The future of productivity looks a lot like a crazy person talking to themselves in public. We have spent decades tethered to keyboards, convinced that typing is the only professional way to work, but that era is rapidly ending. I recently came across a fascinating update from an AI professional who is betting everything on a completely different workflow for 2026. This industry pro describes wandering through Singapore, looking slightly unhinged while muttering into a phone, yet actually performing high-level prompt engineering via voice notes.
This isn’t about asking Siri to set a timer; it is about fundamentally changing how we bridge the gap between human thought and machine output. The author explains that they are “doubling down” on this method because the voice is the most efficient mechanism for capturing ideas in their rawest, most honest form. When we sit down to type, we instinctively self-edit. We worry about grammar, structure, and flow before the idea is even fully formed. By switching to voice, this innovator bypasses that internal filter, capturing the “nuance and noise” that often contains the sparks of genius. It is a bold move toward a frictionless creative process.
🎙️ The Mechanics of Audio-First Engineering
To understand why this works, we have to look at the technology driving it. The expert mentions using Large Language Models (LLMs) like Gemini, which are increasingly multimodal. This means they don’t just read text; they can listen to audio files and process them with incredible accuracy. The workflow described is deceptively simple but technically profound. You record a “long ramble”—a stream-of-consciousness dump of everything in your head regarding a specific topic. Instead of listening to it later and transcribing it manually (which is tedious), you feed the audio or the raw transcript directly into the AI.
Here is where the magic happens. The AI doesn’t just type out what you said; it acts as an intelligent filter. The creator uses the LLM to process that raw data instantly, turning chaotic thoughts into structured, high-quality prompts, blog posts, or solutions. It effectively separates the ideation phase from the structuring phase. You provide the mess, and the AI provides the order. This allows you to “write” at the speed of speech, which is generally three to four times faster than typing, without losing the complex details that usually evaporate when we try to force them into a sentence.
🧠 Capturing Cognitive Nuance
One of the most compelling points the original poster makes is about preserving the “noise in your head.” In traditional writing, we view rambling as a flaw, but in the context of AI co-creation, it is actually a feature. When you speak freely, you often circle around an idea, approaching it from different angles, using analogies, and correcting yourself mid-sentence. A static document cannot easily capture that evolution of thought, but a voice note can.
By feeding this rich, unstructured context into an LLM, you give the AI a much better understanding of your intent. It sees not just the final conclusion, but the logic you used to get there. The expert notes that this method helps generate training topics and solutions instantly. By allowing the AI to hear the hesitation or the emphasis in the phrasing (via the transcript’s syntax), you are effectively providing “few-shot prompting” examples within your narrative. You are showing the AI how you think, not just telling it what to write.
⚙️ The Refinement Workflow
The real power lies in what happens after the recording stops. The author implies a workflow where the voice note is the input for a prompt engineering task. This is a critical distinction: you aren’t just dictating an email; you are dictating the instructions for the email. This savvy professional creates a loop where they can verbally iterate on a problem. If the output isn’t right, they don’t have to rewrite a complex prompt; they simply record another quick note saying, “Keep the second paragraph, but make the tone more authoritative and fix the logic in the conclusion.”
To replicate this, you can use a system prompt that tells the AI how to handle your transcripts. For example, you might instruct the AI: “I am going to provide a raw voice transcription. Your job is not to summarize it, but to extract the core arguments and format them into a LinkedIn post using a specific viral structure.” This turns a five-minute walk to the coffee shop into a productive content creation session. The expert uses this to build posts and solutions on the fly, effectively turning downtime into high-leverage work time.
🚀 Speed and Context Switching
The final major insight from this post is the sheer efficiency of context switching. Sitting down at a laptop signals a “deep work” session, which requires mental preparation and a specific environment. Voice notes, however, are fluid. The creator mentions doing this while visiting Singapore, implying that location is no longer a constraint. You can be in a cab, walking through a park, or waiting for a meeting.
This fluidity allows for what is often called “interstitial productivity.” By utilizing the pockets of time between major tasks, the original poster generates significant output without adding hours to their workday. It also prevents the “blank page syndrome.” It is much easier to just start talking about a problem than it is to stare at a blinking cursor. Once the audio is captured and processed by the AI, you are no longer starting from scratch; you are entering the editing phase, which is cognitively much easier to handle. This shift from creator to editor is the secret sauce behind the productivity boost.
⚠️ Social and Technical Hurdles
Of course, walking around talking to a machine does have its downsides, which the author humorously acknowledges. There is the “crazy person” factor—the social stigma of talking to yourself in public. While society is getting used to people on phone calls, dictating complex prompt engineering instructions can sound bizarre to bystanders. You also have to contend with privacy; you clearly cannot discuss sensitive client data or proprietary secrets while walking down a busy street.
Furthermore, raw audio transcription isn’t always perfect. Accents, background noise (like Singapore traffic), and technical jargon can confuse even the best models. The user needs to develop a habit of reviewing the AI’s output to ensure it didn’t hallucinate a concept based on a misheard word. However, as the author suggests, the technology in 2026 is robust enough that the benefits of capturing the “nuance” far outweigh these minor friction points.
If you want to see exactly how this innovator structures their day or connect with them for more futuristic insights, check out the full post via the link below!