99% of people are still prompting Claude, but the 1% are actually designing loops.

99% of people are still prompting Claude, but the 1% are actually designing loops. But what the heck is a loop? Well, let me show you. The old way to do things is you prompt Claude, Claude works, it gives you an output, you check that output, and then you re-prompt Claude with some revisions. And that process goes on until you get a good output, which can take a lot of time. Now, the new way is to design what's called a loop. So really what you're doing is you're defining a goal, and then Claude then does that work, and when it thinks it's done, it checks to see if it's met the goal. And if it isn't met, then it continues to do the work. So it does quality assurance, it launches its work on a browser, it debugs, all with a series of sub-agents working collaboratively until the goal is accomplished. Now, there's a lot of different ways to run loops, but let's go through the four ways that the Claude Code team actually recommends you do it. So the first one is called turn-based, which is basically an upgrade to the old way of you prompting your Claude. But the difference here is that you're writing a verification skill. In other words, you're making your quality control checklist into a skill.md folder. And Claude then uses that skill file to check its own work so that you don't have to do that step. And second is slash goal, which is where you give Claude a stop condition and it's not going to stop until that condition is met. And this is perfect if you know what finished work looks like. So if you have tasks that Claude has already helped you to do perfectly, then document that process and give Claude the reins following that same system. But don't do this with tasks that have undefined goals or lack the necessary integrations because it'll eat your tokens extremely fast. Then you've got time loops, aka scheduled loops. And you can trigger these with a slash loop command which basically reruns on a timer. So I could tell it, hey, Claude, do a quick scrape on my niche and put some trending topics in my content ideas board. And it'll do this by itself as often as I'd like. And loops like these can be done both locally and also on the cloud. And then the final one is proactive loops, which actually combines all of the previous three. It's basically turn-based, goal-based, and time-based. But the key difference is that there's no start button. And creating one of these loops is basically like hiring a new employee with a narrow job description. And it just continues to do that narrow task for less than a penny a day in most cases. So in case it wasn't obvious, prompt engineering is no longer the highest leverage skill. Now, it's loop designing. Why? Because each new loop is another AI soldier in your army. And each one that you make adds more leverage for your work. And for a highly detailed guide on how to set up your first loop in your AI army, comment LOOP and shoot me a follow and I'll send it your way.