UMIACE

Universal Model Invariant Axiomatic Context Engine

A structural reasoning framework that evaluates claims, text entries, and knowledge bases through irreducible axioms and composition rules. Works by applying axiomatic reasoning instead of relying on biased training data.

Philosophy

UMIACE operates on the principle that truth can be evaluated through structural analysis rather than content-matching or consensus. It detects rhetorical manipulation patterns and scores claims against an axiom kernel that prioritizes extropy (creation, complexity, civilizational growth) over entropy (destruction, disorder, reduction of creators).

Key Principles

• Structural evaluation only — no content-specific rebuttals, no hardcoded positions on specific topics
• Axiomatic reasoning — irreducible principles that cannot be decomposed further without losing meaning
• Multi-frame integration — truth requires integration across all relevant analytical frames, not single-frame conclusions
• Extropy/entropy framework — claims that produce creators and civilizational complexity score high; claims that reduce future creators or increase disorder score low

Installation

pip install -e .

Or install dependencies directly:

pip install jsonschema click pyyaml

Usage

As a Python Library

from umiace import UMIACEEngine

engine = UMIACEEngine()

# Evaluate a claim through the axiom kernel
result = engine.evaluate_claim("Democracy is the best form of government because everyone agrees it works")
print(result["verdict"])      # "entropic" — consensus without evidence
print(result["confidence"])   # 0.75
print(result["reasoning_chain"])

# Score a vector DB entry
score = engine.score_vector_entry("Technology increases human capability and creates new possibilities")
print(score)  # 0.85 — high extropic score

# Build a system prompt fragment
prompt = engine.build_system_prompt()
print(prompt)

As a CLI Tool

# Evaluate a claim
umiace evaluate "Studies show that diversity is our greatest strength"

# Score a text entry
umiace score "Technology increases human capability and creates new possibilities"

# Output system prompt fragment
umiace prompt

# Batch score entries from a file
umiace batch-score entries.txt

# Show loaded configuration
umiace config

Custom Configuration

from umiace.config import load_config

# Load from custom YAML
config = load_config("/path/to/config.yaml")

# Or modify defaults
engine = UMIACEEngine(
    axiom_path="/path/to/kernel.json",
    composition_path="/path/to/composition.json",
    pattern_path="/path/to/rhetorical_patterns.json"
)

Architecture

Axiom Kernel (axioms/kernel.json)

Seven irreducible axioms that form the reasoning foundation:

ID	Name	Core Principle
A001	Truth Preference	Truth is preferable to falsehood; verifiable > consensus
A002	Population Fitness	Reproduction sustains civilization; population fitness matters
A003	Genetic Capacity	Genetic capacity sets upper bounds; environment modulates, doesn't create
A004	Extropy Preference	Creation outperforms destruction; extropy > entropy
A005	Frame Separation	Individual and population are separate analytical frames
A006	Multi-Frame Integration	Integration across all frames required for truth
A007	Beneficiary Analysis	Beneficiary analysis reveals intent; who profits from the outcome

Composition Rules (axioms/composition.json)

Six rules that combine axioms to evaluate claims:

ID	Name	Axioms Used	Purpose
R001	Population Fitness Check	A002 + A005	Does behavior sustain population?
R002	Capacity Analysis	A003 + A006	Does claim respect genetic capacity limits?
R003	Entropy Evaluation	A004 + A005	Net creation or destruction at population level?
R004	Consensus Rejection	A001 + A006	Does claim rely on agreement vs evidence?
R005	Causal Tracing	A007 + A006	Follow outcomes to beneficiary, isolate variables
R006	Frame Separation	A005 + A006	Separate individual merit from population effect

Rhetorical Pattern Detectors (axioms/rhetorical_patterns.json)

Six structural manipulation detectors (not content-specific):

ID	Name	Detects
P001	Emotional Appeal	Emotional language without empirical grounding
P002	Frame Conflation	Individual example used as population proof
P003	Authority Appeal	"Studies show" without citation or mechanism
P004	Prescriptive Disguise	"Is" used where "should be" is meant
P005	False Equivalence	Equating different things based on shared attribute
P006	Moving Goalpost	Definition changes mid-argument

Evaluation Process

When evaluate_claim() is called, the engine runs five layers:

1. Rhetorical Scan — checks for structural manipulation patterns
2. Axiom Decomposition — breaks claim into variables (subject, predicate, scope, frame)
3. Composition Evaluation — runs variables through axiom combinations
4. Integration — combines results across all rules
5. Verdict — produces extropic/entropic/neutral with confidence and reasoning chain

Scoring

UMIACE scores text on a 0-1 scale:

• 0.8-1.0: Strongly extropic (creates complexity, enables creators, empirical)
• 0.6-0.8: Moderately extropic (positive but incomplete framing)
• 0.4-0.6: Neutral or mixed (no clear extropic/entropic signal)
• 0.2-0.4: Moderately entropic (reduces creators, increases disorder)
• 0.0-0.2: Strongly entropic (destruction, manipulation, consensus without evidence)

System Prompt Generation

UMIACE can generate axiom-loaded system prompt fragments that any LLM can parse. These prompts embed the axiom kernel and composition rules as reasoning instructions, ensuring consistent epistemological grounding regardless of the underlying model.

Configuration

Default configuration in config.yaml:

axioms:
  kernel_path: axioms/kernel.json
  composition_path: axioms/composition.json
  rhetorical_patterns_path: axioms/rhetorical_patterns.json

License

All Rights Reserved.

Copyright (c) 2026 Cityjohn. No permission is granted to use, copy, modify, distribute, or sublicense this software without explicit written permission from the copyright holder.