@derrickfromzero: The creator of Claude Code said they don't use prompts anymore… they build loops. I don't fully get it yet, but I'm spending the next week building them using OpenClaw, Hermes, and Claude Code to find out if this is actually the future. Day 1 starts now. Have you tried building loops? Let me know in the comments and follow along to learn along with me!
its also a way to totally trust that it will hand you what you asked for. so many times ive gotten lazy with debugging and built a huge chunk without checking even after having CC run internal checks. ive always been worried CC will pass it 100% just to say its green. then you go to use the app and the sign in button doesnt even work.
2026-06-17 06:46:49
0
LearningAIwithLaureen :
"Loop engineering is just a fancy name for what agentic AI has been doing for months I've been building these for my company and the shift in HOW you think about AI is real. Happy to break this down in plain English for anyone confused 👇 @LearningAIwithLaureen"
2026-06-14 19:05:20
16
It's Mister Jordan :
Hype, Claude one to two shots most things. Why would you need a loop for that.
2026-06-15 19:47:28
1
0 :
Loops now. Next up, it's switches and loops. After that, it's method calls and OOP. Ultimately it will end up in programming again.
2026-06-15 16:57:06
7
Ryan_G :
They have unlimited tokens. Dont do this.
2026-06-15 14:58:01
1
CylonJon33 :
That sounds like some BS to get you to burn through all of your credits in 1 day.
2026-06-17 04:26:21
0
J J :
Isn’t this what coding agents already do? You tell it what you want and it builds and tests until it’s complete
2026-06-17 02:38:26
0
Valentino_has_arrived 🍉 :
I’ve been using loops, but I’m just the middleman to make sure it goes right
2026-06-16 04:40:59
0
Zach Owen :
This should be called - how to blow all your funding 101. API costs are already crazy for Claude, this will cost a fortune quickly
2026-06-15 22:29:59
0
The Bucky Verse :
perfect timing! Just started with loops 😁😁😁
2026-06-14 15:52:09
0
Universal :
Look up ooda loops
2026-06-16 01:43:36
0
user9174797216694 :
Was this supposed to teach anything? Was it supposed to accomplish anything other than sharing terminology that many of us may not be familiar with?
2026-06-14 16:24:31
0
humor_me_biz :
loop engineering is a domain of harness engineering. I've been running one in Claude code for months. It takes linting and gating through code to keep it deterministic because prose instructions are sometimes ignored in the flow.
2026-06-15 13:45:09
0
P. E. :
😂 Non software dev learning software developer practices. You just need a sticky name
2026-06-17 03:54:05
0
yuvalito1 :
Peter is all hype. Just another employee at OpenAI at this point.
2026-06-17 03:47:16
0
Alvaro Ramírez :
We need fractal structures to lower token usage and scaling algorithms
2026-06-15 14:15:18
0
Thias :
to simplify - use Claude agent SDK or openai agent SDK for agent loops. skip do I prompt or loop or something else and learn what to use, why to use, when to use.
2026-06-15 10:56:16
0
Omegazilla :
Loops are legit. Use them with local AI where you are not constant token burning. Leverage cloud when the problem is very complex by giving your AI 1-3 cloud prompts a day. Keeps spend in check but leverages higher intelligence
2026-06-16 00:48:58
3
Green Tardis :
Ralph loops used to be needed to make progress on tasks to make it iterate and improve until it meets the goal. Today the agents basically do that themselves. They will try something, evaluate it, and iterate to improve things if it's not meeting the criteria.
2026-06-15 07:25:32
1
WhyteNation :
loops are great until after 100 iterations it is still trying the same 5 solutions. e. g trying to connect to a non existing server.
2026-06-14 22:38:16
0
eli the ai guy :
I realize most of my workflows aren’t deterministically able to be validated so I can’t find anything that I can loop. If you’re doing strictly dev work, it makes sense. Anything you got looping?
2026-06-15 16:43:52
2
Dr. Ron Jones | Idaho Agent :
For me, it's creating the prompt and deploying it. Then asking at the end what could have gone better...leading to a new, improved prompt...storing that prompt. Then the next time you run the (improved) prompt, you loop the pattern. That way your prompting is looping into a better version every time it runs.
2026-06-15 04:25:44
1
Neal :
Been doing this for a while… insane power and efficiency
2026-06-15 05:40:06
2
To see more videos from user @derrickfromzero, please go to the Tikwm
homepage.