Your First Conversation

Every time you click "New Chat," you're starting a fresh conversation with a blank slate. ChatGPT has no memory of previous chats (unless you've turned on the Memory feature, which we'll cover on Day 5). Here's what's happening behind the scenes:

The context window — ChatGPT can "see" everything in your current conversation up to a certain length. Think of it as the AI's working memory. On GPT-5.2, this window is enormous — roughly 128,000 tokens, which is about 300 pages of text. That means it can hold a long, detailed conversation without forgetting what you discussed at the beginning.

Message threading — Each message you send builds on everything before it. If you ask a question, get an answer, then ask a follow-up, ChatGPT reads the entire history to understand your follow-up in context. This is what makes it feel like a real conversation.

Tokens, not words — ChatGPT processes language in "tokens," which are roughly 3/4 of a word. The word "hamburger" is two tokens ("ham" + "burger"). You don't need to think about this much, but it explains why there are limits — every word you write and every word ChatGPT writes uses tokens from the context window.

Key insight: Because ChatGPT reads your entire conversation history, the quality of your early messages shapes every response that follows. A well-started conversation stays useful for a long time.

Let's start with the most important lesson in this entire course: the quality of your output depends on the quality of your input.

Here's a weak first prompt and what it gets you:

Understanding a little about how ChatGPT generates responses will make you a much better user.

ChatGPT doesn't "know" things the way you do. It doesn't have a database of facts it looks up. Instead, it predicts the most likely next word based on patterns it learned during training on billions of documents. It does this one word at a time, thousands of times per response.

This has important implications:

It can be confidently wrong. Because it's predicting likely words rather than retrieving verified facts, it can produce plausible-sounding text that contains errors. This is sometimes called "hallucination." Always verify critical facts.

It's remarkably good at patterns and structure. Need a business email? It has seen millions of them. Need code in Python? It has been trained on vast quantities of it. The more common the pattern, the better it performs.

Context dramatically improves accuracy. When you give ChatGPT more context — who you are, what you need, what format you want — you narrow down the prediction space. Instead of guessing from millions of possibilities, it's choosing from a much smaller, more relevant set.

It gets better as the conversation goes on. Every message you send gives it more context. A vague first message followed by clarifying follow-ups will still lead to good results — just less efficiently than starting with a clear prompt.

One of the biggest mistakes beginners make is starting a new chat every time they're unhappy with a response. Instead, iterate within the same conversation. ChatGPT is designed for back-and-forth refinement.

Here are phrases that work well for steering ChatGPT:

- "Make it shorter" or "Make it more detailed"

- "Use a more casual tone"

- "Focus more on [specific aspect]"

- "That's good but change the second paragraph to emphasize..."

- "Give me 3 alternative versions"

- "Rewrite this for an audience of [specific group]"

Think of it like working with a human collaborator. You wouldn't fire someone after their first draft — you'd give feedback and iterate. ChatGPT works the same way, except it never gets offended and revises instantly.