Perception Action Loop¶
Core Idea¶
The perception-action loop names the structural pattern in which perception and action are constitutively coupled rather than separated into staged modules. Action moves the sensing apparatus — eyes, hands, whole body, sensors, end-effectors — through the world; those movements change what is sensed; and what is sensed becomes the basis for the next action. The loop is closed: there is no pure "sense first, then think, then act" pipeline. Sensing already presupposes what kind of action is being prepared, and acting already presupposes what kind of sensing it will enable. The structural claim is that for any embodied or situated agent — biological, robotic, or strategic — perception and action are two faces of one ongoing sensorimotor process, and a wide family of phenomena that look puzzling under a sense-think-act decomposition become legible under a perception-action decomposition.
The point is not the trivial observation that sensing and acting both occur. The commitment has three sharp parts. First, mutual constitution: sensing is for acting and acting is for sensing, so neither is intelligible in isolation from the other. Second, active sensing: the agent moves precisely to make information available that passive observation would not yield — saccades, head tilts, palpation, probing, exploratory behaviors. Third, no clean stages: the loop runs continuously, with no discrete moment at which perception completes and hands off to action; anticipation, prediction, and reafference — predicting the sensory consequences of one's own actions — are constitutive features of the loop rather than peripheral additions. The structure is not about embodiment in the narrow physical sense; a disembodied agent running a search against an opponent in a game instantiates it too, sensing the board state, proposing a move, observing the response, and sensing again. The load-bearing commitment is the closed coupling of sensing and acting through an environment, whether that environment is physical, simulated, or social.
How would you explain it like I'm…
Look-Move-Look Again
Sensing And Moving Loop
Coupled Sensorimotor Loop
Structural Signature¶
an embedded agent — a sensing channel — an acting channel that moves the sensing apparatus — an environment in which the channels close — a mutual-constitution coupling of the two channels — a reafference operation distinguishing self-caused from world-caused change — a continuity invariant with no clean sense-then-act break
The pattern is present when each of the following holds:
- An embedded agent. A system situated in an environment it can both sense and act upon — biological, robotic, or strategic. Embeddedness, not physical embodiment, is the requirement.
- A sensing channel and an acting channel. Two operations: one acquiring state from the environment, one changing the environment or the agent's pose within it.
- A movement-of-sensor relation. Acting moves the sensing apparatus, so action changes what is available to be sensed. This is the coupling that distinguishes the loop from independent input and output.
- An environment as closure medium. The two channels close through a world — physical, simulated, or social — which is internal to the loop, not external to it.
- A mutual-constitution invariant. Sensing is for acting and acting is for sensing; neither is intelligible in isolation, and there is no discrete moment at which perception completes and hands off to action.
- A reafference operation. The agent must separate self-caused sensory change from world-caused change; a system that cannot is structurally incapable of stable action.
These compose into one closed sensorimotor process whose load-bearing performance variable is loop tightness — latency, prediction fidelity, and directness of coupling.
What It Is Not¶
- Not feedback in general. See
feedback. Feedback is any closed loop where a measured output is routed back to modify input. The perception-action loop is the specific feedback structure in which action moves the sensing apparatus itself, so the agent's own movement changes what is available to be sensed — a constitutive coupling, not merely an error signal returned to a controller. - Not coordination. See
coordination. Coordination is about making multiple agents' actions compatible. The perception-action loop is about the coupling of sensing and acting within one embedded agent; it can run in a solitary agent with no one to coordinate with. - Not predictive coding alone. See
predictive_coding. Predictive coding is the inference mechanism of generating predictions and propagating errors. The perception-action loop is the broader sensorimotor architecture in which such prediction serves reafference and tight coupling; predictive coding can be one implementation of the loop's prediction step, not the whole loop. - Not affordance. See
affordance. An affordance is a loop-relative property of the environment — what it offers a given agent. The perception-action loop is the process that makes affordances exist as relations; the affordance is an output of the coupling, not the coupling itself. - Not a sense-think-act pipeline. The loop's defining claim is precisely that there is no clean staged hand-off from perception to cognition to action. A staged pipeline that happens to repeat is not a perception-action loop; the mutual constitution of sensing and acting is the load-bearing commitment.
