Code Quality, Context Engines, and Tribal Knowledge

Making AI agents work in your team πŸ€– β€” with Dennis Pilarinos

Topic 3 of 6

Published Sep 15, 2026 Β· Clip 17:34–33:41 (16:07)

Denis argues code quality is a consequence of the context the person or agent has β€” a newcomer without organizational best practices regresses quality, while a veteran's tribal knowledge yields better, faster work. He explains agents shouldn't require humans to change behavior; they should derive context from existing artifacts (Slack, docs, bug trackers). He walks through the content curation trap: stale claude.md files, then MCP servers, then the realization that information must be pre-computed/reconciled before reaching the agent. He covers which sources matter (docs, messaging, source code as ground truth, and increasingly runtime systems like Sentry, Datadog, Snowflake) and how Unblocked's context layer updates near-real-time as systems of record change.