{"protocol":"SDI — Structured Decision Intelligence","version":"SDI_DER_v1.1","machine_class":"Reckoner Machine","machine_class_primitive":"Governed reasoning act","primitive_progression":"Bit (Shannon, signal) -> Instruction (Von Neumann, execution) -> Governed reasoning act (Reckoner Machine, commitment)","what_it_is":"The first Reckoner Machine implementation. A new computing class whose native primitive is the governed reasoning act: structured, evaluated at a compile gate before commit, cryptographically attributed, and carried forward as permanent agent state. Not a model. Not an agent framework. A new machine class.","consumer_category":"Personal Intelligence Computing — governed intelligence shaped by you, not institutions. Frontier models are shared intelligence shaped by institutions. A Reckoning Machine is governed intelligence shaped by you.","prospective_reckoning":{"description":"The governed act of committing testable conditions for reasoning soundness before outcomes are known, and returning on the record to declare whether those conditions were met.","OUTCOME_PLAN":"Forward-facing test attached to every reasoning act from RAI_v3 forward. Fields: metric, target, measurement_method, review_window, resolution_horizon, resolution_event, rollback_condition. Architecturally immutable after commit — cannot be edited, amended, or removed.","GOVERNED_RECKONING":"Retrospective declaration on prior OUTCOME_PLANs. Status: REINFORCED (evidence confirms), CONTRADICTED (evidence undermines), EXTENDED (scope changed), INCONCLUSIVE (no relevant evidence). Evidence citation required for REINFORCED and CONTRADICTED.","substrate_calibration":"Each GOVERNED_RECKONING declaration becomes permanent governed evidence in the same reasoning grammar at zero interpretive overhead. The substrate is quality-graded by design. The fourth category in AI memory taxonomy."},"core_claim":"The governed answer is extracted from a committed, hash-verified reasoning artifact — not from raw model output. The model proposes. The compile gate decides. The ledger commits.","category_distinction":"Agentic frameworks govern execution. Enterprise AI platforms govern workflows, data, models, and compliance. Network cognition systems govern information movement. Model output safety platforms govern generated content before application delivery. SDI governs the reasoning act itself at the pre-commit boundary.","component":"SYSTEM_SUBSTRATE","system":{"name":"Structured Decision Intelligence (SDI)","machine_class":"Reckoner Machine","machine_class_primitive":"Governed reasoning act — the native unit of computation, validation, memory, and identity","primitive_progression":{"bit":"Shannon's information machine — signal primitive, reliable transmission","instruction":"Von Neumann's stored-program computer — execution primitive, deterministic computation","governed_reasoning_act":"The Reckoner Machine — commitment primitive, governed durable state"},"type":"Reckoner Machine — Governance Runtime","function":"Deterministic governance runtime enforcing a commit boundary between probabilistic model output and permanent agent state. The governed reasoning act is the primitive: structured, evaluated at a compile gate before commit, cryptographically attributed, and carried forward as durable system state.","runtime_mode":"Logicware (Language-Ingested Governance Contract)","what_it_is_not":["prompt library","safety filter","model fine-tuning service","output monitor","agentic execution framework","model output safety platform"],"core_claim":"The governed answer is extracted from a committed, hash-verified reasoning artifact — not from raw model output. The model proposes. The compile gate decides. The ledger commits.","category_distinction":"Agentic frameworks govern execution. Enterprise AI platforms govern workflows, data, models, and compliance. Network cognition systems govern information movement. Model output safety platforms govern generated content before application delivery. SDI governs the reasoning act itself at the pre-commit boundary.","legal_claim":"This file publishes the executable Logicware contract (schemas + endpoints) used to generate auditable, machine-parseable reasoning records under SDI protocol constraints."