The Best Editor is the One in Your Pocket: My AI-Driven Writing Workflow

TLDR:

Stop waiting for the "perfect" desk setup to write. I’ve developed a workflow that captures raw ideas via mobile dictation and uses a custom Gemini "Gem" to handle the heavy lifting of cleanup and translation. It’s all about leveraging the tools you have on hand to bridge the gap between human creativity and professional output.

The Evolution of "Vibe Coding"

My blog, Motyl.dev, started as a personal playground for experimenting with AI. Early on, I used tools like v0 to see if I could build entire systems using nothing but "vibe coding" (the practice of prompting and generating without touching the underlying logic).

My conclusion? Pure vibe coding isn't a viable path for building robust, valuable solutions. However, there is no going back to the era of 100% manual coding. The middle ground is where the magic happens: using AI as a powerful assistant while maintaining human oversight. This philosophy doesn't just apply to code; it applies to how I write these very articles.

The Problem: Messy Mobile Transcriptions

This article was born while I was speaking into my Samsung Galaxy S23 Ultra. I love the freedom of dictating a monologue while I'm away from my desk—perhaps walking the dog or just moving around.

The challenge is that built-in mobile STT (Speech-to-Text) is often subpar. It struggles with technical context, hallucinates words to fill gaps, and—since I speak in Polish but want to publish in English—the initial transcription is usually a mess of phonetic errors and linguistic "noise."

The Photography Analogy: The Best Tool is the One You Have

You might wonder why I don't use a superior transcription tool like SuperWhisper. While SuperWhisper is incredibly accurate, it often requires a desktop setup or a specific environment to run efficiently.

It’s like photography. I have a professional camera, but it’s heavy and I don't carry it everywhere. My phone, however, is always in my pocket. If I see a beautiful scene with perfect lighting, the "best" camera is the one I have with me at that moment. Similarly, my phone is the best writing tool because it's available when inspiration strikes. I’d rather have a messy transcription of a great idea than a perfect transcription of an idea I forgot because I wasn't at my computer.

The Solution: A Custom Gemini "Gem" Editor

To bridge the gap between a chaotic mobile transcript and a professional English article, I built a custom Gem (a specialized GPT-like persona) in Google Gemini. This Gem acts as my technical editor.

The workflow is simple but effective:

1. Context Injection: The Gem first asks for the topic and specific keywords (like "LLM" or "Markdown") to avoid phonetic confusion.
2. Raw Input: I paste the "hallucinated" Polish text from my Samsung.
3. Refinement: Gemini analyzes the text, fixes technical terms, removes fillers like "um" and "uh," and translates the entire piece into fluent, professional English.
4. Human Review: I do a final pass to ensure the "soul" of my message hasn't been lost in translation.

AI as the Editor, Not the Author

In this workflow, AI handles the tedious, manual labor of editing and formatting. It doesn't generate the ideas, those come from me during my walks or quiet moments away from the screen.

While I do have a "News" section on my blog that utilizes more generative AI for speed, my main articles remain human-centric. The goal isn't to let AI think for me; it's to let AI clear the path so my thoughts can reach the reader more clearly.