Peter Werry contrasts scattered organizational knowledge with task-specific context assembled from code, pull requests, documents and conversations. His examples compare coding-agent sessions with and without Unblocked and trace a regression to a model change. The reported cost reduction is an illustrative session comparison, not a universal benchmark.
Peter Werry explains why a vector searchSemantic search finds information by the meaning of a query rather than relying only on exact matching words. alone cannot reliably answer structural questions such as which pull requests a person merged last week. His document-query-engine demonstration discovers a schema, maps identities across systems and supplies the model with the information needed to produce a constrained query.
Peter Werry places validation outside the model, covering dangerous operators, hidden fields and tenant constraints. A bounded feedback loopAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior. asks the model to correct rejected queries, while full-text search avoids an unhelpful exact-title match. He closes with a local context-engineAn AI context engine gathers, reconciles and ranks relevant information so an AI agent can act with the history, decisions and permissions surrounding a task. simulator for comparing tasks with and without prepared context.
Watch on YouTube
