Loop Engineering for the Enterprise

In his 7 June 2026 essay on loop engineering, Addy Osmani discussed systems that discover work, delegate it, check results and retain state. He cited Peter Steinberger and Boris Cherny as practitioners emphasizing loop design. The essay is a dated discussion of an evolving practice; it does not establish who coined the term or when agent loops were invented.

The distinction is not cosmetic. A prompt is a single turn. A loop is a control structure; the agent acts, observes the result, decides the next step, and repeats, ideally until a goal is met rather than until the operator runs out of patience. The unit of design moves from the wording of one request to the shape of the process around it: what the agent steers toward, what signals it steers against, and when it is allowed to stop.

When an agent acts over multiple turns, the surrounding process needs a reachable goal, trustworthy feedback, permission boundaries and a stopping rule. A capable model alone does not supply those controls. A well-designed loop can reduce some risks, but it does not guarantee convergence or compensate for a model that cannot perform the task.

This paper offers a design checklist for bounded agent work: goals, feedback, failure recovery, budgets, approvals and handover. Use it to define a finite deliverable and the evidence needed for an owner to accept it.