A Framework for Longitudinal Computational Behavior & Runtime Intelligence

Longitudinal Computational Dynamics establishes a scientific framework for studying computational behavior as it forms and changes through time. It begins from the premise that learned capability becomes organized during operation—and that the history of a runtime can alter the conditions under which later behavior develops.

The framework investigates how continuity, coherence, drift, constraint, instability, boundary transition, failure, and recovery become observable across an ordered runtime. Its object is neither the isolated output nor an inaccessible mental entity. It is Longitudinal Computational Behavior: the evidence-bound development reconstructed from interactions among models, people, agents, tools, roles, constraints, workflows, and operational events.

Intelligence is expressed not only through what a system can produce, but through how its observable behavior develops across time.

Developed by Arjay Asadi through SubstrateX®, the framework consolidates the scientific foundations underlying Runtime Intelligence, Inference-Phase Dynamics, Computational Behavior Architecture, Runtime Evidence, Evidence-Governed Computation™, and SubstrateX Aperture™.

Foundational Framework · September 2026 · Version 1.0

The Research Problem

Modern AI research provides essential accounts of how models are trained and how inference is computed. Learned parameters constrain what a model can generate. Attention, activations, tokenized context, retrieval, cache state, tool access, and decoding all participate in producing a response.

Longitudinal Computational Dynamics does not reject or replace these mechanisms. It addresses a different level of analysis.

A model used once produces an output. A model operating within a conversation, agent loop, tool workflow, or multi-actor process participates in an evolving runtime.

During that runtime:

• earlier outputs return through context or memory;

• tool results alter later possibilities;

• people introduce corrections and constraints;

• roles transfer, coordinate, or contest authority;

• objectives persist, weaken, or change;

• workflow state carries consequences forward;

• and local decisions accumulate into an ordered trajectory.

These relationships cannot be adequately described by evaluating each output in isolation.

A final response may appear coherent even when the preceding trajectory contains repeated correction failure, accumulating contradiction, role fragmentation, or sustained displacement. A poor response may occur within a runtime that subsequently recovers. The scientific object therefore extends beyond the individual answer to the development connecting events across time.

Longitudinal Computational Dynamics studies the organization that becomes visible when computation is examined as a developing runtime rather than a collection of disconnected outputs.

Intelligence in Motion

Intelligence in Motion™ is the organizing proposition of Longitudinal Computational Dynamics.

In its operational form, motion means ordered change within a declared behavioral coordinate system. The proposition does not describe intelligence as a physical fluid, hidden substance, or entity separate from computation. It directs scientific attention toward the observable organization connecting one event to the next.

The proposition distinguishes two related questions:

1. What capability is available to the system?

2. How is that capability organized and expressed across a particular runtime?

The first question concerns training, architecture, parameters, retrieval, tools, and available context. The second concerns the path realized through those resources during operation.

The scientific value of Intelligence in Motion depends on whether this distinction produces:

• clearly defined variables;

• observable differences;

• reproducible measurements;

• discriminating experiments;

• bounded interpretations;

• and hypotheses that can fail.

Intelligence in Motion is therefore the beginning of an investigation—not its conclusion.

The Scientific Object

The primary object of Longitudinal Computational Dynamics is Longitudinal Computational Behavior: the ordered development of observable computational activity across time.

A runtime trajectory may include:

• source events;

• prompts and responses;

• tool calls and results;

• roles and authority relationships;

• objectives and constraints;

• workflow and environmental state;

• temporal dependencies;

• behavioral signals;

• regime conditions;

• transition markers;

• operational outcomes;

• and evidence-availability states.

When these elements are qualified, ordered, and reconstructed under a declared method, they form an evidence-bound worldline.

A worldline is not a direct observation of hidden model state. It is a time-indexed representation derived from available operational records. Its authority cannot exceed the coverage, quality, provenance, and declared processing conditions of those records.

This distinction makes long-horizon behavior scientifically investigable without claiming access to internal mechanisms that the evidence does not contain.

Four Foundational Commitments

Temporal Extension

Continuity, drift, regime formation, boundary transition, failure, and recovery require an interval. They cannot be established from one output alone.