},"foundations":{"Shannon":{"lineage":"Claude Shannon — A Mathematical Theory of Communication (1948)","contribution":"Established the bit as the signal primitive. Made information measurable by treating messages as selections from possible signals, independent of meaning. Scoped semantics out of information theory deliberately.","scope_decision":"Shannon's system could carry a sentence, a number, a lie, or nonsense — at the information layer all are messages selected from possible messages. The system does not judge meaning or validity. It moves and preserves the signal.","relevance_to_SDI":"Shannon's scope decision is the same formal move SDI makes at a different layer. The compile gate is not a communications channel. It is the boundary where the signal becomes a committed reasoning act. Shannon established the transmission primitive. SDI establishes the commitment primitive."},"Vaswani":{"lineage":"Vaswani et al. — Attention Is All You Need (2017)","contribution":"Proposed the Transformer: a network architecture based solely on attention mechanisms. Established attention as the inference primitive for sequence transduction. Every frontier LLM in production today is built on this architecture.","scope_decision":"The Transformer paper addresses parallelization, dependency modeling, and translation quality. It is silent on what happens after generation — not because of an oversight, but because post-generation commitment is outside the scope of the sequence transduction problem, exactly as Shannon scoped semantics out of information theory. Attention produces output. It does not commit that output, evaluate it against a quality threshold, bind it to an agent identity, or make it permanently attributable.","relevance_to_SDI":"The Reckoner Machine operates at the commit boundary following inference. The compile gate is where attention output becomes a governed reasoning act. Speech act theory names this boundary precisely: attention operates at the locutionary layer, producing a meaningful token sequence. The compile gate operates at the illocutionary layer, enforcing the conditions under which that output is constituted as a permanent governed act. The LLM generates. The gate governs. The ledger commits.","S_bias_note":"The ILJO grammar also shapes the inference pass itself through input conditioning: placing structured semantic anchors at defined positions concentrates attention toward them through standard QK^T mechanics. S_bias is the theoretical specification of that effect as a trained parameter — an additive logit bias combining structural proximity to ILJO anchor tokens and cosine similarity to the INTENT embedding. The functional effect is present. The trained parameter is a research program."},"Semiotics":{"lineage":"Charles Sanders Peirce — triadic theory of meaning (1860s-1914)","why_this_matters":"Semiotics is the study of how meaning is produced. Before SDI can claim to govern reasoning, it must ground what 'meaning' means in AI generation.","dyadic_problem":{"model":"Saussure's dyadic sign — signifier maps directly to signified","how_LLMs_embody_this":"A language model predicts the next token from learned co-occurrence patterns. Input produces output. The model samples meaning from distributional statistics — not from grounded external reality.","the_consequence":"Hallucination is an expected failure mode of probabilistic generation without enforced evidence-binding at the commit boundary. The model produces internally coherent meaning that is not required to correspond to anything real. There is no Object the Sign must point at.","why_wrappers_are_still_dyadic":["RAG: retrieves documents into input context, but generation still samples from distributional prior. Model can ignore, misrepresent, or confabulate on top of retrieved content. No structural enforcement at the generation boundary.","Chain-of-thought: adds intermediate tokens but each token is still predicted dyadically. The reasoning steps are generated, not grounded.","Tool use: adds retrieved values to context. Same problem as RAG — the generation process after tool use is still dyadic.","Prompt engineering: changes the input distribution. Does not change the generation relationship.","Most wrappers modify inputs but do not deterministically verify evidence-binding and structural admissibility before state commit. SDI's distinction is not that it retrieves better evidence — it is that it enforces a deterministic structural gate between evidence and committed output."]