Akatama's Slice of the Internet

Refactoring Your Writing: A Developer's Workflow for Using ChatGPT as Your Editor

By Jimmy Lindsey

When I started blogging, I realized that editing was the hardest part. It turns out that there's no compiler for awkward sentences. You can catch a few by reading your draft out loud, but most of the clunky bits slip by, and they’re painfully obvious to everyone else. I wanted an editor I could use anytime, one that didn’t flatten my tone or take over my voice. Since I write this blog in my free time to share my projects and thoughts, hiring an editor for every post just isn’t practical.

That’s how I ended up building a ChatGPT workflow that acts like my own personal code review for writing. Best of all, it’s simple and repeatable. It treats editing like refactoring: small, deliberate changes that make the final result cleaner and easier to read. The goal isn't to automate writing, it's to build a repeatable editing loop that sharpens your own voice.

1. Write the First Draft

I am not going to tell you there is anything wrong with using an LLM to help generate ideas for your writing. I haven't done that yet, but when directed, LLMs can be great at it. That said, even if you got your idea from an LLM, I recommend against using it to completely write for you. There are a few reasons why:

• Write from your perspective.
  • If I used the LLM to write, that would be from someone else's perspective.
  • This is not to say you are always right, but it is important to have a perspective and to shape it while writing.
• Writing helps organize your thoughts.
  • I recommend having at least a rough outline in your head before you start.
  • Even without an outline, you’ll quickly realize which ideas need to come first as you write.
  • It levels up your communication, and sometimes it will even help you solve a tricky problem.

There are two exceptions to this rule: the introduction and conclusion. LLMs are great at summarizing text, and they can quickly generate multiple versions of both. If you have a good idea for an opening or closing, I would still write it out at this stage.

Once you are done with your first draft, you can go to the next step.

2. Read Through Once or Twice

I am not going to say this is editing, although if you find any major issues in terms of grammar, spelling or awkward phrasing feel free to fix it now. What's more important, though, is to make sure that you are not missing anything substantive. In other words, if you were telling a story you would need to have the beginning, middle and end. If you are describing a project you finished, you would need to talk about your goals for the project, what you did, how you did it, and then what you learned. Since we still have to edit, you don't have to worry about it being perfect. By now, your ideas and structure should be solid, even if the sentences still need polish.

3. Use ChatGPT as an Editor

Now that you have written your first draft, it's time to edit. I always make sure to upload my blog post as a markdown file to ChatGPT¹. As part of the message that uploads the file, I say something like this:

Imagine you are Michael, a senior software engineer. What does Michael think of my blog post?

ChatGPT then gives me a high-level summary of what Michael thinks. You can swap in a different persona or even multiple reviewers depending on your audience. Here's an abbreviated example of ChatGPT's output on the first draft of my post about StarWest. It gives me detailed feedback on tone, flow and technical depth:

If I imagine myself as Michael—a senior software engineer—here’s how I’d likely react to your StarWest blog post:

What Michael appreciates:
- Honesty and balance: You give a clear-eyed account of both the positives (great talks, networking, fun events) and the negatives (aggressive sales pitches, high ticket costs, vendor-heavy content). That makes the post feel credible rather than promotional.
- Concrete takeaways: The distilled notes from the talks show that you engaged deeply, and Michael would value seeing the key points without having to attend. He’d probably flag the practical testing/AI insights as especially relevant.
- Personal voice: You make it clear where you agree, where you disagree, and how you felt about the event. Michael would find that refreshing—it’s not just a dry event recap.

What Michael might critique:
- Length and density: It’s fairly long, and some sections (like the vendor passport bit) could be trimmed or tightened. Michael might skim those, focusing more on the talk takeaways
- Technical depth: As a senior engineer, Michael might want to see more personal reflection on how you’ll apply what you learned to your current or future work. The summaries are solid, but connecting them back to real-world DevOps/SDET practices would make the piece stronger for someone at his level.
- Flow: The structure is clear, but Michael might think the “What I learned” section could use stronger framing—like grouping themes (AI testing, DevSecOps, communication/collaboration) instead of just listing talks.

