I want to be honest about how my first two essays came together. The stories and ideas were mine, but AI wrote a lot of the prose.
I’m proud I published them. I’ve been wanting to write for years, and I finally communicated two stories that felt true to my experience. But when I read the experiment reports describing how much of the prose AI had written, I felt embarrassed. My immediate reaction was: okay, great, then what did I do? Are people going to dismiss this as AI slop?
That reaction brought me back to why I built Voice Kit and what I’m hoping to learn from using it.
Why I’m using these tools
After the layoffs at work, I had to figure out how to replace capacity we’d lost while the expectations for the work remained. I’d already been coding with agents. Now I needed to understand how to use them across knowledge work.
That meant having agents help me research, develop requirements, connect information, and communicate. They could build up an incredible amount of context across different sources. Working with them, I could understand that context too. But getting it into a form that other people could understand was another job.
This is where I ran into the feeling of being a meat proxy. I was coordinating between agents, colleagues, documents, and meetings. Sometimes I understood the work. Sometimes I realized I didn’t understand it nearly well enough.
I also ran into the AI slop problem. These tools can produce a lot of communication, but someone still needs to understand what’s being said, why it matters, and whether it makes sense to the audience.
I need to learn how to use this technology to be more productive and effective while continuing to learn, understand, and develop my own skills. That’s the problem I’m trying to work through.
What Voice Kit was supposed to help me do
I’ve wanted to become a better communicator, both verbally and in writing. Voice Kit is an open source project where I’m trying to codify how AI could help me do that.
Could it help me write more? Could it help me express ideas in a way that sounds like me? Could the process help me understand my thoughts clearly enough to explain them to someone else?
The questions helped me figure out what I wanted to say. In the second essay, they helped me look at my experience from the perspective of a product leader and consider what a company would need to do to use these tools effectively.
But seeing how much of the prose came from AI made me question whether I was getting enough practice expressing those ideas myself. Taking the jumble of thoughts in my mind and making it logical and linear feels like an important part of understanding.
Finding a process I can keep using
My first reaction was that I should write or dictate every first draft myself, then use AI as an editor.
But I’m also acknowledging that I won’t always have time to put that much effort into writing.
What feels workable is answering questions, dictating my thoughts at length, and having AI help turn that material into a connected draft. Then I can read it, see what’s missing or unclear, and dictate changes.
I don’t know whether writing 100% of every essay is the right goal. I do want the tools to feel like an extension of me. I want the communication to sound authentic, make sense to the audience, and express something I understand.
Being able to explain my ideas without rereading the essay would be one sign of progress. Being able to dictate a coherent first draft would be another.
Even this note came together with AI. I dictated my thoughts, worked through questions, and had AI assemble and edit the draft. Reading it helped me see what was missing and explain it further. I’m still embarrassed about how much of the prose in the first two essays came from AI, but I’m also glad those stories exist. Over time, I hope more of the language comes from me, and that I get better at explaining my ideas both on the page and in conversation.