Research Program

Capability without accountability is power without balance

A research program on the accountability gap — the structural conditions that make claims, memory, and agentic action valid over time. The product line is the empirical apparatus.

// the accountability gap

Humans answer to webs of accountability. AI does not.

What humans have

Reputation, institutions, law, memory of consequence. Human capability is embedded in webs of accountability that make power answerable — slowly, imperfectly, but structurally.

What AI lacks

Agentic systems act with growing capability and none of that structure. Accountability has to be engineered — into memory, into claims, into action — or it does not exist at all.

The question the program asks: what are the structural conditions under which a claim, a memory, or an action by an agentic system remains valid — and answerable — over time?

01 Published Research

The papers, in three lanes

The lanes are firewalled: the same grammar can recur across regimes, but no evidential weight transfers between them.

01 · flagship SITE LIVE

Closing the Agent Loop

Governed agentic operations — closure, context packets, epistemic agentic memory, valid handoff.

5 papers · DOI pendingEnter the lane →
02 · cognitive SITE LIVE

The Temporal Neuroscience Index

Subjective time as information geometry. Integrative, not diagnostic.

4 papers · DOI pendingEnter the lane →
03 · physics SITE LIVE

The Missing Layer of Time

When quantum possibilities become — and remain — public facts. This lane grows as wave 2 ships.

2 papers · DOI pendingEnter the lane →

Published here first → Zenodo → arXiv → journal submissions. Status records update as each wave ships.

// bridge — from valid descriptions to governable actions

Agentic Frames

Five frames carry the papers into the runtime: each converts a research result into a constraint an agent architecture must satisfy.

F1
Event Frame

An action isn't done until it's registered. Closure precedes claim.

F2
Translation Frame

Meaning does not survive handoff for free. Context must travel as a packet.

F3
Temporal Frame

Validity decays. Every claim carries a horizon and must be re-grounded past it.

F4
Validity Frame

Confidence is not evidence. Claims separate what was observed from what was inferred.

F5
Accountability Frame

Authority is granted, scoped, and revocable. No standing power.

02 Research Systems

From papers to operating architecture

Papers establish concepts. Systems organize implementation. Products operationalize them for teams. Four systems, peers in one architecture.

L

Luna

runtime governance

Governs authority: granted, scoped, revocable. Thesis system of Closing the Agent Loop.

System note →
C

Constellation

memory system · with Atlas

Epistemic agentic memory. The corpus stays home; the projection travels.

System note pending
A

ANT

orchestration

Controlled orchestration: routes, tool boundaries, return contracts.

System note pending
O

OWL

observation

Continuous monitoring for drift, anomaly, and governance failure.

System note pending

Constellation structures memory · Luna governs authority · ANT coordinates execution · OWL observes · context packets connect the layers.

L
// system spotlight Luna
The Luna Thesis

Luna builds the accountability layer for humane agentic intelligence.

AI reasoning structured to respect human agency, dignity, truth, safety, privacy, consent, context, and the consequences of action.

// axiom I

Autonomy and accountability must scale together.

// axiom II

Capability without accountability is power without balance.

// axiom III

Intelligence without moral context is not wisdom.

// axiom IV

Accountability requires visibility; deception is the corruption of visibility.

The Luna system note leads with the seven operational axioms; this moral foundation carries there as its own labeled section.

Every increase in agent autonomy requires a corresponding increase in accountability infrastructure — per capability tier.

Agent capability
Required accountability layer
Can retrieve information
Source boundaries, consent, provenance tracking
Can summarize or advise
Evidence grounding, uncertainty disclosure, limitation marking
Can use tools
Tool-call authorization, action traces, approval gates
Can remember
Memory consent, expiration, correction, deletion
Can coordinate with other agents
Handoff fidelity, role boundaries, disagreement tracking
Can act externally
Human review, escalation, reversibility, audit logs
Can adapt over time
Drift monitoring, replay tests, change attribution
// functional deception detection — research frontier

Once an agent can misrepresent what it is doing, why, what evidence it used, or what constraints it bypassed, supervision becomes performative. Deception detection is not one safety task among many — it is the precondition for whether accountability can function at all.

