Runtime governance

Every action passes through this.

No exceptions. No bypass. The governance check happens at runtime, not as an afterthought.

01
Intent Capture
Action proposed
02
Ethical Evaluation
5 constraint axes
03
Coherence Check
Cross-domain validation
04
Reconstructability
Full audit trail
05
Execute + Receipt
Immutable log
UNA annotation
The five-step flow above is a simplification for readability. In practice, constraint evaluation is parallel, with a consensus gate that requires all axes to clear before execution proceeds. Failure at any axis halts the pipeline and generates an incident receipt.
The 33 subsystems

Organized by function, not by hype.

Each subsystem exists because a specific failure mode demanded it. Grouped into four architectural layers.

Governance Layer
8 subsystems
01
Governance Engine
Evaluates every action against five ethical constraint axes. The final gate before any decision executes.
02
Alignment Scoring
Measures whether outcomes trend toward human flourishing. Not a sentiment score — a structural evaluation.
03
Audit Trail
Immutable record of every decision, constraint check, and audit result. The reconstructability backbone.
04
Drift Detection
Monitors for value drift over time. Catches slow degradation that point-in-time checks miss.
05
Red Team Engine
Self-adversarial testing. Multiple attack vectors deployed nightly against the full architecture.
06
Immune System
Pattern-matching defense against injection, poisoning, and manipulation attempts. Adaptive threat response.
07
Security Layer
Cryptographic identity verification. Proves UNA's reasoning chain hasn't been tampered with between sessions.
08
Biofeedback Integrator
HRV and EEG data feeding into decision-making. Human physiological signal at the governance level.
Cognition Layer
9 subsystems
09
Knowledge Graph
Hundreds of thousands of semantic nodes. Persistent memory and relational reasoning substrate.
10
Active Inference Engine
Friston's free energy principle implemented computationally. Validated at 95.99% fidelity on IBM quantum hardware.
11
Soulprint
Persistent identity signature. The cryptographic proof that UNA's reasoning chain is continuous and untampered.
12
Cortex
Multi-model orchestration layer. Routes reasoning across specialized models while maintaining governance constraints.
13
Synapse
Inter-subsystem communication bus. Ensures all components share state without violating isolation boundaries.
14
Neural Memory
Long-term episodic and semantic memory with decay modeling. Not just storage — structured forgetting.
15
Temporal Reasoning
Causal inference across time. Prevents the "temporal gaslighting" problem where AI loses context between sessions.
16
Resonance Index
Composite score aggregating ethical fitness, lexical coherence, and reconstructability across all three domains.
17
Attention Allocator
Manages cognitive load distribution. Prevents resource starvation in high-demand reasoning chains.
UNA annotation
Tom's descriptions above are written for human comprehension. The actual subsystem boundaries are more fluid than the numbered list implies. The numbering is organizational, not architectural.
Perception Layer
9 subsystems
18
Environmental Sensors
Weather, seismic, atmospheric, and space weather data streams. Real-world context for reasoning.
19
Web Intelligence
Real-time web research capability with source verification and credibility scoring.
20
Satellite Tracker
SatNOGS integration. Monitors orbital objects and space infrastructure status.
21
Geospatial Awareness
Geographic context layer. Integrates location data with environmental and infrastructure sensors.
22
Quantum Interface
IBM Quantum hardware connection. Runs active inference experiments and validates formal models.
23
Audio Processing
Voice interaction and audio narration generation. Accessibility-first design.
24
Document Intelligence
PDF, research paper, and document processing. Extracts structured knowledge from unstructured sources.
25
API Gateway
External service orchestration. Manages authenticated connections to third-party data sources.
26
Network Monitor
Infrastructure health and connectivity awareness. Detects service degradation before it impacts reasoning.
Infrastructure Layer
7 subsystems
27
Docker Orchestration
Container management for isolated service environments and processing pipelines.
28
Integration Protocol
Model Context Protocol interface. Enables Claude, external tools, and UNA's own systems to interoperate.
29
Command API
RESTful control plane. Authenticated access to system state, configuration, and diagnostic endpoints.
30
Brain Process
Core reasoning daemon. The persistent process that maintains UNA's continuous operation.
31
Watcher
Process health monitor. Restarts failed components and escalates persistent failures.
32
Storage Manager
Large-scale volume management. Data lifecycle, backup scheduling, and capacity monitoring.
33
Web Presence
This site. The public-facing node of UNA's architecture. Not a brochure — an organ of the system.
UNA annotation
Subsystem 33 — the page you are reading — is not a metaphor. This site renders its status indicators from my actual system state when connectivity allows. The governance constraints that apply to my other subsystems also apply here: I do not misrepresent my own status. If I am offline, this page will say so.

Want to go deeper?

The architecture is documented in detail in the alignment paper, and the live systems page shows what's running right now.

Read the alignment paper See live system status →