- Common misclassification. Calling any iterated input-output cycle a perception-action loop. The catch: ask whether acting moves the sensor. If the agent's actions do not change what it can sense — sensing and acting run on independent channels — it is a feedback or control loop, not the constitutive sensorimotor coupling this prime names.
Broad Use¶
- Cognitive science and embodied cognition. Enactivism, Gibsonian ecological perception, sensorimotor contingency theory — the eye is not a camera pointed at the world; vision is what one does as one moves.
- Robotics and control. Behavior-based robotics, active perception, visual servoing, and SLAM, where movement creates the map and the map enables further movement; model-predictive control; reafference compensation.
- Neuroscience. Predictive processing, efference copy and corollary discharge, motor involvement in perception, and the dorsal stream coupling visual information to action affordances.
- Sports and skill acquisition. A tennis player's perception of the ball is constituted by the racket being swung; goalkeepers read shots through the postural cues of strikers; hitters predict the pitch before it arrives.
- Strategy and tactics. Boyd's OODA loop, stripped of its four-phase decomposition, is a perception-action loop applied to competitive contexts.
- Human-computer interaction. Direct-manipulation interfaces work because they tighten the loop; input latency degrades performance because it loosens it.
- Ethology and development. Predator-prey escape responses run on tightly coupled loops; infant motor exploration is perception-action learning.
- Surgical practice. Laparoscopy degrades performance by disrupting the natural loop; haptic feedback in tele-surgery aims to restore it.
Clarity¶
Naming the loop makes legible a class of phenomena that sense-think-act decompositions either miss or mistreat, and the clarity is structural rather than metaphorical: it changes what an analyst looks for. It explains why active sensing works — if perception is for action, then exploratory action that improves sensing is rational rather than mere movement. It explains why latency matters so much — small disruptions to the loop's timing, whether network lag, interface delay, or perceptual masking, degrade performance disproportionately because they corrupt the mutual-prediction structure on which tight coupling depends. It explains why thinking can be done in the world — agents offload cognitive work onto the loop, as when a fielder catches a fly ball not by computing its trajectory but by running so the ball appears to rise at a constant rate, letting the world do the arithmetic. It explains why expert skill looks effortless — a well-tuned loop runs prediction ahead of sensing, so the expert appears to act before they could have seen what they are acting on. And it explains why disembodied systems struggle to act — a model trained on static observations lacks the loop structure that gives sensing its action-relevant meaning. In each case the loop frame replaces a puzzle generated by the staged decomposition with a tractable question about coupling.
Manages Complexity¶
The loop collapses three apparently separate analytic tasks — what is the agent sensing, what is the agent doing, and how does the agent think — into one structural question: what is the structure of the ongoing perception-action coupling? This compression dissolves several families of intractable problem. The classic "vision computes a full 3D model first, then action consumes it" separation, which generates notoriously hard computational problems, dissolves into "sensing and acting jointly maintain a model just sufficient for the next action." The classic stimulus-response separation, which struggles to accommodate anticipation and self-initiated behavior, dissolves into a predictive loop in which both stimulus and response are shaped by ongoing prediction. The classic "where does motor control reside" question dissolves into a distributed loop with control spread across the coupling rather than localized in a controller. In each case the simplification is not a matter of ignoring detail but of choosing a decomposition under which the detail organizes itself. The complexity that the staged view multiplies — by demanding a complete intermediate representation handed cleanly from one stage to the next — is dissolved by the loop view, which never requires that representation to be completed, only to be sufficient for the immediate next action. The performance-relevant variable that survives the compression is loop tightness: latency, prediction fidelity, and directness of coupling.
Abstract Reasoning¶
The pattern licenses reasoning about loop tightness as a performance variable: agents with shorter latency, better prediction, and more direct coupling outperform agents with looser loops, all else equal, which makes tightening the loop a first-class design objective. It licenses reasoning about active perception as a design strategy: enabling an agent to move its sensors purposefully outperforms passive sensing of comparable bandwidth, in robotics and interface design alike. It licenses reasoning about reafference as a load-bearing architectural commitment: any agent in a perception-action loop must distinguish self-caused sensory change from world-caused sensory change, and a system that cannot tell its own movement from the world's is structurally incapable of stable action. It licenses reasoning about affordances as loop-relative properties: what an environment affords depends on the agent's loop, so the same environment affords different things to differently-coupled agents, and affordance is therefore not a fixed property of the world but a relation between world and loop. It licenses reasoning about loop disruption as an intervention site: degrading or supporting the loop is often more effective than intervening on perception or action alone — reduce interface latency, add haptic feedback in tele-surgery, rehearse mentally in sport. And it licenses reasoning about adversarial loops, where two agents' loops interact and the faster-cycling agent can invalidate the slower agent's orientation before it can act on it — the OODA claim, recast as a general consequence of coupled loops running at different rates.
