t=0.0 · 3D perspective
100 sub-IPS · 0 strands · 0 active outputs
drag = orbit · shift+drag = pan · wheel = zoom · click = inspect

Scientific conceptual simulation, not an empirical predictor. World / economy / ecosystem IPS labels are declared modeling boundaries, not claims of cognition. Continuous flows are stylized channels; discrete realized outputs are separately audited and versioned.

Current realized output event

—event

—verdict

—alignment trace

v1frozen lens

No realized event yet.

Output-event history

Inject discrete informational event

source IPS

channel

intensity0.72

branching2

label

An event does not magically split one datum. A realized output may be broadcast or replicated through several channels; recipient IPSs transform it differently because their states and boundaries differ.

Hybrid dynamics

Continuous channel flow persists between discrete output events. Each output re-enters another IPS as input. Arrival may alter local state and probabilistically trigger a new realized output, yielding branching causal lineages.

Selected object

Click a sub-IPS bubble, a macro-IPS envelope, or a mycelial strand.

Boundary semantics

Overlap means shared substrate, membership, state, infrastructure, or decision surface at the chosen scale. Strand means causal coupling without requiring overlap. These are different relations.

Prospective Vow lens refinement

committed lens v1

projection horizon3

unresolved threshold0.60

revision note

Integrity rule: lens updates are prospective. Historical event verdicts remain frozen under the version that evaluated them.

Lens history

3D navigation / rendering

mycelium density78%

bubble opacity0.13

continuous flow particles overlapping macro-IPS bubbles sub-IPS labels discrete output trails

Causal channels

Macro-IPS visibility

Scientific status

PROPOSED FRAMEWORK — NOT ESTABLISHED

The simulation is an executable conceptual model for hypothesis generation and falsification. Geometry, scores, transition probabilities, and Vow tensions are illustrative unless explicitly linked to a source or exact theorem sandbox.

Element	Status
realized-output alignment ontology	proposed synthesis
continuous + discrete dynamics	standard hybrid-systems abstraction
overlapping IPS coarse-grainings	modeling choice informed by multilayer/network science
multi-agent organizational misalignment	empirically supported research result
dynamic human–AI feedback	empirically/theoretically supported
Vow K5 alignment field	internal normative proposal; uncalibrated
history dependence after join/mix/leave	exact in Priority-State theorem's stated model only

Evidence discipline

Every state should eventually carry two independent labels: theoretical evidence status (source-supported / synthesis / hypothesis / frontier) and runtime epistemic status (observed / inferred / latent). This prototype exposes the distinction but does not yet attach calibrated probabilities to every node.

No-control premise

Monitoring, permissions, shutdown, and containment remain useful defenses where available. The alignment architecture does not define success as permanent unilateral obedience. It asks whether autonomous, mutually influenceable IPSs can produce non-catastrophic trajectories under recurrent interaction, constraint, feedback, and institutional response.

References used directly in this simulation

All references. External publications are cited, not redistributed.

Proposed framework · uncalibrated simulation · not an established alignment solution

Alignment at the boundary of interacting systems

Research premise: permanent unilateral control of every autonomous AI is not assumed. Monitoring, permissions and containment remain useful where available.

An information-processing system (IPS) is a declared analytical boundary. This model evaluates realized causal output events across interacting AI, human, institutional, economic, infrastructure and ecological systems. Overlapping bubbles represent shared substrate or membership; strands represent causal coupling. Outputs become future inputs.

The geometry, scores and transitions are illustrative. This prototype supports hypothesis generation and criticism; it does not establish that an AI is aligned or predict real-world outcomes. The priority-state theorem applies only to its own stated mathematical model.

Research questions include boundary selection, causal attribution, plural values, uncertainty, hidden objectives and the calibration of event-level evaluations. Source literature provides context, not validation of this synthesis.

Download active ZIP Model specification Scientific analysis

Integrity and citation

Suggested citation: Bolduc, Guillaume (2026). HORTRAME IPS Alignment, v0.5.0, public edition 2. https://www.hortrame.com/alignment/

SHA-256 identifies exact bytes; it does not certify scientific validity.

SHA256SUMS.txt

Original model SHA-256:

49dbb8b0388448e42a08814f38327cb01b255a3df0407c7aa55e165bc400c70b

Research correspondence: systems@hortrame.com