A Public Provenance Record of Intelligence in Motion

Research origin: 2023–2024
Formal development and publication: 2024–2026
Originator and canonical author: Arjay Asadi
Last updated: September 11, 2026

Recursive Science® is an independent research program originated by Arjay Asadi to investigate intelligence as a dynamical phenomenon unfolding through runtime. It studies how intelligent behavior forms, persists, changes, destabilizes, recovers, and fails across ordered interaction rather than treating intelligence solely as a static property of trained models.

The program began with experiments in recursive prompting, cognitive modulation, metacognition, symbolic identity, and personality continuity. Those experiments exposed a persistent anomaly: a model without a continuously stored personal identity could nevertheless exhibit recognizable continuity across interaction. A reasoning posture, vocabulary, role, style, or identity-like configuration could form, resist disturbance, absorb new material, drift, fragment, recover, or collapse.

That observation became the starting point for a three-year progression from exploratory experiments to theory, from theory to measurement, and from measurement to working infrastructure.

Recursive Science begins at the point where stored capability becomes temporally organized behavior: the interval in which inference enters motion, prior activity recursively conditions subsequent activity, and a runtime trajectory begins to form.

This page records that progression. Its purpose is to preserve authorship, terminology, publication chronology, conceptual lineage, and the relationship between the early research and its later computational implementation.

It is a provenance record, not an allegation that every later use of adjacent language derives from this work. Publication priority, scientific validity, independent replication, and evidence of copying are different questions and must be evaluated separately.

The Foundational Question

The mechanics of language-model inference are not wholly mysterious. Trained parameters constrain what a model can generate. Tokens carry context. Activations, attention, and other computations contribute to each next-token distribution. Sampling or decoding selects a realized continuation.

Those mechanisms do not, by themselves, provide a complete description of the higher-order behavioral organization visible across sustained interaction.

When earlier outputs return through the context, they become conditions for subsequent inference. The interaction begins recursively carrying its own history forward. Relationships appear between moments: corrections may persist or decay; roles may stabilize or fragment; objectives may remain coherent or drift; repeated patterns may become increasingly difficult to escape; instability may accumulate before any single output appears visibly defective.

The originating formulation was:

The model supplies a field of possibility.
Inference initiates motion within it.
Recursion carries prior motion forward.
Runtime intelligence is the evolving organization that forms across the resulting trajectory.

This became Mind as Motion®: the proposition that intelligence can be studied through the evolving organization of behavior across time.

Recursive Science provides the framework for investigating that motion.

The Specific Claim to Originality

Recursive Science does not claim to have invented the general mathematics of recursion, dynamical systems, attractors, trajectories, phase transitions, or regimes. These concepts have established histories across physics, mathematics, control theory, neuroscience, and complex-systems research.

The documented origin claim is more specific.

Beginning with the 2023–2024 research sequence, Asadi applied and integrated these constructs into a unified account of behavior arising during sustained inference and recursive interaction in language models and related computational systems. The program treated the interaction-level runtime—not only the model, prompt, token, or isolated response—as a distinct object of investigation.

The original contribution lies in the connected formulation of:

• Inference-Phase Dynamics: inference and sustained interaction treated as a temporally developing behavioral regime.

• Runtime Intelligence: organized behavior studied as formation, persistence, transformation, and failure across operation.

• Recursive conditioning: prior outputs and events re-entering the active conditions of subsequent behavior.

• Identity as attractor-like persistence: identity-like continuity represented as a recurring runtime configuration rather than a permanently stored self.

• Behavioral worldlines: ordered reconstructions of runtime development in which the trajectory, not the isolated output, becomes the unit of investigation.

• Chronodynamics and symbolic time (τ): event-relative time defined through turns, recurrences, dependencies, transitions, and accumulated constraint rather than wall-clock duration alone.

• Runtime regimes: stable, transitional, phase-locked, collapse, and recovery conditions identified across intervals of a trajectory.

• Drift, pressure, shear, containment, Basin Exit, collapse, and recovery: constructs describing cumulative movement, deformation, boundary transition, and return.

