Forget founder mode, we’re all in editor mode now
AI has democratized editing. Is that a good thing?
“Posting AI wordslop is fundamentally disrespectful,” argued Lulu Cheng Meservey on Monday. If you couldn’t be bothered to write, why should anyone bother to read?
But flip it around: Why bother to write if you don’t have to? What if the LLM writes better and faster than you? Why is it worth your time if you can have the result in seconds?
For most of history, you couldn’t separate the process of writing from having the writing exist. The artifact and the practice were fused. Unless, of course, you were rich enough to pay someone to separate them for you — or, now, unless you have a chatbot.
Before AI, I was that someone. I signed NDAs preventing me from sharing who I ghostwrote for, precisely because those people were afraid they’d get accused of not having spent the time to do the thinking and writing. Even people who constantly used speechwriters, like President Barack Obama, undermined his wife by saying her memoir was ghostwritten. (Guess what: all of President Obama’s books besides Dreams of My Father were ghostwritten. Of course they were!)
When ghostwriting is done well, you don’t know it’s been done at all. But when it’s done poorly, people notice. Politicians sound wooden. CEOs sound like their lawyers, and celebrities like their publicists. And we’ve seen plenty of backlash to what’s perceived as a lack of authenticity. Notice that both President Donald Trump and Representative AOC eschew speechwriters and teleprompters.
Still, there are times when we’re more forgiving, when we understand that having the thing written matters more than who wrote it. I wrote 250 speeches a year as a speechwriter for the U.S. ambassador to the UN, who was also a member of President Joe Biden’s cabinet. I did it because it would have been a colossal waste of her time to decide whether we should say we were happy that the rules-based international order existed or glad that it was around (RIP). No one cared that I wrote her speeches, so long as they truly represented her thoughts and views and the stance of the United States government.
Writing is really two activities bundled into one: the generation of new text and ideas, and the refinement of existing text into something better.
But speechwriters are rare. A privilege of the powerful. For everyone else, as my writing teacher Tony Tulathimutte liked to say: No one is going to do your writing for you.
Well, now someone will. Or something will. Instantly, and for peanuts.
Except, not really. Because I don’t think AI has democratized writing. I think it’s democratized editing.
That’s because writing is really two activities bundled into one: the generation of new text and ideas, and the refinement of existing text into something better. Creation and curation, drafting and revising.
Most people, when they imagine “writing,” picture the first thing — the terror of the blank page, the taunting of the blinking cursor. But what AI offers us is the opportunity to start from the second. From now on, you never have to start from scratch. You can always have writing and ideas to react to.
Forget founder mode. We are all in editor mode now.
You’ll see this everywhere once you start looking. Vibe coders and Claude coders ‘edit’ apps into existence. RLHF is just: Did you like this, or did you like that? Musicians using Suno or Udio aren’t playing instruments, they’re selecting generated tracks. Same with the movie-makers using Runway or Midjourney. Lawyers review AI-drafted contracts instead of crafting them from scratch. Your emails are pre-written by AI, and it’s your job to edit them.
Even the phenomenon of calling out AI writing, as Meservey did and as so many others tagging Pangram on X have been doing, is a way of exercising editorial judgment. These slop-callers are performing taste, announcing they can discern the difference and that it matters to them. Pangram itself, a machine that somehow does a better job of judging the Turing test than a human, is just the editorial function made algorithmic. It’s not writing anything new. It’s judging what has already been written.
There’s something freeing about ubiquitous editor mode. It’s democratizing. It helps, for example, dyslexic people (who are significantly overrepresented among entrepreneurs) get their ideas written out. So too with the person whose first language isn’t English, or whose background is primarily in STEM. The blank page, for centuries the great equalizer and the great barrier, is now optional.
But we should also be concerned. Because this is all so new, I worry we’re still figuring out which parts of the writing process are valuable to us and why.
The linguist Emily M. Bender has a nice line: “Teachers don’t ask students to write essays because the world needs more student essays.” The science fiction writer Ted Chiang extended the metaphor. “Using ChatGPT to complete assignments,” he wrote, “is like bringing a forklift into the weight room; you will never improve your cognitive fitness that way.” The lesson is, you don’t go to the gym because the weights need to be lifted. The point is for you to lift the weights.
There’s a difference between generally caring where other ideas come from and caring whether your own ideas were given to you or worked out by you.
So the case for the student is obvious. They’re learning how to write and think. But what about us adults in the real world, those of us who presumably already know how to write and think? Are we lifting weights, or do we just want the weights to have been lifted? Are we writing to improve our cognitive fitness or to have people read our ideas? Are they even “our” ideas if the LLM came up with them?
Danny has already argued that we’ll stop caring about who “owns” ideas anyway. He might be right. But there’s a difference between generally caring where other ideas come from and caring whether your own ideas were given to you or worked out by you. Did you encounter the idea or did you earn it? Do you just agree with it, or did you arrive at it? And is the difference important?
I learned to write using what my mentor Vinca LaFleur calls the “Frankenstein” method. You barf onto the page everything you know — research, stray lines, half thoughts — and then you tinker and shape. The fragments cohere into paragraphs, paragraphs into arguments. You discover what you think through the struggle of making the mess cohere.
That’s editing, of course. But the difference is whose mess you’re editing. In the Frankenstein method, the fragments are yours. They’re incoherent because you haven’t figured it out yet. The editing is the thinking.
With AI, the coherence comes pre-installed. If you use the tool with nuance and clarity of thought, like Yoni Rechtman’s proposed “bicycle method,” then you might just wrestle your own confusion into clarity. But if you are essentially evaluating a stranger’s draft and saying “Sure, I guess” then the mess truly happened elsewhere, distributed across a trillion tokens of other people’s prose.
This is what the slop-detectors suspect: that there’s no mess, that no one is home.
So we’re all in editor mode now, and there’s no going back. The question is whether you’re editing yourself, the AI, or a mix of the two. The first is thinking. The second is outsourcing. And the third is the new thing we’re all going to have to figure out.
The weights don’t care if they get lifted. But you might.
