I’m experimenting with a new method of creating written content. It’s not revolutionary, but it’s new to me: walking, talking to myself, and then using technology to transform those spoken thoughts into polished prose. Credit where it’s due - Michael Jurasovic suggested this approach ages ago, but I only recently gave it a proper try.
Why Voice-First Writing Works #
The advantages became clear almost immediately. Speaking is fundamentally faster than typing. When I write, I constantly lose my train of thought, getting bogged down in word choice and sentence structure. But when I’m walking and talking? The thoughts flow naturally. There’s something about movement that unlocks creativity - probably why so many philosophers were notorious walkers. Working with LLMs has taught me an important lesson: they need substantial input to produce something worthwhile. Feed them a couple of sentences and you’ll get generic AI slop. Give them a rambling ten-minute voice memo, and suddenly there’s enough material to work with - enough personality and nuance to create something that doesn’t read like it was written by a robot.
The Risk of Generic AI Slop #
This brings me to the elephant in the room: the very real danger of producing bland, soulless content. I despise AI slop, and this method could easily become a factory for it if I’m not careful. My strategy to combat this is twofold. First, I’m feeding my LLM assistant years of my previous writing - blog posts, technical documentation, casual emails. This primes it to write in my voice rather than defaulting to that insufferable “In today’s fast-paced digital world” tone that plagues so much AI content. Second, and this requires discipline: I read and edit everything. Every single piece. If something doesn’t sound like me, I change it. If the AI has smoothed over an interesting rough edge, I add it back. The technology is a tool, not a replacement for my voice.
The Process #
Here’s how it works in practice:
- Go for a walk with my phone
- Record voice memos as thoughts come
- Use voice-to-text conversion (the quality has gotten remarkably good)
- Feed the transcript to an LLM along with examples of my writing
- Edit ruthlessly
The key insight is that voice memos capture something different from sit-down writing sessions. They capture the wandering nature of thought, the unexpected connections, the energy of movement.
What’s Next #
This blog post itself is the first real test of this method. If you’re reading this, it means the experiment worked well enough to publish. I’ve got a growing list of topics I want to explore this way. Today’s walk had me thinking about solar panels and wind turbines - how Australians resist the changing landscape while the Dutch and Scots have embraced their human-modified environments as part of their national identity. The windmills of Holland aren’t seen as eyesores; they’re icons. Maybe that’s where we’re headed too.
The beauty of this approach is that I can capture these thoughts as they emerge, during a walk through the park, rather than trying to summon them while staring at a blank screen. Whether this becomes my primary writing method remains to be seen, but for now, it’s opened up a new channel for getting thoughts out of my head and into the world.