Basically: Don't crystallize too early, have a primordial soup of notes that you coagulate/congeal bit by bit. Take little iterative steps on local slices, don't try to construct the final product from the get-go.
This method came quite naturally to me during my writing-based personal projects since I have no deadline or anything and am literally just collecting thousands of little A6/A7 notes that I capture as they pop into my head. I can take all the time I want to stew on them and have a structure bubbling up all on its own.
I suppose some of the beauty of this is that the "notes" are easy. Whereas writing is hard. So the more you can leverage these fragments, the better.
I wonder how the non-fiction element plays into it. I guess if you have a lot of fragments the hard part is organizing and then glueing them together. But in non fiction a lot of this is done or you. You are given the order and a lot of the glue. The glue you are missing is probably relatively obvious ("why did they do this?") and you can get that information and include it, or you cant and there is nothing to be done about it ("the motive is unknown" etc).
Whereas in fiction these are all unknown. You have to decide. And you have to make it compelling, and believable, and maybe astonishing and otherworldly at the same time.
In some limited experience in fictional writing, this is the hard part. I have all these fragments (this happened, this happens, there is this dynamic with 2 characters, some broad themes I want to hit on, etc). Makes me think about if non-fiction stories could be used as a sort of "seed" for the glue. I feel like the content is pretty highly coupled to that sort of pattern though.
> But in non fiction a lot of this is done or you. You are given the order and a lot of the glue. The glue you are missing is probably relatively obvious ("why did they do this?") and you can get that information and include it, or you cant and there is nothing to be done about it ("the motive is unknown" etc).
I published a long time ago and ended up doing some deep reporting on a couple of issues. And honestly, it’s a lot closer to writing fiction than I would have thought going in. In ways, it’s actually a lot harder.
Starting out, I believed roughly what your quote said - that I could get into a deep piece and all the glue would be there. In practice, it’s a lot messier because you try to tell the truth related by witnesses with unpredictable self interests. So as you go through the process, you find multiple truths that can each be corroborated with other sources.
It’s a lot more like building with Lego. A piece could be the nose of a horse or the gutter of a house. It’s where judgment, ethics and good note taking practices come in. Witness problems are a big reason why journalists develop their own note taking style to code interviews - an outlining style gives the freedom to consider each fact as a piece of a larger build.
Edit - If it makes you feel better about writing fiction, I wouldn’t worry so much about plausibility. I wasn’t a great journalist but have heard (and verified) stories that the most creative minds would call impossible. Good characters can make the implausible into magic.
If there were another writer of non-fiction as deeply researched I'd compare McPhee to, it would be Robert Caro. I already knew from Caro's memoir Working that Caro did not use a tape recorder in his interviews with subjects, and from this article about McPhee's method, I learned that McPhee does not either. I'm a bit surprised: I'd have thought for such deep research one would want a recording to refer back to, but both seem to feel that the drawbacks of influencing their subjects outweigh their benefit.
McPhee was recommended as someone whose writing "makes boring things interesting". I did enjoy The Curve of Binding Energy (nuclear science) and to some extent Coming out of the Country (Alaska). Both of those featured interesting vignettes and colorful characters which propelled along the narrative.
However, I then turned to his magnum opus on geology, Annals of the Former World. That was a long slog which, although I enjoyed moments of it, now I wonder if my time wouldn't have been better spent reading something more interesting.
I own every book McPhee published and have read each one at least twice. He is, without question, the finest writer of non-fiction I know. Annals, as you may know, was originally published as 4 separate volumes, each covering a particular US region. Assembling California is my absolute favorite McPhee work. I have a layman’s interest in geology and plate tectonics that I developed specifically because of this book.
> I should say that in those nightly encounters I rarely go straight from notes to full sentences. I bridge the gap with a kind of pseudowriting. In effect it’s an outline that takes the form of paragraphs, mixed in with the raw notes. I write good full sentences, broken half-sentences, non-sentences that also include todos in them.
Back when i wrote a lot, i often did something like this. I used a program called Acta [1], which was described as an outliner. Basically a text editor where paragraphs form a tree, and nodes can be collapsed. I'd start with the high-level structure and recursively flesh it out, and just keep doing that until i had nodes at the level of individual sentences. Then i'd just write a sentence for each node, and concatenate them. Often i'd take a shortcut and just write a whole paragraph. But a lot of the time, i wouldn't - a difficult passage becomes easy when you're incrementally breaking it down, reordering it, revising it, etc.
