RAG + LLM: A Superpower Combination→
How Retrieval Augmented Generation turns an LLM from a storyteller into a grounded fact-checker.
Before ChatGPT, you opened 10 tabs. Now you just ask. But what's actually happening underneath?
Before we begin: There's already a lot of content about AI. This is my attempt to make it clearer and simpler. Better late than never.
Remember how you researched your final-year project during college?
November 30, 2022, changed all of this.
ChatGPT was launched to super-simplify all of that. Now, ChatGPT, Gemini, Claude, and Perplexity are fundamentally the same. Then why do so many tools exist? Simply because big firms chose to compete rather than collaborate.
You can swap the word "ChatGPT" in this post with any of those tools and it will mean the same thing. What's important is the underlying concept.
Now, let's dive in.
It is just an application. But what it uses under the hood is what matters — it uses a Large Language Model (LLM), for example, GPT-5.
One important distinction right away:
ChatGPT is not the LLM itself. It's the interface that sits on top of the LLM, connecting you to it.
We'll talk about LLMs in depth in the next article. But first, let's understand how it all works.
When you type something into ChatGPT — what you type is called a prompt — it routes your request to the LLM plugged into it.
The LLM then does three things:
Fundamentally, it's as simple as that.
Sometimes it predicts incorrectly — and produces wrong or made-up information. That's called hallucination.
If you closely observe, the response never appears all at once. It streams out word by word.
This is not a CSS typewriter effect. It's just how ChatGPT works — it's predicting and outputting one token at a time, live.
Just because content — text, images, audio — is being generated by AI, it's called Generative AI.
It's that simple.
We'll break down Large Language Models in the next article, including why ChatGPT doesn't use Small Language Models.
Actually — does the term "Small Language Model" even exist, or am I just making it up? Think about this 😉
ChatGPT crossed 1 million users in just 5 days — more than the entire population of Iceland.
No product in history had reached that kind of adoption that fast. That's how badly people were waiting for something like this.
Let me know your thoughts, questions, or suggestions. I'm eager to learn and happy to help!