Claude God Tip #10: The Mind-Reading Feature You're Probably Not Using
Published: November 28, 2025 • 5 min read
Welcome to another episode of Claude God Tips! Today, we'll look at a small but powerful tip.
This feature is one of those where if you've used AI a lot, you just conclude that surely everyone knows what "Extended Thinking" is. So I was really hesitant about creating a blog post about it. However, let me tell you a story...
Once upon a time, I was having a conversation with a business major and I used the word "array." She asked what that meant. In that moment, I was shocked because I was convinced that word was a part of everyone's vocabulary, but I was wrong. I realized that I overestimated the popularity of that word, and I don't want to make the same mistake with the Extended Thinking feature.
What Is Extended Thinking?
This feature allows you to see what the LLM is thinking and how it's processing your prompt. This gives you the freedom to stop the LLM from completing its response when you see that it's going down the wrong path.
How to Turn It On in Claude
In Claude, you can turn on this feature by clicking on the clock icon on the left side of the input box area.
If you don't see this icon, you should see a plus icon and another icon with two horizontal lines labeled "Search and Tools" when you hover over it. Clicking on this Search and Tools icon will open up a menu where you can toggle on the "Extended Thinking" feature.
How to Turn It On in ChatGPT
In ChatGPT, which is another really popular LLM, this feature is simply called Thinking. You can turn it on by clicking on the plus icon on the left side of the input box and selecting "Thinking."
What You'll See
When this feature is turned on, after you send a message, you should see a tab which, if you click on, shows you the model's thought process behind the response it is about to provide. As you watch its thought process, you can predict whether or not the response you're about to receive will actually solve your problem.
The Teacher Analogy
Imagine being a teacher standing in front of a classroom of students. You ask a question and a student raises their hand. But now you're not an ordinary teacher. You have a superpower that allows you to read minds.
This superpower allows you to see the thought process behind how that student with their hand raised plans to respond to your question.
This also means you can tell if the student who wants to respond really understood your question. If you see that they didn't, instead of calling on them to respond, you simply paraphrase the question to make it clearer by:
- Expanding on it
- Adding examples
- Being more specific about constraints
Then you ask the question again. Once a student raises their hand, you can read their mind to see their thought process. If the thought process is valid, you can call on them now to provide their response.
This is exactly how the Extended Thinking feature works.
Why This Matters for Building AI Intuition
This is by far one of the most effective methods for building intuition when it comes to working with AI. You get to see:
- When the AI fully understands your question even when you're being oversimplistic
- When you really need to expand on the prompt to get the response you want
When you do this time and time again, you get really good at knowing the best ways to structure your prompts.
I should also mention that there have been a number of times where the response I needed was found in the thinking process itself, not exactly in the final output. So yeah, this is a really powerful feature, and if you're not already using it, you should be!
Hey, Future Prisca Here! (December 8, 2025)
The section below contains inaccuracies. Verbose Output in Claude Code is NOT the same as Extended Thinking. They are different features. Read this follow-up post for the corrected explanation.
Extended Thinking in Claude Code (Terminal)
Now you may be wondering: Is there an equivalent of Extended Thinking in Claude Code in the terminal?
Yes, there is! This is actually where it gets interesting.
One of the slash commands you have access to is /config. You might see it show up as /config (theme). If you select this command, it opens up the config modal with 3 tabs at the top: Status, Config, and Usage.
The Config tab is selected by default, and you'll see a list of options. This is what it looks like:
Settings: Status Config Usage (tab to cycle)
Configure Claude Code preferences
Auto-compact true
Show tips true
Rewind code (checkpoints) true
❯ Verbose output false
Default permission mode Default
Respect .gitignore in file picker true
Theme Dark mode
Notifications Auto
Output style default
Editor mode normal
Model Default (recommended)
Auto-connect to IDE (external terminal) false
Enter/Space to change · Esc to exit
Simply looking at the options, you can probably guess what most of them do. However, what we're interested in is the Verbose Output option.
How to Enable It:
- Notice that by default, Verbose Output is set to
false - Navigate to it and hit Enter or Space to toggle this option to
true - Exit the config modal by hitting the Escape key
Now the moment you do this, Claude in the terminal will start showing its thought process before executing a prompt. Can you imagine how useful this is, especially when you're instructing Claude to make big changes in your project?
The Bottom Line
This feature can be easy to overlook but will save you a lot of time when working with LLMs. It helps you:
- Catch wrong directions before wasting time on bad outputs
- Learn how to write better prompts
- Find answers in the thinking process that don't make it to the final response
As always, thanks for reading!