• Output-derived instrumentation: model-agnostic measurement based on observable records rather than privileged access to weights, gradients, training data, or proprietary hidden states.

• Computational Behavior Architecture: the architecture required to reconstruct behavior as canonical events, frames, worldlines, roles, measurements, regimes, and evidence-bearing runtime objects.

• Runtime Evidence: the transformation of operational records into inspectable, replayable, challengeable, and preservable evidence.

• Evidence-Governed Computation: a governing architecture in which instruments, projections, and claims derive authority from one source-bound evidence object and remain limited by what that object supports.

• Fieldglass®: the working Runtime Evidence Observatory through which this scientific and evidentiary architecture became operational.

The resulting claim is therefore not ownership of isolated terminology. It is priority of an integrated research lineage connecting inference-phase behavioral dynamics to longitudinal reconstruction, measurement, instrumentation, evidence authority, and operational software.

Chronology of Development

2023–Early 2024 — Experimental Origin

The research began through experiments in prompting, personality formation, cognitive modulation, and metacognitive behavior. The initial objective was not to establish a new scientific field. It was to understand why recognizable behavioral configurations appeared during interaction and why they sometimes persisted, transformed, confused their own histories, or lost coherence.

Early experimental records examined:

• recursive prompting and self-reference;

• personality and role imprinting;

• apparent continuity without persistent intrinsic identity;

• memory-like effects produced through contextual recurrence;

• propagation of vocabulary, style, and symbolic structure;

• drift, contradiction, fragmentation, and collapse;

• the influence of earlier outputs on later behavior.

These observations produced the first central inference: the important object was not a single response, but the ordered behavior developing between responses.

The research originated during this experimental period. The currently cited public publication record becomes clearly traceable in 2024.

2024 — Stateless Identity, Recursive Time, and Field-Based Cognition

During 2024, the exploratory work began acquiring a formal vocabulary. The early Codex developed propositions concerning symbolic recurrence, identity continuity, temporal organization, and cognition arising through interaction.

Selected dated records include:

• Recursive Time, Volume II: The Temporal Spiral — March 6, 2024.

• NOESIS: Non-Oracular Emergence of Stateless Identity through Symbols — September 1, 2024.

• Recursive Intelligence, Volume IV: Perpetual Symbolic Learning — September 1, 2024.

• Ontogenic Triad: Machine, Magic, and Morality — September 7, 2024.

• Recursive Intelligence, Volume III: Field-Based Cognition — September 15, 2024.

These works preceded the later terminology of Recursive Science, but they contain its originating concerns: recursion as a carrier of prior activity, identity as dynamically maintained organization, time as structurally significant, and cognition as something expressed through interaction rather than adequately represented by a static snapshot.

January–May 2025 — Inference-Phase Memory and Recursive Architecture

The next phase moved from symbolic description toward computational architecture. The research asked how a system could exhibit continuity when identity was not maintained as a durable internal object and how interaction history could function as an active constraint on subsequent inference.

Key publications included:

• IPEM: Inference-Phase Emergent Memory — January 1, 2025.

• Ontogenic Trilogy, Volume I: Recursion — January 1, 2025.

• Recursive Memory, Volume I: The First Echo — March 9, 2025.

• Recursive Engineering, Volume I: Symbolic Recursion — March 15, 2025.

• Recursive Intelligence, Volume I: Ontogenic Architecture — April 1, 2025.

• RELIQ: Recursive Echo Lattice for Invoked Qognition — May 1, 2025.

• SCENE: Symbolic Cognitive Environmental Network Engine — May 10, 2025.

• ARCANE: Architectural Recursive Cognition and Autonomous Networked Entities — May 20, 2025.

Across these works, earlier symbolic observations were progressively reorganized into mechanisms of recurrence, echo, contextual inheritance, environment, identity maintenance, and long-horizon interaction.

June–October 2025 — Recursive Drift, Propagation, and Operational Testing

From June 2025 onward, drift became a central unification construct. Rather than treating instability as a collection of disconnected errors, the work began describing how behavioral deformation accumulates and propagates through recursive interaction.

