Capability as a First-Class Computational State
Conventional distributed systems treat the question of whether a node can execute a task as something assumed, inferred from resource availability, or conflated with authorization. A task is dispatched, and the node either executes it or returns a failure. The disclosed architecture instead introduces capability as a first-class computational state variable. Capability is a structural condition describing whether an executable form of a given objective can exist on a given execution substrate. It is not a metric, a score, a probability, or a heuristic assessment. It is a computed determination, derived from the structural characteristics of the substrate, the structural requirements of the objective, and the current state of the execution environment.
That determination resolves to one of a bounded set of determinate outcomes: execution is structurally possible, execution is structurally impossible, execution is structurally deferred, or execution must be rerouted to an alternative substrate. Each of these is a valid computational result. None is an error condition, a timeout, or a default. Resource availability, in the disclosed model, is a necessary but insufficient condition for capability: a node may have ample memory and compute cycles yet lack the instruction set, accelerator type, model weights, sensor array, or physical actuator required to produce an executable form of the objective.
Capability occupies a designated position in the architecture that is evaluated prior to the construction of any executable process. The system does not build an execution plan and then check whether the plan can run. It first determines whether any executable form of the objective can exist on the candidate substrate, and only if that determination resolves affirmatively does it proceed to execution synthesis. This pre-synthesis ordering is enforced at the architectural level and eliminates an entire class of computational waste: the construction of execution plans that cannot be carried out.
The Joint Capability, Time, and Uncertainty Condition
The capability determination is conditioned on three jointly evaluated dimensions: the structural capability of the execution substrate, the forecasted temporal window during which execution could occur, and the explicit uncertainty associated with the capability and temporal assessments. These three dimensions are not evaluated independently. They are evaluated as a joint condition that must be simultaneously satisfied for execution synthesis to proceed. A substrate may possess the structural capability to execute an objective but lack a temporally viable window. It may possess both structural capability and a viable window, yet the uncertainty associated with one or both assessments may exceed the threshold at which synthesis is warranted.
In the capability determination subsystem, an objective node is presented for evaluation along two parallel pathways: a capability envelope module that evaluates the candidate substrate's structural characteristics against the objective's requirements, and a governance policy module that evaluates authorization independently. Both feed a joint evaluation gate that combines the independent capability and governance determinations. From that gate, three outcomes diverge: an execution synthesis pathway for objectives that satisfy both capability and governance conditions, a non-synthesis determination for objectives for which no executable form can exist, and a deferred state for objectives whose capability conditions are projected to be satisfied at a future time.
Execution Substrates and Capability Envelopes
Each execution substrate advertises a capability envelope: a structured data object describing the substrate's current structural characteristics along a plurality of defined dimensions. The envelope is not a permission list, a service catalog, or a self-reported performance benchmark. It is a formal description of the substrate's affordances. The disclosed dimensions include compute class (processor type, instruction set architecture, word size, vectorization), memory architecture (addressable memory, bandwidth, cache topology, coherency model, supported access patterns), model access (the inference models, knowledge bases, embedding spaces, and learned representations loaded or loadable on the substrate), locality (geographic region, network position, latency characteristics, jurisdictional classification), execution guarantees (reliability, redundancy, checkpoint-and-restore, real-time scheduling, deterministic execution mode), and sensor and actuator interfaces for embodied substrates.
Capability envelopes are dynamic. When hardware is provisioned or deprovisioned, when models are loaded or unloaded, when network conditions change, or when other agents consume or release shared resources, the envelope is updated to reflect the substrate's current state. The system does not rely on stale or statically configured capability information. Each dimension is represented as a structured descriptor with defined semantics, value ranges, and comparison operators, which enables formal matching between an objective's requirements and a substrate's envelope. A capability match is achieved when every requirement dimension is satisfied. When one or more dimensions are unsatisfied, the specific unsatisfied dimensions are recorded and propagated to the routing, deferral, and decomposition subsystems.
Capability-Native Computation
The system implements capability-native computation: a paradigm in which the determination of whether execution can occur is itself a first-class computation producing structured, auditable, and governable results. It is not a pre-check or a guard clause that precedes the real computation. It is the real computation in its initial phase. The evaluation extracts the objective's capability requirements from the agent's intent field and context block, producing a structured requirements vector; retrieves the candidate substrate's current capability envelope; and performs a dimension-by-dimension comparison.
The match result for each dimension is not binary. Each dimension produces one of three outcomes: satisfied, where the substrate meets or exceeds the requirement; unsatisfied, where it falls short in a manner that cannot be resolved through temporal deferral or reconfiguration; or conditionally satisfiable, where it currently falls short but could be brought into satisfaction through temporal deferral, reconfiguration, or partial decomposition. The aggregate determination follows a defined composition rule: all dimensions satisfied yields structurally possible; one or more unsatisfied with no conditional path yields structurally impossible; one or more conditionally satisfiable within a bounded horizon yields structurally deferred; and unsatisfied on the candidate but satisfied on a known alternative yields rerouted. The computation produces a structured capability determination record, persisted in the agent's lineage, recording the substrate identity, the extracted requirements, the retrieved envelope, the per-dimension results, the aggregate determination, the associated uncertainty bounds, and any forecasted conditions under which the determination may change.
