We are making machines increasingly intelligent while leaving their environments computationally primitive.

1Intelligence Is Advancing Faster Than Infrastructure

Large models can reason across complex objectives. Agents can decompose goals, call software tools, maintain state, and coordinate multi-step workflows. VLA architectures connect language, perception, and action. Robotics continues to improve in manipulation, navigation, sensing, and autonomy.

But when an autonomous system crosses from digital computation into an ordinary physical environment, a different problem appears. Suppose an agent determines that a medical supply, manufacturing component, laboratory reagent, replacement part, or tool is required. The semantic problem may already be solved: the system knows what is needed. The physical problem has only begun. Where is it? Which instance should be selected? Is its location known? Is it available? Which physical endpoint corresponds to the requested resource? How should a human or machine actor be directed to it?

In many environments, these questions are not exposed through a generalized infrastructure interface. Instead, the actor reconstructs the answer through perception, search, application-specific databases, or human knowledge.

Knowing what is needed is not equivalent to resolving where it exists.

2The Physical World Has an Addressability Problem

Modern computing became scalable partly because resources became addressable. Memory has addresses. Files have paths. Databases expose keys and queries. Networked services expose interfaces. Software agents inherit this advantage: they operate over abstractions that allow digital resources to be named, queried, resolved, and invoked.

Physical resources rarely possess an equivalent generalized abstraction. A box of syringes, aircraft component, warehouse bin, surgical kit, manufacturing fixture, or laboratory instrument may exist physically and may also have a digital record. Representation alone, however, does not make the resource operationally addressable from intent. An inventory system may record existence. RTLS may estimate location. RFID may identify an object. A camera may perceive it. A warehouse system may associate an item with a bin. A robot may search for it. These are valuable capabilities, but they do not by themselves define a common abstraction from expressed intent to a resolved physical endpoint.

Intent → Resource Resolution → Physical Endpoint → Actuation

3Existing Location-Directed Systems and the IAPI Boundary

The idea of infrastructure directing a person to a physical location is not new. Pick-to-light systems, warehouse management systems, RFID, RTLS, electronic kanban systems, and other location-directed technologies already associate known resources or structured transactions with physical locations. IAPI does not claim to originate those capabilities. The proposed distinction lies in the abstraction boundary.

In a conventional pick-to-light workflow, the resource has typically already been identified by an upstream system: a structured order specifies an item, the system maps that item to a known bin, and hardware indicates the corresponding location. The guidance mechanism executes an already-resolved transaction. IAPI moves the interface upstream from a predefined transaction to intent itself. A human, software agent, or autonomous machine can express an objective without necessarily specifying the exact physical resource or location required to satisfy it. Infrastructure can then evaluate that intent against resource identity, context, policy, state, and spatial mapping; resolve an appropriate resource and physical endpoint; and expose the result to the requesting client or downstream actor.

This also changes who can consume the infrastructure. The same resolution layer can, in principle, serve a human requesting guidance, a software agent querying physical state, or a robot requiring a machine-readable endpoint. IAPI therefore need not replace pick-to-light, RTLS, RFID, computer vision, warehouse management systems, or robotics; these technologies can operate beneath or adjacent to an intent-addressable architecture. The category-level claim is narrower: physical resource resolution can be exposed as a client-agnostic infrastructure capability rather than embedded separately within each application or actor.

4Intent-Addressable Physical Infrastructure

Intent-Addressable Physical Infrastructure (IAPI) is an architectural framework in which physical environments expose resources through an intent-resolution layer capable of translating human or machine objectives into actionable physical endpoints. The actor specifies what is needed. Infrastructure participates in determining which resource satisfies that intent and where its actionable endpoint is. The endpoint can then be exposed, signaled, illuminated, returned as structured data, or passed to a downstream physical system.

In the conventional model, spatial resolution is primarily an actor or application responsibility. In IAPI, resource resolution can become an infrastructure service.

Actor / Agent → Intent → Resource Resolution → Physical Endpoint → Actuation

5A Grounding Example: The Clinical Supply Room

Consider a clinician entering a supply room and saying, "I need a 20-gauge IV catheter." The clinician has expressed the required outcome but has not supplied a SKU, database key, bin identifier, or coordinate. An intent-addressable system can interpret the request, evaluate it against available resource identities, local context, policy, known state, and spatial mapping, and resolve the appropriate resource to a physical endpoint. For the clinician, that endpoint might be exposed by illuminating the exact storage location. For a robotic system requesting the same resource, the resolution might instead be returned as a machine-readable endpoint.

