ð The era where ‘people good at prompting’ were celebrated is actually coming to an end. In this installment, we will explore the quiet shift occurring in AI utilization in 2026 and the four-stage mapâ’Prompt â Context â Harness â Loop’âthat this series will cover.
Introduction
ð Erica: ‘Hello everyone. Today, we are starting a new series. The theme is “From Prompt to Loop.” For those of you thinking, “Wait, isn’t prompting still relevant?”âthat is actually the point.’
‘Being good at prompting’ was a compliment until recently. It meant someone who could give precise instructions and extract the desired answers from AI. I myself have introduced many such techniques in my English learning articles.
‘You are an accounting expert. Please answer from that perspective.’âI’m sure many of you are familiar with this type of role-assignment technique. While this is still not meaningless, the era where knowing just this makes you powerful is coming to an end.
However, entering 2026, different voices have begun to emerge from the forefront of AI development. It is the sentiment: “Let’s stop writing prompts.”
ð The Quiet Shift Happening in the Field in 2026
ð¹ The voice saying, “Let’s stop writing prompts”
OpenAI engineers are saying that instead of typing prompts directly into coding agents, we should be designing the very mechanisms that continuously send prompts to agents. An engineer working on development tools at Anthropic also stated that they no longer give direct instructions themselves; their job is to write the systems that keep issuing instructions.
In other words, a shift from “people who write good instructions” to “people who build systems that keep issuing instructions” is already beginning to happen in the development field.
ð¹ From Human in the Loop to Human on the Loop
This change is also referred to as the transition from “Human in the Loop” to “Human on the Loop.” The former is a state where humans are inside the work cycle, intervening in every single step. The latter is a state where humans stand outside the cycle, overseeing the whole process.
It is just changing “in” to “on.” But this one-word difference completely changes how we interact with AI.
ð The Evolution of Interacting with AI in 4 Stages
ð¹ Prompt â Context â Harness â Loop
Breaking down this shift further, we can organize the evolution of how we interact with AI into the following four stages.
First is “Prompt Engineering.” This is the stage of refining how you ask AI for things. Next is “Context Engineering.” This is the stage of designing not just a single request, but the entire set of information the AI looks at when making decisions. Furthermore, there is “Harness Engineering.” This is the stage of creating a scaffold that allows AI to work safely and with reproducibility. And finally, there is “Loop Engineering.” This is the stage of designing the very mechanism that automates everything up to issuing instructions to the AI, allowing work to continue autonomously.
ð¹ Each layer “encapsulates” the previous one
What is interesting is that these do not replace each other; they are nested. A prompt is part of the context, the context is part of the harness, and the harness is a component that operates within the loop. It is not that any of them have “died,” but rather that new layers are being stacked on the outside.
In this series, we will look at these four stages one by one, in order.
ð This applies directly to English learning as well
ð¹ The limitations of ‘typing good English composition prompts every day’
If you use AI for English learning, you might have some idea of what this means. Every day, you type prompts like ‘Correct this English composition’ or ‘Make this expression more natural.’ The more you refine them, the better the quality of the corrections becomes, but the effort of ‘I have to type a prompt myself again today’ never goes away.
ð¹ The goal of this series
The ultimate goal of this series is to create a small learning loop where the AI understands your weaknesses, builds practice menus, and keeps records for you, without you having to type prompts yourself every day. From the next installment, we will look at prompts, context, and harnesses in order, and finally, we will build that small loop together.
It might seem a bit forced, but this is a series where you will also understand the evolution from prompt to loop through English.
Conclusion
ð Erica: ‘Next time, we will go back to basics and look at prompt engineering. From here on, I’ll leave a few Japanese x English prompt sets that you can use starting today. There’s also a private, cute, and cunning bromide image of Erica cheering you on (vertical), so take a look. It’s a size you can even use as your smartphone wallpaper.’
From here on is a set of ‘Japanese prompts x English prompts + private Erica images’ that you can use as you read through the series. As a way to practice English, please take a look at both prompts together.
â¼ Paid content starts here (200 yen) â¼
