Introducing the Agential Coupling Drive

Written By William Brown

New on Projects · An open research proposal

Can a machine’s physical history become part of how it chooses what happens next?

I have made the Agential Coupling Drive (ACD) research proposal freely available on the NovoSciences Projects page. It sets out a proposed physical interface for investigating agency in hybrid thinking machines: systems in which a language model or embodied platform interacts with a material substrate whose evolving state can influence subsequent action.

The central question is how an unpredictable event might become more than a passing fluctuation. Could it leave a physical trace, become part of a system’s retained history, and help shape its future responses through a continuing exchange with its environment?

Research status: proposed architecture · No demonstrated ACD device

When an event leaves a history

The ACD brings together three principles: stochastic events, persistent material memory, and reciprocal feedback. A quantum transition supplies a possible event. Its consequences would need to persist in a physical state, such as a change in optical response or material bias. That retained change would then affect how later events unfold.

The loop becomes especially interesting when the hardware participates in action. Its measured state would contribute causally to a decision, while sensory input and the consequences of that decision would alter the substrate’s later state. The proposal asks whether this ongoing relationship can support measurable features of agency.

Randomness alone does not establish agency. Neither does memory by itself. The research problem is to determine what their organized interaction contributes, and under which conditions it matters.

Conceptual illustration 01 · Event, memory, feedback

An event–memory–feedback loop, with agency as a question to investigate Stochastic events produce retained material changes. Reciprocal feedback affects later dynamics. A separate dashed connection marks agency as an experimental question, not an established consequence of the loop. Retained history changes later dynamics Stochastic events A transition occurs Material memory A physical trace persists Reciprocal feedback Consequences return AGENCY? A question to test Events, material memory, and feedback form a recurring loop A vertical sequence links stochastic events to material memory and reciprocal feedback. An amber return path reconnects feedback to later events. Agency appears below as a question to test. History returns Stochastic events A transition occurs Material memory A physical trace persists Reciprocal feedback Consequences return AGENCY? A question to test
The proposed loop links events to retained material changes and later dynamics. “Agency” marks the target of investigation; the animation represents a hypothesis, not measured performance.

A physical interface for a thinking machine

The candidate architecture explores protein-hosted quantum dots within a DNA-templated dielectric network. In accessible terms, it combines light-responsive components with a structured material scaffold. A hierarchy of resonators, distributed optical interference, and optical or material memory would provide possible routes for events to influence a persistent physical state.

A recurrence controller would connect that substrate to a language model or embodied platform. The intended relationship is reciprocal: hardware measurements help shape action, and the results of action feed back into the hardware. Establishing that relationship experimentally would require showing that changes in the substrate make a reproducible difference to what the whole system does.

These are proposed design elements. Their integration, stability, and usefulness remain questions for research. The publication makes the architecture available for examination before any claim of successful implementation.

Conceptual illustration 02 · The proposed interface

A material substrate coupled to a recurrence controller and an acting system A network of light-responsive components retains a changing material state. Measurements pass through a recurrence controller to a language model or embodied platform. Action and sensory consequences return through the controller to affect the material substrate. Light-responsive material substrate Events · Optical response · Retained physical state Read state Change state Recurrence controller Influence action Sensory / action consequences Language model / embodied platform A material substrate exchanges influence with an acting system A light-responsive network with retained state is linked through a recurrence controller to a language model or embodied platform. Teal arrows show reading state and influencing action. Amber arrows return consequences and change the material state. Material substrate Light response + retained state Read state Change state Recurrence controller Influence action Consequences Language model / embodied platform
A conceptual architecture, with the substrate enlarged for clarity. Moving traces indicate proposed exchanges of information and influence; they do not depict a measured signal, a fabrication layout, or a demonstrated quantum advantage.

Turning an idea into questions we can test

Alongside the broader proposal, I am developing a focused companion theory paper on event integration through persistent memory and feedback. Its aim is to make the central claim narrow enough to evaluate: when does an earlier physical event remain causally relevant to a later response?

One candidate test is a delayed-cue task. Information arrives, a delay follows, and a later cue determines the appropriate response. The question is whether retained physical changes help connect those separated moments in a way that matters for performance.

The essential comparisons include removing memory, interrupting feedback, and testing a matched adaptive classical surrogate. Each asks a different question: does persistence matter, does reciprocal coupling matter, and could a comparable classical system account for the observed behavior? Such comparisons are necessary before assigning a distinctive role to the proposed quantum components.

The key question is whether a system’s physical history makes a demonstrable difference to its future action.

Agency, experience, and an open invitation

The proposal also considers more speculative possibilities concerning subjective experience, including coupling to a sentient substrate or a postulated scalar field. These remain open hypotheses requiring independent evidence. Demonstrating useful memory and feedback would not, by itself, demonstrate consciousness.

My purpose in releasing the document is to make the reasoning, architecture, and unresolved questions accessible. I welcome careful criticism, alternative explanations, and suggestions for experiments that can distinguish the proposed mechanism from simpler accounts.

The Projects page now brings together the current research proposal and its historical origin. The revised document develops the ACD framework; the original Pneumanous Engine contest submission is preserved so readers can follow how the idea evolved.

Explore the project Download the ACD proposal · PDF

Historical document: Pneumanous Engine — original contest proposal, 15 October 2025 (PDF).

William Brown
Next

Water, Light, and Life: Presenting New Research at Aquaphotomics 2026