How to Talk to Claude — Your First Conversation Guide

Just ask something — anything

In the previous article, we set up a Claude account. Now let's actually ask Claude a question. Open the Claude Desktop app — or the web version works too.

Then, just like texting a friend, try asking something like this:

Type that in and hit Enter. Claude will ask you about your mood and preferences.

That's the basics of talking to AI. You type a question, you get an answer. That's it.

You might be thinking, "Wait, that's all there is to it?" Yep. The important thing is to experience this input → response cycle first.

Try a few more. Anything goes.

• "What's the weather like in New York tomorrow?"
• "Cats or dogs — which is easier to take care of?"
• "Explain space to a 5-year-old"

You'll notice it answers pretty much anything. No need to pick the right keywords like a Google search, no clicking through links. Ask, and you get an answer. That's the appeal of AI.

Change how you ask, change the answer

Here's where it gets interesting. The same topic can give you completely different answers depending on how you ask.

Early on, I once asked Claude something like "Build me a nice website" — super vague — and the result was nothing like what I wanted. But the problem was how I asked, not the AI itself.

Let's use "dinner" as an example and try three different patterns.

Pattern 1: Ask vaguely

You'll get something like this:

Not bad, but Claude has no idea what's in your fridge. You'll probably end up thinking, "OK, but which one should I actually make?"

Pattern 2: Add conditions

The response is totally different.

See the difference? You could walk into the kitchen right now and start cooking. Just adding a few conditions made all the difference.

Pattern 3: Assign a role

Just adding "you are a professional chef" changes the depth of the answer.

Heat control tips, which way to slice the onion — stuff you didn't even ask for. That's the power of role assignment.

How did the three patterns compare? Same topic — dinner — but the quality of the answers was completely different.

Here's the takeaway:

• Ask vaguely → You get broad suggestions, but they tend to be generic
• Add conditions → You get answers tailored to your situation
• Assign a role → You get deeper, more expert-level responses

You don't need to craft the perfect question every time. Start vague, and if you think "I wish it were a bit more like this," add conditions as you go. Being able to adjust as you go is one of AI's best features.

What AI is good at — and not so good at

After trying all that, you might be thinking "AI is amazing." I felt the same way at first. But the more you use it, the more you'll notice some weak spots.

Good at

• Writing and editing — Email drafts, summarizing text. Personally, this is what I use it for the most
• Brainstorming — "Give me 10 ideas for X" — you'll get some surprisingly good ones mixed in
• Explaining things simply — It's great at breaking down complex concepts. Adding "explain it like I'm 5" makes it even better
• Translation — Not just between languages, but also turning jargon into plain language
• Writing code — It's surprisingly capable at programming (we'll try this in a later article)

Not so good at

• Latest news — Its training data has a cutoff, so it doesn't know yesterday's headlines
• Math — It sometimes gets even basic math wrong. A calculator is more reliable
• Factual accuracy — It can confidently state things that are flat-out wrong. This is called hallucination

Hallucination is the one to remember. AI isn't great at saying "I don't know," so it'll answer with full confidence even when it's making things up.

So here's one important rule:

Treat AI answers as a starting point, not the final word. Don't take them at face value. Double-check anything important. Keep that in mind, and AI becomes an incredibly useful partner.

Try it with your own questions

Now that you've made it this far, all that's left is to try it yourself. If you're not sure what to ask, here are some patterns I use all the time.

In daily life

• "Give me 5 fun things to do with my wife this weekend"
• "Make me a moving checklist"
• "Write a casual birthday message for a friend"

At work

• "Implement the xxx feature"
• "Create a meeting agenda. Topics are X and Y"
• "Make this text sound more polite" (paste your text)

For learning

• "What's an index fund? Explain it like I'm in middle school"
• "Give me polite ways to decline in a professional email"
• "How do I use pivot tables in Excel?"

Remember the three patterns from earlier. Start vague, then add conditions when you think "I want something a bit more specific." That alone will dramatically improve the quality of the answers.

And don't worry if it doesn't work perfectly. You can always say "Make that shorter" or "Make it more casual," and Claude will adjust. That's the beauty of AI — you can always try again.

Summary

Here's what we covered:

• Just ask anything — think of it like texting a friend
• How you ask changes what you get — vague → conditions → role assignment
• AI isn't perfect — watch out for hallucinations (confident but wrong answers)
• Think of AI answers as a first draft, not the final word

In the next article, we'll use an editor called Cursor to start building things with AI. Even if you've never written code before, don't worry. Just tell Claude what you want in plain English, and it'll build something that actually works. Give it a try.