Long Running Agent(s)?

I came across this blog post and am interested to know if anybody has tried or implemented workflows for it: Scaling long-running autonomous coding · Cursor

I’m not knowledgeable enough to know if this blog post refers to spawning multiple agents or whether it all works off of one agent with large quantities of sub-agents.

The concept I’m thinking of is similar to this: Multi-Agent Workflows (Pull request like) - #3 by davideasaf