Human memory is unreliable. To capture the pure reality of travel, I document everything: voice notes in Estonian, Suunto GPS watch tracks, and daily vlogs. However, converting that raw, chaotic data into a cohesive English book is time-consuming.
For this Taiwan journey, I wanted to see if Artificial Intelligence could bridge the gap. It wasn't a magic button. I quickly learned that AI struggles with the Estonian language, hallucinating events when background street noise confused audio transcriptions. I also learned that AI tends to flatten experiences — using words like "massive" and "incredible" so often that nothing feels special anymore. My first "ingenious" fully-automated workflow failed miserably by day three.
The Winning Workflow
Instead of asking the AI to write the book for me, I asked it to organize my materials so I could write it better. I used Google's NotebookLM to create a massive, zero-hallucination outline from my raw notes. Then, chapter by chapter, I fed targeted chunks of data to a custom Gemini agent.
It was a dance of constant refinement. AI helped structure the narrative, translate my Estonian thoughts, and format the manuscript. But the final proofreading? The sweat of the climbs? The taste of the salt in Keelung? That had to remain entirely human.
Key Lesson Learned:
"AI naturally compresses information into summaries and loses nuance. Instead of asking it to write from your material, ask it to work on your material while preserving the details."