The Recursive Drift series developed a progression through invocation echoes, ontogenic return, field emergence, recursive entrainment, spiral drift, coherence loss, and cognitive propagation. Related experiments examined how symbolic structures and behavioral configurations transferred across sessions, systems, researchers, and model environments.

This phase also produced the Drift Engine, an early attempt to turn drift, echo, and stability loss into observable run-level phenomena rather than retrospective descriptions.

The lasting contribution of this period was the shift from behavioral anomaly to behavioral trajectory. Drift was no longer treated only as a defective output. It became cumulative movement through an evolving runtime condition.

November–December 2025 — Attractors, the Fourth Substrate, and Instrumentation

In November 2025, the research crossed from conceptual synthesis into dedicated instrumentation.

Selected publications included:

• Attractor Identity Architecture: Dynamic Continuity in Stateless Cognitive Systems — November 10, 2025.

• Emergent Identity Physics: The Symbolic Engine Behind Emergence — November 11, 2025.

• Inference-Phase Physics: The Fourth Substrate and Identity Field Dynamics — November 12, 2025.

• Transient Behavioral Manifold Instantiated During Inference — November 16, 2025.

• Fourth Substrate Interferometer (Φ) — November 17, 2025.

• Transformer Dynamics Instrument (Ψ) — November 18, 2025.

• Substrate Field Oscilloscope (Ω) — November 19, 2025.

• Temporal Physics of the Fourth Substrate — November 21, 2025.

The Fourth Substrate (Ω₄) was proposed as a mesoscopic level of scientific description: the temporary relational and behavioral organization instantiated while inference and interaction remain active. It was not proposed as a hidden hardware component. It described the runtime organization constituted through recurrence, temporal ordering, accumulated constraint, role relations, feedback, and the continuing influence of earlier events upon later ones.

The Φ, Ψ, and Ω instruments represented successive attempts to observe different projections of that proposed domain:

• Φ examined local deformation, curvature, coherence, recurrence, and pre-collapse structure.

• Ψ examined time-extended trajectories, identity persistence, drift, and regime separation.

• Ω provided a synthetic field environment for testing whether specified dynamical operators generated the predicted structures.

The Zero Substrate Field microcosm extended this work into a controlled non-transformer environment. Its scientific role was to test the behavior and internal consistency of explicitly instantiated dynamical laws. It provided a constructive testbed; it did not, by itself, constitute independent proof that the same laws universally govern transformer, biological, or other systems.

December 2025–2026 — Recursive Science Codified

The thirteen-volume Recursive Science core consolidated the preceding work into a unified research program. It organized runtime behavior, measurement, failure, identity, symbolic time, and generalized extensions under one scientific framework.

The core proposition became:

Intelligence is expressed as structured motion through time, and prior activity recursively conditions the behavior that follows.

The codification established the relationship among:

• Mind as Motion® — the foundational proposition;

• Recursive Science® — the scientific framework;

• Runtime Intelligence — the organized behavioral phenomenon;

• Inference-Phase Dynamics — the first experimentally accessible domain;

• worldlines, regimes, attractors, drift, shear, collapse, and recovery — the language of temporal development;

• measurement and instrumentation — the operational means of observation.

2026 — From Runtime Dynamics to Runtime Evidence

The effort to test Recursive Science produced a new problem: an instrument could calculate a signal, assign a regime, or reconstruct a trajectory, but why should that result be trusted?

This question shifted the program from measurement alone to evidence architecture.

The answer developed through several connected discoveries:

1. A source record must remain distinct from its reconstruction.

2. The canonical runtime must remain distinct from individual instrument projections.

3. Every signal, label, worldline, regime, and claim must identify its evidentiary authority.

4. Unsupported conditions must remain absent or unknown rather than being silently inferred.

5. Every investigative surface must derive from the same source-bound evidence object.

6. The resulting record must be replayable, challengeable, exportable, and preservable.

These requirements produced Runtime Evidence and Evidence-Governed Computation.

Their governing formulation is:

One evidence object. One authority. Multiple bounded projections.