> McPhee usually had one person at the center of each piece, so he would aim to spend a lot of time with that person … stay at their cottage for a season
Even back when every household received a morning paper I cannot fathom how a single article could command such a high pay.
For a magazine like the New Yorker, there was money. You might be interested in Bryan Burrough's experience writing for Vanity Fair in the 90s and 2000s.
> For twenty-five years, I was contracted to produce three articles a year, long ones, typically ten thousand words. For this, my peak salary was $498,141. That’s not a misprint—$498,141, or more than $166,000 per story. Then, as now, $166,000 was a good advance for an entire book. Yes, I realized it was obscene. I took it with a grin.
> Even back when every household received a morning paper I cannot fathom how a single article could command such a high pay.
He wrote for the New Yorker, which is a magazine rather than a newspaper. The number of long-form literary nonfiction pieces that the New Yorker runs every year is drastically fewer than the number of news articles produced to fill a daily newspaper in just a couple weeks.
Roger Zelazny, the science-fiction author, relates a story of how he wrote one story that fulfilled the briefs of three anthologies/magazines, sold it three times, and it was enough to pay for a cruise.
This is such a great piece and I admire Somers Atlantic writing. I just want to say, this process feels similar to the process Michael Lewis used in his later books, particular with "Going Infinite", where he embedded deeply and for a long time with FTX, emerging to write just as the story shifted from FTX legitimizing the world of cryptocurrency to becoming its chief villain.
So I do wonder, when you dig in that deeply before coming close to getting any work product, how do you ensure that you'll be able to cut your losses, as opposed to what Lewis did, trying (and failing) to salvage a narrative that had been demolished by events and other reporting?
My immediate reaction is to notice that this method is actually closely mirrors Barbara Minto’s Pyramid Principle, a go-to tool for McKinsey consultants for the last 40 years or so. One way to build so-called Pyramids is to go bottom up: gather raw facts, aggregate them into themes, themes align into arguments, and only then does the polished narrative (or the key idea) emerge. As far as I can tell, McPhee recommends a very similar approach.
What I didn’t realize until a few months ago about writing, James summarized as
> while “writing a first draft” is intimidating, “reading through all your notes” or moving note cards around on a table, contemplating structure, is not. In fact these tasks are kind of delightful.
John McPhee is a treasure. If you haven't read any of his work, I would. And if you don't want to dive into a full book, he has a number of collections of short stories.
It's the "too much information" style of writing. The subject has to be worth it. I've read his "Basin and Range" (geology), "The Deltoid Pumpkin Seed" (aircraft design), and "The Curve of Binding Energy" (nuclear weapons). Only the last is worth reading. He went off on a years-long geology tangent, interesting only if you're really into geology.
William Langewiesche had a similar style. He died recently, and leaves behind some good writing on aviation. He was the son of Wolfgang Langewiesche, known for writing "Stick and Rudder", a very well written book for pilots on how airplanes really work.
The interesting part is not so much that it’s KEDIT (not that surprising, when these things get mentioned it’s kedit, one of the other xedit editors, xywrite, or wordstar) but that he had someone write note management macros for it with what sounds like a really idiosyncratic workflow.
This is about journalism, in an age when journalism is being bulldozed by cheap content, AI hallucinations and other clickbait tactics. Journalism still has the idea of finding facts, things that correspond to something in reality.
It’s been shown that people in fact prefer to have their beliefs confirmed, rather than challenged by investigative reporting, hence the decline in real journalism, because it is both more difficult and in the end less popular across a large audience.
I disagree with the cynical take in your second sentence.
I think people prefer to be right. Any animal that wants to walk the Earth successfully has an intrinsic need to understand the world as it is, not as they wish it to be. Wanting the fruit tree to be right outside the cave doesn't make it so.
But that desire to have correct beliefs is complex in humans where we also have beliefs about ourselves and others have beliefs about us. We don't want to just be right, but believe ourselves good at being right, and be known to others as someone who is usually right.
That puts some hysteresis in our belief system where we have a tendency to disregard evidence that goes against out beliefs for a while until it's sufficiently compelling.
But, also, no one wants to be a gullible rube or an out-of-touch idiot. We will update our beliefs when in a social and information environment that permits us to do so. The problem today is largely that we have so much choice of our social and information environment, then it's much easier to select an echo chamber that lets us think we're already right then be confronted with the reality that we're wrong.
