[Trust Infrastructure]
Trust as Policy and Evidence Across Boundaries
Trust in a software system is a measurable property: the policy you enforce and the evidence you keep at every boundary where control changes hands.
Artemis Owl Systems
A starting map for trust, semantics, inference engineering, and middleware.
Research and working artifacts on the layers that determine whether software stays legible, governable, and dependable in production.
An independent research practice by Will Corbett — written for teams shipping agents into production systems that have to prove what they did.
Overview
What this site studies
Artemis Owl focuses on the seams where meaning, control, identity, routing, and evidence meet the runtime. Start with the pillar that best matches the pressure you are trying to name.
- Trust is runtime evidence: permissions, policy boundaries, provenance, auditability, and bounded behavior.
- Semantics is operating structure: ontologies, schemas, contracts, and explicitly typed context.
- Inference Engineering is execution control: context brokerage, semantic caching, and dynamic routing.
- Middleware is the mediation layer: routing, context assembly, and policy enforcement between models, tools, and applications.
The Four Pillars
Start with the domain that best matches the constraint.
Each pillar acts as a routing page into the main questions, references, notes, and artifacts for that part of the work.
trust
Trust
Policy, provenance, boundaries, and runtime evidence for systems that need explainable control.
semantics
Semantics
Ontologies, schemas, contracts, and context models that keep meaning stable across interfaces and time.
inference
Inference Engineering
Context brokerage, semantic caching, and dynamic routing to treat inference as a governed operational layer.
middleware
Middleware
Mediation, routing, and control between models, tools, and applications, where multi-step work stays legible or breaks down.
Recent research
Notes on trust, semantics, inference, and middleware.
[Trust Infrastructure]
Middleware as Mediation and Control
Middleware is the mediation and control layer of a system: the place where translation, routing, context assembly, and policy become operational.
[Inference Ops]
Semantic Caching Is an Authorization Boundary
Semantic caching reuses model responses by meaning rather than exact match. That turns the cache into a place where authorization decisions are made, not just a latency optimization.
[Semantic Ops]
Semantics Belongs in the Runtime
Semantics is shared meaning made operational: the contract stack that lets systems interpret, exchange, and preserve meaning without drift.
Artifacts
Selected builds, prototypes, and working systems.
Semantics
Ontology Ops Kit
A starter kit for treating ontologies as living, version-controlled operational assets alongside your code.
Inference
Enterprise DecisionOps Control Plane
A local-first reference architecture for Enterprise DecisionOps, demonstrating how AI agents operate safely by mapping runtime setup to verifiable audit evidence.
Middleware
Context Broker Patterns
A developing pattern library for mediating context, tools, and policy across software workflows.
Operating principles
A working posture for systems that need to hold up in public.
Make systems legible
Clarify boundaries, responsibilities, and evidence so behavior can be understood instead of inferred.
Name the real constraint
Distinguish whether the problem sits in trust, semantics, middleware, platform design, or operations.
Work from structure
Schemas, contracts, routing logic, and runtime conditions shape outcomes as much as any single component.
Prefer evidence over posture
Provenance, observability, and enforceable controls matter more than claims of reliability.
Get in touch
Work grounded in real system constraints
If you are working through these layers in a real platform, reach out.