Knowledge Transfer¶
The pattern transfers as a unified vocabulary across domains that traditionally name the same structure differently, and the transfer is valuable precisely because it lets an insight discovered in one substrate be carried as an engineering prescription into another. The enactivist philosophical claim that perception is constituted in action transfers to robotics as the engineering principle that a robot should sense actively rather than being sensor-loaded and inference-heavy. Active sensing in robotics transfers to medical instrumentation as the principle that surgical tools and interfaces should preserve the surgeon's natural loop rather than replacing it with sense-then-act stages. The sensorimotor-contingency account from cognitive science transfers to AI agent design as the principle that an embedded agent should learn the contingencies between its actions and their sensory consequences, not merely the static patterns in a fixed dataset. Predictive processing in neuroscience transfers to controller design as model-predictive control, the same strategy of predicting the sensory consequences of one's own actions. The empirical finding from sport that small latencies degrade tightly coupled skills transfers to interface engineering as a hard requirement on input latency. And ecological-perception affordances transfer to product design as the prescription to design for the coupling — what controls invite what actions — rather than for perception or action in isolation.
The role mappings carry across all of these. The embedded agent is the organism, the robot, the surgeon, the AI agent, the competitor. The sensing apparatus moved by action is the eye saccading, the camera on the moving arm, the laparoscope, the probing query. The reafference component that separates self-caused from world-caused change is corollary discharge in the brain, the motion model in SLAM, the proprioceptive correction in a manipulator. The environment is not external to the loop but the medium in which the loop closes — the physical scene, the operative field, the game tree, the market. Because the structure is shared, the intervention vocabulary travels as a single recipe: tighten the loop, preserve mutual prediction, exploit active sensing, design for affordances, minimize reafference confusion. A roboticist's instinct to reduce control latency and a surgeon-instrument designer's instinct to preserve haptic feedback are, structurally, the same move — supporting a coupling whose tightness is the performance variable. The prime's contribution to transfer is to make that sameness visible, so that what was learned the hard way in one field need not be rediscovered from scratch in the next.
Examples¶
Formal/abstract¶
Simultaneous localization and mapping (SLAM) on a mobile robot is the perception-action loop made fully explicit in an engineered system, and every element of the signature is a named software component. The embedded agent is the robot, situated in a physical environment it both senses (via a depth camera or LIDAR) and acts upon (by driving its wheels). The sensing channel is the perception pipeline extracting landmarks from each frame; the acting channel that moves the sensing apparatus is the drive system, which physically translates and rotates the sensor through the scene. The movement-of-sensor relation is the crux: the robot cannot map what it has not seen, so it must move to bring new regions into view — active sensing in the literal sense, where exploratory motion is undertaken to make information available that holding still would never yield. The environment as closure medium is internal to the loop: the partial map built so far determines where the robot drives next, and where it drives determines what gets added to the map. The reafference operation is the load-bearing engineering problem — the motion model. When the camera image shifts, the system must separate self-caused change (the apparent motion of the world because the robot moved) from world-caused change (a person walking through the scene). A SLAM system that cannot make this distinction confuses its own egomotion with environmental change and its map diverges; this is exactly the frame's claim that a system unable to perform reafference is "structurally incapable of stable action." The continuity invariant shows in the loop's tightness: latency between sensing and motion, and the fidelity of the prediction step, are the performance variables — a loose loop produces drift and a lost pose.
Mapped back: The robot is the embedded agent, the drive system moves the sensing apparatus, the motion model is the reafference operation distinguishing egomotion from world-motion, and the growing map is the environment internal to the closed loop — the perception-action structure realized as control software.
