Architecture diagram + dev workflow

Matt Dancho posted a detailed production‑grade agent architecture diagram covering planning, execution, memory, tools and evaluation loops — a useful blueprint for scaling reliable agents shared. Damián demonstrated an experimental Vertex AI agent dev workflow that included adversarial review, security checks, GKE deployment and observability‑based rollback, highlighting concrete failure modes you’ll hit in production demonstrated.