LUDA Research Group

Is It Really Necessary to Instruct AI Every Single Time?

Are you still typing things like 'summarize this document' or 'write this email' into ChatGPT or Claude every time you need something done? That's a bit like telling a head chef exactly what to do for every single dish. It works — but you're still the one who has to keep giving orders.

In June 2026, a single post by developer Peter Steinberger racked up 2.3 million views and sent shockwaves through the global tech industry. The message was simple: 'Stop telling AI agents what to do. Instead, design loops that make AI work on its own.'

That was the spark that ignited Loop Engineering.

What Is Loop Engineering? Think of It Like a Factory Assembly Line

In one sentence, loop engineering is: the practice of designing automated systems where AI judges, acts, checks its own results, and repeats — without needing constant human instruction.

Think of a car factory's production line. Once it's properly set up, the machines automatically assemble parts, filter out defects, and pass everything on to the next stage — all on their own. Loop engineering is about building that kind of automated pipeline, but with AI.

The person who gave this concept its official name is Addy Osmani, a Google Cloud engineer. He mapped out the evolution of how we use AI like this:

Prompt Engineering: The art of asking good questions (what you send to the AI)
Context Engineering: Structuring the information the AI can see and work with
Loop Engineering: Designing systems where AI repeatedly acts, observes, decides, and loops back

To put it simply: if prompt engineering is 'knowing how to explain a recipe well to a chef,' then loop engineering is 'designing the entire kitchen system so the chef automatically plans the menu, orders ingredients, cooks the food, and checks the taste — all by itself.'

How Does It Actually Work? A Real-World Case: A 2 AM Payment System Crash

The mechanics of loop engineering come down to a four-step cycle: Act → Observe → Decide → Repeat

Here's a real-world scenario to make it concrete. At 2 in the morning, an online store's payment system goes down.

The traditional approach: The on-call engineer gets a notification in the middle of the night, drags themselves awake, digs through logs, figures out the cause, and manually fixes the issue.

With a loop engineering system:
1. The AI automatically detects the error (the engineer is still asleep)
2. The AI analyzes the root cause and writes a fix
3. A separate 'verification AI' independently checks whether the fix is correct
4. If everything looks good, the fix is automatically deployed
5. The engineer wakes up in the morning to a simple notification: 'Error at 2 AM automatically resolved'

This isn't theoretical. PostHog, a software analytics company, implemented this kind of system and — without any human involvement — automatically uncovered a three-year-old bug and improved performance by 11%.

How Are New Zealand and South Korea Responding to This Shift?

What's Happening in New Zealand

New Zealand's government made its position clear in July 2025, publishing an AI strategy titled 'Investing with Confidence' and officially declaring that adopting AI is no longer optional — it's essential for staying competitive. Engineering New Zealand has also been running its 'Engineering and AI' programme to actively help practitioners integrate AI into their work.

One event worth watching: on October 21–22, 2026, the Agentic Engineering Day conference will be held at the Auckland Aotea Centre. It's New Zealand's first dedicated conference focused on AI agent systems — including loop engineering — and will feature hands-on workshops for building AI automation systems. Tickets start at NZ$190.

The Buzz in South Korea

Interest in loop engineering is spreading fast across South Korea's IT community. Posts on major communities like PyTorchKR and GPTers, as well as specialist platforms like Elancer and TechTalk, have been multiplying rapidly. Among Korean developers, loop engineering is increasingly being called the most important AI trend of the second half of 2026, and the idea that 'how you design your operations matters more than how you craft your prompts' is quickly becoming conventional wisdom.

The global numbers put this momentum in perspective. The AI agent market is projected to grow from roughly $7.6 billion in 2025 to around $183 billion by 2033. And by the end of 2026, an estimated 40% of enterprise applications are expected to incorporate AI agents — a dramatic leap from less than 5% in 2025.

Key Takeaways

Loop engineering is the practice of designing automated systems where AI handles repetitive tasks on its own — rather than waiting for human instructions each time
The core principle is a four-step automated cycle: Act → Observe → Decide → Repeat
The way we use AI is evolving from prompt engineering (asking good questions) to loop engineering (designing automated systems)
New Zealand is actively embracing this shift through its national AI strategy and specialized events like Agentic Engineering Day 2026
South Korea's IT community is rapidly expanding its discussion of loop engineering, with practical know-how being shared at an accelerating pace

Wrapping Up