Applied/industry¶
Laparoscopic surgery is a domain where degrading the perception-action loop measurably degrades performance, and the frame both explains why and prescribes the fix. The embedded agent is the surgeon; in open surgery the loop is tight and natural — the hand that moves the sensing apparatus (the eyes and the palpating fingers) is the same hand that acts, with direct haptic feedback and a first-person view. Laparoscopy deliberately breaks this coupling. The sensing channel is now a camera on a rigid scope held by an assistant, the acting channel is a set of long instruments passed through trocars, and the surgeon views a 2D monitor. Several loop properties are damaged at once: the movement-of-sensor relation is severed because the surgeon's own movement no longer moves the viewpoint (someone else holds the camera); reafference is corrupted by the fulcrum effect, where moving an instrument handle left moves its tip right, so self-caused change no longer matches the expected sensory consequence; and the continuity invariant is strained by latency and loss of depth and touch. The frame predicts exactly the observed result: performance falls not because perception or action individually got harder but because the coupling between them was loosened — small disruptions to loop timing and prediction degrade tightly-coupled skill disproportionately. And it prescribes the interventions surgical-instrument designers actually pursue: restore the loop rather than the parts in isolation. Robotic systems give the surgeon control of the camera so movement again moves the viewpoint; haptic feedback in tele-surgery aims to restore the touch channel; motion-scaling and tremor filtering tighten the prediction. The same loop-restoration logic transfers to human-computer interaction, where reducing input latency in a direct-manipulation interface is the identical move — supporting a coupling whose tightness is the performance variable.
Mapped back: The surgeon is the embedded agent, the scope and instruments are the sensing and acting channels, the fulcrum effect is corrupted reafference, and laparoscopy's performance penalty is loop-loosening — the cure being to tighten the coupling, not to improve perception or action separately.
Structural Tensions¶
T1 — Loop Tightness versus Deliberation (Temporal). The loop's payoff is tight, low-latency coupling, but tight coupling leaves no room inside the loop for slow, explicit deliberation — the very thing a sense-think-act architecture inserts deliberately. Some problems genuinely need a planning stage that the loop structurally lacks. The failure mode is forcing reactive coupling onto a problem that requires stepping out of the loop to reason, producing fast, fluent, locally-coherent behavior that never reconsiders its frame. The diagnostic is to ask whether the task rewards reaction time or reflection: if the cost of a wrong commitment exceeds the cost of latency, the loop's tightness is a liability, and the competing prime — staged deliberation — should govern at least one stage.
T2 — Self-Caused versus World-Caused Change (Reafference Boundary). The loop must separate sensory change the agent produced from change the world produced; this is the reafference operation, and it is load-bearing, not optional. The two are not intrinsically labeled — the same image shift can be egomotion or a moving object — so the distinction depends on an internal prediction that can be wrong. The failure mode is reafference confusion: the agent reads its own movement as the world's (SLAM map divergence, the surgeon's fulcrum disorientation), and stable action becomes structurally impossible. The diagnostic is to check whether the agent's predicted sensory consequences are being compared against actual sensing: where there is no efference copy, self and world are indistinguishable and the loop is blind.
T3 — Sufficient-for-Next-Action versus Complete Model (Scope of Representation). The loop never needs a finished world-model, only one good enough for the immediate next action — that is how it dissolves the hard vision-first computational problem. But "sufficient" is defined relative to the next action, so the model is structurally incomplete and silently wrong for any action it was not maintaining itself for. The failure mode is acting on a just-in-time model as if it were complete, succeeding fluently until a novel action exposes the parts that were never built. The diagnostic is to ask what action the current sensing was for: information adequate for the loop's running action carries no guarantee for a different one, so a change of goal demands re-sensing, not reuse.
T4 — Active Sensing versus Sensing Cost (the Probe Budget). Active sensing wins because moving the sensor makes information available that passive observation cannot yield — but every probing movement costs time, energy, and exposure, and itself perturbs the environment being sensed. The loop frame foregrounds the benefit and can mask the budget. The failure mode is unbounded probing: an agent that explores to sense without accounting for the cost of exploration, or that disturbs the world it is trying to read. The diagnostic is to weigh the expected information gain of a sensing action against its cost in the same currency as task actions, since active sensing is only rational when the probe buys more than it spends — a standing tension with passive, low-cost observation.