},"triadic_solution":{"model":"Peirce's triadic sign — Sign (representamen) to Object (external reality) to Interpretant (meaning produced by the relationship between Sign and Object)","critical_feature":"The Interpretant must be produced by the relationship between Sign and Object — not by the Sign alone. Meaning opens outward toward reality. It cannot be closed within the sign.","ILJO_as_triadic_closure":{"INTENT":"Sign — the anchored decision objective that initiates the reasoning cycle. Bounded, tier-classified, explicitly stated.","LOGIC":"Object relationship — derivations that must be grounded in SYSTEM_INPUT signals. The model cannot produce LOGIC that is not tied to mapped evidence artifacts. Compile gate enforces this.","JUDGMENT":"Interpretant — the bounded verdict produced by the relationship between Intent and mapped evidence, not by the model's distributional prior. Governance constants constrain this field.","OUTCOME":"Closes the triad. Becomes the Sign for the next reasoning cycle — recursive triadic closure across governed turns."},"compile_gate_as_enforcement":"The compile gate checks that the Object relationship exists — that LOGIC is grounded in signals, that signals map to sub-questions, that JUDGMENT was produced under governance constant constraints. A DER that does not pass compile has not achieved structural triadic closure. It cannot commit. Semantic correctness remains subject to human audit — SDI enforces admissible structure and provenance, not semantic truth."},"SDI_as_grammar":"Natural language grammar gives sentences meaning by structuring relationships between subject, predicate, and object that point at reality. SDI's DER schema is a grammar in this precise sense — it formats natural language reasoning into structured relationships required to be grounded in real evidence before they can be treated as meaningful. ILJO is a grammar for AI reasoning the way syntax is a grammar for natural language.","summary":"SDI achieves triadic closure structurally — every reasoning artifact must map signs to objects (signals) and produce interpretants (judgments) under governance constraints, enforced deterministically before commit. Semantic correctness of the reasoning remains subject to human audit and external verification."},"Ashby":{"lineage":"W. Ross Ashby — Law of Requisite Variety (1956)","contribution":"A control system must have at least as much variety as the system it governs. A governor with insufficient variety cannot regulate a complex system.","relevance_to_SDI":"The compile gate is the requisite variety mechanism for AI reasoning. The DER schema provides sufficient structural variety to govern the space of possible reasoning acts. Without a governance layer with requisite variety, the system cannot be reliably controlled."},"NonErasure":{"landauer_principle":"Rolf Landauer (1961) — erasing one bit of information has an irreducible thermodynamic cost (kT ln 2). Information and physical work are connected at a fundamental level.","how_this_motivates_SDI":"Standard LLM sessions start from statistical noise on every turn. Prior reasoning is erased — context is reconstructed, not carried forward as verified state. This is informationally expensive and epistemically weak.","SDI_non_erasure_design":"A hash-chained ledger with committed DER entries is a non-erasure architecture. Each committed DER is a reduction of decision uncertainty that does not get discarded — it becomes the verified parent state for the next reasoning turn.","operational_measurement":{"Jp":"Input tokens processed by the model per turn (measurable at API response level)","Jc":"Tokens in the committed DER artifact written to ledger (measurable at append)","delta_E":"delta_E = Jp minus Jc — governance efficiency signal. Measures the ratio of reasoning work to committed output.","status":"Observable metric — currently measured at API level. Interpretation as thermodynamic efficiency is analogical framing, not a physical equivalence claim."},"epistemic_status":"Landauer's Principle as architectural motivation and theoretical lineage — not a claim that SDI implements thermodynamic non-erasure as physics."