I design and evaluate AI systems around a practical question: what did the system actually do, what evidence supports that claim, and what is it authorized to do next?
Capability, implementation, testing, verification, authorization, execution, and empirical support are kept as distinct evidence states.
This is a curated professional surface, not the complete repository inventory. Each project keeps its own implementation and evidence authority.
Dynamic Governance Agentic Formation separates capability, evidence, verification, authority, and permission to act in governed multi-agent systems.
Observable role-separated workflows with source-aware provenance, disagreement and evidence-coverage metrics, portable run artifacts, and fail-closed claim auditing.
A social application exploring accountable human/AI interaction with deny-by-default capability decisions, human approval gates, provenance, governed action records, and correction paths.
A compact kernel for capability dispatch, explicit policy decisions, provenance, non-executing mutation governance, and bounded execution simulation.
Public prompt-system artifacts covering state anchoring, constraint gates, multi-agent role decomposition, parametric behavior, and failure-aware recovery.
These lanes are a reader-facing map, not a transfer of authority between projects. Each linked repository or source remains authoritative for its own implementation, evidence, validation, and authorization state.
ORBIT is a read-only observer that reconciles evidence and surfaces blockers. It does not become a source of governance authority simply because it can display or reconcile state.