Transcription
The guy who runs Claude code at Anthropic just said that he doesn't actually prompt Claude anymore and his whole job is writing loops. So, let me break down what that means and how you can set it up, too.
For the last few years, the most valuable AI skill was writing good prompts. But, here's the catch. You're stuck in endless back and forth babysitting your agent through every single step. That's why the skill everyone's shifting to now is loop engineering.
A loop is a system that prompts the agent for you. You hand it one goal and it runs by itself. It finds the work, does it, checks its own result, remembers what it finished, and keeps going until the job is actually done. And all of this gets built inside Claude code in five pieces.
Automation, so it kicks off on its own. Work trees, so two agents can work side by side without ever colliding. Skills, so it stops re-explaining your whole project every single time. Connectors, so it can actually reach out and touch your real tools. And sub-agents, so the agent writing the code isn't the same one checking and grading it.
Wire it up like this and it stops feeling like a chatbot you talk to and starts feeling like a team of employees working for you around the clock. If you comment AI, I'll send you the full step-by-step on setting this up yourself, so you can start loop engineering, too.