Anthropic is Pushing Structured Prompts

What Anthropic's latest video about prompt engineering told us about what lies ahead for foundation models and complex prompts

Welcome to another edition of the best damn newsletter in AI. Here we’ll break down AI topics that matter, open your mind to use cases, and keep you ahead of the curve.

Our #1 goal is to be useful. So please shoot us an email 📩 if you have questions or feedback, and especially if you implement something we share!

Here's what we're covering today:

  • What Anthropic’s latest video told us about the future of prompt engineering

  • Apple is getting into large language models (although very very tight lips on in what form)

... and if someone forwarded this email to you, thank them 😉, and subscribe here!

Let’s get to it! 👇


What Anthropic Told Us About The Future of Prompt Engineering

Anthropic recently shared a video with 5 tips for prompt engineering. And they shared a tip so important we’re going to write a whole newsletter on it.

The tip?

'Markup your prompts with XML tags. Claude has been fine tuned to pay special attention to their structure'.

Just like ChatGPT has specifically been trained for chat; Anthropic has now specifically trained Claude to pay attention to XML tags... 👇️ 

they look like this

Anthropic’s example of XML tags in the video

They've done this to allow Claude to differentiate between various parts of the text, with the goal of helping the LLM really understand what you are asking it to do.

If you've seen our prompts use '##', this is another form of markings called Markdown.

Again, adding structure to your prompts.

Does Markdown in your prompts make your head spin? You're not alone. We get comments all the time from people wondering why they should use so much structure in their prompts.

Well, here's what we’ve always believed and experienced ourselves: Writing in a structured format forces you to think in a structured format. And when you think in a structured format, you write better prompts.

But potentially more important, if Anthropic is already training their models to detect specific prompt tags or syntax structures, that likely means the benefits will compound from writing structured prompts.

If you’re like, wait what? Markdown? Structured prompts? We really believe prompting is a foundational skill for people who are looking to delegate their work to these systems. In fact, it’s the first course we launched in our AI Ops series. If you haven’t signed up yet, learn more here.


For your reading list 📚

The big players keep playing…

AI is going mainstream, and that means AI education is paramount…

That's all!

We'll see you again on Tuesday. Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected].

... and if someone forwarded this email to you, thank them 😉, and subscribe here!


🪄 The AI Exchange Team