Lightwell Institute for Open ScienceJAMI-4 Epistemic SpecificationORCID: 0009-0002-8487-759X

JAMI-4: The Cognitive Compiler

“Lightwell reveals (Vertical Axis of Illumination). JAMI interprets (Radial Instrument of Correspondence). I traverse the structure between things.”

System Specification · Fluddian Cosmological Instrument · 2026 Architectural Reference

Core Epistemic Thesis & Fluddian Cosmology

JAMI-4 does not operate as a conventional Q&A chatbot or superficial consumer wrapper. It functions as an interactive cognitive compiler that receives messy, ambiguous, or complex multi-domain observations, decomposes them into explicit claim ASTs, formalizes underlying generative equations, subjects hypotheses to adversarial verification, and compiles verified structures into executable open-access research systems.

Vertical Axis (Lightwell):Continuous illumination from primal noise to verified structural clarity.
Radial Axis (JAMI):Interactive astrolabe mapping correspondence between macrocosm & microcosm.
Section 1.0

Daily Systems Research & Collaboration Profile

Research DomainFunctional System ExecutionFrequency
AI / Systems ArchitectureDesigning engines, pipelines, compilers, agents, epistemic systems★★★★★
Research & AnalysisTesting claims, decomposing arguments, comparing theories, finding contradictions★★★★★
Prompt / Specification EngineeringTurning an architecture into precise prompts/specs for Gemini, Claude, coding agents★★★★★
Codebase / Product EngineeringDesigning features, debugging architecture, planning implementations★★★★☆
Epistemic / Verification WorkEvidence audits, provenance, falsification, causal reasoning, anti-smuggling checks★★★★☆
Technical WritingSpecs, reports, system documentation, research artifacts★★★★☆
Exploration / SensemakingTaking complex/ambiguous material and identifying underlying structure★★★★☆
General Factual ResearchHistory, science, technology, astronomy, biometeorology★★★☆☆
Section 2.0

Dual-Layer System Architecture & Execution Runtime

Figure 2.1 — Agent Harness & Cognitive Workspace TopologyJAMI-4 System Blueprint
                         ┌──────────────────────┐
                         │       USER           │
                         └──────────┬───────────┘
                                    │
                                    ▼
                    ┌────────────────────────────┐
                    │       CHAT INTERFACE       │
                    │  Conversation · Trace UI   │
                    └────────────┬───────────────┘
                                 │
              ╔══════════════════╧═════════════════════╗
              ║             AGENT HARNESS              ║
              ║  Context Assembly  │  State Manager    ║
              ║  Planner           │  Tool Router      ║
              ║  Executor          │  Verifier (SLA)   ║
              ╚══════════════════╤═════════════════════╝
                                 │
                 ┌───────────────┼────────────────┐
                 ▼               ▼                ▼
             ┌────────┐    ┌──────────┐     ┌─────────┐
             │ MODEL  │    │  TOOLS   │     │ MEMORY  │
             └───┬────┘    └────┬─────┘     └────┬────┘
                 │              │                │
                 └──────────────┼────────────────┘
                                │
                                ▼
                 ╔════════════════════════════╗
                 ║   COGNITIVE WORKSPACE      ║
                 ║  Working State · Hypotheses║
                 ║  Observations · Evidence   ║
                 ║  Decisions · Uncertainties ║
                 ╚══════════════╤═════════════╝
                                │
                                ▼
                 ╔════════════════════════════╗
                 ║       TRACE COMPILER       ║
                 ║  private computation       ║
                 ║          ↓                 ║
                 ║  structured observations   ║
                 ║          ↓                 ║
                 ║  provenance & SLA audit    ║
                 ╚══════════════╤═════════════╝
                                │
                                ▼
                 ┌────────────────────────────┐
                 │      USER-VISIBLE TRACE    │
                 └────────────────────────────┘
Layer 01

Model Internal State

Attention matrices, activation weights, latent tensor representations, and internal activation spaces. Private neural computation.

Layer 02

Generated Scratchpad

Step-by-step reasoning text generated prior to emitting tool calls or final outputs. Formulates hypothesis candidates.

Layer 03

Harness Execution State

Tool call dispatches, external API observations, retrieved vector chunks, compiled artifacts, and SLA verification checks.

Layer 04

User-Visible Thought Process

An inspectable, stable contract displaying the system’s computationally meaningful state transitions, evidence graphs, and provenance.

Section 3.0

The Global Workspace Schema & Observable Views

Active Workspace State
ACTIVE GOAL

Determine whether H1 (Geomagnetic Nocturnal SUNDS Amplification) survives verification.