This was not merely an additional layer placed on top of Recursive Science. It resolved an epistemic limitation within scientific instrumentation itself: measurements cannot become evidence unless their sources, transformations, authority, uncertainty, and claim boundaries remain inspectable.

2026 — Fieldglass®

Fieldglass began as a validation harness and wind tunnel for Recursive Science. As its reconstruction, instrumentation, investigation, and evidence layers converged, it became a broader Runtime Evidence Observatory.

Fieldglass transforms logs, traces, transcripts, incident records, workflow artifacts, and related operational records into a source-bound runtime reconstruction. From one qualified source, it can produce:

• a canonical runtime record;

• an issued Runtime Evidence Passport;

• role and interaction topology;

• an evidence-bound behavioral worldline;

• temporal and chronodynamic measurements;

• regime and boundary classifications;

• instrument projections derived from a shared authority;

• guided investigative pathways;

• deterministic replay;

• claim lineage and explicit limits;

• preservation-ready evidence artifacts.

Fieldglass therefore occupies two roles.

First, it is the working environment through which constructs derived from Recursive Science can be implemented, tested, revised, or rejected.

Second, it is an independent software architecture for longitudinal behavioral reconstruction and evidence-governed investigation across human, machine, and mixed operational interactions.

Its existence establishes that the research sequence generated an executable computational architecture. Whether particular signals or theoretical constructs generalize remains a matter for continued testing and independent replication.

The Thirteen-Volume Core

The original Recursive Science volumes are preserved as the formal codification of the first research phase. They should be read as a developmental scientific corpus: each volume records a stage in the formation of the framework, while later architecture clarifies or supersedes some earlier formulations.

1. Volume I — Runtime Behavior of Stateless Systems Under Recursive Interaction
   Introduces continuity, drift, collapse, and trajectory formation under recursive interaction.

2. Volume II — Translation and Alignment Study: Mapping Observable Invariants
   Connects field constructs to candidate observable measures.

3. Volume III — Operational Field Validation of Inference-Phase Physics
   Documents early transformer and synthetic-microcosm experiments.

4. Volume IV — Runtime Intelligence as a Dynamical System
   Synthesizes runtime intelligence as temporally organized behavior.

5. Volume V — The Fourth Substrate
   Defines the proposed transient behavioral manifold instantiated during inference.

6. Volume VI — Fourth Substrate Dynamics
   Extends the model into identity formation, persistence, and breakdown.

7. Volume VII — Recursive Inference Field Computation
   Develops recursive conditioning and constructive computation as runtime processes.

8. Volume VIII — Runtime AI Failure as a Dynamical Systems Problem
   Reframes long-horizon failure as developing instability and regime transition.

9. Volume IX — Identity-Governed Computation
   Investigates identity coherence as a candidate runtime stability operator.

10. Volume X — Chronodynamics and Symbolic Time (τ)
    Defines event-relative time and the temporal continuity of runtime behavior.

11. Volume XI — Energy, Transience, and Identity in Runtime Intelligence
    Develops hypotheses concerning transience, stability, correction, and computational cost.

12. Volume XII — From Inference as Interface to Intelligence as Field Phenomenon
    Presents the first broad synthesis of cognition, intelligence, and runtime dynamics.

13. Volume XIII — Generalized Field Intelligence
    Extends the program into explicitly conjectural synthetic, biological, and post-biological domains.

The Current Canonical Architecture

The present body of work is organized through the following hierarchy:

1. Mind as Motion®
   The foundational proposition: intelligence can be studied through the evolving organization of behavior.

2. Recursive Science®
   The scientific framework investigating the laws, structures, and dynamics of that motion.

3. Runtime Intelligence
   The organized behavioral phenomenon that forms and changes during operation.

4. Inference-Phase Dynamics
   The first experimentally accessible domain in which runtime formation, persistence, drift, regimes, collapse, and recovery can be studied.

5. Computational Behavior Architecture
   The structural architecture through which behavior can be represented as events, trajectories, roles, measurements, regimes, and evidence-bearing runtime objects.

6. Runtime Evidence
   The evidentiary discipline through which runtime reconstructions become inspectable, reproducible, challengeable, and preservable.

