On Friday, Boris Cherny, the creator of Claude Code, took the stage at Meta’s @Scale conference. Surprisingly, the very first audience question centered on loops. “Are loops the next hype cycle,” the attendee asked, “or are they for real?”

Cherny’s response was unequivocal: “Yes, they’re for real.” He explained, “Two years ago, we wrote source code by hand. We started to transition so agents write the code. And now we’re transitioning to the point where agents are prompting agents that then write the code. As big as the step from source code to agents was, loops are just as important and as big a step.”

Later in the talk (around the 32-minute mark in the YouTube video posted above), Cherny got specific about the loops he personally runs. One agent constantly searches for ways to improve the code architecture, while another hunts for duplicated abstractions that can be unified. They submit pull requests just like any human coder, and since the codebase is always changing, they never stop running. This is a powerful concept, especially coming from a figure as influential as Cherny.

With the shift to agentic AI, most users have focused on managing their agents effectively: setting clear goals, checking progress in discrete steps, and keeping them tightly bound to their prompts. The loop takes this further by authorizing a swarm of agents to work continuously in the background, endlessly. That’s a lot of trust to place in AI—but as models improve rapidly, it could represent the next leap in getting AI to handle real-world work.

The first thing to recognize is that this isn’t entirely new. Recursive loops—functions that call themselves to repeat an action, along with a condition that stops the loop—are a staple of introductory computer science courses. These AI loops follow a non-deterministic logic, meaning a sub-agent decides when to stop rather than a clear condition, but the same basic approach is at work. As soon as programmers began using AI to complete tasks, some version of the recursive loop, with AI overseeing AI, was inevitable.

Unlike classic computing, agentic loops can be surprisingly simple. One of the most popular tricks is the Ralph Loop (named after Ralph Wiggum), which essentially sums up all the work the model has done and asks whether it has accomplished its goal. It’s a way to handle AI models that get lost when running too long—basically bouncing the model back and forth until the task is complete.

Another way to think of loops is as part of the broader push for more test-time compute. As OpenAI researcher Noam Brown noted earlier this month, modern models can solve nearly any problem if you throw enough compute at them. That means one way to ensure a problem gets solved is to just keep throwing compute at it until it’s finished. This is especially true for hill-climbing problems like improving a codebase, where the model can keep making incremental improvements until it reaches a given threshold. Or, as in Cherny’s example, it can simply keep making incremental improvements for as long as there’s compute available.

If that sounds expensive, it should. Like agentic AI before it, AI loops burn through tokens much faster than simple Q&A chatbots—and since the point is to keep the loop running constantly, there’s no ceiling on how much you can spend. That’s fine for Anthropic, which is ultimately in the token-selling business, but for everyone else, it could be a pricey way to work. Still, depending on the problem the agentic loop is trying to solve, and with the right setup for overseeing token spend, drift, and other classic AI issues, the benefits could be staggering enough to outweigh the costs.