Tyler Cowen
Economist, author and blogger
Where to hear and read about Tyler Cowen3 episodes and tests
How I WriteIs AI Killing Writing? 12 Different Answers
How I WriteHow to Write with AI (Tyler Cowen Breaks it Down)
How I WriteHow to Teach Millions of People with Your Blog — Tyler Cowen & Alex TabarrokTyler Cowen is an economist and author who has co-written the Marginal Revolution blog with Alex Tabarrok every day since 2003, alongside a shared textbook and free videos, and who has become one of the most practical heavy users of AI among writers. The thing most worth learning from him is his discipline about what only he can do: read enormously, publish daily, keep his voice his own, and use AI hard as a reading partner and critic but never as the author.
What you're writing had better be more interesting than the question box.
How to Write with AI (Tyler Cowen Breaks it Down)
Tyler Cowen: the story and the lessons
1How Tyler Cowen built Marginal Revolution
Cowen told Tabarrok to start a blog before writing their textbook. Tabarrok proposed a textbook; Cowen said to write a blog first. The first post ran in August 2003, before Twitter and the iPhone, and passed 10,000 visitors within weeks.
The early posts were practice, and they did not know anyone was reading. The two still tell the founding story differently.
He hoped for 5,000 academic readers and found a far wider audience. The blog now has 50,000 to 100,000 readers. Cowen had pictured about two-thirds academics and was surprised how much people outside academia care about ideas.
Students now arrive at George Mason saying they have read it since age 12, and some are getting PhDs.
Blog, textbook and free videos work as one product line. Years of explaining ideas on the blog improved the textbook, and the textbook sends readers back. Marginal Revolution University is free on YouTube with no ads and aimed to be in 25% of American high schools by 2025.
The videos began as a narrated PowerPoint made with a $5 iPad app and grew with funding. Cowen admits he is not natural on camera.
The audience is increasingly abroad, especially India. India is the number two market and likely to become number one; strangers there greet Cowen as "the MRU guy." Google's automatic subtitles let students learn English and economics at once.
A built audience lifts every later project. Cowen says the blog may never be the main thing, but it makes other ventures succeed faster. Twenty years of small posts led to things like Fast Grants and the COVID response.
Emergent Ventures applicants have to put up with reading the blog, which works as a filter.
He chose writing because he finds it fun, and most people don't. Cowen treats enjoying the work as a competitive advantage. They are not paid at the margin and could earn more elsewhere, so it has to stay fun.
He met his mentor Walter Grinder at 14, and seeing someone who had read so many books made that life seem real.
2Tyler Cowen's rules for publishing every day
Set a publishing rule and you will meet it. Cowen and Tabarrok have not missed a day in almost 20 years, with four to five posts essentially every day.
Cowen likens daily posting to a radio station readers can turn on, and warns of slippery slopes if you relax the rule.
He keeps a stock of evergreen posts for dry days. A few posts that do not depend on the news cycle sit ready, and posts go out on autopublish at odd hours.
He says he has never been upset about how he would fill a day.
He kills drafts freely and rewrites from scratch. Cowen drops posts he thinks are wrong, stale or on topics he is sick of, and has about 2,500 pages of unpublished book material.
Tabarrok learned this when a chapter Cowen shared came back weeks later completely different. Cowen also holds back some good posts because it is better to leave people wanting more than to pander.
Short posts are worth the crafting time. With so many readers, shaving a second or two off each post is a social good. They call the format "dim sum for the mind."
Readers are one click from leaving, so the internet taught them to get to the point fast.
Cowen refuses to look at analytics. He believes that once you know what gets hits you start chasing it, even subconsciously, so he picks topics he is intrinsically interested in.
He leaves openings for readers to push back. A post arguing Shaquille O'Neal might be the greatest player ever became one of the most read, because everyone has an opinion on rankings.
Cowen says a big reason people read online is to take the writer down a peg, so you have to separate ego from the exchange. Expert economics posts draw fewer comments.
