Stanley Druckenmiller published a Wall Street Journal op-ed so full of AI-isms that a detector scored it 100% — and when the mob came, his answer was 'of course I used AI.' With the Journal backing him up, the question stops being whether AI writing is legitimate and becomes how to do it well. NLW offers five rules, plus a use-case-by-use-case crib sheet for emails, memos, social, marketing copy, and op-eds.
The purity test for AI writing will die. The quality test won't.
Druckenmiller's unapologetic 'of course I used AI' — and the WSJ's shrug — signal that AI writing is now simply a fact of life. But acceptance of AI writing is not acceptance of bad writing: the easier the inputs get, the higher the burden on the outputs. Quality reads as effort, effort's absence is glaringly legible, and writing is still thinking. Different types of writing need different rules — and laziness in work will always come across as laziness in work.
Stanley Druckenmiller — whose fund famously had zero losing years from 1981 to 2010 — published "Let the Bond Market Speak," a full-throated WSJ critique of Treasury Secretary Scott Bessent. Within hours, nobody was talking about monetary policy; they were talking about the fact that the piece was very, very clearly written by AI.
The op-ed was replete with every egregious AI tell, especially "it's not this, it's that" constructions ("that isn't a crisis, it's an invoice"). AI detector Pangram put the chance it was AI-generated at 100%, and critics piled on — calling it plagiarism, slop, and demanding mandatory disclosure from the Journal.
— Stanley Druckenmiller, responding to the op-ed backlash. No sheepish apology. Druckenmiller's full response: "Bro, of course I used AI... There's a reason I moved from an English major to being an economics major. I'm not embarrassed by it."
— Paul Gigot, Wall Street Journal opinion editor. The WSJ opinion editor backed Druckenmiller completely: "AI is a fact of modern life... no one can doubt that his op-ed is his genuine opinion." The Journal's position, clear as day: what matters is the sincerity of the ideas, not the unique human flavor of the words expressing them.
Bloomberg's Joe Weisenthal noted that pre-AI, plenty of high-profile op-eds were entirely written by underlings without controversy. Journalist Sharon Goldman confirmed it from the inside: from 2015 to 2020 she ghostwrote dozens of executive op-eds off one conversation and some notes, never wanting a byline.
Jason Calacanis called publishing AI-written pieces without a first-sentence disclosure "unforgivable for a public figure." His All In co-host Chamath Palihapitiya bit back: you don't disclose every article that shaped your opinion — "If he puts his name behind it, it's his opinion." Others predicted the whole debate is "boomer coded" and nobody will care in two years.
Spectra Markets' Brent Donnelly framed it best: is writing art, or functional communication like code? "There's no right answer." The pro-AI camp says the value is in the thinking, not the words; the nuanced camp says AI-assisted writing is fine when it reflects real thinking — and deserves mockery when it substitutes for it.
The sharpest critique conceded that Druckenmiller and Gigot are "100% true descriptively." The real issue: publishing a visibly low-effort AI draft expresses contempt for the op-ed format itself and for its audience. If you couldn't be bothered to make your argument convincingly, why should readers care about your argument?
CNBC's Deirdre Bosa revived Balaji's framing: the beginning (what's the idea, what are you actually trying to say?) needs a human; the middle (research, organize, draft, tighten, poke holes) is AI territory; the end (is this any good? true? do I buy it?) needs a human again. As Balaji put it: "AI doesn't do end-to-end. It does middle-to-middle. The new bottlenecks are prompting and verifying."
OpenAI's recently released enterprise data shows the preponderance of ChatGPT usage across almost every department is writing — not just comms, but recruiting, marketing, customer support, legal, sales, and policy. AI writing is here and it's here to stay.
An email is not a strategy memo, a strategy memo is not a LinkedIn post, and a LinkedIn post is not an op-ed. They're trying to achieve different things, so how you engage AI around each will differ — which makes "can or should AI write this" mostly the wrong question.
The people saying this whole conversation will feel quaint in a couple of years are probably right. But the increased easiness of the inputs of writing will increase the burden on the outputs even more — accepting that people use AI to write doesn't mean accepting bad writing.
The backlash to Druckenmiller wasn't really about prose quality. The AI-isms were so glaring and so fixable that it seemed like he didn't care enough to change the syntax on a couple of sentences — and if he didn't care to do that, how much thought went into the argument underneath?
We have the phrase "work slop" because models have been better at saying a lot than saying the right thing succinctly. Great writers always knew this — Pascal in the 17th century: "I've made this longer than usual because I have not had time to make it shorter." The first wave of AI writing was defined by eye-bleeding length; the next generation won't be.
Deciding your thesis, supporting arguments, and narrative construction is an exercise whose real product is the substance underneath the words. Using AI doesn't automatically mean outsourcing your thinking, but the risk has to hover in your field of view because it's so easy to fall into. NLW's own example: he built this episode's five rules manually in Notion rather than handing the concept to Claude — slower, but right for this context.
No one has ever judged the vast majority of emails on prose quality, and emails are such small units of thought that the intellectual-outsourcing risk is low. But for interactive, Slack-style email, AI may be overkill — dictation tools like Whisperflow can speed you up more effectively.
Summarization is compression, not composition. The failure mode is an AI that's great at summarizing what everyone said but bad at surfacing the one or two things that actually mattered. Outsource the summary, then come in over the top with your own clarification of what's important.
Unless you're hyper-precise about the strategy, AI goes off the rails — adding its own inventions or generic training-data advice instead of your organization's actual context. The fix is splitting the work: exhaustive personal thinking (with AI supporting the bulleting, outlining, and iteration), then letting AI write the words only once the outline is tight.
AI handles an old-school one-line X post better than a full LinkedIn essay, which has more room for AI-isms to shine through. The bigger problem: social media no longer rewards pithy writing — it demands engagement, commenting, and human signal that automated posting can't provide.
He built a full pipeline from episode disaggregation on aidailybrief.ai to generated X and LinkedIn posts — and it worked, increasing sharing versus doing zero social before. But within weeks it slammed against the walls of what AI writing can do for social. The lesson: realistic expectations.
Marketing wants uniqueness and distinction, so the cost of common AI-isms is highest here. The maddening failure mode: tell the model "stop presuming so much about the reader" and the next draft includes a headline reading "We assume nothing about the reader." The better use for now is rapid iteration — ask for 20 taglines to spark your own.
The whole point of an op-ed is to convince — and when readers can instantly tell it was low effort, they switch off. Get extremely clear on the thesis and supporting points before turning it over to AI ("the five-paragraph essay from grade school, baby"), then go back through and scrub the incredibly obvious AI-isms.
AI writing isn't going anywhere, and it will unlock opportunity — people who never used writing for communication will start to. What it won't change: there are no shortcuts for doing things well, even with tools that help us do things well faster and better.