The Idea is Your Job. The Rest is Negotiable.
You bring the thought. The machine does the heavy lifting. That split works but only if you're honest about which side you're on.
There's a certain kind of person who spends three hours building a tool that automates a 20-minute task. Everyone knows one. They're not wrong, exactly. They just got the division of labor backwards.
Machines are fast, cheap, and tireless. They don't have good days. They don't have bad ones either. You hand them a clear instruction and they execute it without complaint, at 2am, in under a second. That's genuinely remarkable. But it only matters if the instruction is worth executing.
That part, deciding what's worth doing, is still yours.
"Most people treat AI like a faster Google. The ones getting something real out of it treat it like a collaborator with unlimited time and no fucking ego."
The creative process has always had two modes. There's the generative, messy, uncertain part where you figure out what you want. And there's the production part where you make it real. For most of human history, both were slow. Now only one is.
That changes things. If you're still treating your ideas like rough drafts waiting for the right moment, you're wasting the machine. If you're letting the machine generate the ideas too, you're just automating mediocrity at scale. Neither works.
The mental model that does work: you are the editor-in-chief, not the factory floor. You decide what gets made. You set the standard. You kill the bad versions. The machine runs the press.
A rough split
You own this
The initial idea; raw, half-formed, weird
Knowing when something is good enough
Context that doesn't exist anywhere in writing
Taste. What to cut. What to keep.
The judgment call when the answer isn't obvious
Hand it off
First drafts of anything
Formatting, reformatting, restructuring
Research and synthesis
Repetitive production work
Variants; 5 versions of the same thing
None of this is precious. The machine doesn't care that you had a different vision. It doesn't have feelings about the third revision. That's actually useful. You can iterate without social cost. Try something bad. Scrap it. Try again. The feedback loop is tight enough that you can actually learn what works.
What you can't hand off is the moment before. The bit where you look at a problem and think: this shit matters, let's figure it out. That's still a human act. It still requires something the machine genuinely doesn't have. Stakes, judgment, a reason to care.
So bring the idea. Make it a real one. Then let the machine do what machines are good at.