Assistant Words: Agent, RAG, Fine-Tuning, and How to Decode Any New Word
AI keeps changing, and new words appear every few months: agent, MCP, RAG, embeddings, vibe coding. This last lesson explains the words you will meet when people talk about smarter AI assistants. Then it gives you a simple method to understand any new AI word on your own, so you never feel lost again.
What You'll Learn
- The difference between a chatbot, an assistant and an agent
- What tools, memory and automation mean in AI apps
- What RAG, embeddings and fine-tuning are, in one line each
- What MCP and vibe coding mean
- A three-step method and a prompt to decode any new AI word
Chatbot, assistant and agent
These three words are often mixed up. A simple way to tell them apart:
- Chatbot: answers your messages. You ask, it replies.
- Assistant: a chatbot with extra abilities, like reading your files, searching the web or remembering your preferences.
- Agent: an AI that can take actions in steps to reach a goal, like searching several websites, filling a form or editing files, checking its own work as it goes.
Example: "Write me a travel plan" is a chatbot task. "Find three hotels under my budget, compare them and draft an email to book the best one" is an agent task.
Learn more: How AI Agents Work or What Are AI Agents?
Tools
Meaning: in AI, a tool is an ability the AI can use, like web search, a calculator, running code or sending an email. An agent decides when to use which tool.
Memory
Meaning: an AI feature that saves facts about you between chats, like your name, your course or how you like answers written. Without memory, each new chat starts from zero.
Tip: most apps let you see and delete what is stored in memory in their settings.
Automation and workflow
- Workflow: a set of steps done in order to finish a job. "When a form is submitted, save the answers, then send a thank-you email."
- Automation: making a computer run that workflow for you, without clicking each step.
You'll see it: in tools like Zapier, Make and n8n. AI Automation Fundamentals is a good start.
Embeddings
Meaning: a way to turn text into a list of numbers that captures its meaning, so a computer can find texts with similar meanings even if they use different words.
Example: a search for "cheap laptop for students" also finds "budget notebook for college" because their embeddings are close.
You'll see it: with the word semantic, which just means "about meaning". Semantic search is search by meaning, not exact words. Embeddings Explained goes deeper.
RAG
Meaning: "Retrieval-Augmented Generation". The AI first looks up relevant information in your documents or a database, then writes its answer using what it found.
Example: a college chatbot that answers questions from the college's own rule book, instead of guessing.
Why it matters: RAG helps reduce hallucinations because answers are based on real sources. What Is RAG?
Fine-tuning
Meaning: training an existing model a bit more on your own examples, so it gets better at one specific job or style.
Compared with prompting: prompting changes what you ask; fine-tuning changes the model itself. Most people never need fine-tuning. Good prompts, custom instructions or RAG usually do the job. RAG vs Fine-Tuning vs Prompt Engineering compares them.
Custom instructions, GPTs and Gems
Meaning: saved instructions that shape how an AI answers you every time, like "always answer in simple English with examples."
- In ChatGPT, a custom version with its own instructions and files is called a GPT.
- In Gemini, a similar feature is called a Gem.
- In Claude, Projects keep instructions and files together.
Learn more: Build a Custom GPT
MCP
Meaning: "Model Context Protocol". A shared standard that lets AI assistants connect to other apps and data, like a universal plug. One MCP connector can work with many different AI apps.
Learn more: What Is MCP?
Vibe coding
Meaning: building an app or website by describing what you want to an AI in plain language and letting it write the code, often without reading the code yourself.
Be careful: it is great for quick prototypes, but for anything important, someone needs to check the code. What Is Vibe Coding?
How to decode any new AI word
New words will keep coming. Use this three-step method:
- Find the plain meaning. Ask: "What does this word mean in one sentence, for a beginner?"
- Get an everyday example. Ask for a comparison to normal life, like the waiter for an API.
- Ask "do I need this?" Many words matter to developers but not to everyday users. Knowing that saves time.
Here is a prompt you can reuse in any free chatbot. Change the word and the language at the top:
You can also change "My language" to Hindi, Tamil, Spanish or any language you think in. Understanding a word in your own language first, then in English, often makes it stick.
Inside FreeAcademy lessons, you can ask Byte, the lesson tutor, the same kind of question: "What does tokens mean here? Explain simply." Byte answers based on the lesson you are reading.
Where to go next
Now that the words make sense, these courses will feel much easier:
- AI Essentials: a full beginner's tour of AI
- ChatGPT for Complete Beginners: hands-on, step by step
- Prompt Engineering Micro Course: use the prompt words from lesson 1
- How LLMs Work: the how-AI-works words in more depth
Key Takeaways
- A chatbot replies, an assistant has extra abilities, and an agent takes actions in steps toward a goal.
- Embeddings capture meaning as numbers, and RAG looks up real sources before answering.
- Fine-tuning changes the model; prompting changes what you ask, and prompting is usually enough.
- MCP connects AI assistants to other apps, and vibe coding builds apps from plain-language requests.
- To decode any new AI word: get the plain meaning, an everyday example, and whether you need it.
Quiz
Questions & Answers
Public questions about this lesson. The AI Tutor answers right away in your language, the FreeAcademy team checks answers, and every learner can read them.