Reference · v2.0.0 · Reference Implementation

Documentation

A complete reference for AetherRuntime v2.0.0 — architecture, formal definitions, failure conditions, and reference implementation

A research article plus reference implementation (v2.0.0) of the runtime-assurance architecture for the Aether family. The manuscript is published on Zenodo and Figshare. The OSF registration is published. A reference implementation is built (16 modules, 60 tests) and 3 experiments are completed on simulated data. No real-world deployment has been evaluated.

Table of Contents

Overview

AetherRuntime is a structured investigation of whether a runtime-assurance gap genuinely exists in the Aether family, or whether it is already covered by the existing five dimensions: AetherGit (storage), AetherFlow (integration), AetherProof (verification), AetherMesh (decentralization), and AetherGovern (governance). It is the sixth functional dimension (provisional) of the Aether family. v2.0.0 adds a reference implementation; no real-world deployment is claimed.

Project identity
FieldValue
Full titleAetherRuntime: A Falsification-First Investigation of Runtime Assurance in the Aether Family
SubtitleA Falsification-First Investigation of Monitoring, Detection, Containment, and Recovery in Multi-Agent Software Ecosystems
Version2.0.0 (paper v0.1.0 + reference implementation)
StageReference Implementation (preliminary results, simulated)
LicenseMIT (code) · CC-BY-4.0 (docs, manuscript)
DomainDistributed Systems / Software Engineering / Institutional Analysis
DOI (Zenodo)10.5281/zenodo.24078477
DOI (Figshare)10.6084/m9.figshare.34955746
OSF10.17605/OSF.IO/KZU7P
PyPIPublished
Record IDAETHERRUNTIME-02

"AetherRuntime is a candidate, not a result."

Installation

The PyPI package (2.0.0) is the reference implementation. It is not a production system. The v0.1.0 paper remains the authoritative hypothesis record.
# From PyPI (published)
pip install aetherruntime-arch

# From source (development)
git clone https://github.com/gitdeeper15/aetherruntime.git
cd aetherruntime
pip install -e ".[dev]"

# Run tests
pytest

# Run experiments
python experiments/detection_accuracy/run.py
python experiments/halt_window/run.py
python experiments/recovery_correctness/run.py

# CLI
aetherruntime info
aetherruntime verify

Requires Python 3.9+. Runtime dependency: cryptography >= 41.0.

Architecture — The Five Runtime Layers

All of this is hypothesis: a specification-level description of what a runtime-assurance dimension might contain.
The five layers of AetherRuntime
LayerQuestionStatus
1 — MonitoringWhat is the agent doing?Hypothesis
2 — DetectionIs this behaviour anomalous?Hypothesis
3 — ContainmentHow do we isolate the agent?Hypothesis
4 — HaltHow do we stop the agent?Hypothesis
5 — RecoveryHow do we restore the system?Hypothesis

Runtime lifecycle: observe → detect → classify → contain → halt → recover. Schematic, not a protocol specification.

Runtime state machine: each state has defined entry and exit conditions; exception transitions represent recovery paths.

Formal Definitions

Definition 1 — Telemetry Stream. A time-stamped sequence of observations from a running agent. Categories: CPU usage, memory usage, network I/O, syscall stream, and semantic action log.

telemetry = [(t, category, value), ...]

Definition 2 — Anomaly Predicate. A deterministic function of the telemetry stream and a baseline model that returns a boolean verdict (anomalous / not anomalous) together with a confidence score.

anomaly(telemetry, baseline) → (bool, confidence)

Definition 3 — Containment Policy. A record (target, level, restrictions, expiry) that specifies the containment level (L1–L4) applied to an anomalous agent, together with the restrictions (resource limits, access scope) and the expiry of the containment.

containment = (target, level, restrictions, expiry)

Definition 4 — Halt Order. A signed record (agent_id, reason, timestamp, proposer, approvals) that, when approved by a quorum of authorized nodes, requires every honest node to refuse further actions from the halted agent.

halt = (agent_id, reason, timestamp, proposer, approvals)

Failure Conditions (F1–F5)

Falsification status of the five failure conditions
IDClaim under testStatus
F1Non-redundancyNot falsified (R3, R4, R5)
F2Orthogonality to verificationNot falsified (narrowly)
F3Boundary with governanceNot falsified (decentralized setting)
F4DecentralizabilityNot falsified (with residuals)
F5Observable benefitUntested
"Not falsified" is a weaker status than "confirmed". F5 (observable benefit) remains untested. Neither the implementation nor the simulated experiments promote "not falsified" to "confirmed".

Propositions (P1–P2)

Propositions and their proof status
#PropositionStatus
P1Determinism of Local Anomaly DetectionProof sketch (not machine-checked)
P2Irreducible Halt WindowProof sketch (not machine-checked)
Both propositions are stated with proof sketches; neither is machine-checked. They are conditional on assumptions A1–A6 (inherited from AetherMesh) and bear directly on F4 (decentralizability).

Proposition 1 (mathematical form).

S_i(t+1) = S_i(t) ⊔ ⊔_{m ∈ M_i} m

∀ i, j, A_i(S_i) ∧ A_j(S_j) ⇒ (S_i ⊔ S_j) = (S_j ⊔ S_i)

Proposition 2 (mathematical form).

Δt_window = t_halt_enforced − t_detect
          = D_gossip(N, f) + t_queue + t_verify

Resource monitoring metric.