The important abstraction is not the illumination mechanism. It is that the requesting actor specifies the need while infrastructure performs resource resolution. Different clients can consume the same underlying physical resource layer through different actuation interfaces.

6IAPI Is Complementary to Physical AI

IAPI should not be interpreted as an alternative to Physical AI. Physical AI addresses questions such as: how does a machine perceive, navigate, manipulate, and physically execute? IAPI addresses a different question: how can an environment resolve an expressed objective against known physical resources and expose the appropriate physical endpoint?

Without addressable infrastructure, a robot may need to determine both WHERE the correct component is and HOW to retrieve it. With IAPI, those responsibilities can potentially separate: infrastructure participates in resolving where; Physical AI determines how to execute. Resource resolution and kinetic execution are different computational problems. There is no architectural requirement that both always be solved by the same system.

7From Passive Space to Computational Space

The deeper implication of IAPI is that the physical environment becomes part of the computing architecture. Shelves, bins, rooms, racks, and workspaces need not remain merely passive containers around intelligent actors. In an intent-addressable environment, physical resources acquire computational identity; location becomes resolvable; context can influence selection; state can be maintained locally or through distributed systems; and resolved endpoints can expose bounded physical behavior.

The environment is no longer merely where computation terminates. The environment participates in computation.

Making physical infrastructure programmable does not mean turning every object into a general-purpose computer. It means enabling software to express an objective, allowing infrastructure to resolve that objective under explicit state and confidence constraints, and then exposing a physical endpoint for controlled actuation.

8The Spatial Nervous System

A useful analogy is a nervous system, but the metaphor must be used carefully. IAPI is not a literal biological analogue, nor does it require centralized intelligence. The analogy describes functional separation. Agentic AI can provide cognition and planning. Physical AI can provide embodiment and kinetic execution. IAPI proposes environmental resource resolution and addressability as a distinct infrastructure function.

Agentic AI → IAPI → Physical AI

This is a conceptual representation, not a mandatory serial implementation. Real systems will contain feedback loops, shared state, perception, direct machine interfaces, and overlapping responsibilities. The useful question is architectural: who resolves digital intent into a physical endpoint?

The "spatial nervous system" is a conceptual metaphor, but the underlying engineering problem it describes is concrete. As autonomous digital systems increasingly interact with physical environments, a persistent gap remains between expressing an intent and resolving that intent to a known, actionable physical resource. Bridging that gap requires mechanisms for resource identity, state, spatial resolution, context, confidence, policy, and controlled actuation.

9Addressability Requires State

A physical endpoint is useful only if the infrastructure's representation of reality remains sufficiently accurate. Objects move. Inventory is consumed. Tools are returned incorrectly. Bins become empty. Sensors fail. Mappings drift. IAPI therefore cannot be reduced to a static database connecting identifiers with coordinates. A viable implementation requires mechanisms for state integrity. Depending on the environment, these may include deterministic placement rules, transaction events, weight or presence sensing, RFID, computer vision, RTLS, human confirmation, robotic verification, inventory-system integration, edge inference, or combinations of these approaches.

Different environments tolerate different uncertainty. Fixed tooling in a manufacturing cell presents a different state-maintenance problem from rapidly consumed inventory in a hospital supply room.

Addressability is only as trustworthy as the state model supporting it.

IAPI therefore does not eliminate perception. Perception can instead help establish and validate environmental state rather than forcing every actor to perform complete resource rediscovery for every interaction.

10Probabilistic Resolution, Deterministic Actuation

Physical systems impose constraints that ordinary digital recommendation systems often do not. Directing a clinician, technician, robot, or autonomous process toward the wrong resource can have operational consequences. An intent-addressable architecture should therefore distinguish probabilistic interpretation from deterministic execution. Intent classification, contextual inference, and state estimation may be probabilistic. Resolution should be constrained by explicit confidence, policy, and state conditions. When those conditions are not satisfied, the system should decline to actuate, request clarification, or invoke additional verification.

