Typing is officially the slowest way to communicate with artificial intelligence. We often obsess over the perfect written syntax, carefully editing our words before we even hit enter, but this process inherently filters out the raw creativity of our initial thoughts. I recently encountered a fascinating post from an AI professional based in Singapore who is completely rethinking this dynamic. This innovator describes a future—specifically 2026—where he navigates the city like a “crazy man in black,” talking to his phone with a suspicious smile, not to a person, but to a Large Language Model.
🎙️ The Mechanics of Voice-First Engineering
The core concept the author presents is shifting from “text-based prompting” to “stream-of-consciousness engineering.” When we sit down to type, we subconsciously edit ourselves. We worry about structure, grammar, and clarity, which often strips away the nuance and the “noise” that actually contains the most valuable context. The expert explains that the voice is the most efficient vessel for capturing ideas in their original, unpolished form.
This approach relies on the capabilities of multimodal models like Gemini. The process is simple yet profound: you record a long, unstructured ramble—everything running through your head about a specific topic. You don’t try to organize it; you just dump the data. Then, you feed this raw audio into the AI. Because the model can process the transcript instantly, it acts as a hyper-intelligent filter. It takes the chaos of your verbal brainstorming and structures it into usable assets. The creator is essentially using the AI to bridge the gap between human messy thought and machine-readable structure.
💡 Insight 1: Instant Content Architecture
The first major application this industry pro highlights is the generation of content and posts. Think about the friction involved in writing a LinkedIn post or a blog article. You have an idea while walking the dog or commuting, but by the time you sit at a computer, the spark is gone, or you struggle to find the right words.
By adopting the author’s method, you remove the friction of the blank page. You simply open your recording app and speak for three to five minutes. You describe the story, the lesson you learned, and the key takeaways you want the audience to have. You can even include instructions like, “Make the tone professional but witty.” The AI takes that transcript, removes the “ums” and “ahs,” organizes the rambling thoughts into a coherent narrative, and generates a draft that is 90% complete. This turns “dead time” like walking or commuting into your most productive creative hours.
⚙️ Insight 2: The Meta-Prompting Strategy
Perhaps the most compelling use case mentioned by the original poster is doing “prompt engineering via voice notes.” This sounds counterintuitive because we usually think of prompt engineering as a precise, code-like discipline. However, this savvy professional uses voice to describe the intent of a prompt rather than writing the prompt itself.
For example, instead of struggling to write a complex system instruction for an AI agent, he might say, “I need a prompt that acts as a senior Python developer. It should critique my code, look for security vulnerabilities, and offer three alternative solutions for every function I provide. It needs to be strict and not overly polite.” The AI listens to this verbal request and generates the high-quality, technically structured text prompt needed to achieve that result. You are effectively using natural language to code the behavior of the model.
🚀 Insight 3: Rapid Solution Prototyping
The third pillar of this strategy is using voice for complex problem solving and “solution” generation. The LinkedIn user notes that this method helps him generate training topics and solutions instantly. This is essentially “rubber duck debugging” on steroids. When you are stuck on a difficult business problem or a technical blocker, explaining it out loud often helps you visualize the issue.
By recording this explanation and feeding it to the LLM, you provide the AI with a massive amount of context that you would never have the patience to type out. You can describe the variables, the constraints, your current frustrations, and your desired outcome. The AI analyzes this rich data set and proposes structured solutions, training modules, or strategic roadmaps. It turns a ten-minute rant into a clean, actionable project plan.
⚠️ Nuances to Consider
While this method is incredibly powerful, the author hints at the social awkwardness involved. Talking to your phone with a “suspicious smile” in public might draw some looks. There is also the challenge of privacy; you need to be mindful of where you are recording if you are discussing sensitive proprietary information. Furthermore, relying on voice requires trust in the model’s transcription and reasoning capabilities. It works best with advanced models that can understand context and intent, rather than simple speech-to-text tools that only dictate words.
If you want to see exactly how this creator is planning to double down on this strategy in 2026, I highly recommend checking out his full post for more context.