A news summary based solely on the provided article body: Loops, where an agent autonomously takes actions until it hits a verifiable goal, are becoming popular because the AI now has "eyes," "hands," "memory," and a "brain." Multimodal models and stronger computer use let the loop inspect its own output through screenshots and UI states instead of guessing through text-only proxies. Bash, MCPs, CLIs, plugins, and harnesses gave the agent reach to move through systems like GitHub, Notion, Slack, and terminals. Context windows got big enough that memory became a given, and reasoning models got better with more data on agentic computer use. The article notes that early loops like AutoGPT in March 2023, which decomposed a goal into subtasks and looped until done, became a cautionary tale about autonomy without verification, as they wandered, hallucinated, repeated themselves, and set bills on fire. Ralph loops revived the idea with more discipline by bringing tests, builds, and external checks into the loop. The article emphasizes that the heart of building a good loop is verification, with the new frontier being verification for work that used to require a human pair of eyes.