Research Systems
Implemented and experimental systems supporting the research programme. Philosophical proposals, implemented software capabilities, computational approximations, and unverified research hypotheses are distinguished throughout.
Distributed Operational Research Architecture
Distributed Operational Research Architecture
The Ziran System is a distributed research architecture for tracing how heterogeneous configurations generate, retain, transform, propagate, and reintroduce differences across time, scale, boundaries, and address domains. It conditionally fires, stops, disconnects, reconnects, and revises operative systems without requiring a unified theory of what exists.
The architecture provides methodological responsibility for compiling modules from heterogeneous domains under specifiable conditions. It does not unify domains under a single explanatory framework; it tracks the conditions under which modules can be activated, suspended, and revised.
The Ziran System as described in this section is an architectural and philosophical proposal. Proposed capabilities and fully implemented capabilities are distinct. See the Simulator page for the four-level theory–implementation distinction.
Organization of Differences Through Transitions
An operational shape concerns the organization of differences through transitions, couplings, separations, retention, branching, cessation, and historical re-entry. Operational units and their boundaries may themselves change during these processes.
Fixed objects, agents, categories, or network nodes are not assumed as universal prerequisites. Relevant phenomena include transitions, coupling, differentiation, retention, transformation, re-entry, branching, closure, amplification, attenuation, real effects, and reality penetration.
These are not universally mandatory stages. Their applicability must be established independently for each configuration rather than presupposed.
Specifying the Conditions of an Operation
Every claimed ignition condition must specify the relevant configuration. An abstract description of ignition independent of its enabling conditions is insufficient. The following parameters are relevant where applicable:
- Target configuration
- Observational apparatus
- Boundary and scale
- Temporal interval
- Connections and resources
- Institutional conditions
- External memory
- Feedback mechanisms
Configuration-specific ignition does not require that every parameter be fully known before a research operation begins. It requires that claimed ignition conditions not be described as if independent of all conditions.
Occurrence vs. Actual Change in a Configuration
A real effect is an actual change in a specified target configuration. The occurrence of an operation must be distinguished from whether a change has taken place in a physical, informational, institutional, or social configuration.
A recorded event does not automatically establish that the relevant configuration has changed. The claim of a real effect requires specification of the target configuration and evidence of change within it.
Local Effects and Heterogeneous Transmission
Local effects must be distinguished from their propagation into heterogeneous operational domains. Specifying propagation requires identifying the transmission pathway, the receiving configuration, the temporal interval, and the conditions under which propagation occurs.
Cross-domain propagation must not be inferred merely because a local operation succeeded. A successful operation in one domain is evidence of a local change; it is not automatically evidence of change in a different domain.
Retention and Reuse of Prior Operational States
Historical re-entry designates the reuse of a retained prior operational state in a subsequent operation. The following five phases are distinguished:
- Generation of a difference
- Retention of that difference
- Subsequent operational reuse
- Changes resulting from reuse
- Propagation into other domains
Historical re-entry may occur without institutional authorization. Credit scoring, human recognition, and institutional acceptance are not universal prerequisites for a prior operational state to be reintroduced into a current operation.
Simultaneous and Heterogeneous Configuration Change
Operational configurations may undergo splitting, coupling, overlap, recombination, attenuation, amplification, cessation, and renewed ignition. These changes may occur simultaneously across heterogeneous configurations.
A universal, mutually exclusive sequence of states is not imposed. The applicable sequence is determined by the configuration under investigation.
Distinguishing Absence of Detection from Absence of Difference
The following conditions must be held distinct at all times:
- No difference detected
- Difference not measured
- Instrument incompatibility
- Scale incompatibility
- Temporal-window incompatibility
- Boundary-induced invisibility
- Evidence supporting the absence of a specified difference
Failure to observe a difference must not automatically be interpreted as the absence of that difference. This constraint applies to both theoretical descriptions and experimental interpretation.
Multiple Operational Rates and Historical Dependences
The Ziran System investigates operations across heterogeneous temporal scales, which may include cosmological, geological, biological, technological, institutional, computational, and social processes. Operational speed, latency, irreversibility, synchronization, and historical dependence differ across domains.
Where the simulator uses a common computational time step, this is an implementation convention. The simulator's computational time should not be equated with physical or historical time.
Experimental Identification Infrastructure
TII (Transition-Ignition Identifier) is an experimental infrastructure for minimal append-only identification and operational audit. It assigns persistent identifiers to configurations, transitions, and operational events without requiring a centralized registry.
TII is distinct from earlier identifier proposals. Identifiers issued by the current implementation are provisional and non-authoritative. Status: experimental.
TII ArchiveOperational Transitions, Autonomous Science, and Computational Research
The following experiment records are those for which verified execution records and outputs exist. Conceptual visualizations, simulation outputs, executed experiments, and unverified research hypotheses are distinguished in each record.
Post-Active Inference Benchmark
Research Question
Does active inference — as formalized by Friston and collaborators — constitute a sufficient framework for operational configurations under heterogeneous conditions? Does it presuppose observational and model conditions that cannot be taken as fixed across all configurations?
Configuration
Computational experiment within the Xeno-Ziran Runtime environment. One world graph, one control task, one discrete Active Inference comparator family (fixed and adaptive variants), one model-light controller, one tracker. Results are not about active inference in general nor about brains, organisms, or any system outside the simulation. Runtime v0.2.0.
Experimental Stages
Stage 1 — Exploratory development: 10 seeds, 47 perturbation subtypes in 8 families, 1,890 cells, 9,120 runs. Identified candidate crossing thresholds. This stage is frozen as exploratory evidence; results were superseded on confirmatory questions by Stage 2.
Stage 2 — Pre-registered confirmatory: 30 fresh seeds (disjoint from development), 12,750 cells, 55,680 runs. Pre-registration and analysis rules frozen before execution.
Confirmatory Results (Stage 2, authoritative)
10 of 19 development crossings replicated under the pre-registered criterion. All 8 perturbations reaching POST_ACTIVE_INFERENCE are observation- or interface-side. World-structure perturbations (topology, coupling, memory, structural) did not reach POST in the tested range.
The central hypothesis that active inference “remains effective sometimes beyond the interval” was not supported: in POST cells, both AIF variants performed worse than no control (fixed −0.028, Holm p = 0.0085; adaptive −0.058, Holm p = 0.0005). Validity and control are positively coupled (Spearman ρ = 0.44): control fails with the primitives.
Inside the compatible regime, AIF's control advantage over no control is small: +0.03 [0.02, 0.03], d_z = 0.10. The development figure (+0.11) did not replicate.
Strongest defensible formulation
Under the tested task, world, controller and outcome measures: the representational primitives of this discrete AIF comparator lose stability reproducibly, mostly gradually and severity-dependently, when its observation map or action interface is disrupted. Disruptions of world structure do not push it out of the compatible regime in the tested range. Where the primitives fail, control fails with them.
Not claimed: that active inference is falsified; that operational-shape tracking is a superior controller in general; that other active inference implementations share these limits; that a general Post-Active-Inference theory has been demonstrated. Experimental findings establish results only for the specified computational configuration and do not confirm the theoretical framework of Xenoscience or the Ziran System.
Runtime repository (GitHub)A structured reference index of the theoretical framework, organized into formal categories: top-level theory, primary research branches, core principles, reference concepts, methods, and experimental implementations.
View Theory Index →