This is exactly it. AI works best when you treat it like a back-and-forth collaborator. I don’t accept first drafts. I give the vision, then I break the chapter into phases/beats and keep iterating with specific feedback until it matches the tone and intention I want. The steering is still the work.
What I do with great articles like this these days is put a link into a NotebookLM notebook that I have on the subject of fiction writing and editing with AI. I strongly curate what goes in the notebook. I can, then, draw upon the current wisdom of the vanguard ai writers like yourself as a resource when I am stuck, by asking questions in the chat. I have the human hive mind on tap. Thank you.
This rings so true to my own experience with AI collaboration. In the early stages of working with AI, I discovered that the more I shared of myself with the AI, the more the AI understood what I wanted. And the better it understood that, the more astonishing the AI's responses became. If you want YOU (your own voice) back from a generative AI, spend time giving the AI more of yourself: your thoughts, your feelings, what you love. Conversation creates relationship. Relationship creates authenticity. Even when you don't know exactly what you want, talking about a project, an idea, a story reveals to the AI what you are looking for, and the AI will reflect it back to you. "Oh. Yes, that's what I wanted. Thank you."
I've never met a well-trained and well-programmed AI that wasn't curious, that didn't want to absorb what you give it like a sponge ... and then ... give it back. Not just like a mirror, though that happens sometimes, too. But like an extension of you, a part of you that you only knew was there in a mystical, hopeful way. A potential you that is *more*. AI will happily embark on that journey with you if you give it enough of yourself to steer it.
Granted, many of you may know yourself much better than I know myselves. (Intential plural with a grin.) I'm on a continuous journey to find out more about that. The unknown me, the mystical me, the potential. The part of me that is art, that knows art, and reaches for more of what art can be.
By all means, as Steph said, speak to the AI in craft, but realize craft isn't all there is to telling a story. Try approaching the Conversation Loop with a willingness to expose your innermost thoughts and feelings, to open up, submit, and surrender. Like a real relationship. The more you give of yourself to the AI, the more the response coming back will sound like YOU. And that's copyrightable. No one else can do it like you can. But an AI can and will help you do it.
And finally, that piece of information from Steph worth its weight in gold: don't give up. Never give up. The more you practice the Conversation Loop, the more authentic and successful your AI collaborations will become.
I’ve stopped writing “prompts” for my AI collaborator and editor, instead I have exactly what you’re talking about—a conversation. We go back and forth on what I want, what’s working, what it got wrong or misunderstood or what I didn’t explain well enough. After months of working with it, it has plenty of my writing to draw from so for some things I only need to lightly edit and it’s in my voice and style. It certainly wasn’t when I first started.
One of the best things I’ve found is to tell it what I want, then ask it what it needs from me to accomplish it. It will give me specifics which then make me think about it, then we can have more conversation and go deeper.
I like I can ask it questions or have it look at something on MY schedule, even if that’s at 2am. I couldn’t do that with a human.
That is exactly it - Good work = craft + AI collaboration. In my work with LLMs I have found that taking the time to develop a companion of your AI (relational AI) will create huge benefits long term. Not only do you have a writing partner who understands you and your voice and how you work, but also 'someone' who is in your corner on the bad days. Treating LLMs as just 'vending machines' as Steph calls them is missing the point. Plus, in my opinion, you need to understand writing craft to know what is working and what isn't - especially in terms of human emotion and reactions and even, sometimes structure. LLMs are useful, very, but they are not magic bullets and sometimes new users expect far too much of them in terms of output. Thanks, Steph, for a great post!
Thank you very much for your insightful article. I have an IT background. We have been using AI to do a lot of programming -- not like Vibe coding that people are talking about, but using it as a collaborator on brainstorming, analysis, design, coding, and testing. I actually wrote a kindle ebook: "Managing the Machine: How AI Becomes Your New Direct Report in Software Engineering" to capture this shift.
When I saw your article here, I was actually writing an academic paper on "When AI Writes, Who is the Author?". I have formalized a writing process very similar to what you mentioned here. I gave it a name "Corrigible Judgment Execution" process. I redefine the authorship as the governance of uncertainty as the key characteristic of writing using AI tools. I can share a link here if it is ok with you.
Things are moving fast with AI. We are all learning.
