LLM Illusion of Memory

🧠 The Illusion of Memory

part3.ipynb— Continuing in level1 folder*

Here's one of the most mind-bending truths about AI chat systems:

💡 LLMs don't actually remember anything — they just make it feel like they do.

⚡ Every LLM Call Starts Fresh

Each API request is completely stateless — meaning the model wakes up with zero memory of:

• 💬 Your previous messages

• 🪪 Your name

• 📜 Any earlier conversations

Unless you manually send the history along with every new request, it's as if the conversation never happened.

✅ Lets gets the hands on Experience

Env Variable Import

▶️ Create a new code cell and run it:

import os
from dotenv import load_dotenv

load_dotenv(override=True)
api_key = os.getenv('OPENAI_API_KEY')

if not api_key:
    print("No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!")
elif not api_key.startswith("sk-proj-"):
    print("An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook")
else:
    print("API key found and looks good so far!")

Creating new OpenAI Instance

▶️ Create a new code cell and run it:

from openai import OpenAI

openai = OpenAI()

Setting Message dicts for OpenAI

▶️ Create a new code cell and run it:

messages = [
    {"role": "system", "content": "You are a helpful assistant"},
    {"role": "user", "content": "Hi! I'm Bhaskar!"}
    ]

Making the API call and Printing the Response

▶️ Create a new code cell and run it:

response = openai.chat.completions.create(model="gpt-4.1-mini", messages=messages)
response.choices[0].message.content

# Example Response
# 'Hello Bhaskar! How can I assist you today?'

Now, Lets ask a follow up question

▶️ Create a new code cell and run it:

messages = [
    {"role": "system", "content": "You are a helpful assistant"},
    {"role": "user", "content": "What's my name?"}
    ]

Again Making API call and printing response

▶️ Create a new code cell and run it:

response = openai.chat.completions.create(model="gpt-4.1-mini", messages=messages)
response.choices[0].message.content

# Example Response
# 'I’m sorry, but I don’t know your name. Could you please tell me?'2

😲 Did You Notice That?

In the first call you told GPT your name — yet in the second call, it had absolutely no idea who you were! 🤯

🔄 Here's Why

Every single call to an LLM is completely stateless — it's a brand new conversation every time, with zero recollection of what came before.

🧑‍💻 This is where we come in!
As AI engineers, it's our job to build techniques that simulate memory — crafting the illusion that the LLM actually remembers. 🎭