Ω_i(t) = ∫_0^t ( CPU_i(τ) + Memory_i(τ) + NetworkIO_i(τ) ) dτ > Θ_max

Validation

Validation checklist (v2.0.0)
CheckResult
Formal Definitions4 (all defined)
Propositions2 (stated with proof sketches)
Failure Conditions5 (F1–F5)
Requirements6 (R1–R6)
Runtime Layers5 (Monitoring, Detection, Containment, Halt, Recovery)
Figures6 (all generated)
References25 (in paper); 15 (in Zenodo record)
Open Questions20 (Q1–Q20)
Evaluation Metrics9 (M1–M9)
Baselines5 (B1–B5)
Scenarios6 (S1–S6)
Threats to Validity7
Python Modules16 (reference implementation)
Tests60 passed
Experiments3/3 completed (simulated)
PyPI PackagePublished
Real-world deploymentNot evaluated
Causal ClaimsNone (argument only)

Experimental Results (v2.0.0)

Results from experiments/*/results.json (simulated data)
#ExperimentMetricsResult
1Detection AccuracyM1, M2M1 = 109.67 (S1–S5 total); M2 = 7.33 (S6 FP)
2Halt WindowM510/10 enforced; mean 0.000243 s; median 0.000220 s
3Recovery CorrectnessM610/10 verified; M6 = 1.0000
Simulation only. The halt window is a property of the simulation's time step, not a real network measurement. M6 = 1.0 is a ceiling on real recovery quality, not a lower bound. No generalization beyond simulation is claimed.

Scope & Limits

What AetherRuntime v2.0.0 supports:

What AetherRuntime does not support:

Design philosophy. AetherRuntime is a research investigation with a reference implementation, not a production system. Its purpose is to specify an architecture precise enough to be tested against the failure conditions — and to make the limits of that specification explicit. The final project structure (single, dual, triple, or quadruple project) is explicitly left undetermined.

Threats and Limitations

Selected threats to validity
TypeThreatStatus
InternalSame author proposes candidate, failure conditions, baselines, thresholdsPre-registration; invite external review
InternalWeak or straw-man baselinesInclude B2, B4, B5; report B3 as ceiling
Construct"Anomaly" and "harm" simulated by the experimenterStated explicitly
ConstructM6 measured against ground truth that exists only in simulationReported as upper bound
ExternalSimulated network, simplified adversary, synthetic workloadNo generalization claimed
ConclusionFew scenario classes; thresholds are proposalsAll outcomes reported
LiteratureRuntime-verification and anomaly-detection literature not reviewed in depthAcknowledged

Citation

APA (paper): Baladi, S. (2026). AetherRuntime: A Falsification-First Investigation of Runtime Assurance in the Aether Family (Version 0.1.0). Zenodo. https://doi.org/10.5281/zenodo.24078477

APA (software): Baladi, S. (2026). AetherRuntime v2.0.0: Reference Implementation of the Falsification-First Runtime Assurance Architecture (Version 2.0.0). PyPI. https://pypi.org/project/aetherruntime-arch/

@techreport{baladi2026aetherruntime,
  author       = {Baladi, Samir},
  title        = {AetherRuntime: A Falsification-First Investigation of
                  Runtime Assurance in the Aether Family},
  institution  = {Ronin Institute / Rite of Renaissance},
  type         = {Research Paper},
  year         = {2026},
  month        = {November},
  version      = {0.1.0},
  doi          = {10.5281/zenodo.24078477},
  url          = {https://doi.org/10.5281/zenodo.24078477}
}

@software{baladi2026aetherruntime_v2,
  author       = {Baladi, Samir},
  title        = {{AetherRuntime} v2.0.0: Reference Implementation of the
                  Falsification-First Runtime Assurance Architecture},
  year         = {2026},
  month        = {November},
  publisher    = {PyPI},
  version      = {2.0.0},
  doi          = {10.5281/zenodo.24078477},
  url          = {https://pypi.org/project/aetherruntime-arch/},
  note         = {Reference implementation. Available via pip install aetherruntime-arch.}
}

@misc{baladi2026aetherruntime_figshare,
  author       = {Baladi, Samir},
  title        = {AetherRuntime: A Falsification-First Investigation of
                  Runtime Assurance in the Aether Family},
  year         = {2026},
  month        = nov,
  publisher    = {Figshare},
  doi          = {10.6084/m9.figshare.34955746},
  url          = {https://doi.org/10.6084/m9.figshare.34955746},
  note         = {Preprint. Archived version of the AetherRuntime manuscript.}
}

@misc{baladi2026aetherruntime_osf,
  author       = {Baladi, Samir},
  title        = {{AetherRuntime: A Falsification-First Investigation of
                  Runtime Assurance in the Aether Family}},
  year         = {2026},
  month        = nov,
  day          = {9},
  publisher    = {OSF},
  doi          = {10.17605/OSF.IO/KZU7P},
  url          = {https://doi.org/10.17605/OSF.IO/KZU7P},
  note         = {OSF Preregistration. Associated project: https://osf.io/ft7ez}
}
Status note (v2.0.0): This is a research article plus reference implementation (v2.0.0) of the runtime-assurance architecture for the Aether family. A reference implementation is built (16 modules, 60 tests) and 3 experiments are completed on simulated data. No real-world deployment has been evaluated. F1–F4 are not falsified by the implementation or the experiments, but they are also not confirmed. F5 (observable benefit) remains untested. Propositions P1–P2 remain proof sketches (not machine-checked) and are conditional on assumptions A1–A6. The paper is not peer reviewed. AetherRuntime is a candidate, not a result.