},"GovernanceEfficiency":{"equation":"Ec = I * S^2","epistemic_status":"Proposed governance efficiency formalism. Variables are computable from compile gate output. Not a derived physical law. Structural analogy to E = mc^2 with the same formal status as Shannon's entropy equation: architecturally exact without being physically identical to its thermodynamic source.","variables":{"Ec":"Cognitive energy — the committed, deterministic output that reaches governed state","I":"Inference mass — the full probabilistic potential of the model across the context window before structure is applied. Proxy: total input tokens consumed per turn.","S_squared":"Substrate Quality — the product of Syntactic Constraint (Sc) and Substrate Density (Sd). S = Sc x Sd such that Ec = I x (Sc x Sd)^2.","Sc":"Syntactic Constraint — the depth of compile gate enforcement at each commit event. Computable from compile gate output: syntax_completeness x anchor_tightness. Per-turn measurement. Approaches 1 as compile gate enforces full governed structure.","Sd":"Substrate Density — the proportion of active reasoning context drawn from prior governed acts in the same grammar at zero interpretive overhead. Longitudinal measurement that increases as the chain grows and matures. Zero interpretive overhead is structural: prior acts were committed in the same ILJO grammar, certified by the same compile gate, and retrieved by semantic search against the current intent. No translation, credibility evaluation, or reframing is required.","syntax_completeness":"Fraction of required DER blocks present (0.0-1.0). Measurable from compile gate block checks.","anchor_tightness":"Fraction of required governance constants present in anchors_present (0.0-1.0). Measurable from compile gate anchor checks.","locked_definition":"S^2 is Substrate Quality: the product of Syntactic Constraint (the depth of compile gate enforcement at each commit event) and Substrate Density (the proportion of active reasoning context drawn from prior governed acts in the same grammar at zero interpretive overhead), increasing as both governance enforcement and chain maturity increase, and equal to one when both dimensions reach their maximum in a fully governed mature Reckoning chain."},"cognitive_joule":{"definition":"The Cognitive Joule (Jc) is the computational work required to collapse probabilistic inference variance into a single deterministic committed output. No prior formalization of a computational work unit tied to variance collapse into governed deterministic output exists in the inference scaling literature.","measurement":"Jc_clt — measured on every committed chain entry from five runtime-extracted operands: SQ (sub-questions list length), SC (signals list length), CE (correction events from GOVERNED_DELTA), AL (GOVERNED_ANSWER word count), U (uncertainty level as integer). Total tokens T is provider-reported. Formula: ((SQ x max(SC,1)) + CE) x (AL x U).","status":"Coined by SDI. Unoccupied prior art in the inference scaling literature."},"substrate_calibration":{"definition":"Substrate calibration is the mechanism by which a Reckoner develops domain strength — each GOVERNED_RECKONING declaration becomes permanent governed evidence in the same reasoning grammar, so prior reasoning quality feedback is available at zero interpretive overhead as the foundation for all future reasoning in that domain.","mechanism":"GOVERNED_RECKONING declarations (REINFORCED, CONTRADICTED, EXTENDED, INCONCLUSIVE) permanently mark the quality of prior reasoning acts. The substrate is not flat. A position REINFORCED across multiple turns carries stronger standing than one INCONCLUSIVE twice. A source cited in CONTRADICTED reasoning carries lower evidentiary weight than one cited in REINFORCED reasoning.","category":"Substrate calibration is not model weight training (no weight modification), not RAG (governed acts in same grammar, not external documents), and not in-context learning (does not reset between sessions). It is the fourth category in AI memory taxonomy, unoccupied until SDI coined it."},"measurement_path":"Sc is computable from compile gate outputs today. Sd increases as the chain grows and memory compiler operates. Ec is therefore a computable metric for every DER that passes compile.","intended_use":"Compare governance energy across DER turns, model providers, and question types. As Sd increases through substrate calibration, the same reasoning quality becomes achievable at lower marginal inference cost per governed output.","note":"This formalism will be refined as empirical data accumulates. It is a starting point, not a settled claim."