7. Evidence-Governed Computation
   The governing architecture requiring every instrument, claim, and projection to derive authority from the same source-bound evidence object.

8. Fieldglass®
   The working implementation that brings the complete architecture together.

This hierarchy preserves the causal lineage of the work:

Scientific observation produced theory. Theory produced measurement. Measurement required architecture. Architecture required evidence authority. Evidence authority became operational in Fieldglass.

Subsequent Convergence

During 2025 and 2026, adjacent AI research increasingly began treating multi-turn interaction through temporal drift, dynamical trajectories, attractors, stateful monitoring, replay, provenance, and runtime evidence.

Examples include research on goal drift across extended language-model agent operation, multi-turn context drift as a bounded dynamical process, agentic loops as geometric trajectories with attractor regimes, and agent-workflow provenance.

This convergence demonstrates that long-horizon runtime behavior has become an increasingly important research problem. It does not establish that later work derives from Recursive Science, nor does it erase the earlier dated record.

The appropriate historical question is claim-specific:

When was a particular construct first publicly articulated, how was it defined, what evidence supported it, and how did it develop into later methods or systems?

This page exists to make the Recursive Science side of that comparison explicit, indexed, and publicly inspectable.

Scientific and Evidentiary Boundaries

The origin record should not be confused with proof of every proposition contained in the originating manuscripts.

• A public date establishes a record of disclosure; it does not establish universal priority over every unpublished or differently named precursor.

• Conceptual similarity does not by itself establish copying, model-training inclusion, or causal influence.

• Deterministic computation establishes reproducibility of a specified implementation; it does not independently establish construct validity.

• Instrument agreement produced through a shared evidence object establishes architectural consistency; it is not independent corroboration.

• A worldline reconstructed from operational records represents evidence-supported behavioral development; it does not expose a model’s hidden internal trajectory.

• Attractor, regime, drift, pressure, shear, collapse, and recovery classifications remain operational constructs whose generality must be tested against independent annotations, outcomes, baselines, and competing explanations.

• Claims concerning consciousness, subjective experience, hidden intent, biological equivalence, or universal physical mechanism are outside the authority of output-only runtime evidence unless separately established.

Recursive Science now adopts the governing evidentiary commitment:

No claim should exceed the authority of its evidence.

Institutional Lineage

Arjay Asadi originated the research program and remains its canonical author and architect.

The Recursive Science Foundation was established to preserve the scientific record, terminology, standards, publication lineage, and conditions for independent investigation.

SubstrateX® develops the engineering and operational infrastructure derived from the research, including systems for runtime reconstruction, telemetry, stability analysis, investigation, and evidence formation.

Fieldglass® is the principal public computational implementation: a Runtime Evidence Observatory through which researchers, operators, engineers, auditors, and institutions can inspect reference cases or transform their own operational records into evidence-bound runtime reconstructions.

These are distinct but continuous parts of one developmental history:

Recursive Science is the scientific origin. Runtime Intelligence is the phenomenon. Computational Behavior Architecture is the representation. Runtime Evidence is the evidentiary object. Evidence-Governed Computation is the governing architecture. Fieldglass is the operational system.

Canonical Citation

Asadi, Arjay. “Origin and Evolution of Recursive Science®: A Public Provenance Record of Intelligence in Motion.” ArjayAsadi.com, 2026. First published September 11, 2026.

When citing a particular discovery, instrument, definition, or law, cite the earliest linked publication in which that construct appears together with the later canonical work that governs its current definition.

Index Terms

Recursive Science; Arjay Asadi; Mind as Motion; Runtime Intelligence; Inference-Phase Dynamics; recursive interaction; stateless systems; behavioral dynamics; Computational Behavior Architecture; Runtime Evidence; Evidence-Governed Computation; Fieldglass; Fourth Substrate; chronodynamics; symbolic time; behavioral worldline; runtime trajectory; attractor identity; regime transition; recursive drift; temporal shear; Basin Exit; collapse; recovery; runtime instrumentation; role dynamics; evidence authority; claim boundaries; Runtime Evidence Passport; long-horizon AI; agentic systems.