Mastering Chatbots
Have you ever tried to create something with AI and ended up frustrated because the chatbot just wasn't giving you useful answers? You're definitely not alone. A lot of people think AI is the problem, when in reality it's often the way we communicate our ideas. In this session, we'll learn how to write better prompts, explain our ideas more clearly, and get much better results from AI tools. Whether you're using ChatGPT to study, work, or build projects, a few simple techniques can make a huge difference.
Series: Legendary Prompts (1 article)
- 1Mastering Chatbots This article
The Power of Real-World Examples
If there's one thing that helps everyone understand an idea the same way, it's a real-world example. So instead of just talking about prompts, let's look at one in practice and see the difference between a basic prompt and a well-written one.
For this demonstration, we'll be using Claude Sonnet 4.6 with low reasoning effort.
Let's be honest, everyone wants to make more money. So let's see what happens when we ask our AI how to do exactly that.
Okay, that wasn't very specific. Honestly, it feels like I could have typed the exact same thing into Google and gotten similar results.
On top of that, the response was incredibly long. In fact, I couldn't even fit the entire screenshot into this article. Just imagine how many tokens were used to answer such a simple question.
The problem isn't that the AI gave a bad answer. The problem is that we asked a very broad question, so the AI had to make a lot of assumptions about what we actually wanted.
So, what exactly did we do wrong?
First of all, not every simple prompt is a bad prompt. The problem with this one is that it was way too broad. We asked a question that could have hundreds of different answers, and the right answer depends entirely on the context.
For example, I could have asked:
"How can I make money online?"
That small change alone would have solved half of the problem. Come on, Claude. Maybe I don't own a motorcycle or a bicycle to start delivering food for Uber Eats!
The more context you provide, the fewer assumptions the AI has to make. That usually means more relevant answers, less unnecessary information, and fewer tokens being used.
So, let's move on to Example 2. We'll ask the same question, but this time we'll make it more specific, provide some context, and get a much more useful response while saving a lot of tokens in the process.
In the first prompt, "How can I make money?", the AI had to guess:
- Which country you live in.
- Your age.
- Your level of education.
- Whether you're looking for online or in-person income.
- How much money you want to make.
- Whether you own a car, motorcycle, or bicycle.
- Whether you're looking for a job, a business opportunity, or an investment.
- How much time you have available.
Because there wasn't enough context, the AI had to generate a generic response that could apply to dozens of different situations.
In the second prompt, you started providing context, but there were still some missing pieces.
In the third prompt, you did something very important: besides providing context, you controlled the output.
You specified:
- Who you are → a Computer Science student.
- Where you are → Brazil.
- When → 2026.
- What you want → ways to earn extra income online.
- How many answers you want → no more than three.
- How you want the answers organized → from easiest to most difficult.
- The desired length → a maximum two-line summary for each option.
- What you don't want → unnecessary comments.
In practice, you turned a vague question into a clear specification.
Series: Legendary Prompts (1 article)
- 1Mastering Chatbots This article
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article