The AI world is getting "loopy" — swarms of agents running continuously in the background

TechCrunch reports that the next step beyond agentic AI is the "loop": authorizing a swarm of agents to work continuously in the background, endlessly, against long-running coding and operations tasks. The shift turns coding agents from on-demand tools into persistent teammates t

The pattern is already visible in GitHub's Copilot cloud agent, OpenAI's Codex sandbox acquisition (Ona, for persistent execution environments), and Anthropic's Fable 5 / Mythos 5 long-context agents. The "loop" framing reframes software development as a continuous async activity rather than a sequence of human-driven sessions — and it's pushing the cost conversation toward token rationing (see entry #2) and AI-credit billing.