As developers, we spend a lot of time thinking, debugging, and explaining. velog gives us a clean, Markdown‑based space to share what we’ve learned, but actually turning raw ideas into solid posts takes effort: reproducing bugs, trimming code examples, and writing explanations that other developers can follow.
Over the last year, I’ve started using a multi‑model AI chat platform alongside velog. It hasn’t replaced my thinking, but it has become a powerful assistant that helps me ship better posts, faster.
In this article, I’ll share how I use AI in my writing workflow on velog—from debugging code to polishing the final article—so you can decide if a similar setup might help you too.
Turning rough ideas into clear outlines
Most of my posts start as a messy note: screenshots, stack traces, and half‑written thoughts. Before, I would stare at that mess and procrastinate. Now I open a multi‑model AI chat tool and do this:
-
Paste my rough notes and code snippets.
-
Ask: “Help me structure this into a blog post for intermediate developers. Suggest an outline with sections and key points.”
-
Compare outlines from several different AI models side by side.
Seeing multiple outlines at once helps me notice patterns: if three different models propose a similar structure, I usually adopt that as a starting point. I still decide what’s correct, but I skip the “blank page” problem.
For this, I use an AI chat platform that lets me compare AI models side by side in one interface, instead of jumping between different tools and browser tabs.
Using AI as a debugging and explanation partner
A lot of velog posts are about bugs: “I got this error, here’s how I fixed it.” AI is surprisingly good at helping with both the fix and the explanation.
Here’s how I use it:
-
Reproduce and explain errors
I paste the error message and the minimal code snippet into the chat and ask the models to:
-
Explain the root cause in simple language.
-
Suggest a minimal reproducible example I can share in the post.
-
Propose a safer or more idiomatic solution.
-
-
Compare multiple explanations
Because I can run several models at once, I see different ways of explaining the same bug. One model might focus on the framework, another on the language details. I then combine the best parts into a clear explanation for my readers.
-
Check edge cases
After I think I’ve solved the problem, I ask the AI: “What edge cases am I missing here?” This often surfaces extra details I can add to the post, like performance concerns or version‑specific behavior.
AI doesn’t replace reading docs, but it does help me move from “I fixed it once” to “I understand it well enough to teach it.”
Drafting and polishing the final article
Once the outline and example code are ready, I draft the article directly in velog’s Markdown editor. I still write the first pass myself, but I use AI to tighten it up:
-
Ask for alternative titles and pick one that’s clear and searchable.
-
Paste a section and ask, “Rewrite this to be clearer for a junior developer who just learned this framework.”
-
Check for confusing sentences or missing steps in tutorials.
One thing I always do: I never paste AI output directly into my post without editing. Instead, I treat it like a reviewer’s suggestion. I keep my own voice, but I let AI point out what might be unclear.
If you’re curious how different models handle technical writing, you can look at detailed comparison posts that run the same prompts through multiple models—for example, a DeepSeek R1 vs GPT‑4 comparison that focuses on reasoning and explanation quality.
👉 https://chatmultipleai.com/blog/deepseek-r1-vs-gpt4
Final thoughts
velog is already a great place for developers to share knowledge. Adding a multi‑model AI chat companion to your workflow doesn’t replace your expertise—it just helps you:
-
Turn rough notes into clear outlines.
-
Explain bugs and fixes more carefully.
-
Polish your writing for readers at different levels.
If you’ve been meaning to write more on velog but get stuck between “idea” and “published post,” experimenting with AI as a thinking partner might be the missing piece.