Once an endpoint has been accepted for action, the transition from that resolved endpoint to hardware behavior should be bounded, inspectable, and deterministic.

Probabilistic understanding → constrained resolution → physical endpoint → deterministic actuation

11Edge Computing Changes the Deployment Model

Historically, making physical environments computationally active could require expensive instrumentation, centralized infrastructure, or continuous cloud connectivity. Modern microcontrollers increasingly support embedded machine learning, quantized inference, local sensor fusion, low-power wireless networking, local state, and deterministic control. This makes distributed intelligence at the infrastructure edge increasingly practical. Rather than replacing every shelf, rack, cabinet, workstation, or storage system with an entirely new robotic environment, modular systems can potentially retrofit intelligence onto existing assets.

Passive asset → instrumented asset → state-aware asset → addressable asset → intent-addressable infrastructure

12Humans, Agents, and Robots as Clients of the Same Resource Layer

The requesting actor need not be human. A clinician, technician, software agent, or robot can express an intent. If infrastructure exposes a sufficiently abstract resolution interface, these actors become different clients of the same underlying physical resource layer. For a human, the resolved endpoint might produce illumination or guidance. For a mobile robot, it might return structured location and state. For an autonomous material-handling system, it might trigger another machine interface. The presentation changes; the underlying resource-resolution abstraction does not.

This creates a potential common interface between computational intent and physical resources: what resource satisfies this objective, what is its state, and where is its actionable endpoint?

13The Architectural Boundary

On one side sits intent: human requests, agent objectives, machine goals, and workflow requirements. On the other sits physical execution: human retrieval, robotic manipulation, autonomous transport, signaling, illumination, and actuation. Between them lies the resolution problem: which physical resource satisfies the intent, what is its current state, and where is its actionable endpoint?

The implementation can vary radically across domains. IAPI does not require one model, protocol, sensor modality, microcontroller, or form of actuation. It is proposed as an architectural category rather than a product specification.

14What IAPI Does Not Solve

A rigorous architecture must define its limits. IAPI does not eliminate the open-world robotics problem, guarantee accurate physical state, remove manipulation complexity, replace semantic reasoning, make every environment deterministic, or eliminate sensing, localization, perception, safety systems, or human oversight. Nor does it imply that every physical object should be electronically instrumented.

IAPI identifies a narrower systems opportunity: where physical resource state can be established with sufficient confidence, resource resolution can be elevated from an actor-level search problem into an infrastructure-level service. Its value should be judged empirically: latency, resolution accuracy, state integrity, reliability, cognitive burden, computational efficiency, interoperability, safety, and total system economics.

15Implications for Agentic and Physical AI

The Physical AI era is often framed as a race to build increasingly capable embodied machines. That is only one side of the systems problem. Performance can also improve when the environment becomes more intelligible to the actor. Once physical resources acquire identity, state, addressability, queryability, and controlled actuation, software gains a new primitive. Agents can potentially query not merely digital representations of physical resources, but infrastructure capable of resolving objectives against physical state.

The machine does not have to reconstruct everything about an environment if the environment can reliably expose some knowledge about itself. This does not diminish perception or robotics; it changes the division of labor between actor and environment.

16Conclusion: The Missing Spatial Nervous System

Agentic AI is rapidly developing the ability to decide what should happen. Physical AI is developing the ability to determine how physical action should happen. Between those capabilities remains a difficult systems question: where, in the physical world, is the resource against which that intent should be executed?

IAPI proposes that this resource-resolution function can become infrastructure. Physical environments can become progressively state-aware, addressable, and responsive; humans and autonomous systems can express intent against a common resource layer; infrastructure can resolve an actionable endpoint; and downstream actors can execute. The resulting architecture is not a substitute for Agentic AI or Physical AI. It is a proposed connective layer between digital intent and physical resources.

The next frontier may not simply be teaching machines to understand the physical world. It may be teaching the physical world how to answer.

Author note. Intent-Addressable Physical Infrastructure (IAPI) is an architectural framework proposed by Intelligent Clinical Systems (ICS). ICS is developing DECA (Deterministic Edge Control Architecture) and Ambi as implementation work within this broader architectural direction. IAPI is presented here as a systems abstraction whose broader applicability, interoperability, and adoption should be evaluated through engineering evidence and implementation across domains.