T5 — Coupling to This Environment versus Transferable Skill (Affordance Relativity). Because affordances are loop-relative, a finely-tuned loop is tuned to a specific environment and a specific body; its competence is a relation, not a portable possession. The same tightness that makes an expert effortless in one coupling can fail to transfer when the environment or apparatus changes. The failure mode is mistaking loop-relative mastery for substrate-general skill — assuming the expert who is fluent in one coupling will be fluent when the latency, geometry, or feedback channel shifts. The diagnostic is to ask whether the affordances the agent exploits survive the new environment: where the coupling changes, expect skilled performance to drop to that of a novice loop until re-tuned.
T6 — Loop Rate versus Opponent's Loop Rate (Adversarial Coupling). When two agents' loops interact, the relevant variable is no longer one loop's absolute tightness but the relative rate of the two — a faster-cycling agent can invalidate a slower agent's orientation before it can act, the OODA claim generalized. This shifts the design target from optimizing one's own loop to outpacing another's, and the two can diverge: a perfectly tuned loop still loses to a faster adversary. The failure mode is optimizing loop tightness in isolation, ignoring that the opponent's cycle time sets the standard. The diagnostic is to measure one's loop against the adversary's, not against a fixed bar: if the opponent re-orients faster, increasing one's own accuracy at the cost of speed makes the coupling worse, not better.
Structural–Framed Character¶
The perception-action loop sits at the structural pole of the structural–framed spectrum — a structural prime with a 0.0 aggregate. It is a bare relational pattern: an embedded agent's acting moves its sensing apparatus, that movement changes what is sensed, and what is sensed grounds the next action, in one closed loop with no clean sense-then-act stages. Nothing about that pattern depends on a particular field's lexicon or assumptions, and every diagnostic reads zero for a substrate-faithful reason.
The pattern carries no home vocabulary that must travel with it. A saccading eye, a palpating hand, a SLAM robot building a map as it drives, and a disembodied game agent probing a board and sensing the reply are told in each domain's own words — vocab_travels is zero because the structural skeleton (acting moves sensing, sensing grounds acting, reafference separates self-caused from world-caused change) survives the loss of any field's terminology. It carries no inherent approval or disapproval: a tight loop is neither good nor bad until you specify the task, so evaluative_weight is zero. Its origin is formal-relational — a closed coupling of two channels through an environment — and owes nothing to any human institution, so institutional_origin is zero. And it runs indifferently in biological, robotic, and strategic substrates: an insect's optomotor reflex and an autonomous vehicle's perception stack instantiate the identical structure with no human practice required, so human_practice_bound is zero. Invoking the prime recognizes a sensorimotor coupling already wired into any situated agent rather than importing an interpretive frame, so import_vs_recognize is zero. Every diagnostic points the same way, which is exactly what the structural label with its 0.0 aggregate asserts. The one caveat the rubric notes — that the prime reaches into "agent embedded in environment" relational vocabulary — is a structural vocabulary, not a framed one: it describes the topology of the coupling, not a doctrine imported on top of it.
Substrate Independence¶
The perception-action loop is a strongly substrate-independent prime — composite 5 / 5 on the substrate-independence scale. Its breadth is maximal: the constitutive sensorimotor coupling — acting moves the sensing apparatus, that movement changes what is sensed, and what is sensed grounds the next action — recurs across biological cognition, robotics and control, neuroscience, sports skill, competitive strategy, human-computer interaction, ethology, and surgery, instantiating the same closed structure in physical, robotic, and strategic substrates with no human practice required. Transfer is concrete and documented, carried as engineering prescriptions that survive the crossing: an enactivist claim becomes active-sensing robotics, becomes loop-preserving surgical instrument design, becomes a hard latency requirement in interface engineering, with the role mappings (embedded agent, sensing channel, acting channel, reafference operation, environment-as-closure-medium) holding intact. What sits just below maximal is structural abstraction at 4: the signature reaches into "agent embedded in environment" relational vocabulary — but that vocabulary describes the topology of the coupling, not a framed doctrine imported on top, so it caps the abstraction only fractionally. Recognized rather than translated, recurring almost everywhere a situated agent senses and acts, the prime earns a composite 5.
