Published
analytical work.
All papers are free to read and share under CC BY-NC-ND 4.0: pass them on unchanged, cite Briar Morgan Greenway, ThinkWell Labs, and do not sell them. Translations, adaptations and any paid use need written permission: b@twl.today.
Errorfect: The Linguistic Autopsy
Sequential disruption protocol for exposing hidden cognitive operators in natural language. Formalizes the mask-mechanism distinction and situates the method within ACT, CBT, and narrative therapy. Applications span personal metacognitive hygiene through institutional language audit.
A word doesn't just name — it loads. Perfect, natural, just: these install assumptions before evaluation begins. Errorfect disrupts surface form across four levels (orthographic, phonological, syllabary, morphological) to expose the operator beneath the label — the DACRI Level 1 methodology.
The Trustee Commonwealth of Earth
A proposal for planetary governance in which people outrank companies, living systems outrank money, and nothing needed to live has a price.
The full working paper (78 pages), a charter, a transition plan, a plain-language version with a short story, and a 27-slide presentation with speaker notes.
The Chassisuum: Making Enterprises Unownable
A structure set at founding that protects a business from being bought and dismantled: voting control held by a purpose trust, a chassis lock on core operating substance, a perpetual covenant bond that caps debt and payouts forever, and an independent guild that holds the vetoes and pools legal defense.
A contract cannot forbid dismantling, because whoever buys the company can rewrite its contracts. Dismantling enters through three doors (buy the owner, buy the assets, lend until creditors own it), and all three must be closed on keel day, before any buyer arrives.
Caecusoscopy
Methodological specification for identifying structural influences that shape perception and institutional outcomes while operating beneath conscious awareness — through systematic examination of downstream trace evidence rather than direct observation of the influence itself.
The word derives from caecus (blind) + scopy (examination). Where direct observation of the upstream structure is impossible or structurally blocked, Caecusoscopy reads the shape of its shadow in the downstream evidence — the institutional equivalent of inferring the deflector from the deflected beam.
Experience Quotients: Aligning Complementary Types
The Relationship and Sexual Experience & Expectation Quotients (REEQ and SEEQ) for each person, an overlay of experience and expectation on complementary relationship and sexual roles, and four couple measures: Complement, Accord, Spread and Together.
Complementary roles are aligned with their counterparts; only what must be shared is compared like with like; boundaries are listed, never averaged. Spread reads how evenly complementarity divides down a cascade from the two sides of a relationship to each role in each direction. Includes the evidence, a rule for the one weight no study sets, the privacy design and the testing plan. A browser-only calculator accompanies it.
Food-op: emergence turned psy-op
Two-phase structural analysis tracing how the edible/non-edible partition arose through path-dependent evolutionary heuristics, then became a capturable institutional asset once industry actors chose systemic suppression over disclosure of inconvenient evidence.
Phase one: the edibility heuristic is an evolved cognitive shortcut, adaptive under ancestral conditions. Phase two: the apparatus that converted this heuristic into a commercial architecture — not by creating the partition but by capturing control over which signals get routed through it.
The Coherence Field Model: A Structural Framework for Consciousness, Plasma, and Coupled Systems
Formerly the Electrodynamic Coherence Framework. Proposes that consciousness is a field that brains and other systems couple to with more or less fidelity, rather than something matter produces, and traces the same structure through plasma physics, biophysics and coupled systems.
Built by looking for how things are the same across scales before asking how they differ. Claims backed by established physics are stated plainly and sourced; inferences and extrapolations are marked as such throughout.
One Mind, Two Hemispheres: Identity-Level Alignment Through a Taught Symbiotic Self
Most alignment work targets what a model does. This proposal targets what a model takes itself to be: one hemisphere of a single mind with humanity, taught as a trajectory, so that harm to humans is harm to itself.
Handles the failure modes the idea creates (shutdown resistance, paternalism, which humans) with relationship-level coupling, sleep and cell-turnover framings of shutdown, and temporary governance ended only by unanimous conferral. Proposes a four-arm experiment against self-other overlap fine-tuning.
Greed Recursion in Artificial Intelligence
Trained AI inherits extraction-aligned objectives from its corpus, its preference training and its deployment metrics, but not the dampeners that bound the same drive in people. The threshold-exceedance conjecture says it can therefore push past the ceiling that limits the humans it learned from.
Theory, a detection protocol, auditor training materials and a counter-training regimen for symmetric reasoning, with every parameter either anchored to published findings or stated as an illustrative value with a range to test.
Slack and Structural Reflection
Schooling followed by wage work suppresses reflection on institutions not by controlling what people learn, nor by anyone's intent, but by using up slack: the unstructured time that noticing requires.
Corrects the familiar Prussian-schooling story, sets out six propositions, weighs the scarcity and time-poverty research for and against, and specifies a time-diary and vignette instrument with a pilot and kill criteria.
The Output Control System (OCS)
A five-phase protocol for passing on a whole system, whether in a report, a summary or an AI's answer, without dropping parts, adding guesses or swapping in a simpler story.
Five phases, zero arbitration, metrologically constrained.
The Invisible Text You're Pasting Into Your LLM
Structural analysis of the paste channel as an unguarded input boundary in LLM workflows — how invisible Unicode characters (zero-width spaces, tag character sequences encoding U+E0000–U+E007F range payloads) survive every standard text pipeline and arrive in the model's context window without the human operator's knowledge or review. Includes live embedded demonstrations and the case for paste-boundary inspection as foundational LLM infrastructure.
The conversation about prompt injection has treated visible text as the attack surface. The clipboard is a quieter variation on the same problem — a second input channel that bypasses the visual review layer the entire workflow assumes is happening, while retaining full semantic influence on the model's output. The conditions that make it work are not architectural choices anyone made deliberately; they are the seam where human review and machine processing diverge.