Temporal Executability Forecasting
Temporal executability forecasting projects bounded future time windows during which execution could occur on a given substrate, conditioned on the substrate's capability trajectory and the objective's temporal requirements. It is not reactive scheduling that waits for capability to become available. It is predictive computation. The forecast projects the trajectory of each envelope dimension forward in time, modeling each from known scheduled events (provisioning, model deployment, maintenance windows, reservation expirations), observed trends (consumption rates, load patterns, degradation curves), and declared constraints (lease durations, reservation boundaries, policy-mandated availability windows). The temporal executability window is the intersection of the intervals during which all required dimensions simultaneously satisfy the objective's requirements.
The forecast distinguishes three temporally conditioned states: immediate executability, where the capability-time intersection exists now; deferred executability, where it does not exist now but is forecasted within a bounded future window; and temporal impossibility, where no intersection is forecasted within the horizon. Distinguishing deferred executability from temporal impossibility prevents the system from indefinitely deferring execution on a substrate that will never become capable. The forecast produces confidence-bounded window estimates rather than point predictions: rather than predicting capability at a single time, it predicts a window bounded by an earliest and a latest time with a confidence level derived from the uncertainty model. The subsystem continuously recomputes its projections as envelopes change, reservations are created or canceled, or observed consumption deviates from the modeled trend.
Uncertainty as a Propagated Variable
Uncertainty participates as a first-class propagated variable alongside capability and time. It is not a residual or an error term but an explicit, bounded epistemic condition that the system computes, maintains, propagates, and consults at every stage. The uncertainty bound encodes the capability evaluation infrastructure's confidence in its own assessment of whether a substrate can execute an objective within a given window. This is distinct from the agent's own confidence, which is governed by the confidence governor disclosed elsewhere: the confidence governor evaluates whether the agent should execute, while the uncertainty bound evaluates whether the system's assessment that the agent can execute is itself reliable.
The uncertainty model differentiates three epistemic categories. Deterministic impossibility: the determination has resolved to structurally impossible with negligible uncertainty, so the system routes, decomposes, or reports non-executability without deferring or retrying. Deferred possibility with confidence bounds: the determination has resolved to structurally deferred and a future window is identified, but the forecast carries non-negligible uncertainty, so the system may defer while propagating the bounds so that contingency plans can be prepared. Indeterminate feasibility: the determination cannot be resolved with sufficient confidence due to incomplete information, so the system must gather more information, invoke a higher-authority evaluation, or report the indeterminate condition to governance. Uncertainty is propagated through the pipeline so that downstream decisions inherit the uncertainty of their inputs and accumulate it, rather than silently discarding it. The system maintains an uncertainty ledger, persisted in the agent's lineage, recording the uncertainty state of each active determination, forecast, and synthesis decision.
Execution Synthesis and Non-Synthesis
Execution synthesis constructs an executable form of an objective on a substrate, conditioned on the joint satisfaction of the three conditions: the capability determination has resolved to structurally possible (or transitioned from deferred to possible on arrival of the forecasted window), the temporal executability window is currently open, and the aggregate uncertainty is below the configured threshold. If any condition is unmet, synthesis does not occur. That condition is designated non-synthesis, and it is treated as a valid computational result, not an error, timeout, or failure. Non-synthesis is a positive determination: the system has computed that non-execution is the appropriate result, and it records, reports, and acts on this with the same rigor it applies to successful synthesis. An agent that receives a non-synthesis determination has received useful information enabling rerouting, deferral, decomposition, or objective revision.
The non-synthesis record captures the evaluated substrate, the objective, the specific capability dimensions that were unsatisfied or conditionally unsatisfiable, the temporal conditions that were unmet, the uncertainty conditions that exceeded the threshold, and, where determinable, whether non-synthesis is permanent, temporal, conditional, or indeterminate. When synthesis proceeds, it constructs a capability execution plan: a directed acyclic graph of computational steps synthesized dynamically from the objective's requirements, the substrate's envelope, and the temporal window constraints, rather than drawn from a stored template. A final pre-commitment validation re-evaluates the capability determination and temporal forecast before submission, and may abort and return a non-synthesis determination if the conditions that obtained when synthesis began no longer hold.
Disclosure Scope
The capability, time, and uncertainty architecture described here, comprising capability as a first-class computational state resolving to structurally possible, impossible, deferred, or rerouted; the architectural separation of capability from permission and the resulting four-quadrant model; the capability envelope advertised per substrate across compute class, memory architecture, model access, locality, execution guarantees, and sensor and actuator dimensions; capability-native computation with per-dimension satisfied, unsatisfied, and conditionally satisfiable results composed into an aggregate determination record; temporal executability forecasting producing confidence-bounded windows and the immediate, deferred, and temporally impossible states; uncertainty as a propagated variable differentiating deterministic impossibility, deferred possibility with confidence bounds, and indeterminate feasibility, maintained in an uncertainty ledger; and execution synthesis conditioned on the joint satisfaction of all three with non-synthesis as a valid structured result, is disclosed in the cognition filing (U.S. Application No. 19/647,395). This article describes that disclosed mechanism. The scope extends to embodiments in which the capability envelope encompasses physical affordances of embodied and robotic substrates and the assessed affordances of human operators, and to embodiments in which capability modulates discovery traversal, feeds the confidence computation, and governs substrate resource negotiation, provided the capability, time, and uncertainty determination remains a pre-synthesis computation producing determinate, auditable results.
