Title here
Summary here
Course
Loop Engineering: Getting AI to Finish the Job
Go past one-shot prompting — learn to run an AI in a loop that attempts, checks its own work, corrects, and stops. Beginner-friendly, no coding required.
At a glance
Level
Beginner · no coding required
Lessons
20 lessons across 4 modules
What you build
A complete, documented loop for real work
Cost
Free course · works with the tools you already pay for
What you'll learn to do
You'll turn everyday AI fumbling into a repeatable method built around one idea: a loop is only as good as its check. You'll write goals a loop can actually finish, build checks that catch a false "done", run loops by hand in ChatGPT and Claude, hand loops off to file agents like Claude Code and Codex (for non-developers too), design loops that stop safely instead of running up a bill, and keep a vocabulary of loop patterns so you pick instead of improvise. The running example throughout is Harborlight, a small community bookstore, and every project is a task a real person actually has. No coding is required to complete the course.
Course syllabus
Work through the modules at your own pace. Each lesson is a self-contained, hands-on read.
1 Why One Prompt Isn't Enough
2 Writing Goals a Loop Can Finish
3 The Check: How a Loop Knows It's Working
Before you start
All you need is regular access to a chat AI — ChatGPT or Claude — and everyday familiarity with using one. You do not need to code, and you do not need a technical background. The later modules use two free command-line tools, Claude Code and Codex, and the course walks you through installing and using them safely; if you'd rather stay entirely in a browser, Modules 1–4 stand on their own. When you're ready to see the same loops written as real Python, our Building AI Agents in Python course is the natural next step, and LLM Evaluation & Observability turns Module 3's "check" into something you can measure.
What you'll need
You can do this whole course with tools you likely already have. Here's the short list:
- A chat AI account — a free or paid ChatGPT or Claude account is enough for Modules 1–4.
- Optional, for Modules 5–7: Claude Code and/or Codex — free command-line agents you point at a folder. The course shows you how to set them up with safe permissions.
- The practice files we provide — a deliberately messy Harborlight customer-feedback export and a jumbled shared-folder fixture — so you can run the data and file loops for real.
AI tools change fast. If a menu, button, or command flag has moved since a lesson was written, the loop ideas still hold — adjust the click or the command to match what you see. Every tool-specific step is marked so it's easy to update.
Ready to make AI finish the job?
Start with Module 1 and discover you've been running loops by hand all along — then learn to run them on purpose.
Start the first lessonWant this taught live to your team?
Mehdi runs tailored corporate workshops on this exact material — hands-on, in-person or remote.
