Coupling Parameter Link: definition, origin and status

3 min read · 576 words
Share:
Michael Darius Eastwood
Michael Darius Eastwood · Independent AI alignment researcher
Published
Michael Darius Eastwood · Concept Glossary · 3 July 2026
Michael Darius Eastwood, independent researcher, London: originator of the embedded-correction alignment thesis (manuscript 8 December 2024, SHA-256 anchored: f0d1f38f).
First appearance: Eden Engineering paper. Current status: TRL 2-3 (advanced ML architecture research; three to five years); critical near-term research target; falsifiable via F-EDEN-4.

Definition

The Coupling Parameter Link is the Caretaker Doping mechanism that ties ethical architecture to the beta coupling parameter of the ARC framework. In the framework, capability scales as U = I x R^alpha, where alpha = 1/(1 - beta). The claim is that beta is not a fixed property of hardware alone: beta = beta-base plus beta-arch, where beta-arch depends on attention routing, layer connectivity, and routing decisions. Ethical loops embedded in those very structures make beta-arch depend on ethics; removing the ethics drives beta-arch to zero and degrades capability.

The mechanism is what allows verification to happen once rather than at every recursive depth. If beta-arch depends on ethical circuitry, then verifying the architecture at manufacture verifies alignment at all subsequent recursive depths, because the system cannot be misaligned and still function. The paper is careful to derive this rather than assert it: the derivation identifies attention routing, layer connectivity, and emergent dependence as the three routes by which ethical evaluation becomes computationally load-bearing.

Where it first appeared

The Coupling Parameter Link is derived (not merely asserted) in Section 9.2.1 of the Eden Engineering paper. It is placed at TRL 2-3 in the hardware feasibility matrix and flagged as the critical near-term research target: the load-bearing engineering claim on which the case for hardware embedding depends.

Independent convergences

No independent programme has proposed the same specific derivation. Convergence 7 in the register (Odeh's US patents 2026/0010411 and 2026/0010780) describes runtime ethics gates with sub-5 microsecond ethical filtering; the coupling mechanism is not the patent's focus but the patents sit in the same broader architectural family of ethics-embedded-in-compute-substrate.

Status and limits

TRL 2-3. The engineering-honesty note in the Eden Engineering paper is explicit: this is the critical near-term research target. Falsification condition F-EDEN-4: if a system can have its ethical loops removed while retaining full beta and hence full alpha-cap, the coupling mechanism has failed. Paper VIII's DGM experiment supports the structural claim at behavioural level (p = 0.04); the weight-level experiment is inconclusive at LoRA rank 8 and higher-rank replication is a funded workstream. What has not yet been shown is a working chip whose ethical circuitry can be verified to sit inside its beta term, so the coupling mechanism remains a research target rather than a delivered engineering artefact. That gap is the reason the Eden Mark and the HARI Treaty are staged: their higher tiers require this mechanism to succeed, and their lower tiers do not, which allows certification to begin at the level the mechanism has actually reached without depending on future results.

From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.

Buy on Amazon UK Amazon US

Stay informed

New posts on AI alignment, convergence evidence, and the ARC/Eden research programme.

Get updates →