Improving my writing with GPT

Apr 2023

As a child, I spent countless hours in the library reading. I even dared to imagine myself becoming a writer. But my lackluster writing skills dashed those hopes. Writing was the domain of a few blessed individuals. That was what I believed until GPT arrived.

There are many articles that focus on its ability to automate writing. I’m not interested in that. I want to improve my own writing.

Here I’ll show you a real example of how I improved my article on ”B-trees and Database Indexes”. Before you think it’s not relevant to you, if I can transform a dry technical article into one that is enticing to read, it can definitely help you 😉 .

1. Write your a rough draft first

We’re not here to automate our writing, but to improve it. We can only improve our writing once we have a draft.

2. Get general feedback

Now that we have our draft, let’s get some feedback on the whole article first. We will move from general to specific feedback.

How should we ask ChatGPT for feedback?

It’s useful here to remember that it’s been trained on practically every piece of text imaginable. If you just ask it something vague like “Improve this piece of writing to make it sound more interesting”, it’ll give you something vague. It won’t be as useful compared to priming it with what you want.

Priming? Like from Robert Cialdini’s Influence? Yes, surprisingly many psychology concepts are useful when you interact with ChatGPT.

Here’s what I prompted with:

You are a technical writing editor GPT trained by William Zinsser. Please provide specific feedback to improve the below article. The target audience is software developers. The goal is to get the audience to sign up to my email newsletter.

(Yes, I say “please” when I interact with it)

I pasted my article below that prompt, and received this response.

[ChatGPT]: Overall, the article is informative and well-organized. Here are some specific suggestions to improve it:

• The title could be more attention-grabbing and specific. Something like “Boost Your Database Performance with B-Tree Indexes” might be more effective in getting the target audience’s attention.
• Provide specific examples or use cases on how B-trees are used in real-world database systems. This will help the audience understand the practical benefits of using B-trees and why they should sign up for the newsletter.
• Define technical terms like “self-balancing” and “logarithmic growth” in simple language to make it easier for the audience to understand.
• Add an introduction that briefly summarizes what the article will cover, and a conclusion that reminds the audience of the benefits of B-trees and encourages them to sign up for the newsletter.

Wow thanks for the compliments ChatGPT :)

That’s a great point, my title could definitely be more specific.

Hmm, that’s true, those technical terms like “logarithmic growth” are not easily understood. I should give a simple definition.

Yes, a good introduction is key since that’s the first thing people read.

Note that sometimes GPT may not give you suggestions at all and reply with something like GPT can definitely help improve your writing, but ultimately the skill is still yours to hone and develop. If that happens, don’t feed it your whole article. Just feed it section by section (see section 4. Specific feedback ). But do read the next section since it will be crucial to helping you understand section 4.

3. How to create a good prompt to get the feedback you want

I’d say I found the feedback useful and actionable. How do you actually get good feedback?

Since ChatGPT’s been trained on so much data, you want it to be primed with the right perspective. Let’s break down the prompt sentence by sentence.

a. Priming Perspective

You are a technical writing editor GPT trained by William Zinsser.

The first sentence of your prompt narrows down the perspective. In the case of this particular post, I want a perspective of a technical writing editor because my content is technical. If your writing is fiction, you could start with: You are a fiction writing editor…

The next part of perspective is what I consider the “worldview”. In this case I chose William Zinsser because I admire his clear and succinct writing. Choose your own heroes.

b. The “ask”

This part is straightforward. You’re asking it to give you what you want. In this case, specific suggestions. I want to emphasize these two words: specific suggestions (or feedback). Without saying you want something specific, you may end up with general suggestions that are not useful. Even something like this that may seem obvious is easily missed.

Please provide specific feedback to improve the below article

This “ask” is relatively simple. You can be more detailed.

c. The audience

Much like how you’re writing for a specific audience, ChatGPT will also benefit from knowing who you’re writing for.

The target audience is software developers.

You will receive drastically different feedback based on the audience.

d. The goal

What do you want your writing to achieve? Here my goal was straightforward, so the feedback I got was straightforward (eg. … encourages them to sign up for the newsletter.)

The goal is to get the audience to sign up to my email newsletter.

Some examples of different goals I could’ve prompted with:

• ensure the audience thoroughly learns about database indexes
• ensure the audience understands the different querying strategies
• keep the reader interested throughout the article

e. Template

Here is everything above summarized into a ‘template’ you can use.

• priming perspective
• “ask”
• audience
• goal

If you really want an acronym to remember, some variation of “PAAG” or “PAGA” might work :)

4. Specific feedback

Now that we’ve gotten some general feedback, we can work on getting specific feedback. We can vary the prompt around, using different “asks” or goals.