3How Tyler Cowen thinks and gets ideas
Cowen takes in information at a pace people don't believe. He reads four or five books a day. Tabarrok, as his research assistant, was sent to fetch ten library books and return them the next day.
Later Cowen read Tabarrok's paper on the spot and cited Arrow 1958. Cowen says this intake is what keeps a daily writer from repeating himself.
His ideas come from dialogue, not solitary thinking. Cowen says he is bad at thinking up ideas in the shower and needs interaction with a text or, better, a person.
Reader comments and direct email extend that conversation; he finds about 80% of his email smart enough to answer. Comments, he notes, are good for facts and bad for theory.
He writes high context on purpose, as a filter. Cowen leaves room for interpretation and wants the blog hard enough to read that it selects better readers, bringing better email and a better network.
He calls this maybe the biggest plus of the blog. "Context is that which is scarce."
Readers remember how you think, not one clever point. Cowen says no one writes in about one brilliant point on corporate tax; the sum of posts showing how you reason draws the best inbound.
He wants to embody the life of an infovore and put it out unashamedly.
Writers should edit more and indulge their ego less. Much of the blog is curating and amplifying others' ideas. Cowen's father was a professional editor, not a writer.
He worries Substack writers get more personal and emotional with each post, and then boring.
He encircles his biases rather than pretending to drop them. Living in several countries, plus travel and art, humbled him; he wants readers of every politics and religion to feel welcome, not handed a take it or leave it line.
Their refusal to take sides first makes enemies; people hate you more if you are partly on their side.
4How Cowen and Alex Tabarrok collaborate
Cowen starts complex and cuts back; Tabarrok starts stark and adds. They meet in the middle, which makes their textbook both rich and clear.
Tabarrok's joke: if a post confuses you it is Tyler's; if you understand it and are furious, it is Alex's.
Cowen imagines Tabarrok reading as he writes. For the textbook he thinks "Alex is going to read that" and simplifies. They write every chapter themselves, with no junior authors.
Each chapter still goes to about 40 referees. Cowen compares the integrated result to Seinfeld being on every part of his show.
Each returns shared drafts fast so both keep working. Doing the collaborative piece first means the partner is busy while you do your own work, roughly doubling output.
Tabarrok's rule is to do a task now or say no, because no one is less busy in three weeks.
They avoid conflict by detaching ego from outcomes. They argue constantly but focus on the right answer, and Cowen says true partnership is rarer than people who can be bossed.
Cowen sometimes writes in Tabarrok's voice on purpose. His pandemic post on giving first vaccine doses first was a deliberate homage, and Tabarrok called it one of Cowen's best.
5How Tyler Cowen uses AI to read and learn
Cowen treats AI as the new secondary literature. Rereading Richard II or Wuthering Heights, he asks the model what happens in each chapter and what puzzles to ponder. He now buys two or three books on a topic instead of 20 to 30.
Prepping for podcasts with historian Helen Castor this way lets him do many more episodes.
People who find AI mid ask questions too general. Cowen says they will not put in their own time generating context. He keeps hitting it with facts, questions and interpretations.
Asking what to ask a guest is, to him, the worst question; asking what was special about Wycliffe's Bible translation gives real material.
His first prompt is long and dictated; follow-ups stay short. He speaks for a minute and a half to three minutes to give the model context, then plans short follow-ups instead of one giant prompt.
His rule is to prompt as if speaking to an alien or a non-human animal, which forces literalness. He says the hard part is emotional, not cognitive.
He matches models to jobs and stays on the frontier. He cites o1 Pro for queries, deep research for 10-page reports, Claude for writing, Gemini for thick files and Perplexity as his Google replacement. The top model costs $200 a month.
He uses DeepSeek for colorful, emotional prose but not research. To keep current he asks Perplexity what came out last month and spends 15 minutes trying it.
