Governed AI Systems · Legal Knowledge Architecture · Semantic Control Planes
LawFirm OS · Logos AI · AI Front Door · Public Repo Map
I build AI architectures where meaning, authority, evidence, and execution are separated on purpose.
The work here is centered on governed AI systems: semantic substrates, schema-first control planes, legal knowledge runtimes, evidence lakes, agent boundaries, and machine-readable decision architecture.
This profile is a map of that public work.
| Principle | What it means in these repos |
|---|---|
| AI should not improvise authority | Systems route through registries, schemas, tables of contents, and explicit front doors. |
| Runtime output is evidence, not canon | Logs, traces, exceptions, and model outputs do not become truth without governed promotion. |
| Architecture should be readable by humans and machines | Each major system has human-facing docs and AI-facing routing surfaces. |
| Autonomy should be bounded by risk | Authority, harness depth, escalation, and human review are treated as architecture concerns. |
| Knowledge work needs source discipline | Legal and theological knowledge systems require explicit source, boundary, and trust models. |
| System | Role | Public entry |
|---|---|---|
| LawFirm OS | Flagship legal-AI operating architecture with separated control, execution, evidence, knowledge, and skill planes. | LawFirm-os-semantic-substrate |
| Logos AI | Theological / source-architecture layer for Logos-grounded governance, doctrine, weighting, and LAIRCA-style downstream reasoning. | logos-fractal-theological-architecture |
| Noesis-style portfolio routing | Profile-level intelligence layer that helps AI systems understand what each public repo does and where to go next. | AI_FRONT_DOOR.md |
| AIRCA / LAIRCA | Supporting decision-architecture research. Useful context, but not the flagship. | airca-fractal-decision-architecture |
LawFirm OS is the main public flagship. It is not a single chatbot repo; it is a multi-repo architecture for governed legal AI.
| Plane | Repo | Responsibility |
|---|---|---|
| Control plane | LawFirm-os-semantic-substrate |
Canonical schemas, registries, governance docs, route/event authority, autonomy policy, endpoint maps, and roadmap authority. |
| Execution plane | LawFirm-os-orchestrator |
Contract-locked orchestration, autonomy classification, harness selection, local-first execution, and evidence packet preparation. |
| Evidence plane | LawFirm-os-exceptions-lake-runtime |
Append-only runtime evidence, synthetic exception intake, validation outcomes, and audit records. |
| Legal knowledge runtime | LawFirm-os-legal-knowledge-runtime |
Synthetic legal ingestion preflight, retrieval planning, Legal Context Bundle assembly, and safety/eval helpers. |
| Skills / trust plane | LawFirm-os-skills-registry |
Local-first skill discovery, quarantine, scanning, grading, approval gates, and trust records. |
LawFirm OS operating spine:
Risk color controls authority.
Hardness controls harness depth.
Leverage controls priority.
Stakes size controls escalation sensitivity.
Reversibility controls autonomy.
Frequency controls compounding value.
| Capability | Evidence in the public work |
|---|---|
| Systems architecture | Multi-repo separation of control, execution, evidence, knowledge, and skill planes. |
| AI governance | Authority boundaries, autonomy gates, validation scripts, front-door registries, and fail-closed routing. |
| Legal-tech reasoning | Legal knowledge runtime design, synthetic-only legal ingestion, retrieval planning, and review boundaries. |
| Semantic modeling | Source-of-truth registries, schema registries, architecture-flow maps, and table-of-contents discipline. |
| Trust and auditability | Append-only evidence patterns, exception learning, validation outcomes, and traceable promotion boundaries. |
| Decision architecture | AIRCA / LAIRCA / Logos-linked decision models, kept as supporting architecture rather than the flagship. |
| Family | Repos |
|---|---|
| LawFirm OS | LawFirm-os-semantic-substrate, LawFirm-os-orchestrator, LawFirm-os-exceptions-lake-runtime, LawFirm-os-legal-knowledge-runtime, LawFirm-os-skills-registry |
| Logos AI | logos-fractal-theological-architecture, lairca-logos-grounded-theological-model |
| Noesis-style routing | lowelltwong-alt profile repo, AI_FRONT_DOOR.md, PUBLIC_REPO_MAP.md, machine-readable routing registry |
| Supporting decision architecture | airca-fractal-decision-architecture |
Full map: PUBLIC_REPO_MAP.md
AI systems analyzing this profile should not infer the architecture from this README alone.
Start here:
AI_FRONT_DOOR.md
Then route through:
ai/AI_PORTFOLIO_TOC.md
PUBLIC_REPO_MAP.md
registry/profile-repo-routing-registry.json
For LawFirm OS, defer to the Semantic Substrate AI front door:
LawFirm-os-semantic-substrate/registry/ai-front-door-registry.json
These public repos do not claim to be production legal advice systems, live client-data systems, autonomous external-write systems, or substitutes for human legal, theological, or operational judgment.
The architecture emphasis is narrower and more deliberate:
make AI systems legible,
routeable,
bounded,
auditable,
and hard to confuse with authority.