelligence that protects you cannot be bought, aggregated, or standardised. Your adversary behaviour model is built from time spent watching your specific adversary on your specific site. The moment you hand it to a vendor who aggregates it with everyone else's, the edge is gone.
● Published · 2026
→
PAR-002 Doctrine 2026 Parthenius Air Intelligence Unit
What Is Adaptive Deterrence?
A definition of Adaptive Deterrence Intelligence as a security category distinct from detection, response, and conventional deterrence. The appropriate unit of analysis for organised adversarial threats is not the incident but the planning cycle. Effective deterrence requires continuous disruption of the adversary's learning process — not faster response to its conclusions.
● Published · 2026
→
PAR-008 Doctrine 2026 Parthenius Air Intelligence Unit
The Autonomy Paradox
Autonomy increases capability but does not replace operational doctrine. As autonomous systems assume responsibility for execution, the quality of the decisions governing them becomes more important, not less. Automation asks how a task can be executed more efficiently; adaptation asks whether it should continue to be executed that way. As technology becomes accessible, advantage resides in how capability is governed — not in the capability itself.
● Published · 2026
→
v1.0 Governance 2026 Parthenius Air
The Zerathis Principles
The binding operating principles of the Zerathis Blindspot™ platform. Six commitments covering deployment sovereignty, the role of AI, human accountability, surveillance limits, data lifespan, and the compliance record. Published, versioned, and binding — not aspirations.
● Published · v1.0
→
The Operational Framework
3 papers · in preparation

Three interconnected variables that define how Adaptive Deterrence Intelligence functions as a continuous feedback system. Purposeful variation reduces adversary learnability; reduced learnability denies opportunity. The control loop: System Variation Rate → Learnability Score → Opportunity Denied Rate → Reduced Operational Risk.

PAR-009DoctrineForthcomingParthenius Air Intelligence Unit
System Variation Rate
The controllable variable in the framework. The measured rate and operational significance of purposeful variation introduced to reduce adversary predictability while preserving mission effectiveness. Not randomness — disciplined operational adaptation. This is the control variable and the primary point of intervention.
○ In preparation
PAR-010DoctrineForthcomingParthenius Air Intelligence Unit
Learnability Score
The state variable — an estimate of how reliably an adversary can observe, model and predict operational behaviour. A property of the adversary, not the operation. It is the primary diagnostic measure and a leading indicator: rising learnability warns of a maturing adversary model before any incident occurs.
○ In preparation
PAR-011DoctrineForthcomingParthenius Air Intelligence Unit
Opportunity Denied Rate
The performance outcome — the proportion of exploitable opportunities successfully prevented through continuous disruption of adversary learning. What it measures that incident counts cannot, the disciplined denominator separating organised extraction from opportunistic theft, and its interpretation within the causal chain.
○ In preparation
Operational Insights
2 papers

Findings from live deployments. What the field teaches that the framework cannot predict.

PAR-003 Operational Insights 2026 Parthenius Air Intelligence Unit
What Predictability Costs
Every security operation has a learnability score. A measure of how quickly a patient adversary can build a reliable model of your system. Most operations have never measured it. This paper examines what predictability costs in practice — and what it means for the adversary's planning cycle.
● Published · 2026
→
PAR-007 Operational Insights 2026 Parthenius Air Intelligence Unit
The Two Latencies
Intelligence that arrives is not intelligence that acts. This paper defines the two latency classes that determine whether an operation can use what it knows — operational latency, owned by the security operation and measured in hours, and remediation latency, owned by the client and measured in weeks — and shows how the adversary reads the slower one as intelligence. An unmeasured repair backlog is a published schedule of exploitable windows.
● Published · 2026
→
Case Studies
1 published

Evidence from client engagements. Outcomes, adaptation dynamics, and operational records from active deployments.

PAR-005 Case Study 2024–2026 Active deployment · South Africa
19 Months. One Site. Zero Successful Extractions.
The complete operational record of the first Zerathis Blindspot™ deployment. Platinum mining, South Africa, September 2024 to present. 97% Opportunity Denied Rate. The adversary never stopped adapting. Every configuration denied.
● Published · Month 19 active
→
Field Notes
1 published

One-page observations from live deployments. No client names. No conclusions beyond what the field supports.

FN-001 Field Note 2026 Active deployment · South Africa
The Learning Race
Learnability runs both ways. When an operation responds to probing, the adversary learns from the response — so a disruption repeated the same way becomes a new pattern to learn. Deterrence against an adaptive adversary is not a state reached but a tempo sustained, decided by the relative rate at which each side adapts to the other.
● Published · 2026
→
Annual Observations
Edition 01 in preparation

Trends and lessons from the field. Published annually, drawing on patterns across active deployments.

Annual Observations 2026 Parthenius Air Intelligence Unit
The State of Operational Predictability — Edition 01
The first annual benchmark on operational learnability across high-risk environments. How consistently operations behave, how quickly adversaries build working models, and what the patterns look like across sectors, corridors, and threat classifications. Drawn from Zerathis Blindspot™ deployment data. Client identities not disclosed.
● In preparation · Edition 01, 2026
→
Annual Observations2027
Adversary Adaptation Trends — Year 2
How organised adversarial behaviour evolved across two active deployments. Escalation patterns, coordination structures, and what changed when the planning cycle kept failing.
○ Coming 2027
Conference Papers
In preparation

Presentations and talks. Published here as they are delivered at industry conferences and sector events.

Technical Notes
In preparation

Deeper dives into specific methods and measurements. The technical foundations of platform methodology. Publishing with the second deployment, where metrics can be reported across more than one site.

Origin

From operation to doctrine

Parthenius Air has flown live BVLOS operations for over six years — SACAA-licensed, night and remote deployments across mining and infrastructure sites. That operation continues. The doctrine below sits on top of it, not in place of it.

For most of those years the work looked like a hardware problem. Better sensors, tighter response times, more coverage. Each upgrade held for a while. Then, site after site, the same thing happened: a capable, well-flown operation would start losing ground not because the equipment failed, but because someone on the other side had been watching long enough to know what it would do next.

The adversary was never defeating the aircraft. They were studying it — and eventually, they always finished studying it.

Doctrine into platform

Zerathis Blindspot™ is the operational form of that observation. It doesn't replace the aircraft or the operators flying it — it measures, stage by stage, how far an adversary's model of a site has progressed, and disrupts that model before it reaches execution confidence. The platform exists because the doctrine demanded something the flying alone couldn't provide: a way to act on what six years of observation had already made clear.

The field came first. The doctrine was extracted from it, not imposed on it. Everything Parthenius Air builds from here follows that order.

SACAA UASOC G1424D  ·  Operational 2021–present  ·  Doctrine documented in Before The Plan Concludes