Welcome to the AI Series from Internet Farmer! In this series, we’ll attempt to simplify AI concepts and give you the knowledge you need to work with AI in your business and life.
Over the past few months of diving deep into AI, I’ve been carrying around this mental image that I think simplifies what’s really happening under the hood of AI.
So here it is, the Brain of AI…
So let’s walk through these pieces.
LLMs (Large Language Model)
This is what gives AI the ability to reason, respond, and interact in natural language. When you hear names like ChatGPT, Claude, Deepseek, Mistral, or LLaMA, those are all different LLMs.
Each one has its own strengths: some are trained to code, some are better at images or video, and others are more conversational or specialized in a niche or just plain smarter.
Think of these like different engines under the hood, that you can swap out any time or even use them together. And each of them has their pros and cons for any individual.
RAG (Retrieval Augmented Generation)
That’s a fancy word that means ‘i look at these documents to provide answers’. Like memory on your computer the AI pulls the information in your files. Like spreadsheets, SOPs, HR manuals, legal documents, or emails so you can talk to all of it.
The key word here for me is ‘unorganized data’. Unorganized data is when you have files all over the place, spreadsheets, Word docs, email files, etc. As opposed to data being perfectly streamlined in a database, which is organized.
No more need for databases and everything being in perfect format. You put it all in a folder and start asking questions. The end.
Start gathering your docs—intelligence nirvana awaits.
System Prompt/Instruction
This is like the AI’s mission statement. You give it personality, tone, and direction here.
Example: “You are a helpful customer support agent who responds quickly, uses light humor, and sticks to brief, actionable answers.”
You can also shape what it knows, what it ignores, how it behaves, and example Q&A that helps with discussion. This is your AI’s starting context and can act like a “pre-memory” before you plug in documents or retrain the LLM.
You would also put security here. Protecting files, how your AI works, and what to allow or disallow in conversation.
Tools
We can make fire, AI can speak and hear and crawl the web and make nice images.
These are extra functions you can plug in to AI. Like plugins for WordPress or browser extensions.
Tools like:
- Whisper for speech recognition
- DALL·E for generating images
- Search to browse the web
- Speech to talk out loud and to listen to you
When you combine them with everything else—LLMs, RAG, prompts—you get real workflow power for you and your employees.
For instance…
“Hey AI, write my next blog post.”
AI drafts it, formats it, posts it across platforms, creates a podcast, turns it into a YouTube video, and emails your team to let them know it’s live.
It’s like having a team of interns you can talk to. Fast ones.
This is ultimately the power we are all looking for. The new stone tools to live our best lives.
Speech
If you want your AI to talk, or listen like your Alexa or Siri, you’ll need to plug in tools like Whisper for voice recognition and speaking. To make it speak from text, we’ll need a LLM that allows for text-to-speech for replies.
Wrap Up
This is the core I keep returning to: AI feels like magic but it’s a set of small, powerful pieces that plug into each other. LLMs to think and talk with you. RAG to remember. Prompts to guide. Tools to act. And Speech to talk.
Once you see it this way, it becomes less mysterious—and more like building with blocks.
And the best part? You don’t need to be an engineer to start. Just someone curious enough to try.
If you’re ready to explore what AI can do for your work or business, Let’s build something incredible together.