He double-checks AI like any other source. Cowen says hallucinations in the best models fell more than 10x in a year, and they matter less when he is the questioner.
For each book, his research assistant checks every claim, told to "act like you're out to destroy me."
Being non-technical is an advantage for him. He focuses on what AI is good for, not how it works. His book Average Is Over predicted AI's rise from his chess intuition.
He requires his students to use AI and report how, and calls it a scandal that other teachers do not.
6Why Tyler Cowen never lets AI write his words
Cowen writes every word himself, even where AI is better. He calls his writing "my little baby" and would refuse AI text even if his outlets allowed it.
He says if the world stopped reading him and read only AI, he would be at peace.
He wants to stay weird, so he refuses AI smoothing. Smoothing would make his prose easier to read, but his cryptic, parable-like sentences are part of the appeal.
Plain AI output, he says, sounds like a denser and smarter Wikipedia.
He uses AI for background, then writes without pasting it. For a column on declassifying documents he asked o1 Pro for legal background, used none of its text, and felt "not an idiot on the topic."
He asks AI what readers will find obnoxious. On Agnes Callard's suggestion, he ran a draft through AI; it flagged a supercilious, condescending passage and he changed it.
He had considered keeping the passage, and now sees the AI's verdict as a kind of wisdom.
7What Tyler Cowen expects AI to do to writing
AI may turn books into a question box. Cowen expects fewer AI-written books and more people simply asking AI what they would have read in one, so your work must beat the box.
He expects to write fewer books and may stop after his book Mentors.
Books on the near future no longer make sense. A book takes about four years to reach readers, so he covers fast topics like AI through daily blogging and Substack and picks book topics frozen in history.
A history of AI works; a predictive book on AI does not.
Generic corporate writing goes first. He compares it to falling jobs for certain programmers. Memoir, fieldwork and lived experience are the safest categories.
Mentors passes his test because only someone who has been both mentor and mentee can write it credibly. AI cannot interview the high school teacher.
Writers need to personalize more. He agrees a line like eating lobster on a Kerala houseboat is something no AI could supply.
On economics questions he says AI already beats him, and "I'm not going to get sentimental about it."
He writes partly for AI readers now. Having blogged daily for 22 years and written 16 or 17 books, he calls AIs his most sympathetic and best-informed readers and fills in fewer blanks.
He plans about 20 blog posts, such as on his years in Fall River, Massachusetts, so a future AI can write his biography at almost no cost.
He is putting more effort into video. His podcasts now always get a video version, since audiences will want real humans for a while.
His sister could not tell an AI paragraph of his voice from the real one.
8Tyler Cowen on networks, secrets and careers after AI
When everyone has the same AI, your network matters more. Getting things done still means raising money and hiring, so Cowen thinks a good network could be 50x more valuable. He now travels and meets people far more.
His two universal pieces of advice: get more and better mentors, and keep improving your peer network.
Secrets gain value as public information loses it. Using Peter Thiel's sense of a secret, Cowen says skill at trading secrets with friends makes you more productive.
Mentoring works through example more than advice. Cowen hopes some part of how you are rubs off on mentees.
Set career paths are fading and risk is rising. He expects fewer PhDs and more people who manage AIs, with some commanding armies of agents.
He says nervousness about AI's disruption is objectively the correct view, even as he expects large gains.
Is AI Killing Writing? 12 Different Answers
Twelve writers disagree about whether AI belongs anywhere near a draft, but most agree the thinking, the reading and the voice have to stay human. The most useful ideas are specific: use AI as a hostile critic, as secondary literature and as a source of a bad first draft, never as the author.
Cowen treats AI as the new secondary literature. Rereading Richard II or Wuthering Heights, he asks the model what happens in each chapter and what puzzles to ponder. He now buys two or three books on a topic instead of 20 to 30.
Prepping for podcasts with historian Helen Castor this way lets him do many more episodes.