HYPOTHESES

H1: Active (beta = 0.5538, p = 0.0113)

H2: Rejected (False Analogy)

UNCERTAINTY

Causal Attribution: HIGH (99.4%)

Data Completeness: 21 Years NVSS

Section 4.0

Mathematical Formalisms & Cangjie AST Syntax

Equation 4.1 — Fixed-Effects Poisson Regression with HC3 Robust Standard Errors
\[
\log(\mathbb{E}[Y_{it} \mid X_{it}]) = \alpha_i + \gamma_t + \beta_1 \cdot \text{Geomagnetic}_{it} + \beta_2 \cdot (\text{Geomagnetic}_{it} \times \text{Nocturnal}_{it}) + \mathbf{\delta}^T \mathbf{Z}_{it}
\]
Equation 4.2 — Non-Monotonic Belief Revision Calculus (AGM Postulates)
\[
K \circ \phi = \text{ArgMin}_{K' \subseteq \mathcal{L}} \left\{ d(K, K') \mid K' \vdash \phi, \ \text{GateSLA}(K') \ge 0.998 \right\}
\]
Section 5.0

The LIDE 18-Stage Transformation Pipeline

LIDE-00LIDE Core

Create Workspace

Initialize task state, active goals, and null provenance graph.

LIDE-01LIDE Core

Extract Sources

Ingest raw literature, PDFs, and observation signals.

LIDE-02LIDE Core

Extract Claims

Decompose text into atomic verifiable claim objects.

LIDE-03LIDE Core

Build Ontology

Map claims to CIR AST canonical concept nodes.

LIDE-04LIDE Core

Generate Mechanisms

Formulate candidate generative causal equations.

LIDE-05LIDE Core

Cross-Discovery

Surfaces structural analogies across distant domains.

LIDE-06LIDE Core

Causal Analysis

Evaluate structural DAGs and backdoor criteria.

LIDE-07LIDE Core

Verification

Three-Gate Protocol compliance check.

LIDE-08LIDE Core

Falsification

Execute circular-shift null permutation tests.

LIDE-09LIDE Core

Contradiction Audit

Identify circular reasoning and unevidenced leaps.

LIDE-10LIDE Core

Parameter Fitting

Bayesian MCMC HMC/NUTS distribution inference.

LIDE-11LIDE Core

Uncertainty Quantification

Assign HC3 confidence intervals to claims.

LIDE-12LIDE Core

System Synthesis

Compile claims into executable engine spec.

LIDE-13LIDE Core

Code Generation

Produce formal TypeScript/Python instruction set.

LIDE-14LIDE Core

Runtime Mounting

Mount engine into isolated sandbox iframe.

LIDE-15LIDE Core

Telemetry Calibration

Attach real-time event listeners and stream probes.

LIDE-16LIDE Core

Provenance Locking

Generate immutable cryptographic SHA-256 graph.

LIDE-17LIDE Core

Report Compilation

Export standalone HTML/Markdown research paper.

LIDE-18LIDE Core

Audit & Return

Update internal belief base K (Model Revision).

Section 6.0

Deployed Research Engines & Open Source Repositories

biomedicalNo compliance SLA claimed

FEP Metabolic Screening Protocol Engine

Three-Gate screening protocol for First-Episode Psychosis & metabolic vulnerability under nocturnal circadian disorganization.

causalUnder review

SUNDS Geomagnetic Activity 21-Yr CDC Cohort

Fixed-effects Poisson regression modeling of nocturnal geomagnetic storm intensity (Kp/Ap index) against NVSS mortality microdata.

epistemicUnder review

3-6-9 Epistemic OS & Model Revision Calculus

Non-monotonic AGM belief revision engine for autonomous cognitive agents executing structural falsification checks.

epistemicUnder review

Cangjie Structural AST Generator & Ocompressor v5.7

Syntactic structural extractor compressing natural language research papers into formal Cangjie AST graph nodes.

causalDo-Calculus check

First-Principles Causal Attribution Engine

Backdoor criterion DAG engine isolating direct vs indirect causal pathways in multi-scale physiological datasets.

epistemicUnder review

OmniMind RAG & GASOIRE Reasoning Engine

Galilean-Aristotelian Sampling Observational Inductive Reasoning Engine for structural cross-domain analogy extraction.

physicsUnder review

Celestial Blueprint & Joplin Tornado Electrodynamics

Magnetohydrodynamic (MHD) boundary-layer coupling analysis of extreme tornadic plasma discharge channels.

physicsUnder review

Casimir Force SCUFF-EM Boundary Solver

Quantum vacuum fluctuation pressure calculations on dialectic gold-backed ITO pairs using SCUFF-EM boundary element method.

epistemicUnder review

Roko's Council Deliberation Engine

Multi-agent adversarial consensus protocol executing non-cooperative game theory balance on decision topologies.

Section 7.0

The Three Operational Roles of JAMI-4

ROLE 01

Researcher

“Find out what is actually true.” Interrogates source material, separates evidence from speculation, and extracts verifiable empirical facts.

ROLE 02

Adversarial Auditor

“Now try to prove that our interpretation is wrong.” Challenges hypotheses, performs circular-shift null tests, and rejects false analogies.

ROLE 03

Systems Architect

“How do we turn the surviving structure into an engine?” Compiles verified structures into executable code, schemas, and software architectures.

Enter JAMI-4 Frontier Chat & Execution Environment

Inspect real-time reasoning traces, workspace state graphs, live tool calls, and WebGL particle orb.