great question Ozan D., so the way it works is that the Brain becomes the true source of knowledge, when a task is dispatched, Jarvis the chief orchestrator collects all the context needed and then compartmentalizes it while spinning up the sub-agents to accomplish the task. The work is then cross-referenced at a QA subagent level, then at the Jarvis Level. Lastly, it's then put in a human approval status for the last Human approval to offload the risk which is by now highly unlikely. Last but not least, every approval or denial by the human reinforces the loop