Semantic Systems Need State, Not Just Knowledge
Why operational AI needs explicit state, transitions, permissions, provenance, uncertainty and responsibility alongside structured knowledge.
- Semantic Intelligence
- Knowledge Graphs
- State Management
CARM'INK Journal / Publication layer
The lab's publication layer for semantic intelligence, inference engineering, governed agentic systems and the conditions that make technical work useful in the world.
19 published articles
Why operational AI needs explicit state, transitions, permissions, provenance, uncertainty and responsibility alongside structured knowledge.
How to combine probabilistic interpretation with deterministic authorization, validation, state transitions, and human review in production AI systems.
A concise thesis on why production AI must be understood as a governed system of knowledge, context, rules, state, tools, inference, observability, and people.
Inference engineering covers the full path from model and runtime selection to context, routing, output contracts, fallbacks, cost, observability, and controlled execution.
When retrieval finds relevant passages but cannot connect them, context graphs can organize evidence, relationships, and constraints for inference.
CARM'INK’s engineering thesis: make entities, relationships, constraints, and provenance explicit before selecting context and generating an answer.
Why consequential AI actions need independent authorization, meaningful oversight, and clear limits on agent autonomy.
Context engineering starts before prompt writing: structure knowledge, resolve evidence, and select the context that inference can use.