From my experience with editors as an op-ed writer:
--My most frequent frustration is with the headlines they come up with, missing the main point or, now that online publications often begin with a summary paragraph before linking to the actual text, botching the main points of a piece;
--But, if editors tell me they don't understand a point, I'll concede that the average reader won't either, and I'll add clarifications where needed.
Noah doesn't mention self-editing. I rarely trust my first drafts and will set aside a piece for a few hours or overnight, before reading it with an editor's eye, and then clarifying or reorganizing as needed. As my (old-fashioned) journalism professor used to tell us: Writing is re-writing.
“The Economist has done an amazing job of editing all of its writers to sound like they’re the same person — a wry British know-it-all.”
I like listening to The Economist weekly audio edition as a NPR substitute. It’s not literally a single voice; they employ 3 or 4 readers for each addition but they all come across (enjoyably) as British know-it alls
I encountered the Economist when I was managing a small company in England for the parent company I worked for. I was also looking for the best magazines to keep up with world affairs. The Economist came out at the top of the list, and I found the style delightful - higher-quality writing than that for the average American in US news weeklies, and some classic phrases which remain embedded in my mind, like "form swallows function" (architecture) and "towering edifices of supposition" (theories of universe structure).
The problem was that almost everything in the magazine was worth reading. It came out weekly, but took all week to read it! I just couldn't keep up.
That’s why I listen to the audio edition. I make use of the time when I can’t read. The level of research seems higher than any other general news publication.
Unfortunately, I am constantly building a reference library of links, quotes, and illustrations on my computer. Audio doesn't do it for me. Even the print magazines I get are a problem - when I've found a particularly good article, I have to go to the computer and try to find it to bookmark or quote it.
Great article, Noah. I love the insider viewpoint, and I agree with you. As a consumer of this stuff, I am looking forward to the day where all "volume" op-ed writers are independent. The ones that suck would fall off the edge of the cliff while the good ones would prosper like you have. It would also be better for the reader. I like reading Paul Krugman, David Brooks, and Russ Duhat in the Times, but I would like reading them much better on Substack, for the reasons you mention. I'm sure others like their other writers, too, and they would do fine independently. That would leave the Op-Ed section to guest writers like elected officials, Oprah, CEO's, candidates, disgraced former RNC committee chairs or whomever that is involved with a story to provide a unique viewpoint that a journalist may not be able to provide. I'd rather the Times focus on providing me all the news fit to print, accurately and objectively. It may still be the battering ram of the national liberal conversation, but it would be less in-your-face about it. Thanks for what you do, Noah. I read you and Matt regularly over my morning coffee. If you run into Professor Krugman, please encourage him to jump ship.
Facts are snippets we take out of context to support our narrative. Usually this involves discarding or ignoring a greater number snippets that might weaken the narrative.
The universe of facts is very large while the selection of narratives on offer is very limited. Substack helps expand the selection of narratives.
The only "facts" that most all MSM outfits will print these days are ones that fit their narrative. And, of course, just throw in some "sources" and "experts" is if this establishes their cherry picked "facts" as credible. "Facts" that do not support the narrative are discounted or outright ignored. as if a fact in and of itself is dangerous if it is a "bad" fact.
When you mentioned the (delete-block) error I thought you were going to talk about what I call a "wordo" instead of a typo.
When I first started using the word star word processor in about 1988, I loved being able to just let it roll, knowing that I could go back and make Corrections without scribbling all over a yellow pad. But one drawback was that I would make a small change that would change the tense of a sentence and and then I'd miss another word that needed to be changed to match that change.
Your points on the (lack of) value of LLMs to writing are true in my experience, and especially about voice, that most important and subtle aspect of writing. So far I've seen nothing that suggests LLMs can create maintain a consistent writing (or chatting) voice. Historical or literary character AIs like the ones from Khanmigo or deepAI all sound like AI generated art portraits look. They seem like eager fifth grade history teachers putting on a costume: pedantic, eager to please, and unable to get any deeper than an encyclopedia-level answer. I'm not sure with a generative book like Cowan's GOAT, but it seems like he wrote the content, and it is the form that is new. Anyone have examples of LLMs successfully generating a consistent voice?
Nice arguments. I (a tech worker) am persuaded by the analogy. Getting the producer (of content, of an application) closer to the end user almost always improves performance. That was the whole point of Agile. You could say Substack is Agile applied to opinion.
I disagree with your point on the usefulness of LLMs in the writing process however. I think LLMs are most useful in discovery, or pre-writing. They accelerate the ability to organize thoughts and information, especially for novices in certain topic areas or infrequent writers. I agree that LLMs probably subtract voice, but I don’t think that’s the step in the process in which they are optimally employed. Yan Lecunn himself called them “mere writing aids” and I think this “0 to 1” step is what he had in mind.
Writing aids seems correct, especially as you say, for novice writers. Voice is what separates writing produced mostly through LLMs and good human writers. It also why I prefer to read good human writers over novice human writers.
