Hey Warrior,

Claude has been my go-to AI model lately.

If you’ve read my latest posts, you’ll know that in my opinion it performs a lot better than other models like ChatGPT or Gemini. Especially for the tasks I use AI the most for.

Will that always be the case? Probably not, but it is now.

So today, I want to share 5 Claude hacks that have been really helpful for me and will help you level up your usage as well.

Here's what we'll cover:

• How to get faster (and cheaper) responses at almost the same quality

• Why you should use XML tags in your Claude prompts

• How to optimize all your prompts with one prompt

• A plugin that Google Sheets power users will love

• The (overlooked) chat search function

Let's jump in!

Read time: 5 minutes

1. How to get faster (and cheaper) responses at almost the same quality

There are currently three Claude 3 models. The Opus model is the most intelligent one, while Haiku is the fastest.

Additionally, Haiku is also the cheapest (by far). Not only is it free to use on claude.ai, the API pricing of Haiku is also 60x cheaper than Opus. Even compared to OpenAI’s top model, it’s still 24x cheaper.

But here is the cool thing.

Some users have been noticing that Haiku is actually not much less intelligent that Opus. That is, if you prompt it in a certain way…

The trick is to provide enough examples in your prompt.

What do I mean by that?

Let’s say I want to use Claude to write a product description. My prompt might look like this:

Write a compelling product description for my product.

Product context: [context]

Here are some examples of good product descriptions:

- [Example 1]
- [Example 2]
- [Example 3]
...

It seems that the more examples you provide (try up to 10), the better Haiku will perform, reaching a point where it performs just as well as Opus.

So if you’re a developer and want to keep the costs low, or even if you just want to use Claude on the website without upgrading to Pro (which you’d need to to use Opus), this might be a good option for you.

2. Why you should use XML tags in your Claude prompts

According to the official Claude prompt engineering guide, using XML tags in your prompts, not only helps to get more structure into your prompts, but also leads to better outputs.

This might sound a bit technical for non-techies, but all that is, is to wrap certain parts of your prompts into separate sections.

Using the same prompt example for writing product descriptions from above, here’s what that looks like:

Write a compelling product description for my product.

<Product context> context </Product context>

Here are some examples of good product descriptions:

<Example 1> Example 1 </Example 1>
<Example 2> Example 2 </Example 2>
<Example 3> Example 3 </Example 3>
...

As you can see, I’ve wrapped the “context” and “example” sections with xml tags in this format: <tag> </tag>.

Simple, but can make a huge difference.

3. How to optimize all your prompts with one prompt

In a previous post I wrote about the official Claude prompt optimizer that works extremely well.

It basically creates an optimized prompt template for any prompt you enter.

Since it can be a bit technical to use, I've now created a simplified prompt that you can just copy and paste into the Claude UI.

Just get the full prompt from my tweet above, then copy the entire thing into Claude and hit enter.

Then write the prompt that you want to have optimized…

I personally use this prompt optimizer a lot now to improve my prompts.

4. A Plugin that Google Sheets power users will love

There's an official Claude <> Google Sheets plugin that allows you to use Claude directly within Google Sheets.

This opens up a world of use cases since you can now use intelligent language operations within your sheets.

I go into two examples in this thread, covering how to write personalized cold emails and custom product descriptions.

5) The (overlooked) chat search function

I think one of the most requested features of ChatGPT is a chat search (the weird thing is that there is a search in the mobile app, but not on desktop).

But the good news?

Claude actually has a built-in chat search function.

As you type your prompt, Claude searches through all your previous chats, making it easier to find past interactions. Unfortunately it currently seems to only search keywords in the chat title, which doesn’t make it as useful as it can be.

But let’s hope this gets an upgrade soon.

Thanks for reading!

Btw, if you’ve been wondering what’s happening with the AI Growth Kit, here’s a quick update:

After getting over 50+ pre-orders, I have now closed the pre-sales. I’ll be focusing on getting feedback and improving the product now!

We’ve already made a lot of progress and I’m super excited where this is going. I think this will be the best AI resource for any entrepreneur, maker and content creator that want to grow their business out there!

Stay tuned for the official launch in a few weeks!

If you enjoyed this email, consider forwarding this newsletter to a friend or colleague.

See you in the next one!

Moritz