Typing is officially the biggest bottleneck in your creative process. We can speak at roughly 150 words per minute, yet most of us type at less than 40, meaning we are actively throttling our own intelligence every time we sit down at a keyboard. This concept was perfectly illustrated by a LinkedIn creator who plans to double down on voice workflows in 2026, predicting a future where talking to our phones is the standard for high-level work. I was fascinated by this approach because it solves the friction problem that plagues so many professionals trying to leverage AI.
The Mechanism: Capturing the Raw Signal
The core philosophy shared by this expert is simple but profound: your voice is the most efficient vessel for capturing ideas in their purest form. When we type, we tend to self-edit. We worry about sentence structure, spelling, and flow, which engages the analytical part of the brain and dampens the creative side. The author argues that by switching to voice, you bypass that internal editor entirely. You aren’t just dictating; you are performing a brain dump that includes “all the nuances and noise in your head.”
This is where the “engineering” part of the equation comes in. In the past, voice notes were difficult to use because scrubbing through audio is tedious. However, this innovator explains that the game has shifted with the arrival of advanced Large Language Models (LLMs) like Gemini. The workflow isn’t just about recording; it’s about recording long, unstructured rambles and then using the AI to act as the restructuring engine. The AI takes the chaotic raw signal, filled with stutters, corrections, and half-baked thoughts, and transmutes it into polished, high-quality prompts or content. The machine handles the syntax so you can focus entirely on the substance.
📌 The Value of “Nuance and Noise”
One of the most striking points the original poster makes is the value of the noise itself. We are often taught to be concise when interacting with computers, but this contributor suggests that the “messy” parts of our speech actually contain critical context. When you speak naturally, your tone, your hesitation, and your self-correction (“actually, let’s phrase it like this…”) give the AI clues about your intent that a typed query often lacks.
By feeding an LLM a raw transcript of your thought process, you are essentially giving it a map of how you reached a conclusion. The creator uses this method to generate ideas, posts, solutions, and training topics instantly. The AI can sift through the “ums” and “ahs” to find the diamond in the rough, understanding not just what you want to say, but why you want to say it. This results in outputs that sound much more like you because they are derived from your natural cadence and stream of consciousness rather than a sterile text prompt.
📌 Leveraging Multimodal Models like Gemini
The savvy professional behind this post specifically highlights using Gemini for this workflow. This is a crucial detail because of how modern multimodal models handle long context windows. You can record a five-minute monologue detailing a complex problem you are facing at work. Instead of trying to summarize that problem into a text box, you upload the transcript (or the audio file directly, depending on the tool) and simply say, “Here is my situation.”
From there, the author notes you can turn that transcript into anything. You might ask the AI to extract a bulleted list of action items, draft a LinkedIn post based on the key themes, or create a step-by-step solution guide. The heavy lifting of organization is offloaded to the software. This transforms the AI from a mere text generator into a sophisticated synthesizer of your own thoughts. It allows for a “human-in-the-loop” workflow where the human provides the raw creativity and the machine provides the structural rigor.
📌 Recursive Prompting: Talking to Build Prompts
Perhaps the most actionable takeaway from this industry pro is the concept of using voice to build the prompts themselves. The user describes this as “prompt engineering via voice notes.” Instead of agonizing over the perfect syntax for a complex prompt, you can simply explain your goal to the AI verbally. You might say, “I need a prompt that will help me analyze financial data, but I want it to be sarcastic and use metaphors from 1980s action movies.”
The AI can then take that verbal instruction and write the complex, technical prompt for you. You are using natural language to program the model. This recursive approach, using the AI to write the instructions for the AI, drastically lowers the barrier to entry for complex prompt engineering. It means you don’t need to memorize frameworks or cheat sheets; you just need to be able to articulate what you want the end result to look like.
Potential Challenges and Nuances
Of course, adopting this habit comes with social friction. The post’s author humorously notes the image of a “crazy man in black” talking to his phone with a suspicious smile in public. There is a distinct awkwardness to dictating complex ideas in a coffee shop or walking down the street. Beyond the social aspect, there are privacy considerations to keep in mind. Uploading sensitive work discussions to cloud-based LLMs requires due diligence regarding data retention policies. Furthermore, while transcription technology is excellent, it can still struggle with heavy jargon or specific acronyms, occasionally requiring a manual cleanup step or a custom vocabulary setup.
⚙️ Captain YAR’s Practical Application
To replicate the success of this innovator, you need a system to bridge the gap between your voice recorder and your AI model. Here is a simple workflow you can try today.
The “Brain-to-Structure” Workflow:
- Record: Open your phone’s voice memo app and talk for 3–5 minutes about a specific topic. Do not stop to correct yourself.
- Transcribe: Use a tool like the native dictation on your phone, Otter.ai, or upload the file to Gemini/ChatGPT if they support audio processing.
- Refine: Paste the raw text into your LLM and use the following prompt to structure it.
The “Cleanup Prompt”:
“I am providing a raw transcript of a voice note regarding [Topic]. The text contains filler words, redundancies, and unstructured thoughts. Please analyze the transcript and:
- Summarize the core argument in one sentence.
- Extract the key action items or insights into a bulleted list.
- Rewrite the content into a cohesive, professional draft suitable for [Email/LinkedIn/Report].”
The original creator is betting big on this for 2026, and after seeing the efficiency gains, it is easy to see why!
Check out the full post to see the author’s prediction for the future of work.