The AI world is getting ‘loopy’

AI agents are now being organized into continuous loops where multiple agents work together repeatedly in the background to accomplish tasks, with one agent potentially prompting another agent that then writes code. This represents a significant shift from earlier approaches where users managed discrete units of agent progress with clear stopping points. The concept relies on allocating substantial computational resources to keep agents running continuously on problems like code improvement, which can be expensive in terms of token consumption but may deliver substantial benefits if properly overseen.