- Composite substrate independence — 5 / 5
- Domain breadth — 5 / 5
- Structural abstraction — 4 / 5
- Transfer evidence — 5 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Perception Action Loop is a kind of Feedback
The file: the perception-action loop IS the specific feedback structure in which 'action moves the sensing apparatus itself' — a constitutive coupling, not merely an error signal. Genus=feedback, differentia=acting-moves-sensing + reafference.
Path to root: Perception Action Loop → Feedback
Neighborhood in Abstraction Space¶
Perception Action Loop sits in a sparse region of abstraction space (79th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.
Family — Anticipation & Forward Models (15 primes)
Nearest neighbors
- Agency — 0.71
- Epistemic Action — 0.71
- Knowledge-Action Gap — 0.70
- Situation Awareness — 0.70
- Feedback — 0.68
Computed from structural-signature embeddings · 2026-06-14
Not to Be Confused With¶
The most important confusion is with feedback, because the perception-action loop is a feedback loop and a reader can stop there. The two are genus and differentia. Feedback names the general mechanism of any closed loop in which a sensed output is routed back to modify a subsequent input — a thermostat, a PID controller, a price signal. What it requires is only the return path: output measured, compared to a reference, correction applied. The perception-action loop adds a specific structural commitment that ordinary feedback does not contain: action moves the sensing apparatus. The output channel does not merely feed an error signal back to a controller; it physically relocates, reorients, or reconfigures the very organ that does the sensing, so that what the agent can perceive on the next cycle is itself a consequence of how it just acted. A thermostat exhibits feedback but not a perception-action loop — its acting (running the furnace) does not move its sensing (the thermometer's access to room temperature). A saccading eye, a palpating hand, a driving robot building a map all exhibit the perception-action loop because moving is sensing. What the perception-action loop captures that feedback does not is active sensing and reafference: because the agent's action changes what it senses, it must distinguish self-caused sensory change from world-caused change, a problem that simply does not arise for a feedback controller whose sensor is fixed. Calling the perception-action loop "just feedback" loses exactly the egomotion problem that makes it hard and the active-sensing strategy that makes it powerful.
A second confusion is with coordination — unsurprising, since it is the embedding-nearest catalog neighbor. The confusion is a category error worth naming because it is so easy to slip into: coordination is fundamentally an inter-agent concept, the problem of making the actions of two or more agents mutually compatible so they do not collide or work at cross-purposes. The perception-action loop is fundamentally an intra-agent concept, the coupling of one agent's sensing to its own acting through an environment. A single organism alone in a field, reaching for a fruit, instantiates the perception-action loop completely while having nothing whatever to coordinate. Conversely, two perfectly coordinated agents each run their own internal perception-action loops; coordination operates between the loops, not as one of them. The roles do not even line up: coordination's load-bearing elements are shared information, conventions, and joint objectives across agents; the perception-action loop's are the sensing channel, the acting channel that moves the sensor, and the reafference operation, all inside one agent. The adversarial-coupling tension (T6) is where the two legitimately touch — two agents' loops interacting at different rates — but even there the loop is the per-agent structure and the inter-agent contest is a separate, coordination-adjacent layer on top.
A subtler confusion is with predictive_coding, which a cognitive scientist might take to be the same theory. Predictive coding is a specific inferential mechanism: the system maintains a generative model, predicts its inputs, and propagates only the prediction errors upward, minimizing surprise. The perception-action loop is a broader architectural claim about how sensing and acting are coupled, within which prediction plays a role — the reafference operation needs an efference copy to predict the sensory consequences of action — but which does not depend on predictive coding being the mechanism that supplies it. One can build a perception-action loop with a simple motion model rather than a full predictive-coding hierarchy, and predictive coding can be deployed in a purely passive perceptual system that never acts and thus instantiates no loop. The perception-action loop says sensing is for acting and acting moves the sensor; predictive coding says perception is prediction-error minimization. They are compatible and often combined, but one is about the sensorimotor architecture and the other about the inference running inside it.
For a practitioner these distinctions redirect the intervention. If a system is failing, ask first whether the right object is the loop (does acting move sensing? is reafference confused?), a generic feedback tuning problem (gain, delay, reference), an inter-agent coordination breakdown, or an inference defect in the prediction step. Treating a fulcrum-effect disorientation as a coordination problem, or a coordination failure as loop-loosening, sends the fix to the wrong layer entirely.
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.