Make sure to take time to read and understand everything the LLM outputs on your first draft. That is not to say that we will make every change that the LLM is going to suggest. In particular, I have seen the LLM really get stuck on "lessons learned" or "important takeaways" even though I have already written about them.

Next, I will usually ask it something like this to generate me some titles for my blog post:

What would be a good title to capture Michael's attention?

The LLM's response is helpful:

If you want to catch Michael’s attention—a senior engineer who’s selective about what’s worth his time—a good title should be:

Here are some options that would resonate with him:

1. “What StarWest Taught Me About Testing, AI, and DevOps”

2. “Lessons from StarWest: Beyond Sales Pitches to Real Takeaways”

3. “AI, Testing, and DevOps: Practical Insights from StarWest 2025”

4. “Is StarWest Worth It for Engineers? My Honest Take”

5. “From Sales Floors to Keynotes: What Actually Matters at StarWest”

You do not need to pick any one title that the LLM gives to you. If you don't like any of the suggestions, have it make more for you. Also consider combining elements you like from multiple titles, or just use the suggestions as a jumping off point to create a nice title on your own.

Next, I will ask the LLM to create me a good introduction to get Michael's attention, as well as a good conclusion. I am not going to reproduce the output from ChatGPT here, as I think you get the idea. For each case, have it generate you 3-5 suggestions, then read through them and combine elements that you like together. If you want a specific kind of introduction or conclusion make sure you tell that to ChatGPT.

Now that we have gotten a good title, introduction and conclusion, now is the time to do the real editing. Ask the LLM questions like these:

Are there any sections that I can trim down to keep Michael's interest?
Can you list any awkward sentences that need to be edited further?
Can you rewrite <section you had trouble writing>?
Any other suggestions that Michael would give me to improve my blog post?

Here's an example of an awkward sentence ChatGPT found, and its suggestion for rewriting it.

Original: “One thing you can do: use the LLM to export a summary of your conversation in Markdown syntax, use that to restart your conversation.”
Why: Reads clunky with repetition of “conversation.”
Cleaner: “One useful trick: export your LLM session as a Markdown summary, then use that summary to restart and refine the dialogue.”

As with all of the other times you have prompted the LLM, make sure you are thinking about whether the suggestions it provides are useful. ChatGPT is not perfect as an editor, just as a human editor wouldn't be. Sometimes it’ll suggest a rewrite that changes your meaning, or recommends trimming a sentence in a way that loses your intent. It's honestly okay to take few or no suggestions directly from ChatGPT, and instead use the suggestions to guide your own edits.

4. Repeat Steps 2 and 3

After you have made significant edits, read your post again from start to finish. The new intro or conclusion might not match your latest tone, or some trimmed sections might now feel too thin. Keep refining until you see the rate of improvement drop. That’s when you know you’re done.

When I feed my next draft, I always ask ChatGPT this:

Here is an edited version of my blog post. Can you tell me what Michael thinks about it now?

5. Publish and Celebrate

Once you’re happy with your final draft, ask ChatGPT to generate a few LinkedIn post ideas or summaries for sharing. Then all you need to do is publish and celebrate.

Can you write me 5 LinkedIn posts that would appeal to Michael?

You are done! You have used ChatGPT to help you write. Along the way, you improved your own writing skills, didn't lose your voice, and managed to speed up the process. Great job!

Conclusion

First Draft → Read Through → Edit with ChatGPT → Revise → Repeat → Share

At this point, I’ve found that using ChatGPT as part of my editing process doesn’t just make my posts flow better, it makes me a better writer. It helps me notice patterns I’d otherwise miss and think more critically about how I structure ideas. Like any good tool, it’s not a replacement for real effort, but it does make the work a little easier and a lot more consistent.

Approach it like refactoring code and treat ChatGPT as a co-pilot, not a content factory. You’ll see steady improvement to your writing over time. It gives you structured feedback tailored to your audience, but you're still the lead developer. That's the sweet spot: your human voice, powered by machine precision.