The decline in investigative journalism can, I think, be explained largely in economic terms. The news has always been driven by ads. Even when people paid for papers, the subscription cost didn't fully cover the newspapers' expenses. The classified ads filled that gap.
Then the Internet arose and transferred advertising out of classifieds and into Craigslist, eBay, and others. And it transferred human attention out of newspapers and into social media. There simply weren't enough resources left to fund expensive investigative journalism.
You are right, I am being casual in asserting what “people prefer”.
You assert that people prefer to be right, I agree but in practice “being right” requires having to accept changes in your beliefs, and that’s painful.
Also “being right” requires some effort to check what the facts are, and that is arduous when people are busy with life.
As a result I think people prefer to feel like they’re right rather than dig deeper, and so they prefer information sources that tend to confirm their beliefs.
Behavior makes a lot more sense when you realize that humans are trying to have as accurate a view of the world as possible while also conserving cognitive resources. Running a brain is actually astonishingly expensive in terms of calories, so we've evolved to be judicious in how much thought we put into things.
Changing an existing belief is hard because we already spent effort acquiring the belief in the first place. Throwing that out should be expensive because otherwise we risk thrashing where we are constantly vacillating between competing beliefs. It makes more sense in terms of efficiency and being able to take action if there is some hysteresis and beliefs are sticky.
Note that while people don't change beliefs easily, they do acquire them pretty easily. If I tell you something that doesn't directly conflict with an existing belief, it's easy to absorb.
I enjoyed your thoughts on people preferring to be right but having competing personal interests.
I have started and stopped writing several chains of thought related to it. But it makes me think about what categories of things we may prefer to be right about. And how they do or do not fit into the model you described. For example, empirical (the elephant weights more than a mouse), pragmatic ("hot water is good for your health" in Chinese culture), moral (murder is bad), etc. "Truth", "being right" etc. are just loaded terms I guess. The claims about being able to function effectively seem largely true for the more empirical stuff but the sort of information silos discussed online draw more idealogical lines.
Also made me think how a nihilist or social constructivist might behave or model things differently. it just seems like if you have a more relativist interpretation of truth, being right and believing in something is just the same thing. So it would be literally impossible for those things to be in conflict (either naturally or from the perspective of the person).
This approach sounds very similar to the construction of grounded theory in ethnography/anthropology -- something I've always wanted to practise but never had the patience for!
> This was most of what I knew about nonfiction writing when I managed to land an assignment, on spec, to profile Douglas Hofstadter for a piece in the Atlantic’s print magazine.
I suspect a large contingent here will really hate this suggestion but here it goes:
The McPhee method sounds like a great framework for making writing prompts. That is, prompts for LLMs to write things.
>in stage one he accumulates notes; in stage two he selects them; in stage three he structures them; and in stage four he writes. By the time he is crafting sentences the structure of the piece as a whole, and of each section, even paragraph, and the logic connecting them all, is already determined, thanks to the mechanical work done in the first three stages. McPhee is on rails the whole time he writes his first draft. From there it’s all downhill and the standard thing that everybody does: revision, revision again, then refinement—a sculptor with ax, then knife, then scalpel.
I know hackernews kinda hates LLMs but I don think this idea has to be so offensive. Much of the work and value from the author is in collecting these fragments and structuring them. Purely from a communication standpoint, I have no issues whatsoever with an LLM stitching them together and choosing the vocab and grammar.
OP describes this ^ long arduous process and then notes:
> Your writing can only be as good as your taste.
That is, using an LLM to help with "collecting these fragments and structuring them" might be okay--should the writer still be able to deeply immerse--but the "stitching them together and choosing the vocab and grammar" of a faux writer LLM is likely to leave a bad taste in the mind of some readers (e.g. those HN types who "kinda hates LLMs").
Basically: Don't crystallize too early, have a primordial soup of notes that you coagulate/congeal bit by bit. Take little iterative steps on local slices, don't try to construct the final product from the get-go.
This method came quite naturally to me during my writing-based personal projects since I have no deadline or anything and am literally just collecting thousands of little A6/A7 notes that I capture as they pop into my head. I can take all the time I want to stew on them and have a structure bubbling up all on its own.
I suppose some of the beauty of this is that the "notes" are easy. Whereas writing is hard. So the more you can leverage these fragments, the better.
I wonder how the non-fiction element plays into it. I guess if you have a lot of fragments the hard part is organizing and then glueing them together. But in non fiction a lot of this is done or you. You are given the order and a lot of the glue. The glue you are missing is probably relatively obvious ("why did they do this?") and you can get that information and include it, or you cant and there is nothing to be done about it ("the motive is unknown" etc).