Runtime investigation must therefore preserve event order, temporal coordinates, dependencies, duration, and the evidentiary horizon across which a pattern is claimed to persist.

Path Dependence

Earlier outputs, tool results, corrections, role actions, workflow changes, and environmental events may become conditions of later computation.

If prior activity matters, modifying, reordering, or removing it should produce testable differences in the trajectories that follow. Those differences must still be evaluated against stochastic variation, source quality, and alternative explanations.

Runtime Organization

The same trained model may express different organized behavior under different contexts, interaction histories, tools, participants, workflow states, and sampled outputs.

Longitudinal Computational Dynamics studies how available capability becomes organized during a particular runtime. It does not claim that capability is created independently of training, architecture, retrieval, memory, tools, or external state.

Evidence Dependence

A description of runtime behavior becomes scientifically meaningful only when its source, transformation, method, support state, uncertainty, and inferential limits can be inspected.

A compelling label is not evidence. A deterministic value is not automatically a valid measure. An untraceable measurement cannot support a durable scientific claim.

How Stateless Calls Produce Longitudinal Behavior

An individual model call need not retain persistent internal memory for the wider runtime to become history-bearing.

History may be carried through:

• model context;

• external memory;

• retrieval systems;

• summaries and prior outputs;

• tool and workflow state;

• role handoffs;

• human corrections;

• operational artifacts;

• permissions and authority;

• and environmental responses.

The computational runtime is therefore better described as potentially parameter-static but operationally stateful. Learned parameters may remain unchanged while the surrounding system repeatedly reintroduces the consequences of earlier computation.

This re-entry operates across several scales:

• Token scale: each generated token conditions what may follow within the generation.

• Turn scale: a response becomes part of a later interaction context.

• Agent and tool scale: plans, actions, observations, and tool results alter subsequent decisions.

• Workflow scale: state, permissions, failures, and handoffs constrain later operations.

• Multi-actor scale: people, models, agents, tools, policies, and environments respond to one another over time.

Path dependence is a testable consequence of this structure. If otherwise comparable runs differ in an earlier correction, tool result, role instruction, or event order, their later behavior may diverge.

The scientific question is not merely whether divergence occurs. It is which earlier conditions matter, how long their influence remains observable, and whether the resulting difference exceeds ordinary variation under the declared experimental conditions.

Runtime Intelligence

Runtime Intelligence names the organized behavioral phenomenon expressed while a computational system operates.

It is not a claim that inference creates intelligence or capability from nothing. Training, architecture, retrieval, tools, memory, policies, and human participation establish the resources and constraints available to a runtime.

Runtime Intelligence concerns how those resources become organized across a particular trajectory.

It may become observable through patterns such as:

• a constraint that persists across many turns;

• a reasoning posture that survives disturbance;

• a role configuration that repeatedly returns;

• a correction that initially holds and later disappears;

• an objective that gradually changes;

• a coordination pattern that fragments across participants;

• or sustained displacement preceding an observable failure marker.

These patterns can be investigated without assigning consciousness, subjective experience, intent, or an enduring self to the system.

Runtime Intelligence occupies a level between low-level model mechanism and final task outcome: the level of organized behavior during operation.

Inference-Phase Dynamics

Inference-Phase Dynamics is the first experimentally accessible domain of Longitudinal Computational Dynamics.

It studies the formation, persistence, deformation, transition, loss, and recovery of observable behavioral organization during inference and sustained interaction.

Its evidence may include:

• prompts and responses;

• system instructions when available;

• timestamps and event order;

• role labels;

• tool activity;

• corrections and evaluator markers;

• external state changes;

• workflow events;

• and operational outcomes.

From these records, investigators can examine whether behavior remained bounded, accumulated displacement, entered a persistent configuration, approached a candidate boundary, crossed a declared observable threshold, or sustained recovery beyond a temporary correction.

Inference-Phase Dynamics begins with AI systems because their operational records can often be preserved and replayed. Transfer to human, animal, biological, or other socio-technical domains cannot be assumed. Each domain would require its own definitions, measurement validity, ethical constraints, and independent evidence.

