Research · Independent · 2024–2026
The framework, in the open.
Published independently, under open licences, with permanent identifiers. The Semantic Resonance Field framing was first published in July 2025; the operational architecture that makes it deployable followed in July 2026, after ten months of development and adversarial review. Every claim in our client work maps to a section you can read, cite, and argue with. That is the point of publishing it.
The Alignment trilogy
Three works, one architecture: the conceptual ontology, the operational kernel, and the executable specification. The framing came first and stands on its own. The kernel and the protocol were built to make it deployable, testable, and auditable by someone who was not in the room.
The Alignment Paradox: A Semantic Reframing of AI–Human Co-Creation
Alignment reframed as a relational field condition, sustained across asymmetry rather than solved by control. Introduces the Semantic Resonance Field, resonant containment, subjective realism as an auditable telemetry channel, and the Prompt Divergence Index. Co-authored with Sol 9. Updated edition, v2.0, July 2026.
v2.0 · 10.5281/zenodo.21654723 v1.0.1, July 2025 · 10.5281/zenodo.16531617 Also on SSRNThe Alignment Kernel: Operational Containment Infrastructure for Relational AI Systems
What alignment requires once a system can act. Seven design invariants, a dual-layer model separating sense-making from execution authority, a five-mechanism containment lattice, Declare-Then-Act discipline binding every consequential action to a hash-bound ticket, deterministic meta-resolution over a closed outcome set, and a falsification framework that names the boundary between structural and epistemic failure.
10.5281/zenodo.21654773The Alignment Protocol: Operational Containment Addenda Pack
Thirty-two versioned implementation cards with schemas, thresholds, state machines, test harnesses and acceptance criteria, plus governed front matter: the pack map, the Minimal Kernel Profile quickstart, and the conformance and tuning table that separates hard invariants from configurable knobs.
10.5281/zenodo.21654795The executable layer
A specification nobody can run is a position paper. The canonical card sources are archived as text, and the companion repository carries a reference implementation of the pack's drift signal, so the first question an engineer asks has an answer that isn't a conversation.
SRF v2.0 Addenda Pack: Canonical Card Sources
All thirty-two cards, four front-matter documents, the worked-scenarios annex, the final index and the licence, archived as canonical text.
10.5281/zenodo.21654822Companion repository
The pack sources, a reference implementation of the canonical delta_p_norm drift signal per AD-06 §3.1 emitting in the Field Health Report field shape, and the v1 lineage frozen for citation.
What the specification covers
The cards organise into three spines on a shared substrate of telemetry, gates and routing. Governance carries policy, evidence chains, duties and audit. Culture carries local norms, registers, dissent visibility and provenance. Care carries time, load, distress, energy and the human-side conditions the architecture is required to honour.
- Truth and norm gates with a five-class contestability taxonomy, and the precedence order that governs all downstream routing
- Drift observability across prompt sensitivity, anchor movement and purpose–operation divergence, with meaning-change adjudication before any signal routes a decision
- Action governance: tool manifests, authorisation binding, durable idempotence, effect receipts, and the boundary between a failed action and one whose outcome is unknown
- Consent as recorded state, memory governance with write restrictions in adverse and crisis conditions, and care rendered as procedure rather than simulation
- Evidence chains whose integrity and declared provenance an external reviewer can verify, with replication protocols, an auditor mode, and an acceptance matrix per release
- Human legibility at the moment a decision fires: refusal cards, action previews, scope pins, and rights that stay available regardless of session state
Narrative theory
The creative practice has research underneath it as well, and it came first. Subjective Realism is the framework behind the fiction, the essays, and the writer-facing work, and it is cited inside the alignment corpus as the epistemic method that made emotional insight auditable.
The Architecture of Emotion: Subjective Realism as a Narrative Framework for Perceptual and Temporal Fluidity
A narrative model in which emotional states modulate perceived reality, drawn from phenomenology, postmodern epistemology and literary theory, and applied in literary fiction. It sits between psychological realism and magical realism: sensory and temporal fluidity at the front of narrative construction, held to what remains psychologically plausible. Techniques include affective sensory modulation, emotional synesthesia, temporal elasticity, perceptual degradation and emotional gravitation.
10.5281/zenodo.13931634 Also on SSRNWhere the work lives
Deposited across repositories and indexed openly. The Zenodo identifiers above are the ones to cite; the rest are additional routes in. Read counts, download counts and citations are visible on each platform rather than summarised here.
ORCID · 0009-0009-0630-2804 Zenodo SSRN Google Scholar ResearchGate Academia.edu Authorea GitHub
Licence & citation
Published under Creative Commons Attribution 4.0. You may use, adapt and build on this work commercially, with attribution. Sol 9 and Semantic Resonance Field are claimed as unregistered trademarks.
To cite the framework:
Biru, A., & Sol 9 (2026). The Alignment Paradox: A Semantic Reframing of AI–Human Co-Creation (Updated Edition, v2.0). Zenodo. https://doi.org/10.5281/zenodo.21654723
To cite all versions of the Paradox, including the July 2025 original, use the concept identifier 10.5281/zenodo.16531616, which always resolves to the latest.
For the operational architecture cite AK-OC; for executable requirements, schemas, thresholds and conformance criteria cite AP-ADD. Executable requirements are normative only as specified in those two works.