Whereas in fiction these are all unknown. You have to decide. And you have to make it compelling, and believable, and maybe astonishing and otherworldly at the same time.
In some limited experience in fictional writing, this is the hard part. I have all these fragments (this happened, this happens, there is this dynamic with 2 characters, some broad themes I want to hit on, etc). Makes me think about if non-fiction stories could be used as a sort of "seed" for the glue. I feel like the content is pretty highly coupled to that sort of pattern though.
> But in non fiction a lot of this is done or you. You are given the order and a lot of the glue. The glue you are missing is probably relatively obvious ("why did they do this?") and you can get that information and include it, or you cant and there is nothing to be done about it ("the motive is unknown" etc).
I published a long time ago and ended up doing some deep reporting on a couple of issues. And honestly, it’s a lot closer to writing fiction than I would have thought going in. In ways, it’s actually a lot harder.
Starting out, I believed roughly what your quote said - that I could get into a deep piece and all the glue would be there. In practice, it’s a lot messier because you try to tell the truth related by witnesses with unpredictable self interests. So as you go through the process, you find multiple truths that can each be corroborated with other sources.
It’s a lot more like building with Lego. A piece could be the nose of a horse or the gutter of a house. It’s where judgment, ethics and good note taking practices come in. Witness problems are a big reason why journalists develop their own note taking style to code interviews - an outlining style gives the freedom to consider each fact as a piece of a larger build.
Edit - If it makes you feel better about writing fiction, I wouldn’t worry so much about plausibility. I wasn’t a great journalist but have heard (and verified) stories that the most creative minds would call impossible. Good characters can make the implausible into magic.
...and use Emacs. Very important.
If there were another writer of non-fiction as deeply researched I'd compare McPhee to, it would be Robert Caro. I already knew from Caro's memoir Working that Caro did not use a tape recorder in his interviews with subjects, and from this article about McPhee's method, I learned that McPhee does not either. I'm a bit surprised: I'd have thought for such deep research one would want a recording to refer back to, but both seem to feel that the drawbacks of influencing their subjects outweigh their benefit.
McPhee was recommended as someone whose writing "makes boring things interesting". I did enjoy The Curve of Binding Energy (nuclear science) and to some extent Coming out of the Country (Alaska). Both of those featured interesting vignettes and colorful characters which propelled along the narrative.
However, I then turned to his magnum opus on geology, Annals of the Former World. That was a long slog which, although I enjoyed moments of it, now I wonder if my time wouldn't have been better spent reading something more interesting.
I own every book McPhee published and have read each one at least twice. He is, without question, the finest writer of non-fiction I know. Annals, as you may know, was originally published as 4 separate volumes, each covering a particular US region. Assembling California is my absolute favorite McPhee work. I have a layman’s interest in geology and plate tectonics that I developed specifically because of this book.
For me, it’s "The Pine Barrens". It’s boring, but he did make it interesting. For a while, at least.
Sharing a critical opinion? That's a downvote for you! (Sheesh)
> I should say that in those nightly encounters I rarely go straight from notes to full sentences. I bridge the gap with a kind of pseudowriting. In effect it’s an outline that takes the form of paragraphs, mixed in with the raw notes. I write good full sentences, broken half-sentences, non-sentences that also include todos in them.
Back when i wrote a lot, i often did something like this. I used a program called Acta [1], which was described as an outliner. Basically a text editor where paragraphs form a tree, and nodes can be collapsed. I'd start with the high-level structure and recursively flesh it out, and just keep doing that until i had nodes at the level of individual sentences. Then i'd just write a sentence for each node, and concatenate them. Often i'd take a shortcut and just write a whole paragraph. But a lot of the time, i wouldn't - a difficult passage becomes easy when you're incrementally breaking it down, reordering it, revising it, etc.
[1] https://www.a-sharp.com/acta/
Sounds a bit like Gingko, which I used to really love:
https://gingkowriter.com/
> McPhee usually had one person at the center of each piece, so he would aim to spend a lot of time with that person … stay at their cottage for a season
Even back when every household received a morning paper I cannot fathom how a single article could command such a high pay.
For a magazine like the New Yorker, there was money. You might be interested in Bryan Burrough's experience writing for Vanity Fair in the 90s and 2000s.