People who find AI mid ask questions too general. Cowen says they will not put in their own time generating context. He keeps hitting it with facts, questions and interpretations.
Asking what to ask a guest is, to him, the worst question; asking what was special about Wycliffe's Bible translation gives real material.
Cowen writes every word himself, even where AI is better. He calls his writing "my little baby" and would refuse AI text even if his outlets allowed it.
He says if the world stopped reading him and read only AI, he would be at peace.
He uses AI for background, then writes without pasting it. For a column on declassifying documents he asked o1 Pro for legal background, used none of its text, and felt "not an idiot on the topic."
AI may turn books into a question box. Cowen expects fewer AI-written books and more people simply asking AI what they would have read in one, so your work must beat the box.
He expects to write fewer books and may stop after his book Mentors.
Books on the near future no longer make sense. A book takes about four years to reach readers, so he covers fast topics like AI through daily blogging and Substack and picks book topics frozen in history.
A history of AI works; a predictive book on AI does not.
Generic corporate writing goes first. He compares it to falling jobs for certain programmers. Memoir, fieldwork and lived experience are the safest categories.
Mentors passes his test because only someone who has been both mentor and mentee can write it credibly. AI cannot interview the high school teacher.
How to Write with AI (Tyler Cowen Breaks it Down)
Tyler Cowen explains how he uses AI as a reading companion and critic while refusing to let it write for him, so his work stays weird and recognisably his. The technique worth hearing is his habit of feeding AI long context and specific questions, then writing only what a human with real experience could write.
Cowen treats AI as the new secondary literature. Rereading Richard II or Wuthering Heights, he asks the model what happens in each chapter and what puzzles to ponder. He now buys two or three books on a topic instead of 20 to 30.
Prepping for podcasts with historian Helen Castor this way lets him do many more episodes.
People who find AI mid ask questions too general. Cowen says they will not put in their own time generating context. He keeps hitting it with facts, questions and interpretations.
Asking what to ask a guest is, to him, the worst question; asking what was special about Wycliffe's Bible translation gives real material.
His first prompt is long and dictated; follow-ups stay short. He speaks for a minute and a half to three minutes to give the model context, then plans short follow-ups instead of one giant prompt.
His rule is to prompt as if speaking to an alien or a non-human animal, which forces literalness. He says the hard part is emotional, not cognitive.
He matches models to jobs and stays on the frontier. He cites o1 Pro for queries, deep research for 10-page reports, Claude for writing, Gemini for thick files and Perplexity as his Google replacement. The top model costs $200 a month.
He uses DeepSeek for colorful, emotional prose but not research. To keep current he asks Perplexity what came out last month and spends 15 minutes trying it.
He double-checks AI like any other source. Cowen says hallucinations in the best models fell more than 10x in a year, and they matter less when he is the questioner.
For each book, his research assistant checks every claim, told to "act like you're out to destroy me."
Being non-technical is an advantage for him. He focuses on what AI is good for, not how it works. His book Average Is Over predicted AI's rise from his chess intuition.
He requires his students to use AI and report how, and calls it a scandal that other teachers do not.
Cowen writes every word himself, even where AI is better. He calls his writing "my little baby" and would refuse AI text even if his outlets allowed it.
He says if the world stopped reading him and read only AI, he would be at peace.
He wants to stay weird, so he refuses AI smoothing. Smoothing would make his prose easier to read, but his cryptic, parable-like sentences are part of the appeal.
Plain AI output, he says, sounds like a denser and smarter Wikipedia.
He uses AI for background, then writes without pasting it. For a column on declassifying documents he asked o1 Pro for legal background, used none of its text, and felt "not an idiot on the topic."
He asks AI what readers will find obnoxious. On Agnes Callard's suggestion, he ran a draft through AI; it flagged a supercilious, condescending passage and he changed it.
He had considered keeping the passage, and now sees the AI's verdict as a kind of wisdom.
