The fastest-growing role in AI engineering — and what it actually means.
A loop engineer builds the systems that drive AI agents — not the individual prompts. Instead of manually prompting agents, loop engineers design loops that automate iterative decision-making. In 2026, this is the fastest-growing role in AI engineering as companies shift from hand-crafted prompts to autonomous loop systems.
"You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents."
"I don't prompt Claude anymore. I have loops that are running. They're the ones that are prompting Claude and figuring out what to do. My job is to write loops."
A loop is a small program you write that prompts the agent, reads what it produced, decides whether it is done, and if not prompts it again. You stop being the person inside the loop typing prompts — you become the author of the loop, and the model becomes a subroutine. In one line: a loop is cron plus a decision-maker in the body.
Boris Cherny's ladder describes three levels of AI-assisted work:
You type the code, the model suggests completions. The fastest way to write simple code, but you remain in the loop.
You spawn multiple agents to explore different solutions in parallel, then decide which to pursue. You still decide and direct.
You write the decision-maker; the loop and model run autonomously. You only step in when the loop gets stuck or needs refinement.
"In the last 30 days, 100% of my contributions to Claude Code were written by Claude Code — 259 PRs."
Great engineers matter more than ever — the work moved up an altitude, from writing the code to writing the thing that writes the code. Loop engineers are the architects of agent systems, not prompt typists. They design for clarity, scalability, and autonomous decision-making. The problem space is larger, and the skill ceiling is higher.
A loop without clear termination is a loop that runs forever. Great loop engineers define exit conditions upfront: success criteria, error handlers, and finite iteration counts.
Loops should validate their own output before declaring victory. This means building validators and test cases into the loop body — not relying on humans to spot errors.
Runaway loops are a hazard. Great engineers set hard limits on compute, token spend, and retry counts. A loop that respects its budget is a loop you can trust.
If you do something more than once, turn it into a skill or abstraction. Peter Steinberger's principle: reusable components scale better than one-off prompts, and they're easier to debug and maintain.
Browse open loop engineer positions or join our talent network to connect with top companies.
A loop engineer builds the systems that drive AI agents—not the individual prompts. Instead of manually prompting coding agents, loop engineers design loops: small programs that prompt the agent, read its output, decide whether it's done, and re-prompt if needed. It's fundamentally about automating the iterative work between human and AI.
Loop engineers need strong fundamentals in programming, system design, and autonomous decision-making. Key skills include designing loops with clear stopping behavior, building self-verification into loop logic, capping iterations and compute budgets, and turning repeated work into reusable abstractions. Understanding LLMs and prompt engineering is also valuable, but the focus is on the orchestration layer.
Loop engineer salaries vary based on experience, location, and company, but industry trends show roles ranging from $150k to $300k+ annually for skilled practitioners at major AI-focused companies. As the role matures and demand increases, compensation is expected to remain competitive with other senior AI engineering roles.