You can use this method for improving each section of your article. If GPT is not giving you suggestions on the whole article, then you can use the below method. So replace “introduction” with something like “first section” or “conclusion” depending on which section you’re working on.

a. Improving the introduction

One of the most important parts of a blog article is the introduction. Let’s try to make it more interesting. For reference, here’s the original intro:

Database indexes speed up the performance of queries. They don’t always make your DB go brrr though. You can memorize the situations where they don’t work well or you can learn the data structure behind an index and understand it from first principles. There are many different types of index data structures, but the most common one is the B-tree, which is the focus of this post.

Now let’s see how we can improve it. Here’s our prompt:

You are a technical writing editor GPT trained by William Zinsser. Please provide specific feedback on improving the introduction paragraph. The target audience is software developers. The goal is to get the audience to sign up to my email newsletter.

After pasting the introduction and table of contents after the prompt, I get this feedback:

[ChatGPT]: To improve the introduction paragraph, you could consider providing a clear and concise statement about the value proposition of your email newsletter for software developers. This could be followed by a brief preview of the topics you plan to cover in your newsletter that will be relevant and interesting to your target audience. Additionally, you could add a hook or provocative question that sheds light on the challenges software developers face in their work that your newsletter could help address.

This is useful.

I don’t have a clear value proposition in my introduction paragraph, so I should include that. And that last sentence! Adding a hook about the challenges software developers face could really grab their attention.

Here’s an attempt at improving the introduction:

Have you ever wondered why database indexes work and why they don’t work (leading to slower query times)? This post will talk about the B-tree index, the most common type of index (data structure) used to speed up query performance. With a lookup in a B-tree taking O(log n) time compared to O(n) time without an index, you’ll understand from first principles why it works. With that understanding, you’ll also figure out the situations where it won’t work.

I’d say it’s better than my first attempt.

b. Improving the title

Another specific place to get feedback on is the title. I’m going to vary my ‘ask’ and ‘goal’ this time. I want to brainstorm some potential titles. My goal for the title is to get the target audience to continue reading the rest of the article.

As reference, here was the original title:

B-trees and Database Indexes

Our prompt:

You are a technical writing editor GPT trained by William Zinsser. Please provide 5 titles for the below article in the style of William Zinsser. The target audience is software developers. The goal is to get the target audience to read the rest of the article.

Again, if you don’t get an adequate response from GPT, or if your article is too long, you can paste your Table of Contents instead.

The response:

[ChatGPT]:

• “Mastering B-trees: The Key to Efficient Database Indexing”
• “Exploring B-trees: A Self-Balancing Data Structure for Fast Database Lookup”
• “Maximizing Database Performance with B-tree Indexing: Tips and Techniques for Developers”
• “Demystifying B-trees: How They Work and Why They Matter for Database Indexing”
• “Database Indexing 101: Understanding B-trees and their Impact on Query Performance”

None of these are quite what I’m looking for. But that doesn’t matter. It’s given me new ideas for interesting titles. After some thought, I’ve come up with this:

Why understanding B-trees will help you improve database performance

5. Phrases

Let’s try something even more specific.

I’m trying to think of a vivid phrase that describes the recklessness of blindly using database indexes without understanding how they work. Let’s see if GPT has some ideas.

You are a technical writer trained by William Zinsser. Please come up with a few vivid analogies that describe the recklessness of blindly using database indexes without understanding how they work. The target audience is software developers.

The response:

[ChatGPT]:

• Using indexes without understanding them is like driving a sports car at top speed without knowing how to brake or turn.
• Blindly relying on indexes is like building a skyscraper on a shaky foundation without verifying the structural stability.
• Using indexes without understanding how they work is like blindly following a GPS that is leading you to a cliff.
• It’s like trying to fix a clock without understanding how the gears and springs work - you might make it work temporarily, but it’s likely to break again.
• Relying on indexes without understanding them is like using a hammer to fix everything, regardless of whether it’s appropriate or not.

I like the first one. It meshes with how indexes are used to make database queries fast. Thanks GPT.

Closing thoughts

“Writing is hard work. A clear sentence is no accident. Very few sentences come out right the first time, or even the third time. Remember this in moments of despair. If you find that writing is hard, it’s because it is hard.”

― William Zinsser

I’m reminded again by this quote that writing is difficult. But with the help of GPT, I see it now as a skill that can be honed and improved.

You now have a personal editor who can help you write more engaging, nuanced, and vivid prose. Why not try embracing the technology and becoming a more skilled writer?