ai agents / large language models
What Is An AI Agent? (ChatGPT Doesn't Count)
Course walkthrough
AI Agents from Scratch
All lessons in this series
- 1I'll Teach You How To Build AI Agents From Scratch (10 Billion Tokens Proof)
- 2What Is An AI Agent? (ChatGPT Doesn't Count)
- 3The ONE Trick That Turns ChatGPT Into An AI Agent (Tools)
- 4The ReAct Loop: How Cursor & Claude Code Actually Work
- 5The System Prompt That Changes Everything (Prompt Engineering For AI Agents)
- 6How To Make Your AI Agent Truly Autonomous (Without It Going Rogue)
- 7How To Trigger An AI Agent (5 Different Ways)
- 8Subagents Explained: How AI Agents Launch Other AI Agents
- 9Simple Memory: How AI Agents Remember Between Runs
- 10How AI Agents Actually Remember (Memory + RAG Explained)
- 11How To Actually Tell If Your AI Agent Is Good (Evaluation)
- 12The Complete Guide To Building AI Agents (10 Lesson Recap)
What's the actual difference between ChatGPT and a real AI agent? Most people use these terms interchangeably — but they're not the same thing, and understanding the distinction is the foundation for everything else in this course.
In this lesson, we go under the hood: → ChatGPT is an APP — not a language model itself → The raw language model lives in the OpenAI Playground (and it's surprisingly limited) → System prompts: how you actually give an AI context → Why language models are "stateless" and what that really means → The first hint at what makes a real AI agent: TOOLS
By the end, you'll understand the exact distinction every AI engineer at big tech companies takes for granted — and why ChatGPT is just the beginning.
This is Lesson 2 of "How To Build AI Agents From Scratch" — a free 10-lesson masterclass on building production-ready AI agents.
#AIAgents #LLM #ChatGPT #SystemPrompts #AIEngineering
Transcript
0:05In this first course for introduction to AI agents I'm going to start with the first agent that we all know and love and then we'll learn how it's really not an actual AI agent
0:18I'm going to start on ChatGPT So ChatGPT is an application that has a large language model You put in text and then a ChatGPT responds
0:31appropriately in the back and forth conversation Most people are familiar with ChatGPT Now when you're building AI applications think about
0:44ChatGPT as an app The actual AI model is not fully visible to the user If you want to see the actual AI model you're going to
0:56go to the OpenAI Playground Now this is basically an advanced version of ChatGPT but it's still the same thing So for example right here I'm just
1:09going to have a regular developer message I'm not going to put anything here I'm going to just say hey and then I'm going to click submit and we can see as I click submit
1:21here it's going to respond similar to ChatGPT although it's going to be less personalized You know that's because ChatGPT is an app and a lot of things under the hood while the OpenAI Playground
1:35is a language model in its purest form Now with these language models you'll notice several things which are really important So for one
1:49we notice this developer message which traditionally is called the system prompt Two we notice the user messages and three we notice the assistant messages And
2:03so this back and forth is how all chat apps are structured I can say you know a response how are you The chat is going well thanks How can I help
2:17All that makes sense Now let's make it something a little bit different Let me go ahead and erase these messages and I'm going to update the system prompt I'm going to say only respond
2:32with hi and if you see this if I type a message to the AI and have this system prompt then the AI is always going to respond with hi
2:48How are you Hi Why are you saying hi It's doing this because that is what the system prompt tells the AI It tells the AI
3:01what instructions it should follow for this conversation Now let's return back to this system prompt you're a helpful assistant and I'm
3:15going to ask this question what is my name You have to remember that the language model is purely stateless It doesn't know anything about you any real
3:27time events anything recent And all it knows is the historical information that it was trained on and any information that you give it So if I were to repeat this question what is my
3:41name But I instead say here note my name is Austin Then when I ask what is my name it can actually respond with the
3:55correct answer because I gave it all of the necessary information So you can think about how you can expand it Maybe if you Googled
4:07a whole bunch of news and then you dump this news into chat GPT then the AI will know what happened in the news I can say what
4:20happened in the news today Without this system prompt it won't know the answer because you're not giving it any information It doesn't know anything about real time news
4:35real time information But if I were to dump this all this news into the chat and then I re ran this question then it will know
4:48what happened in the news because I told it The system prompt is the way to give the language model information
5:02The key difference between this the OpenAI Playground and ChatGPT is ChatGPT is somewhat agentic It
5:14has access to real world information and it can actually do some things How It's with tools If I ask what happened in the news today
5:28if I ask it in ChatGPT it's going to execute tools like searching the web and then it can respond back with the information That is the key
5:41difference between a large language model and an AI agent Agentic tools like Cursor and Claude
5:55Code take this one step further While ChatGPT has tools it will still just do a back and forth conversation With AI tools like Cursor and Claude
6:08Code it will keep executing tools in a loop writing code searching the web looking at external resources It's going to do this continuously in a loop until
6:21it either finishes the task or it needs your help your input or has a question That's what really makes an AI agent an AI agent The ability
6:34to act on tools and the ability to do things autonomously in a loop in order to arrive at a final answer
Join the conversation
Loading conversation…