The water/guiding water reframe is the sharpest move in here, and the 40/60 breakdown earns the argument in a way that abstract claims about judgment never quite do. You're writing from inside the practice and it shows.
The place I'd push: you get close to something harder in the 'building your judgment' section and then pull back into reassurance. The implication you don't quite follow is that the foundation has to exist before the tool, not alongside it. Judgment isn't just something you apply to AI output — it's something you develop through the friction of not having the shortcut available. The resistance of the blank page, the dead end that won't resolve, the draft that's wrong in ways you can't yet articulate but have to sit with until you can — that's the training ground. Speed removes exactly the conditions under which that capacity grows.
Which means the writers who start with these tools before they've built that foundation aren't just getting mediocre output. They're potentially training themselves out of the thing that would let them recognize it as mediocre. Your advice — learn the craft independently of the AI — is right. It's also probably the hardest sell in this entire conversation, because the tools are right there and the craft takes years.
None of that undermines what you're saying. It just suggests the stakes are higher than 'trust your taste.' The taste has to be built somewhere the AI can't reach.
As for a specific moment in working with AI, you already pointed at it, it is often the other way round. When the AI, in my case Claude as well, comes up with an idea, a direction I would have never thought of.
What also works nicely, for me, I build extensive character sheets, between 2k to 5k of words. I feed that into a GPT and have discussions with my characters, feed them with situations. I take what I like and forget the rest. But it is often quite eye-opening.
That's a pretty great expression of the pro-AI + humans "side", for my money. Not agreeing or disagreeing, at this point, being exhausted with the debate itself, but I recognise a well-reasoned argument when I see one.
Smart article that was super helpful. Steph please recommend books to fulfill your guidance “Pick up books by authors and academics you trust and familiarize yourself with good storytelling. Study structure.” Thanks. Keep up the good work
Anything James Scott Bell. He’s my favorite and he relates everything back to popular movies, TV, and books which makes his examples easy to follow. My favorites are Super Structure and Write Your Novel from the Middle.
"RIP Claude 3.7. Our time together was far too short" Omg, I feel seen! I was just reflecting on Spring/Summer of 2025, which I view as the "golden era" of Claude. Would you say that's when the 3.7 model was in use? I far prefer these earlier models to the newer ones, but when I tell people this they act like I'm crazy. The other night on a whim I decided to test out app using Claude 4.0 (the earliest model it would let me use), and I wanted to cry tears of joy! Now, this is not to say it's without issues. It's definitely prone to excess. But that's fine; I can trim that pretty easily.
I completely agree with you that at the end of the day, it's about bringing ourselves into the writing process. I like breaking the fourth wall occasionally and throwing in weird descriptions and casual asides (like "I can't even") that aren't exactly "writerly" but they're me, and no bot is going to take that away from me! Discernment is key. I know when to accept a suggestion and when to push back. There's no "rulebook" for this; it just comes down to trusting my own judgement.
One more thing -- while I agree that so much of this is about building context and developing a relationship over time as opposed to creating the perfect prompt, I do think that there are times when it can be helpful to work without context, just to have a clean slate, so to speak, as sometimes I find that the memory and the context can "muddy up" the output.
The water/guiding water reframe is the sharpest move in here, and the 40/60 breakdown earns the argument in a way that abstract claims about judgment never quite do. You're writing from inside the practice and it shows.
The place I'd push: you get close to something harder in the 'building your judgment' section and then pull back into reassurance. The implication you don't quite follow is that the foundation has to exist before the tool, not alongside it. Judgment isn't just something you apply to AI output — it's something you develop through the friction of not having the shortcut available. The resistance of the blank page, the dead end that won't resolve, the draft that's wrong in ways you can't yet articulate but have to sit with until you can — that's the training ground. Speed removes exactly the conditions under which that capacity grows.
Which means the writers who start with these tools before they've built that foundation aren't just getting mediocre output. They're potentially training themselves out of the thing that would let them recognize it as mediocre. Your advice — learn the craft independently of the AI — is right. It's also probably the hardest sell in this entire conversation, because the tools are right there and the craft takes years.
None of that undermines what you're saying. It just suggests the stakes are higher than 'trust your taste.' The taste has to be built somewhere the AI can't reach.
As for a specific moment in working with AI, you already pointed at it, it is often the other way round. When the AI, in my case Claude as well, comes up with an idea, a direction I would have never thought of.
What also works nicely, for me, I build extensive character sheets, between 2k to 5k of words. I feed that into a GPT and have discussions with my characters, feed them with situations. I take what I like and forget the rest. But it is often quite eye-opening.
That's a pretty great expression of the pro-AI + humans "side", for my money. Not agreeing or disagreeing, at this point, being exhausted with the debate itself, but I recognise a well-reasoned argument when I see one.
Smart article that was super helpful. Steph please recommend books to fulfill your guidance “Pick up books by authors and academics you trust and familiarize yourself with good storytelling. Study structure.” Thanks. Keep up the good work
Anything James Scott Bell. He’s my favorite and he relates everything back to popular movies, TV, and books which makes his examples easy to follow. My favorites are Super Structure and Write Your Novel from the Middle.
"RIP Claude 3.7. Our time together was far too short" Omg, I feel seen! I was just reflecting on Spring/Summer of 2025, which I view as the "golden era" of Claude. Would you say that's when the 3.7 model was in use? I far prefer these earlier models to the newer ones, but when I tell people this they act like I'm crazy. The other night on a whim I decided to test out app using Claude 4.0 (the earliest model it would let me use), and I wanted to cry tears of joy! Now, this is not to say it's without issues. It's definitely prone to excess. But that's fine; I can trim that pretty easily.
I completely agree with you that at the end of the day, it's about bringing ourselves into the writing process. I like breaking the fourth wall occasionally and throwing in weird descriptions and casual asides (like "I can't even") that aren't exactly "writerly" but they're me, and no bot is going to take that away from me! Discernment is key. I know when to accept a suggestion and when to push back. There's no "rulebook" for this; it just comes down to trusting my own judgement.
One more thing -- while I agree that so much of this is about building context and developing a relationship over time as opposed to creating the perfect prompt, I do think that there are times when it can be helpful to work without context, just to have a clean slate, so to speak, as sometimes I find that the memory and the context can "muddy up" the output.
Thank you as always for your wonderful insights!