> For twenty-five years, I was contracted to produce three articles a year, long ones, typically ten thousand words. For this, my peak salary was $498,141. That’s not a misprint—$498,141, or more than $166,000 per story. Then, as now, $166,000 was a good advance for an entire book. Yes, I realized it was obscene. I took it with a grin.
https://yalereview.org/article/burrough-vanity-fair-graydon-...
> Even back when every household received a morning paper I cannot fathom how a single article could command such a high pay.
He wrote for the New Yorker, which is a magazine rather than a newspaper. The number of long-form literary nonfiction pieces that the New Yorker runs every year is drastically fewer than the number of news articles produced to fill a daily newspaper in just a couple weeks.
Roger Zelazny, the science-fiction author, relates a story of how he wrote one story that fulfilled the briefs of three anthologies/magazines, sold it three times, and it was enough to pay for a cruise.
Things were different back in the day.
How do you think Googlers command such high pay?
This is such a great piece and I admire Somers Atlantic writing. I just want to say, this process feels similar to the process Michael Lewis used in his later books, particular with "Going Infinite", where he embedded deeply and for a long time with FTX, emerging to write just as the story shifted from FTX legitimizing the world of cryptocurrency to becoming its chief villain.
So I do wonder, when you dig in that deeply before coming close to getting any work product, how do you ensure that you'll be able to cut your losses, as opposed to what Lewis did, trying (and failing) to salvage a narrative that had been demolished by events and other reporting?
Another lesson you could take away there: It's possible to do this sort of total deep immersion for months, and also completely miss the real story.
My immediate reaction is to notice that this method is actually closely mirrors Barbara Minto’s Pyramid Principle, a go-to tool for McKinsey consultants for the last 40 years or so. One way to build so-called Pyramids is to go bottom up: gather raw facts, aggregate them into themes, themes align into arguments, and only then does the polished narrative (or the key idea) emerge. As far as I can tell, McPhee recommends a very similar approach.
"McPhee has been my model. He's the most elegant of all the journalists writing today, I think."
— Tracy Kidder, author of The Soul of a New Machine and many other great writings.
What I didn’t realize until a few months ago about writing, James summarized as
> while “writing a first draft” is intimidating, “reading through all your notes” or moving note cards around on a table, contemplating structure, is not. In fact these tasks are kind of delightful.
Thanks for sharing.
I absolutely loved this:
> if you tell someone you’re a journalist they’re going to believe you. Your job is to honor their trust.
A lot of life wisdom wedged into those two sentences, I think, that goes beyond journalism.
John McPhee is a treasure. If you haven't read any of his work, I would. And if you don't want to dive into a full book, he has a number of collections of short stories.
It's the "too much information" style of writing. The subject has to be worth it. I've read his "Basin and Range" (geology), "The Deltoid Pumpkin Seed" (aircraft design), and "The Curve of Binding Energy" (nuclear weapons). Only the last is worth reading. He went off on a years-long geology tangent, interesting only if you're really into geology.
William Langewiesche had a similar style. He died recently, and leaves behind some good writing on aviation. He was the son of Wolfgang Langewiesche, known for writing "Stick and Rudder", a very well written book for pilots on how airplanes really work.
Don't miss the discussion of McPhee's text editor. I would love to get more details about that. Catnip for this crowd.
The interesting part is not so much that it’s KEDIT (not that surprising, when these things get mentioned it’s kedit, one of the other xedit editors, xywrite, or wordstar) but that he had someone write note management macros for it with what sounds like a really idiosyncratic workflow.
I suppose Scrivener would be the modern equivalent.
https://www.literatureandlatte.com/scrivener/overview
This is about journalism, in an age when journalism is being bulldozed by cheap content, AI hallucinations and other clickbait tactics. Journalism still has the idea of finding facts, things that correspond to something in reality.
It’s been shown that people in fact prefer to have their beliefs confirmed, rather than challenged by investigative reporting, hence the decline in real journalism, because it is both more difficult and in the end less popular across a large audience.
I disagree with the cynical take in your second sentence.
I think people prefer to be right. Any animal that wants to walk the Earth successfully has an intrinsic need to understand the world as it is, not as they wish it to be. Wanting the fruit tree to be right outside the cave doesn't make it so.
But that desire to have correct beliefs is complex in humans where we also have beliefs about ourselves and others have beliefs about us. We don't want to just be right, but believe ourselves good at being right, and be known to others as someone who is usually right.
That puts some hysteresis in our belief system where we have a tendency to disregard evidence that goes against out beliefs for a while until it's sufficiently compelling.
