Run this playbook: iris playbook run agentic-loop Steps: plan → build → scout → growth → verify → synthesize → write-memory → schedule → summary

Agentic Loop (loop engineering)

A runnable reference for the "set the goal once, the agents prompt themselves" pattern:

GOAL → DISCOVER/PLAN → EXECUTE (Builder · Scout · Growth) → VERIFY → SHIP/ITERATE
   + MEMORY (next-steps, outside the conversation)   + weekly SCHEDULE

Each specialist below is a prompt step you can later swap for a real agent fanned out across the Hive — iris hive run <node> "iris agents chat <specialistId> '…' --bloq <mem>" — for true parallel execution. See iris how-to view agentic-loops.

All AI steps use gpt-4.1-nano (cheap, closed-loop economics). Memory persists to a local next-steps file (the video's "memory outside the conversation") and, if --bloq is given, is ingested into that knowledge base for recall next cycle.