AI may turn books into a question box. Cowen expects fewer AI-written books and more people simply asking AI what they would have read in one, so your work must beat the box.
He expects to write fewer books and may stop after his book Mentors.
Books on the near future no longer make sense. A book takes about four years to reach readers, so he covers fast topics like AI through daily blogging and Substack and picks book topics frozen in history.
A history of AI works; a predictive book on AI does not.
Generic corporate writing goes first. He compares it to falling jobs for certain programmers. Memoir, fieldwork and lived experience are the safest categories.
Mentors passes his test because only someone who has been both mentor and mentee can write it credibly. AI cannot interview the high school teacher.
Writers need to personalize more. He agrees a line like eating lobster on a Kerala houseboat is something no AI could supply.
On economics questions he says AI already beats him, and "I'm not going to get sentimental about it."
He writes partly for AI readers now. Having blogged daily for 22 years and written 16 or 17 books, he calls AIs his most sympathetic and best-informed readers and fills in fewer blanks.
He plans about 20 blog posts, such as on his years in Fall River, Massachusetts, so a future AI can write his biography at almost no cost.
He is putting more effort into video. His podcasts now always get a video version, since audiences will want real humans for a while.
His sister could not tell an AI paragraph of his voice from the real one.
When everyone has the same AI, your network matters more. Getting things done still means raising money and hiring, so Cowen thinks a good network could be 50x more valuable. He now travels and meets people far more.
His two universal pieces of advice: get more and better mentors, and keep improving your peer network.
Secrets gain value as public information loses it. Using Peter Thiel's sense of a secret, Cowen says skill at trading secrets with friends makes you more productive.
Mentoring works through example more than advice. Cowen hopes some part of how you are rubs off on mentees.
Set career paths are fading and risk is rising. He expects fewer PhDs and more people who manage AIs, with some commanding armies of agents.
He says nervousness about AI's disruption is objectively the correct view, even as he expects large gains.
How to Teach Millions of People with Your Blog — Tyler Cowen & Alex Tabarrok
Two economists who have blogged together every day for over 20 years explain how they write short, clear pieces for busy readers, and how their different instincts (one starts complex, one starts simple) make both better. The habits worth hearing are a daily publishing rule, killing drafts freely, and getting ideas from conversation rather than solitary thinking.
Cowen told Tabarrok to start a blog before writing their textbook. Tabarrok proposed a textbook; Cowen said to write a blog first. The first post ran in August 2003, before Twitter and the iPhone, and passed 10,000 visitors within weeks.
The early posts were practice, and they did not know anyone was reading. The two still tell the founding story differently.
He hoped for 5,000 academic readers and found a far wider audience. The blog now has 50,000 to 100,000 readers. Cowen had pictured about two-thirds academics and was surprised how much people outside academia care about ideas.
Students now arrive at George Mason saying they have read it since age 12, and some are getting PhDs.
Blog, textbook and free videos work as one product line. Years of explaining ideas on the blog improved the textbook, and the textbook sends readers back. Marginal Revolution University is free on YouTube with no ads and aimed to be in 25% of American high schools by 2025.
The videos began as a narrated PowerPoint made with a $5 iPad app and grew with funding. Cowen admits he is not natural on camera.
The audience is increasingly abroad, especially India. India is the number two market and likely to become number one; strangers there greet Cowen as "the MRU guy." Google's automatic subtitles let students learn English and economics at once.
A built audience lifts every later project. Cowen says the blog may never be the main thing, but it makes other ventures succeed faster. Twenty years of small posts led to things like Fast Grants and the COVID response.
Emergent Ventures applicants have to put up with reading the blog, which works as a filter.
He chose writing because he finds it fun, and most people don't. Cowen treats enjoying the work as a competitive advantage. They are not paid at the margin and could earn more elsewhere, so it has to stay fun.
He met his mentor Walter Grinder at 14, and seeing someone who had read so many books made that life seem real.
