Your keyboard might actually be the biggest bottleneck in your creative workflow. We often obsess over crafting the perfect syntactic structure for our AI prompts, carefully selecting every word like a lawyer drafting a contract. I recently saw a forward-thinking post from this industry pro who argues that the future of prompt engineering isn’t textual at all—it’s vocal.
The premise is that by 2026, or perhaps even sooner given the pace of technology, the most efficient creators will look like they are talking to themselves. The author describes a scenario where walking down the street and muttering into a phone isn’t a sign of madness, but a sign of high-level productivity. This approach leverages the natural speed of speech to bypass the friction of typing, allowing for a more fluid transfer of ideas from brain to digital format.
⚙️ The Mechanism: Stream of Consciousness to Structured Gold
The core mechanism described by the creator relies on the evolving capabilities of Multimodal LLMs, such as Gemini. The workflow shifts the burden of structure from the human to the AI. In a traditional workflow, you filter your thoughts before you type them to ensure the AI understands. In this voice-first workflow, you provide the raw, unfiltered data.
You record a “ramble.” This includes the noise, the hesitation, the nuances, and the rapid-fire connections your brain makes. You then feed this audio (or its transcript) into the AI with a specific instruction to process it. The AI acts as an intelligent layer that sifts through the chaos, identifies the key pillars of your argument, and reconstructs them into a polished output. It is essentially using the LLM as a highly skilled editor that listens to your brainstorming session and immediately writes the final report.
💡 Insight 1: Velocity of Ideation Over Perfection
The first major takeaway from this innovator’s method is the sheer advantage of speed. Most people speak at a rate of 120 to 150 words per minute, whereas the average person types significantly slower, often around 40 words per minute. When you rely on typing, your brain has to throttle its output to match your fingers.
By switching to voice notes, you remove that governor. You can capture a fleeting idea in its entirety while walking to get coffee. The expert notes that this is the most efficient way to capture ideas in their “original form.” This prevents the loss of creative spark that happens when you pause to correct a typo or reconsider a sentence structure. The goal is to get the data out of your head as fast as possible and let the machine handle the formatting later.
💡 Insight 2: Capturing Nuance and Tone
There is a depth to spoken language that text often misses. When you write, you inadvertently sterilize your thoughts to make them readable. The post’s author highlights that voice captures “all the nuances and noise in your head.” While “noise” sounds like a negative, in the context of Large Language Models, it can actually serve as rich context.
An LLM can detect urgency, emphasis, or hesitation if the transcript is literal, or if the model processes raw audio. This helps the AI understand the intent behind the prompt better than a sterilized command. For example, rambling about a frustration with a software bug gives the AI more context about the user experience than simply typing “fix this code.” It allows the AI to generate solutions that are more aligned with the emotional and practical reality of the problem.
💡 Insight 3: The Shift to Asynchronous Prompting
This method signifies a shift towards what I call asynchronous prompt engineering. Usually, prompting is a synchronous activity: you sit, you type, you wait, you read. The workflow described by the creator allows for a separation of input and processing.
You can record five different voice notes throughout the day regarding a project. At the end of the day, you upload them all. The expert uses this specifically to generate posts, solutions, and training topics instantly. It turns “dead time”—like commuting or waiting in line—into productive time. You aren’t just recording reminders; you are actively engineering the output for later. The AI becomes a repository for your stream of consciousness, ready to crystallize it into value on demand.
Potential Challenges and Nuances
While this sounds incredibly efficient, there are social and technical hurdles. As the author humorously notes, you might look like a “crazy man” talking to your phone with a suspicious smile. Social acceptance of constant voice dictation in public spaces is still evolving. Additionally, reliance on transcription accuracy is key; if the AI misinterprets technical jargon, the output will suffer. It requires a review phase to ensure the “hallucinations” didn’t sneak in during the translation from speech to text.
Practical Application: The Cleaner Prompt
To make this actionable, here is a framework you can use to process your own voice notes using the author’s strategy.
The Prompt:
“I am going to provide you with a raw, unstructured transcript of a voice note. It is stream-of-consciousness and may contain stuttering, side thoughts, and informal language. Your goal is to extract the core ideas and structure them.
Context: [Insert what the voice note is about, e.g., A LinkedIn post idea, a strategy for Q4, a debugging problem]
Goal: [Insert desired output, e.g., Create a 3-step action plan, Write a 200-word summary, Draft an email to the team]
Transcript:
[Paste your voice note text here]”
This technique allows you to be messy in your input while demanding perfection in the output!
You should definitely check out the original post to see how this savvy professional is planning to dominate 2026.