I Don’t Write. I Synthesize

Every morning I come to the keyboard — I still call it that, though it’s been a screen for almost forty years — and the blank page taunts me with the only question that has ever mattered: what do you actually think? Not what happened. Thirty newsletters have what happened sitting in your inbox by seven. What do *you* think, and can you prove it by fusing three fragments or concepts that nobody else bothered to connect. That’s the job. It was always the job. The typing was never the job.

Stan Druckenmiller handed me the cleanest version of this argument I’ve ever seen, and he did it by accident. He wrote an op-ed in the Wall Street Journal telling the Treasury to stop leaning on the bond market — aimed straight at Scott Bessent, his own former lieutenant from the Soros days. A detector flagged the prose as AI-written. Cornered, Druck didn’t flinch: “Of course I used AI. Same reason I use a calculator when I do math problems.” His name was on it. It was his message.

The internet lost its mind. Jason Calacanis led one wing of the parade: you must disclose “AI wrote this for me” in the first sentence, or you’ve undermined the entire premise of publishing a thought piece — because we can no longer tell what’s yours and what’s “the magic black box’s thoughts.” Charlotte Alter, who formerly wrote for Time, took the other wing and put the instinct even more nakedly: “If he couldn’t be bothered to write it, why should any of us bother to read it.”

They are both completely and utterly off base. Both believe the thoughts live in the sentences — that whoever produced the words is the author, and that the labor of composition is the proof of thought. Calacanis wants it disclosed. Alter wants it ignored. Neither one thought to argue with what Druckenmiller actually said – likely because neither is actually qualified to in any meaningful way. 

Which is the part that should have mattered. Stan Druckenmiller has forgotten more about fiscal policy and the Treasury and the currency markets than his critics will ever learn, and he was handing people insight gleaned over a lifetime of actually doing the thing. The response was to attack the calculator — not the math, and certainly not the mathematician. Which is, of course, idiotic.

What I actually do at my computer, wherever that computer may be located at any given moment, is not type. I read forty stories, a dozen newsletters, a founder’s tweet that everyone else scrolled past, a line about a 1957 walkout at Fairchild Semiconductor, a scene from a movie I watched twenty years ago. Then I do the one thing the machine cannot do for me: I decide which of those things belong together. I find the collision. I throw out the eight connections that are merely clever and keep the one that’s true. I pick the premise and I kill off the bad branches. When I am finished thinking, when I am finished synthesizing a new idea, I put my name on the conclusion and disseminate it digitally. 

I don’t write all the words. I haven’t for quite a while. I think the last time I did was in the 1990’s when I was an II ranked equity research analyst. I know how to write – but that’s not the point. I synthesize ideas. I sharpen them, I stress them, and I hand the raw material back and forth with a machine and model that is very good at sentences and has never once had an original thought — and out of that back-and-forth comes a daily piece that, on a good day, makes a business owner in Ohio understand what is actually happening to her company. It lets someone new to AI understand what everyone is talking about. The judgment about what matters, what’s real, what connects: that doesn’t come out of a box. Those are my synapses firing, the same way the WSJ editorial was ideated by Druckenmiller. How many of the words Stan wrote is immaterial – it is the weight of the argument that is what matters.

So when someone calls this “slop,” they are not insulting the output. They are telling you what they always believed the work was — the typing – the slaving over the exact turn of phrase that made them sound erudite. And this is the uncomfortable part for many, the part that explains why the reaction runs so hot. The people screaming hardest and loudest are the ones with the most to lose when the creation of ideas decouples from their presentation.

For twenty years you could not cleanly separate a person who had ideas from a person who merely sounded like they did, because producing the polished version was expensive enough to pass for proof. If you could write it well, we assumed you could think it well. The model just made the polished version free. That concept is now dead. The market can finally price the two things separately – and one is worth a hell of a lot more than the other.

Take Jason. He’s a rich man, and you have to give him his due — he supposedly turned twenty-five thousand dollars into a reported fortune on Uber, and he has hit a couple of other grand slams in the angel investing arena. But angel returns are a power law, and being early and loud in the right zip code is an edge that has nothing to do with synthesis and everything to do with buying lottery tickets (in a very smart way). One Uber is not a repeatable engine. It’s proximity, a megaphone, and take enough at-bats in a rising market: something was bound to land. Strip the megaphone away and ask the only question that matters: could he source the idea himself, build the argument, string together things that don’t obviously belong together and make a new thought out of them? I doubt it. That’s Travis and Garrett on a street in Paris – taking an idea and running with it. On the evidence of this week — a man who thinks authorship is a function of who pressed the keys — I wouldn’t take the other side of that bet. (If you ask my honest opinion, Jason and Chamath and the All In crew are smart guys and great marketers – and I wouldn’t be surprised if no matter what he thinks personally – this is all a marketing gimmick and a great one).

Alter is the cleaner case, because with her the irony is total. Calacanis is an investor defending the words; you can at least see how he got confused (or used the opportunity to create hype). Alter is a writer. Her entire trade is the words, and she is arguing that the words are the work. That is the one argument a person in her chair should never make, because it is the argument that ends with her own job. If the value was ever in the sentences, the sentence machine was always coming for the sentence-makers first. Druckenmiller outsourced some of the words and kept the part that pays. Alter is defending the words because, for a certain kind of writer, the words were all there ever was and all there will ever be.

And you don’t have to take my word for any of it. Watch the replies under every one of these takes. One trader: attacking the column because Druck used AI is “like disagreeing with someone’s math because they did it in Excel.” Another: “I read Druck for his ideas.” A third cut straight to the bone — the only people who seem to care that a machine styled the piece are the ones who make a living writing. The people who actually do things saw it in a second. Only the people who sell sentences needed it explained, and only they are seething.

This is not a defense. I don’t owe anyone one. I don’t know Stan Druckenmiller, or Jason or Alter – or any of the other people tweeting up a storm today. But it touched a nerve, because today’s kerfuffle is a condemnation of my own work – this piece included.

However, if you want the proof of ideation, it’s sitting in the archive. Go back six months and read every one of my essays, one after another. You will not find a machine’s body of work — an automaton mindlessly producing words on command and forgetting those words milliseconds after rendering them. You’ll find a human mind arguing with itself: tracking a story day after day, revising a call when the tape proved it wrong, ignoring the lead others grabbed, and hunting for the connection underneath the headline that nobody else was able to piece together. That is what synthesis looks like when you watch it happen in public, and no model on earth produces it alone, because no model knows which of the day’s thousand facts is the one that matters. That decision is judgment, and judgment is the entire game.

When it comes to the subject in today’s WSJ, Stan Druckenmiller has more context, more judgement than anyone crying foul. There aren’t a dozen people on earth who could credibly argue the points with him. So let’s save the crocodile tears for a different user of AI – Druck is far beyond caring or engaging. His record speaks for itself quite eloquently. 

The tool got cheaper. It’s going to keep getting cheaper, all the way to free, sitting on an $899 box on your desk. The judgment got rarer, and it’s going to keep getting rarer, because everyone now has the words and almost nobody has the thing the words are supposed to carry.

So tomorrow morning I’ll come back to the keyboard, and the blank page will ask me the only question that matters. Same as it did today. Same as it will on the day the machine writes every last sentence and I still write every idea.

That’s the job. 

It was always the job.