In this post — and throughout the series — I’d like to take you behind the scenes of Redmenta to show you how our AI-based system works and how we build something this complex.
Let’s start with a fundamental question:
How do AI assistants generate content?
What is the content actually built on?
Let’s break it down.
When you open an AI Assistant, you can set the curriculum, subject, grade level — and even upload your own learning materials. Based on this, the Assistant generates its ideas in the form of Outline.
But let’s be clear: those “ideas” aren’t quite like human ideas. What we call an “idea” here is actually a carefully orchestrated outcome of complex instruction chains, or prompts. Behind every AI assistant lies a series of prompts — these are the detailed instructions we give to the AI to tell it what to do. Think of a prompt like a recipe: it tells the AI how to get to the end result and how to produce the content you see.
When I develop a new assistant, the first step is writing the prompts that drive how it works. And writing prompts is no small task — it takes time, because we have to consider things like:
It’s important to understand: AI cannot read your mind — or even think independently.
That’s one of the most crucial things to remember when working with AI. To get high-quality results, I need to write extremely precise and complex prompts. For example, our Practice Task Assistant runs on three full pages of prompt instructions!
This also explains why the content you get from an assistant might not be spot-on the first time. That’s because the initial prompt includes multiple possible directions — it’s designed to generate the most general result first, so you can see the range of what the assistant can do. That’s also why it’s so important to interact with the AI: if the result isn’t quite what you wanted, just tell it.
Any edits or preferences you type in can override parts of the underlying prompt. So even if the base prompt includes, let’s say, a Creative Board, but all you want is a set of true-or-false questions — no problem. You can guide the AI to remove the Creative Board and focus on what matters to you. The core logic and internal connections remain intact — it’s like we’re opening up the last part of the prompt just for you, so you can steer the assistant toward the result that fits your needs.
In the next part of the series, I’ll show you how to write these messages to the AI in a way that helps you get what you want — faster. Since I spend a lot of time talking to the AI — and building the bridges between you and the system — I’ll be using this series to share all the behind-the-scenes knowledge that can help you get the most out of Redmenta AI. 🙂
