# AEGISES

> Intelligence with integrity. Public reference architectures for governable intelligence and sovereign compute.

## Reactor

Author: Ryan Smithright · [Qthonic Labs](https://qthonic.com/) · ryan@smithright.com.
Reactor: A Governable Reference Architecture for Enterprise AI Agent Orchestration.
Public design draft 2.6, 14 September 2026. A proposed governable architecture, not a validated implementation, certified deployment or claim of general alignment.

- [Canonical article](https://aegises.org/agi-reactor/): Complete static HTML, interactive system map and worked journeys.
- [Complete plain-text article](https://aegises.org/agi-reactor/reactor.txt): Generated from the same article, including diagram labels and architecture explanations; readable without JavaScript.
- [Technical specification](https://aegises.org/agi-reactor/architecture.md): Record semantics, concurrency, authority, effect recovery and upgrade contracts in Markdown.

- [Implementation and responsible design disclosure](https://aegises.org/agi-reactor/#implementation-disclosure): Qthonic is implementing the architecture. Recursive software-improvement experience is author-reported; implementation status is distinct from qualification.
- [Executable contract evidence](https://aegises.org/agi-reactor/#qualification): Source-derived local specimen, 14 tests, mutation controls, measured local admission latency and explicit limitations.
- [Download executable specimen](https://aegises.org/agi-reactor/reactor-contract-demo.zip)
- [AI Safety Risks](https://aegises.org/agi-reactor/#ai-safety-risks): Thirty proposed risk-control mappings with service responsibilities, records, accountable owners, qualification exercises and remaining risks.
- [Risk register JSON](https://aegises.org/agi-reactor/ai-safety-risks.json): Structured proposals and risk-framing references; not validated controls or standards conformance.

## State AI legislative maturity atlas

Beyond the Bill Count — AEGISES research edition 1.0, observed September 14, 2026. A representative legislative corpus across all 50 states, with DC separately unclassified. This is not an exhaustive legal inventory, a numeric state ranking or an outcomes evaluation. Missing domain cells mean not assessed, not no law.

- [Publication page](https://aegises.org/ai-legislation/): Complete initial HTML, six original analytical figures and a searchable state ledger.
- [Complete text](https://aegises.org/ai-legislation/ai-legislation.txt): Original analysis, limitations and source links.
- [Structured evidence](https://aegises.org/ai-legislation/data/states.json): Instruments, source types, claims, interpretation boundaries and observation date.
- [CSV ledger](https://aegises.org/ai-legislation/data/states.csv)
- [PDF publication](https://aegises.org/ai-legislation/state-ai-legislative-maturity.pdf)

## Related architectures

- [Sovereign Compute Architecture](https://aegises.org/architecture/)
- [Public Policy Ontology for CHERI](https://aegises.org/cheri/)
- [Typed Policy Reactor](https://aegises.org/pansigna/)
- [CHERI-seL4-NixOS](https://aegises.org/gap-map/)

## Visual specification

The architectural thesis is “Policy-governed service orchestration is all you need”: services compose cognition, memory, work, policy, learning and recovery under explicit contracts. The diagrams distinguish agentic judgment from process and follow ingress, artifact lineage, concurrent work, typed bindings and service enforcement. Their complete labels and explanations are in the plain-text article above.

- [Preserved public draft 1.3](https://aegises.org/agi-reactor/versions/1.3/)