Summary
Boston Tech Week meetup on agentic AI in production: retrieval, drift, and what actually works (talks from Chainguard and Elastic).
Why it matters
Most agent demos fail in the real world because context goes stale and retrieval quality is weak—this is where reliability and costs blow up.
Tokenmaxxing angle
Treat retrieval + drift detection as your cost controls: better context means fewer “try again” loops, fewer hallucinations, and less wasted token spend.
Source takeaway
Agenda includes a talk on DriftlessAF (an open-source agentic reconciliation framework) and another on why the best-retrieving agent wins.