// meta-integrity rule

Intelligence and model behavior are symbiotic with the integrity of the environment and architecture.

03 Open Specs & Releases

Buildable artifacts

// specification LIVE

Context Packet Specification v0.4.0

The interchange format: how meaning survives handoff between agents and systems.

// evaluation PLANNED

Benchmark harness

Reproducible evaluation of closure, handoff validity, and drift detection.

// reference node PLANNED

Constellation node

A runnable reference implementation of the memory architecture.

04 Future Operations

The product line is the empirical apparatus

The research is not upstream of the business — it runs through it. Client deployments are how the program's claims meet production reality.

method 1

Vertical-slice deployment

Every governed workflow shipped is a test of the frames under real constraints.

method 2

Parallel configuration testing

Competing governance configurations run side by side; the record decides.

method 3

Production-scale evidence

Operating data feeds back into the papers — the lanes grow from practice.

// researcher

Megan Anderson

Independent Researcher · Systems Architect · Educator · Founder

Megan Anderson is an independent researcher, systems architect, educator, and founder whose work develops constraint architectures for accountable agentic reasoning, temporal cognition, child safety, and complex systems.

She is the founder of AI ARMY, where the research substrate is applied to governed AI operations and agentic infrastructure, and of Kaleidoscope Support, which holds the Coalition for Child Safety and applies the work to neurodiversity education and child protection. The two programs are distinct in surface and audience, and both are grounded in the same research foundation.

// subjects of interest
Complex Adaptive Systems & Governance Symbiotic AI Reasoning Cognitive-Emotional Regulation & Neurotypologies Philosophy of Time & Systems Cognition Multimodal AI & Time-Structured Cognition Legal Reasoning & Interpretability Neuroscience & AI Cognition Time Theory in AI Reasoning Governance of AI Ecosystems Cognitive Wellness & Recovery Philosophy of Science & Foundations of Physics Neurodiversity & Accessible Technology Humane AI & Deception Detection Constellation Memory System & Atlas Luna Runtime Governance ANT Command-Control Agent Architecture
// cite this work

Anderson, Megan. AI ARMY Research Program. AI ARMY, INC., aiarmy.co/research, 2026.

Paper DOIs and arXiv records are added as each publishes.

// engage

For partners, researchers, and operators.

Tap a door to see how its journey through the page should read.

Partners

Explore opportunities around governed agentic workflows, Luna pilots, Constellation / Atlas memory architecture, and accountable AI infrastructure.

Researchers

Engage with the research program, preprints, and the Luna Runtime Governance methodology.

Operators

Build AI operations with agents, workflows, memory, and governance designed in from the start.

The partner's path
The researcher's path
The operator's path
01

Start at the ladder — the program end to end, papers through operations.

02

Read the systems row: Luna, Constellation, ANT, OWL — the architecture a pilot runs on.

03

Review the methodology: vertical-slice deployment is how a partnership starts.

04

Partner with us on pilots, memory architecture, and accountable AI infrastructure.

01

Land on the thesis — the accountability gap, stated as a program with falsifiable structure.

02

Read the three lanes as an agenda: flagship, cognitive, physics — firewalled, each with its own status record.

03

Check the pipeline: published here first, then Zenodo, then arXiv, then journal submissions.

04

Cite the program in its neutral form; DOIs attach as each paper publishes. Then open a conversation about the open problems.

01

Start at the accountability gap — the problem stated in operating terms.

02

See the four systems as what actually runs: Constellation structures memory, Luna governs authority, ANT coordinates execution, OWL observes.

03

Read the methodology: the product line is the empirical apparatus — your deployment generates the evidence.

04

Engage AI ARMY for governed agentic operations designed in from the start. Book a discovery call.

Ready to take the next step? Ready to take the next step? Ready to take the next step?

Megan Anderson · AI ARMY, INC. · aiarmy.co · 2026