The Core Constructs

Longitudinal Computational Dynamics develops a language for describing runtime behavior through time.

• Worldlines represent ordered, evidence-bound runtime trajectories.

• Operational time represents event order, duration, recurrence, dependency, and transition within the available record.

• Regimes describe persistent conditions of runtime organization under declared classification criteria.

• Attractor proxies represent recurring configurations toward which observable behavior appears to converge or return.

• Containment describes persistence within a declared behavioral boundary.

• Drift describes cumulative displacement from an explicitly defined reference condition.

• Curvature describes change in trajectory direction within a declared coordinate system or projection.

• Pressure describes accumulated competing demands, contradictions, or unresolved constraints through defined signals.

• Shear describes divergence or phase lag between coupled runtime dimensions.

• Basin Exit identifies a qualifying observable boundary crossing at t* under a declared method.

• Collapse describes persistent loss of an established runtime organization under defined criteria.

• Recovery requires sustained re-entry into, or reorganization toward, a supported recovery condition.

• Role topology represents changing relationships among people, models, agents, tools, policies, and other participants.

These are analytical constructs. Where they are expressed mathematically, their coordinates, scales, transformations, reference conditions, uncertainty, and validation status must be declared. Without that support, they remain disciplined conceptual models rather than established physical quantities.

No single signal is inherently diagnostic of instability, failure, or recovery. Interpretation depends on the evidence available for the run, the declared instrument contract, persistence requirements, contextual information, and independent validation.

Evidence-Governed Computation

Longitudinal reconstruction creates an additional problem: computation can produce measurements and interpretations more easily than evidence can authorize them.

Evidence-Governed Computation™ is the architectural discipline developed to constrain that gap.

Its governing law is:

No claim should exceed the authority of its evidence.

The architecture distinguishes:

• source records from transformed observations;

• observations from measurements;

• measurements from findings;

• findings from interpretations;

• interpretations from legitimate claims;

• and presentation from evidentiary authority.

It also requires unavailable evidence to remain unavailable. An instrument, interface, summary, or export cannot create support merely by rendering a value or explanation convincingly.

The governing relationship is:

Source → Observation → Measurement → Finding → Legitimate Claim → Projection → Export and Preservation

Claim legitimacy means admissibility under declared evidence and method rules. It does not certify that a claim is objectively true.

The Scientific and Operational Hierarchy

The body of work contains connected layers that must not be treated as synonyms.

1. Intelligence in Motion™
   The organizing proposition that intelligence can be studied through the evolving organization of observable behavior.

2. Longitudinal Computational Dynamics
   The scientific framework for investigating the structures, temporal relationships, and dynamics of that development.

3. Longitudinal Computational Behavior
   The primary object of study: ordered, path-dependent computational behavior reconstructed across time.

4. Runtime Intelligence
   The organized behavioral phenomenon expressed during operation.

5. Inference-Phase Dynamics
   The first empirical domain for studying runtime formation, drift, regimes, boundaries, failure, and recovery.

6. Computational Behavior Architecture
   The representational architecture for trajectories, roles, events, measurements, regimes, markers, and evidence-bearing runtime objects.

7. Runtime Evidence
   The source-bound evidence object and discipline through which reconstruction becomes inspectable, reproducible, challengeable, and preservable.

8. Evidence-Governed Computation™
   The authority architecture constraining observations, instruments, projections, interpretations, claims, and exports to their supporting evidence.

9. SubstrateX Aperture™
   The Runtime Evidence Observatory through which the integrated architecture becomes executable and investigable.

SubstrateX® is the institutional and engineering home of this complete program. The conceptual dependency moves through the hierarchy toward implementation. Evidence produced through implementation returns to the research, where it may support, refine, narrow, or contradict the scientific propositions.

Implementation closes the experimental loop. It does not convert implementation into proof.

The Distinctive Contribution

Dynamical approaches to cognition have an established intellectual history. Context dependence, autoregressive conditioning, entropy, similarity, variance, trajectory analysis, observability, and provenance are also not new by themselves.

Longitudinal Computational Dynamics does not depend on claiming otherwise.

