Resources
Field notes
On agentic AI governance, data sovereignty, bring-your-own-model, and building a second mind your organisation actually controls.
BYOM: why an agent platform shouldn't lock your LLM
BYOM explained — bring your own model so your agent platform never locks you to one LLM. The three protocol shapes, LLM-as-resource, no token markup, portability.
Read ArchitectureData sovereignty for agentic AI: self-hosted vs cloud
Data sovereignty for agentic AI explained — what self-hosted vs cloud-with-your-keys actually guarantees, and why KMS separation matters more than location.
Read Governance & ReBACHuman-in-the-loop as a first-class primitive (not a workaround)
Human-in-the-loop for AI agents should be a runtime primitive, not a bolt-on approval queue. How Oraclous treats waiting on a human like waiting on a tool return.
Read Open source & portabilityOHM: a portable manifest format for AI agents
OHM is Oraclous's portable manifest for AI agents — a Harness as YAML plus Markdown. What it carries, where it can travel, and the limits the docs state up front.
Read Governance & ReBACPlatform-as-code: separating governance from agent behaviour
Platform-as-code separates governance (enforced in code) from agent behaviour (interpreted from prose). When they conflict, code wins. Why that beats governance-in-app-logic.
Read GovernanceReBAC vs RBAC for AI agents — which fits?
ReBAC vs RBAC for AI agents — why role-based access hits role explosion with autonomous agents, and why relationship-based access fits agentic work.
Read ConceptsWhat is a 'second mind'? Human-AI collaboration
A second mind is an organisation's combined human and AI capacity, governed as one fabric. How it differs from an AI assistant, copilot, or chatbot.
Read