Typing might actually be the bottleneck holding your creativity back. I used to think perfectly crafted text prompts were the only way to get quality outputs, but my perspective shifted entirely after reading this logic. I just saw this incredible post from an AI professional who predicts he’ll look like a “crazy man in black” wandering Singapore in 2026, muttering into his phone. He isn’t talking to himself; he’s mastering the art of prompt engineering through voice notes.
This innovator explains that he started experimenting with this workflow in mid-2025 and plans to double down on it significantly over the next few years. The core philosophy here is that your voice is the most efficient vessel for raw ideas. When we type, we unconsciously self-edit. We worry about sentence structure, spelling, and flow before the idea is even fully formed, which strips away valuable context. The original poster realized that voice grabs thoughts in their “original form,” capturing all the nuance, emotion, and cognitive noise that usually gets filtered out by a keyboard. The magic happens when you pair this raw audio with a Large Language Model like Gemini. It acts as a bridge, taking the chaotic stream of consciousness and structuring it into usable assets instantly.
📌 The Value of the “Ramble”
We have been trained to believe that clarity requires brevity, but when working with advanced AI models, context is actually king. This industry pro highlights the specific value of recording “long rambles” rather than short, clipped commands. When you speak freely, you naturally include background details, edge cases, and specific constraints that you might forget to type out because they feel tangential at the moment.
However, for an LLM, that “noise” is data. The creator suggests that by letting yourself ramble, you provide the model with a richer dataset to work from. The AI can sift through the noise to find the signal much faster than you can curate it manually. It turns your brain dump into a goldmine of context that results in significantly higher-quality outputs because the model understands the intent behind the words, not just the keywords themselves.
💡 From Audio to Assets Instantly
The workflow described by the expert is deceptively simple but incredibly powerful because it collapses the time between ideation and execution. You aren’t just getting a transcript; you are getting a finished product. By instructing the AI to process the transcription immediately, you bypass the “blank page” syndrome entirely.
The author notes that this method allows him to generate ideas, social media posts, solutions to complex problems, and training topics instantly. You go from a fleeting thought while walking down the street to a structured deliverable in seconds. This is crucial for professionals who have their best ideas away from the desk. Instead of losing that spark or jotting down a cryptic note you won’t understand later, you capture the full fidelity of the solution and have the AI do the heavy lifting of formatting it.
🚀 Rethinking the Interface
This isn’t just a productivity hack; it represents a fundamental shift in how we interact with technology. The creator is doubling down on this for 2026 because it liberates the user from the screen. It encourages “thinking outside the box” quite literally by allowing you to work while moving, observing, and experiencing the world.
If you can engineer prompts while walking through a city, your inputs will naturally be more inspired than if you were staring at a cursor blinking on a white screen. The expert implies that the future of prompt engineering isn’t about learning complex syntax, but about becoming a better verbal communicator with your AI tools. It is about becoming comfortable with the messy, human side of thinking and trusting the machine to handle the structure.
Potential Challenges to Consider
Of course, walking around talking to yourself does come with social friction, as the innovator humorously points out regarding his “suspicious smile” and the likelihood of looking like a “crazy” person. You have to be comfortable being that person in public, or find private spaces to record. Furthermore, there is a technical nuance in the “post-processing” prompt. You cannot just feed raw audio text to a model and expect magic without telling it how to process that text. You need a reliable system prompt that instructs the AI to ignore filler words and structure the output according to your specific needs, otherwise, you just end up with a wall of text.
⚙️ Practical Application: The Voice-to-Prompt Workflow
Based on the author’s method, here is how you can implement this strategy today:
- Record: Use a high-quality voice recorder or the dictation feature on your phone. Don’t stop to correct yourself. If you make a mistake, just say “correction” and keep going.
- Transcribe: Use an accurate speech-to-text tool (many LLMs have this built-in now via their mobile apps).
- Process: Paste the messy transcript into your LLM with a specific instruction to format it.
Try this instruction on your transcript:
I am providing a raw voice transcript of my ideas regarding [Topic]. Please analyze the text, ignore filler words and stuttering, and restructure the core arguments into a [Format, e.g., LinkedIn Post/Email/Project Outline]. Keep the original tone but improve clarity.
This perspective on voice-first prompting is refreshing. I highly recommend reading the full post to see exactly how this expert frames his 2026 vision!