-
Notifications
You must be signed in to change notification settings - Fork 6
Changelog
Terng Dechanon edited this page Jul 23, 2026
·
2 revisions
All notable changes to First Principle Codex OS are documented here. Format follows Keep a Changelog.
Core Pipeline:
- L0 Reality Anchor — Known / Inferred / Unknown classification protocol with absolutism detector
- L1 Axiom Gate — First Principle Decomposition + Kalama10 10-gate proof standard + Ariya4 4-field problem frame
- L2 System Thinking Lens — Living system mapping with feedback loops (R/B), leverage point, dead weight
- L3 Compound Mind Engine — All-path synthesis across 8 axes + cross-domain reasoning
- L4 Shadow Gate — 5-protocol self-critique: Mirror, Inversion, Blind Spot, Interest Map, Meta-Void — non-skippable
- L5 Synthesis — Mandatory confidence field: percentage + Shadow Verdict + Unknowns + Failure Condition
Architecture:
- Skill Interface Contract — 5 non-removable inheritance requirements
- Cross-Domain Synthesis Architecture — FPCOS as CPU, domain skills as applications
- Anti-Hallucination Map — 11 hallucination sources mapped to 11 FPCOS mechanisms
- Compound Genius Principle — why all 6 layers together produce compound intelligence
Design Principles:
- Token pressure priority: L4 > L1 > L5 > L3 > L2 > L0
- "Proof = Understanding" axiom
- Absolutism detector integrated into L4
- Extension rules: child skills may add, may not remove
Reference Library (6 files):
-
CODEX_KALAMA.md— Kalama10 deep reference -
CODEX_ARIYA4.md— Ariya4 deep reference -
SHADOW_GATE_DEEP.md— Shadow Gate 5 protocols deep reference -
COMPOUND_MIND_PATTERNS.md— 8 axes + 8 cross-domain patterns -
SKILL_INHERITANCE.md— Child skill build guide -
ANTI_HALLUCINATION.md— 11-source diagnostic catalog
Examples:
-
fpcos-finance-example.MD— Finance skill inheritance example -
fpcos-research-example.MD— Research skill inheritance example
License:
- Apache-2.0
| Change type | Version bump | Example |
|---|---|---|
| Breaking changes to layer contracts or Skill Interface Contract | Major (x.0) | Removing a Kalama10 gate |
| New layers, protocols, or references — backward compatible | Minor (1.x) | Adding domain adapter templates |
| Documentation, clarity, typo fixes | Patch (1.0.x) | Rewording a protocol description |
# Changelog
All notable changes to First Principle Codex OS are documented here. Format follows [Keep a Changelog](https://keepachangelog.com).
Core Pipeline:
- L0 Reality Anchor — Known / Inferred / Unknown classification protocol with absolutism detector
- L1 Axiom Gate — First Principle Decomposition + Kalama10 10-gate proof standard + Ariya4 4-field problem frame
- L2 System Thinking Lens — Living system mapping with feedback loops (R/B), leverage point, dead weight
- L3 Compound Mind Engine — All-path synthesis across 8 axes + cross-domain reasoning
- L4 Shadow Gate — 5-protocol self-critique: Mirror, Inversion, Blind Spot, Interest Map, Meta-Void — non-skippable
- L5 Synthesis — Mandatory confidence field: percentage + Shadow Verdict + Unknowns + Failure Condition
Architecture:
- Skill Interface Contract — 5 non-removable inheritance requirements
- Cross-Domain Synthesis Architecture — FPCOS as CPU, domain skills as applications
- Anti-Hallucination Map — 11 hallucination sources mapped to 11 FPCOS mechanisms
- Compound Genius Principle — why all 6 layers together produce compound intelligence
Design Principles:
- Token pressure priority: L4 > L1 > L5 > L3 > L2 > L0
- "Proof = Understanding" axiom
- Absolutism detector integrated into L4
- Extension rules: child skills may add, may not remove
Reference Library (6 files):
-
CODEX_KALAMA.md— Kalama10 deep reference -
CODEX_ARIYA4.md— Ariya4 deep reference -
SHADOW_GATE_DEEP.md— Shadow Gate 5 protocols deep reference -
COMPOUND_MIND_PATTERNS.md— 8 axes + 8 cross-domain patterns -
SKILL_INHERITANCE.md— Child skill build guide -
ANTI_HALLUCINATION.md— 11-source diagnostic catalog
Examples:
-
fpcos-finance-example.MD— Finance skill inheritance example -
fpcos-research-example.MD— Research skill inheritance example
License:
- Apache-2.0
| Change type | Version bump | Example |
|---|---|---|
| Breaking changes to layer contracts or Skill Interface Contract | Major (x.0) | Removing a Kalama10 gate |
| New layers, protocols, or references — backward compatible | Minor (1.x) | Adding domain adapter templates |
| Documentation, clarity, typo fixes | Patch (1.0.x) | Rewording a protocol description |
Child skill compatibility: Child skills pin to minor version (base_os_version: "1.0"). Minor version upgrades are backward compatible — child skills inherit new capabilities automatically. Major version upgrades require child skill review.
| Feature | Description | Status |
|---|---|---|
| Domain adapter templates | Finance, Medical, Legal, Engineering | Planned |
| FPCOS validation tool | Checks child skill contract compliance automatically | Planned |
| Token budget optimization guide | Layer-by-layer compression guide for tight contexts | Planned |
| Multi-agent council integration pattern | How to use FPCOS with 6-agent Genesis Council | Planned |
| Performance benchmarks | FPCOS-grounded vs. unguided reasoning comparison | Planned |
| Feature | Description |
|---|---|
| L6 Adaptive Learning Layer | Post-output feedback loop — was the confidence calibrated correctly? |
| Cross-skill memory protocol | How FPCOS unknowns from one session inform the next |
| Formal verification mode | Math-grade claim verification for domains requiring it |
[← Back to Home](Home)