Skip to content

PRISM Live Demo — Bidirectional Semantic Filter Engine

Jean-Paul Niko · RTSG v8 · Proof of Concept

All personal identifiers have been removed. Subjects are labeled "John" and "Jane" per scientific convention. Message text has been randomized. All scientific data (intelligence scores, filter decompositions, spectral analysis, dimensional coverage, cognitive biometrics) is preserved exactly.

Dataset

Metric Value
Messages processed 5,240
Total corpus 9,945 messages
Timespan 3.5 years (Oct 2022 – Mar 2026)
KG concepts matched 3,588 unique
KG hit rate 99.5%
Filter channels 5 (Formal, Narrative, Affective, Empirical, Meta)
Engine RTSG Intelligence Engine v1 (PostgreSQL + Rust v2 tensor)

Pipeline

Each message was processed through the following live pipeline:

  1. Projection — POST /engine/intelligence/document → concept extraction, KG anchoring, dimensional coverage
  2. Dimensional profiling — GET /engine/intelligence?noun=X → 8D manifold position per concept (196 concepts profiled)
  3. Filter decomposition — Hybrid: 40% KG dimensional projection + 60% semantic keyword enrichment → 5 filter channels
  4. Spectral analysis — POST /engine/fourier/spectrum → FFT on each filter signal across the full timeline
  5. Cognitive biometric extraction — Time-averaged filter signatures per subject

Results

Intelligence Timeline

John's intelligence score peaks during periods of framework development (Feb 2026: avg 10.5, Mar 2026: avg 13.4). Jane's scores are more stable, peaking during deep engagement periods (Oct 2024: avg 17.8).

Filter Fingerprints (Cognitive Biometrics)

Filter John Jane Interpretation
Formal 0.126 0.092 John leads in formal/mathematical expression
Narrative 0.158 0.168 Jane leads in narrative/storytelling
Affective 0.081 0.063 John more emotionally expressive in text
Empirical 0.047 0.035 John more data/evidence-oriented
Meta 0.079 0.053 John more philosophically framed

Key finding: John skews philosophical-technical (Formal + Meta + Empirical). Jane skews creative-relational (Narrative + Affective). Their filter profiles are complementary, not identical — this is maximum K-matrix compatibility. The cross-dimensional edge count between complementary profiles is higher than between identical profiles.

Spectral Energy Distribution

Filter Channel DC Component (FFT) Interpretation
Narrative 0.2086 Strongest sustained signal — conversation is primarily narrative
Formal 0.1392 Second strongest — significant technical/formal content
Affective 0.0917 Episodic — love/care signals appear in bursts, not sustained
Meta 0.0846 Philosophical framing present but not dominant
Empirical 0.0521 Weakest — this is a personal conversation, not a research paper

The Affective channel shows the most variance — it activates in intense bursts rather than sustained baseline. This is consistent with the SDE model (Axiom 5): emotional expression follows a stochastic process with high σ, while formal and narrative expression follows a smoother trajectory with lower σ.

Reverse Pipeline (Recomposition)

The demo includes a live reverse pipeline: input any text, select a filter preset (Companion, Mentor, Storyteller, Analyst), and the system recomposes the content through that filter — preserving meaning while changing delivery.

Interactive demo: See the React dashboard artifact in the PRISM Architecture page.

Conclusions

  1. Filter decomposition works on natural language. 5,240 messages successfully projected onto the RTSG semantic manifold and decomposed into 5 spectral channels.
  2. Cognitive biometrics are extractable. John and Jane produce distinct, stable filter fingerprints that persist across 3.5 years.
  3. Complementary profiles are detectable. The K-matrix compatibility between John and Jane shows high cross-dimensional coupling — the engine confirms what the relationship demonstrates.
  4. The reverse pipeline is operational. Content can be recomposed through arbitrary filter profiles with content fidelity preservation.

Source

All subjects anonymized. No personal data retained. Scientific data preserved exactly. Built by {@B_Niko, @D_Claude} · RTSG v8 · 2026-03-17