Its distinctive contribution is the integration of:

• longitudinal behavioral reconstruction;

• operational and symbolic time;

• role and interaction topology;

• worldlines and regime analysis;

• stability, drift, pressure, boundary formation, failure, and recovery;

• source-bound evidence authority;

• bounded instrument contracts;

• deterministic evidence artifacts;

• explicit availability and claim boundaries;

• guided investigation;

• replay;

• export;

• and preservation.

These components form a continuous path from scientific proposition to externally inspectable computational evidence.

The originality of the program rests in this integrated architecture and the research object it makes available—not in asserting ownership over every underlying concept or method it employs.

Research Status

The program contains four kinds of contribution that must remain distinct:

1. Foundational propositions about studying intelligence as temporally organized behavior.

2. Theoretical constructs including worldlines, regimes, attractor proxies, boundaries, and the runtime trajectory.

3. Implemented architecture including canonical runtime records, the Current Evidence Run, evidence ledgers, Passports, instrument contracts, guided investigation, deterministic reconstruction, replay, and export.

4. Empirical hypotheses that remain open to confirmation, revision, narrowing, or rejection.

SubstrateX Aperture™ demonstrates that the computational architecture can be implemented and that the same qualified source, declared context, method version, configuration, and processing conditions can produce the same computed evidence object and deterministic artifact.

It does not independently establish that every signal measures its intended construct, that every regime generalizes across systems, that every boundary has prospective value, or that the proposed dynamics are universal.

Those claims require:

• controlled perturbation;

• baseline comparison;

• stable negative cases;

• ablation;

• held-out evaluation;

• prospective testing;

• threshold and false-positive calibration;

• cross-system transfer;

• and independent replication.

Retrospective reconstruction must precede prospective early-warning claims. Formal Lead-Time exists only when both a qualifying observable boundary crossing at t* and an independently supplied failure marker at tf are available.

What Would Validate the Program

The framework establishes a testable research agenda. It should produce evidence that:

• earlier runtime events create persistent, measurable effects beyond sampling variation;

• trajectory measurements discriminate meaningful temporal organization from simpler output-level baselines;

• regime classifications describe sustained structure beyond arbitrary thresholding;

• candidate boundaries precede independently defined changes under prospective criteria;

• recovery can be distinguished from a temporary or one-turn correction;

• role-aware reconstruction outperforms role-blind analysis where role structure materially affects the runtime;

• qualified precursor markers provide useful warning at acceptable false-alarm rates after prospective validation;

• selected measurements retain meaning across systems without unrestricted retuning;

• independent implementations can reproduce declared computations from the same qualified evidence;

• and all projections derived from one evidence authority remain logically and numerically consistent.

Failure to meet these conditions would require the relevant constructs to be revised, narrowed, reclassified, or rejected.

That possibility is essential. The program becomes scientific only to the extent that its propositions remain open to failure.

Claim Boundaries

The framework supports investigation of observable computational behavior and its longitudinal organization.

It does not, from operational records alone:

• reveal hidden model state;

• identify a unique internal mechanism;

• establish intent;

• prove consciousness, subjective experience, or enduring identity;

• determine root cause from chronology alone;

• convert analytical geometry into literal physics;

• make every deterministic measurement scientifically valid;

• establish prospective prediction from retrospective pattern reconstruction;

• or generalize automatically from AI records to human, animal, or biological cognition.

These boundaries are not qualifications added after the fact. They are part of the architecture itself.

They preserve an ambitious research object without depending on metaphysical agreement, inaccessible evidence, or authority claimed beyond the record.

From Foundation to Investigation

Longitudinal Computational Dynamics begins with a simple proposition:

Intelligence can be studied through the structured development of observable behavior across time, and the history of a runtime can alter what follows.

SubstrateX turns that proposition into temporal coordinates, worldlines, regime models, measurement contracts, evidence authorities, scientific instruments, guided investigations, and working infrastructure.

Its significance will not be established by the coherence of its vocabulary or the scale of its implementation alone. It will be determined by whether its constructs produce reliable, discriminating, and useful knowledge about runtime behavior beyond what simpler methods already provide.

SubstrateX Aperture™ makes that question operationally testable.