operational AI
A production readiness review for AI systemsWhy the operating model matters more than the model demoThe evidence an AI roadmap should containcoznix publishing
search the notes ↗field notes.
Practical writing on building AI systems that survive contact with real work.
product · agents · knowledge · operationslatest notes
thinking made useful.
publishing status / pre-launch
the system is ready. the claims still need people.
Public posts appear only after author, reviewer, source, and publication checks pass. Twelve launch briefs are staged across the four topic hubs.
AI product engineering
Designing an evaluation set before the interfaceChoosing a model without designing a dependencyThe vertical slice that makes AI product risk visibleknowledge systems
RAG begins with source governance, not chunk sizeWhat trustworthy citations require from your contentLearning from failed searches inside the organisationagentic operations
Bounded autonomy: a practical agent permission modelDesigning escalation before agent orchestrationWhat to observe when software starts taking actions