Editorial notes for critical systems

We unpack architecture, observability, and performance bets through the same lenses we apply in consulting: reproducibility, observability, and unit economics.

April 22, 20265 min read

OpenViking: a context database for long-running agents

Why a file-system-style memory beats flat vector search for multi-agent workflows.

A look at OpenViking through two lenses we apply to every AI infra bet: architecture and unit economics. The short version: better retrieval, better observability, better task completion, and lower cost at the same time.

AI EngineeringAgentic AIDeveloper ToolsLLMRAG

Read the noteBased on a LinkedIn post by Anton Strukov

Next step

Want the same breakdown for your stack?

We apply these same lenses in consulting: architecture, observability, unit economics. If you want a review with owners and KPIs, start with a fast request.

Send a fast request See services