Personal Intelligence Infrastructure
Artificial intelligence can reason and act, but must never become the authority over itself.
TELHAR decouples memory, cognition, evidence, policy, and execution into distinct, governed planes — preserving human agency across time, surfaces, and models.
Five Planes. Zero Ambiguity of Authority.
Traditional AI systems collapse memory, permissions, reasoning, and network access into opaque black boxes. TELHAR enforces explicit mathematical and cryptographic boundaries.
Mnḗmē
Continuous context, temporal indexing, and belief state. Never a global truth or identity authority.
Praxis & ARK
Parses intent, formulates plans, and deliberates. Proposes actions, but holds zero intrinsic execution authority.
Foedus
Policy Decision Point. Evaluates covenants and deterministic rules prior to any state mutation in the world.
Explorator & Refuto
Protected external discovery and independent falsification harness. "An agent cannot be judge of its own execution."
Experience Platform
Multimodal perceptual projection over a singular core. The modality is an adapter, not a reimplementation.
The 7 Controls of the Effect Gate
No model or agent has direct connectivity to actuators, APIs, or filesystem effects. Every action proposal crosses the 7 controls deterministically, terminating at the first DENY verdict.
From Human Intent to Verifiable Action
A simple human experience supported by rigorous systems engineering. Every consequential step is audited, authorized, and anchored.
1. Person & Identity
A sovereign individual authenticates via cryptographic OIDC 2.1 credentials. Identity is a verified principal, never an ephemeral cookie.
2. Intent Formulation
The individual expresses a goal. The cognitive engine parses intent and retrieves relevant governed context from Mnḗmē.
3. Evidence Gathering
If external facts are needed, Explorator queries sources under strict egress guards and cost ceilings, producing graded claims.
4. Proposed Plan Hash
Cognition proposes a concrete action set. Its cryptographic hash (approvedSetHash) is bound to the specific turn.
5. Policy Authorization
Foedus evaluates the proposed plan against active tenant policy. Missing rules result in an immediate fail-closed DENY.
6. Effect Gate Execution
The singular chokepoint validates rate limits, tenant isolation, and capability tokens before invoking external tools.
7. Evidence Anchoring
The result, causation ID, and decision token are recorded into an append-only ledger for post-execution verification.
State of the System
We do not confuse vision with current capabilities. We classify every component with empirical honesty.
| Component | Domain | Status | Repository Evidence |
|---|---|---|---|
| Mnḗmē (ai/mneme) | Memory Plane | PROVEN | Continuous memory with pgvector, KMS/DEK encryption, FTS, and 214 active contract references. |
| Explorator (ai/explorator) | Evidence Plane | PROVEN | Fail-closed egress guard & SSRF classifier with 133,921 tested addresses (0 leaks, 17.8 ns). |
| Refuto (/refuto) | Assurance & Falsification | PROVEN | Independent falsification harness with append-only ledger and strict 6-status contract. |
| Agent Runtime Kernel (ARK) | Control Automaton | PROVEN | Formally model-checked FSM in TLA+ (TLC 1.8.0, 149 states with zero deadlocks). |
| Foedus (security/foedus) | Policy Plane (PDP) | IMPLEMENTED | OPA/Rego authorization engine issuing DecisionTokens with tenant boundary validation. |
| Gateway (core/gateway) | Exposure Gateway | IMPLEMENTED | Native Axum binary serving as execution chokepoint for governed traffic. |
| Praxis (ai/praxis-api) | Cognitive Plane | IMPLEMENTED | Cognitive agent runtime and bridge (CommandTracker and RemediationEngine with cooldown). |
| Aesthesis (ai/aesthesis) | Perceptual Physics | PROVEN | Perceptual physics engine grounded in physical equations (Sellmeier, Fresnel, APCA-W3). |
| Experience Platform (platform/experience) | Experience Plane | EXPERIMENTAL | Multi-surface monorepo (Web, Tauri, React Native). CLI TUI in production; Web in scaffolding. |
| Protect-OS | Subordination Vision | CONCEPTUAL | Guiding principle subordinating cognition to governance. Implemented piecewise in ARK, Foedus, and Gateway. |
Architecture for Intelligent Systems
TELHAR is governed by an unyielding Brand & Engineering Constitution. It rejects market hype, decorative technology, and anthropomorphic illusions.
Read the 12 Principles →The Three Refusals (Canonical Invariants)
- • We do not sell AI — we design intelligent systems.
- • We do not build isolated products — we build governed capabilities.
- • We do not automate blindly — we architect enduring ecosystems.