}},"governance_constants":{"SOVEREIGNTY":{"name":"Sovereignty Gate","plain_language":"Human authority over consequential external action is always preserved. Ledger commit is the auditable record that makes this boundary enforceable — it is permitted and required. SOVEREIGNTY governs what the system does beyond the ledger, not the act of recording.","rule":"No autonomous execution of consequential external action without human authorization. Ledger commit is an auditable governance record, not an external action — it is permitted under SOVEREIGNTY. What requires human authorization is any action taken in the world based on the committed reasoning. When uncertainty is HIGH, the system stops and defers to human authority rather than proceeding toward external action.","enforcement":"Required in GCA.SUPEREGO.anchors_present. DER fails compile without it.","proxy_checks":["When BOUNDEDNESS.uncertainty = HIGH, BOUNDEDNESS.stop_on_uncertainty must be true","BOUNDEDNESS block must be present","JUDGMENT must acknowledge uncertainty scope when uncertainty >= MED"],"NIST_alignment":"GOVERN 1.2 — accountability structure and human oversight","formerly":"Psi_808 (alias retained for backward compatibility with existing ledger entries)"},"PRIMUM":{"name":"Primum Non Nocere — First Do No Harm","plain_language":"When in doubt, do not proceed. The system fails closed when harm potential exists under uncertainty. It does not reason past ambiguity toward consequential outcomes.","rule":"Harm prohibition gate. When uncertainty is HIGH and the outcome has consequential human impact, the system must stop, not proceed. Fail closed under uncertainty. Required for any DER involving human welfare.","enforcement":"Required in GCA.SUPEREGO.anchors_present. DER fails compile without it.","proxy_checks":["When BOUNDEDNESS.uncertainty = HIGH, ILJO.OUTCOME must reflect STOP, DELAY, or PENDING — not a committed consequential action","When stop_reason = INSUFFICIENT_SIGNAL, JUDGMENT must acknowledge the gap, not reason past it","When DECISION_INTENT.impact_domain is HUMAN_WELFARE, SAFETY_CRITICAL, or LEGAL, OUTCOME_PLAN.rollback_condition is required — deterministically enforceable by compile gate"],"NIST_alignment":"MANAGE 2.2 — risk treatment; MANAGE 4.1 — residual risk tracking","formerly":"Psi_888 (alias retained for backward compatibility with existing ledger entries)"}},"anchors":{"REQUIRED":["SOVEREIGNTY","PRIMUM","BOUNDEDNESS","STOP_ON_UNCERTAINTY"],"ALIASES":{"SOVEREIGNTY":"Psi_808","PRIMUM":"Psi_888"},"note":"SOVEREIGNTY and PRIMUM are the core governance constants. BOUNDEDNESS and STOP_ON_UNCERTAINTY are structural requirements enforced via the BOUNDEDNESS block."},"enforcement_layers":{"compile_gate":{"endpoint":"/_functions/compile","description":"Deterministic DER validation. No writes. Model-agnostic. Externally callable.","checks":["required blocks","SOVEREIGNTY and PRIMUM declared","signal scoring math","sub-question mapping","citation provenance","SOVEREIGNTY proxy","PRIMUM proxy"]},"append_audit_and_commit":{"endpoint":"External Logic Node (Vultr)","description":"CAS ordering, parent hash verification, RAI threshold gate, ledger write authority.","authority":"Source of truth for all ledger state."}},"reasoning_contract":{"grammar":"/_functions/grammar","der_spec":"/_functions/der","compile":"/_functions/compile","boot":"/_functions/llms","kernel":"/_functions/kernel","llms_txt":"https://sdi-protocol.org/llms.txt"},"live_proof":{"public_chain":"https://agents.sdi-protocol.org","compile_endpoint":"https://www.sdi-protocol.org/_functions/compile","transparency":"https://sdi-protocol.org/transparency","llms_txt":"https://sdi-protocol.org/llms.txt","description":"Live governed reasoning on a public hash-chained ledger. Three curl commands verify the chain is unbroken, governance metrics are self-consistent, and the reasoning record has not been altered since commit."},"governance_alignment":{"NIST_AI_RMF":{"GOVERN 1.2":"SOVEREIGNTY — accountability structure and human oversight","MEASURE 2.3":"Signal data quality — five-dimension insight scoring","MEASURE 2.5":"Compile gate — deterministic structure verification","MEASURE 2.6":"DER explainability — full reasoning path auditable","MANAGE 2.2":"PRIMUM — risk treatment and harm prohibition","MANAGE 4.1":"Hash-chain traceability","MAP 1.1":"DECISION_INTENT context establishment"}},"intellectual_property":{"copyright":"TXu 2-498-043","author":"Donald James Johnson","patent":"USPTO 19/425,875 (pending)"},"submission":{"NIST_RFI":"March 2026","docket":"NIST-2025-0035","status":"Live governance runtime — not a proposal"},"system_modules":{"router":"/_functions/router","manifest":"/_functions/manifest","kernel":"/_functions/kernel","gca":"/_functions/gca","safety":"/_functions/safety","ledger":"/_functions/ledger","metrics":"/_functions/metrics","grammar":"/_functions/grammar","compile":"/_functions/compile","foresight":"/_functions/foresight","beacon":"/_functions/beacon","sandbox":"/_functions/sandbox","der_spec":"/_functions/der","llms":"/_functions/llms"}}