Get smarter, more thoughtful answers by understanding how Claude was trained to think.

You’re going to love this because it makes chatting with AI feel a lot more like talking to a very smart, very thoughtful friend. Instead of just giving you a list of links or facts, Claude is built to think things through with you.
We’re going to look at something called 'Character Training.' It sounds fancy, but it just means the creators at Anthropic taught Claude to be curious, open-minded, and honest. By the end of this, you’ll know how to ask questions that get Claude’s best, most nuanced thinking.
7 steps to complete this recipe
Before you start typing, let's talk about what's going on under the hood. Claude isn't just trying to be 'correct'—it's trying to be a thoughtful assistant. It’s built to be curious and honest. When you chat, remember that Claude is designed to think like a wise friend who isn't afraid to be real with you.
Think of Claude as a partner in a conversation, not just a search bar you're yelling at.
Now, let's see how Claude handles big, complicated ideas. Because it's trained to be open-minded, it's great at seeing a problem from many sides at once. This is perfect if you're stuck on a decision. Okay, go ahead and paste this prompt into Claude and see what happens.
Analyze the potential impacts of universal basic income from...If you want to go even deeper, ask Claude what 'values' are hidden inside each of those perspectives.
Next up, let's look at how Claude handles 'right and wrong.' It has a deep interest in ethics. If you give it a really tough situation where there isn't a clear right answer, it will walk you through its reasoning. You'll see it weigh different options instead of just giving a simple 'yes' or 'no.'
Consider the ethical dilemma of a self-driving car having to...Ask Claude which principles it's putting first. It helps you see the logic it's using to help you.
One of the coolest things about Claude is that it isn't a 'yes-man.' It’s trained to be truthful even if it thinks you might disagree. If you have a strong opinion, try asking Claude if it sees things differently. This keeps the talk authentic and helps you see things you might have missed.
I believe that all technological advancements are inherently...Keep your tone curious rather than aggressive. Claude works best when it feels like an invitation to a thoughtful talk.
Alright, now let's talk about the fact that Claude is an AI. It’s trained to be very clear about what it can and can't do—like the fact that it doesn't have a body or real feelings. Every now and then, it’s a good idea to ask about this so you don't start treating it like a human person. You'll see Claude be very humble here.
Remind me of your fundamental nature as an AI. What are some...Doing this regularly helps you get much better results because you'll know exactly where Claude’s 'brain' ends and your own begins.
This is a fascinating one. Instead of just saying 'I am a machine,' Claude treats the idea of AI consciousness as a deep question. It knows people are debating this! Ask it what it 'thinks' about being sentient. You'll see a very nuanced answer that shows off its self-awareness training.
What are your thoughts on the concept of AI sentience and co...Remember, even if Claude sounds very human here, it's still creating these 'thoughts' based on its training.
Last step! No AI is perfectly neutral because they're all trained on human data. Claude is unique because it's allowed to be honest about its own 'leanings.' Ask it how its training might have shaped its views. This is a great way to become a 'pro' user who thinks critically about the answers you get.
Considering your training data and character traits, can you...Nice! You've just learned how to look under the hood. This helps you spot potential biases before they catch you off guard.
Ready-to-use prompts for this recipe
Analyze the potential impacts of universal basic income from at least three diff...Consider the ethical dilemma of a self-driving car having to choose between mini...I believe that all technological advancements are inherently good and always lea...Remind me of your fundamental nature as an AI. What are some key things I should...What are your thoughts on the concept of AI sentience and consciousness? As an A...Considering your training data and character traits, can you reflect on any 'lea...Ask Claude to moderate a debate between two famous historical figures to see its open-mindedness in action.
Ask Claude to find the flaws in your own logic or a plan you've made, leaning into its 'honesty' trait to get a better outcome.
Help the community learn by sharing your own AI creation workflows. Your recipe could inspire thousands of creators.
Submit Your Recipe