But, also, no one wants to be a gullible rube or an out-of-touch idiot. We will update our beliefs when in a social and information environment that permits us to do so. The problem today is largely that we have so much choice of our social and information environment, then it's much easier to select an echo chamber that lets us think we're already right then be confronted with the reality that we're wrong.
The decline in investigative journalism can, I think, be explained largely in economic terms. The news has always been driven by ads. Even when people paid for papers, the subscription cost didn't fully cover the newspapers' expenses. The classified ads filled that gap.
Then the Internet arose and transferred advertising out of classifieds and into Craigslist, eBay, and others. And it transferred human attention out of newspapers and into social media. There simply weren't enough resources left to fund expensive investigative journalism.
You are right, I am being casual in asserting what “people prefer”.
You assert that people prefer to be right, I agree but in practice “being right” requires having to accept changes in your beliefs, and that’s painful.
Also “being right” requires some effort to check what the facts are, and that is arduous when people are busy with life.
As a result I think people prefer to feel like they’re right rather than dig deeper, and so they prefer information sources that tend to confirm their beliefs.
Behavior makes a lot more sense when you realize that humans are trying to have as accurate a view of the world as possible while also conserving cognitive resources. Running a brain is actually astonishingly expensive in terms of calories, so we've evolved to be judicious in how much thought we put into things.
Changing an existing belief is hard because we already spent effort acquiring the belief in the first place. Throwing that out should be expensive because otherwise we risk thrashing where we are constantly vacillating between competing beliefs. It makes more sense in terms of efficiency and being able to take action if there is some hysteresis and beliefs are sticky.
Note that while people don't change beliefs easily, they do acquire them pretty easily. If I tell you something that doesn't directly conflict with an existing belief, it's easy to absorb.
I enjoyed your thoughts on people preferring to be right but having competing personal interests.
I have started and stopped writing several chains of thought related to it. But it makes me think about what categories of things we may prefer to be right about. And how they do or do not fit into the model you described. For example, empirical (the elephant weights more than a mouse), pragmatic ("hot water is good for your health" in Chinese culture), moral (murder is bad), etc. "Truth", "being right" etc. are just loaded terms I guess. The claims about being able to function effectively seem largely true for the more empirical stuff but the sort of information silos discussed online draw more idealogical lines.
Also made me think how a nihilist or social constructivist might behave or model things differently. it just seems like if you have a more relativist interpretation of truth, being right and believing in something is just the same thing. So it would be literally impossible for those things to be in conflict (either naturally or from the perspective of the person).
This approach sounds very similar to the construction of grounded theory in ethnography/anthropology -- something I've always wanted to practise but never had the patience for!
This is a beautiful website. The article and homepage.
It looks directly inspired by the design sensibility and aesthetics of Edward Tufte.
> This was most of what I knew about nonfiction writing when I managed to land an assignment, on spec, to profile Douglas Hofstadter for a piece in the Atlantic’s print magazine.
That sounds like an impressively rapid climb.
I suspect a large contingent here will really hate this suggestion but here it goes:
The McPhee method sounds like a great framework for making writing prompts. That is, prompts for LLMs to write things.
>in stage one he accumulates notes; in stage two he selects them; in stage three he structures them; and in stage four he writes. By the time he is crafting sentences the structure of the piece as a whole, and of each section, even paragraph, and the logic connecting them all, is already determined, thanks to the mechanical work done in the first three stages. McPhee is on rails the whole time he writes his first draft. From there it’s all downhill and the standard thing that everybody does: revision, revision again, then refinement—a sculptor with ax, then knife, then scalpel.
I know hackernews kinda hates LLMs but I don think this idea has to be so offensive. Much of the work and value from the author is in collecting these fragments and structuring them. Purely from a communication standpoint, I have no issues whatsoever with an LLM stitching them together and choosing the vocab and grammar.
OP describes this ^ long arduous process and then notes:
> Your writing can only be as good as your taste.
That is, using an LLM to help with "collecting these fragments and structuring them" might be okay--should the writer still be able to deeply immerse--but the "stitching them together and choosing the vocab and grammar" of a faux writer LLM is likely to leave a bad taste in the mind of some readers (e.g. those HN types who "kinda hates LLMs").
This is a well-written, interesting article. The only issue I have is with this:
> He said, if you tell someone you’re a journalist they’re going to believe you. Your job is to honor their trust.
Sadly, that is no longer true. Today's "journalists" far too often see their job as proselytizing rather than reporting.