Set a publishing rule and you will meet it. Cowen and Tabarrok have not missed a day in almost 20 years, with four to five posts essentially every day.
Cowen likens daily posting to a radio station readers can turn on, and warns of slippery slopes if you relax the rule.
He keeps a stock of evergreen posts for dry days. A few posts that do not depend on the news cycle sit ready, and posts go out on autopublish at odd hours.
He says he has never been upset about how he would fill a day.
He kills drafts freely and rewrites from scratch. Cowen drops posts he thinks are wrong, stale or on topics he is sick of, and has about 2,500 pages of unpublished book material.
Tabarrok learned this when a chapter Cowen shared came back weeks later completely different. Cowen also holds back some good posts because it is better to leave people wanting more than to pander.
Short posts are worth the crafting time. With so many readers, shaving a second or two off each post is a social good. They call the format "dim sum for the mind."
Readers are one click from leaving, so the internet taught them to get to the point fast.
Cowen refuses to look at analytics. He believes that once you know what gets hits you start chasing it, even subconsciously, so he picks topics he is intrinsically interested in.
He leaves openings for readers to push back. A post arguing Shaquille O'Neal might be the greatest player ever became one of the most read, because everyone has an opinion on rankings.
Cowen says a big reason people read online is to take the writer down a peg, so you have to separate ego from the exchange. Expert economics posts draw fewer comments.
Cowen takes in information at a pace people don't believe. He reads four or five books a day. Tabarrok, as his research assistant, was sent to fetch ten library books and return them the next day.
Later Cowen read Tabarrok's paper on the spot and cited Arrow 1958. Cowen says this intake is what keeps a daily writer from repeating himself.
His ideas come from dialogue, not solitary thinking. Cowen says he is bad at thinking up ideas in the shower and needs interaction with a text or, better, a person.
Reader comments and direct email extend that conversation; he finds about 80% of his email smart enough to answer. Comments, he notes, are good for facts and bad for theory.
He writes high context on purpose, as a filter. Cowen leaves room for interpretation and wants the blog hard enough to read that it selects better readers, bringing better email and a better network.
He calls this maybe the biggest plus of the blog. "Context is that which is scarce."
Readers remember how you think, not one clever point. Cowen says no one writes in about one brilliant point on corporate tax; the sum of posts showing how you reason draws the best inbound.
He wants to embody the life of an infovore and put it out unashamedly.
Writers should edit more and indulge their ego less. Much of the blog is curating and amplifying others' ideas. Cowen's father was a professional editor, not a writer.
He worries Substack writers get more personal and emotional with each post, and then boring.
He encircles his biases rather than pretending to drop them. Living in several countries, plus travel and art, humbled him; he wants readers of every politics and religion to feel welcome, not handed a take it or leave it line.
Their refusal to take sides first makes enemies; people hate you more if you are partly on their side.
Cowen starts complex and cuts back; Tabarrok starts stark and adds. They meet in the middle, which makes their textbook both rich and clear.
Tabarrok's joke: if a post confuses you it is Tyler's; if you understand it and are furious, it is Alex's.
Cowen imagines Tabarrok reading as he writes. For the textbook he thinks "Alex is going to read that" and simplifies. They write every chapter themselves, with no junior authors.
Each chapter still goes to about 40 referees. Cowen compares the integrated result to Seinfeld being on every part of his show.
Each returns shared drafts fast so both keep working. Doing the collaborative piece first means the partner is busy while you do your own work, roughly doubling output.
Tabarrok's rule is to do a task now or say no, because no one is less busy in three weeks.
They avoid conflict by detaching ego from outcomes. They argue constantly but focus on the right answer, and Cowen says true partnership is rarer than people who can be bossed.
Cowen sometimes writes in Tabarrok's voice on purpose. His pandemic post on giving first vaccine doses first was a deliberate homage, and Tabarrok called it one of Cowen's best.