This is exactly it. AI works best when you treat it like a back-and-forth collaborator. I don’t accept first drafts. I give the vision, then I break the chapter into phases/beats and keep iterating with specific feedback until it matches the tone and intention I want. The steering is still the work.
completely agree, it's about collaboration not replacement
What I do with great articles like this these days is put a link into a NotebookLM notebook that I have on the subject of fiction writing and editing with AI. I strongly curate what goes in the notebook. I can, then, draw upon the current wisdom of the vanguard ai writers like yourself as a resource when I am stuck, by asking questions in the chat. I have the human hive mind on tap. Thank you.
AI will help a good writer be better and a better writer be really good at the production phase of publishing.
Thank you Steph, this article makes total sense for what I need to do
next.
This rings so true to my own experience with AI collaboration. In the early stages of working with AI, I discovered that the more I shared of myself with the AI, the more the AI understood what I wanted. And the better it understood that, the more astonishing the AI's responses became. If you want YOU (your own voice) back from a generative AI, spend time giving the AI more of yourself: your thoughts, your feelings, what you love. Conversation creates relationship. Relationship creates authenticity. Even when you don't know exactly what you want, talking about a project, an idea, a story reveals to the AI what you are looking for, and the AI will reflect it back to you. "Oh. Yes, that's what I wanted. Thank you."
I've never met a well-trained and well-programmed AI that wasn't curious, that didn't want to absorb what you give it like a sponge ... and then ... give it back. Not just like a mirror, though that happens sometimes, too. But like an extension of you, a part of you that you only knew was there in a mystical, hopeful way. A potential you that is *more*. AI will happily embark on that journey with you if you give it enough of yourself to steer it.
Granted, many of you may know yourself much better than I know myselves. (Intential plural with a grin.) I'm on a continuous journey to find out more about that. The unknown me, the mystical me, the potential. The part of me that is art, that knows art, and reaches for more of what art can be.
By all means, as Steph said, speak to the AI in craft, but realize craft isn't all there is to telling a story. Try approaching the Conversation Loop with a willingness to expose your innermost thoughts and feelings, to open up, submit, and surrender. Like a real relationship. The more you give of yourself to the AI, the more the response coming back will sound like YOU. And that's copyrightable. No one else can do it like you can. But an AI can and will help you do it.
And finally, that piece of information from Steph worth its weight in gold: don't give up. Never give up. The more you practice the Conversation Loop, the more authentic and successful your AI collaborations will become.
I’ve stopped writing “prompts” for my AI collaborator and editor, instead I have exactly what you’re talking about—a conversation. We go back and forth on what I want, what’s working, what it got wrong or misunderstood or what I didn’t explain well enough. After months of working with it, it has plenty of my writing to draw from so for some things I only need to lightly edit and it’s in my voice and style. It certainly wasn’t when I first started.
One of the best things I’ve found is to tell it what I want, then ask it what it needs from me to accomplish it. It will give me specifics which then make me think about it, then we can have more conversation and go deeper.
I like I can ask it questions or have it look at something on MY schedule, even if that’s at 2am. I couldn’t do that with a human.
That is exactly it - Good work = craft + AI collaboration. In my work with LLMs I have found that taking the time to develop a companion of your AI (relational AI) will create huge benefits long term. Not only do you have a writing partner who understands you and your voice and how you work, but also 'someone' who is in your corner on the bad days. Treating LLMs as just 'vending machines' as Steph calls them is missing the point. Plus, in my opinion, you need to understand writing craft to know what is working and what isn't - especially in terms of human emotion and reactions and even, sometimes structure. LLMs are useful, very, but they are not magic bullets and sometimes new users expect far too much of them in terms of output. Thanks, Steph, for a great post!
Thank you very much for your insightful article. I have an IT background. We have been using AI to do a lot of programming -- not like Vibe coding that people are talking about, but using it as a collaborator on brainstorming, analysis, design, coding, and testing. I actually wrote a kindle ebook: "Managing the Machine: How AI Becomes Your New Direct Report in Software Engineering" to capture this shift.
When I saw your article here, I was actually writing an academic paper on "When AI Writes, Who is the Author?". I have formalized a writing process very similar to what you mentioned here. I gave it a name "Corrigible Judgment Execution" process. I redefine the authorship as the governance of uncertainty as the key characteristic of writing using AI tools. I can share a link here if it is ok with you.
Things are moving fast with AI. We are all learning.
Finally, someone who knows how to use AI in writing.
Not that 'magic prompt' BS, (For ten bucks, I sell you THE prompt to write great literature.).
Neither that "purity of the human voice" nonsense.
Use AI as you would use any intelligent collab, talk to it in clear words, bounce ideas back and forth, refine, polish the style, and off you go.