> Anyone have examples of LLMs successfully generating a consistent voice
I’ve seen it done successfully- but it requires fine-tuned training in a reasonably large corpus of prior works (hundreds or even thousands of articles). You can’t (consistently) do it with prompt engineering. Anthropic’s are much better for this than OpenAI’s models.
That's interesting that you describe Matthew Yglesias's voice as "dry and snarky". One of the things I find that I like about his writing on Substack is that he is actually quite earnest, and doesn't seem very snarky at all. When I see his tweets, I'm always much less happy, because they do seem to be on the snarky side (which I think is a feature of the medium).
I think Yglesias enjoys being snarky and knows it drives engagement from Twitter, but I agree, his posts are more earnest exhortations for better political and policy thinking.
Another great feature here is that readers get to have a bit of back and forth with you in the comments. And also you get to write the title of your post instead of someone writing a clickbait title that the original author would not have chosen.
The first thing that comes to mind to describe your voice is "cheerful", which maybe seems dismissive, but, like, your posts almost always make my day better. I am a tech optimist but a bit of a social/political pessimist. (I'm still pretty worried about the core argument of MattY's "American Democracy is Doomed" essay from a while back, based on Juan Linz's work.) Focusing on the bright side helps one to get through the day, and maintain hope that "doing the work" will actually lead to a better future. Politics is a tedious slog. "The slow boring of hard boards." It's easier to keep trudging along when you can lift your eyes and see the goal, now and again.
Oh, and regarding the block delete problem -- dude. Get yourself a desk with a keyboard tray, so you're not typing with your wrists elevated, and/or leaning way down to look at the screen. A laptop is fine for a little while, but if you really are going to work for an extended period, the relationship between the position of the screen and the position of the keyboard is terrible. Your future self will thank you for avoiding carpal tunnel syndrome and pinched nerves in your neck.
(Given I work 8-10 hrs a day in front of a computer, I have a decent desk with tray, and I have a big screen that is positioned above my laptop, and I can drag windows between the small and large screen. Usually I'm keeping my email and teamchat messaging on the small screen, and the big screen has various windows up on it for data visualization and command consoles for our onsite control devices.)
I’m not sure if this is the same topic, but the gap between the article title (which the editor or whoever chooses) and the article itself can often do a disservice to the writer.
This is a massively important point. I can't tell you how many times I've discovered that the headline doesn't properly represent the contents. This has gotten worse in the digital age, where the headlines are competing for eyeballs and clicks. Having the author write the headline is a little-appreciated bonus.
I wonder if it would have been difficult for you to learn on the fly without an editor early in your career. Is it possible that the utility of an editor declines as the writer becomes more knowledgeable but that it's important early on?
As an editor of opinion pieces (though primarily a copyeditor, which gets me something of a pass), I'm going to disagree. If the reading public saw some of the stuff that doesn't get to them, or that doesn't get to them in the form it eventually does, there'd be more appreciation for editors. Granted, I'm thinking about this in aggregate terms; it's certainly possible for an editor to have a negative impact on an article. Some more thoughts on copyediting: https://www.splicetoday.com/writing/meditations-of-a-copy-editor
That is probably the main benefit Noah overlooked: curation. As a writer, slowing down the ideation to publish process is must be annoying- but as a reader, I want both interesting pieces from authors who have a perspective I already value as well as novel works I wouldn’t otherwise encounter. One of the things I value from the Substack writers I subscribe to is when they recommend specific pieces. Most writers are themselves voracious consumer so getting their filter is valuable. On the editorial side- we no longer have limited print pages- so more outlets could publish aggressively and then have editors curate the best stuff to the main page. How else will we fine the next Noah’s?
While I also think your remark about editors often removing voice is spot on, I do think this is going a bit far. The main problem I see is not that we have op-ed editors, but that we have plenty of bad editors, who rarely add value to what goes through them and unnecessarily stifle the voice of op-ed writers.
Thus, I think this post is shifting attention away from the real issue. What we should be doing is not removing op-ed editors, but rather look into why do we have so many poor examples, and what we can do to remedy that.
I strongly believe that having a second opinion on a piece is important, as one can easily fall into a pattern, not seeing the big picture. I am not saying this necessarily applies to you, as you don't take yourself too seriously, and you are capable of looking at your pieces critically. My point is that this is a skill that the majority of people do not have. Of course, the ideal situation would be to have an editor who works with op-ed writers, not over their head, which looks like is not the case at present.
From my experience with editors as an op-ed writer:
--My most frequent frustration is with the headlines they come up with, missing the main point or, now that online publications often begin with a summary paragraph before linking to the actual text, botching the main points of a piece;
--But, if editors tell me they don't understand a point, I'll concede that the average reader won't either, and I'll add clarifications where needed.
Noah doesn't mention self-editing. I rarely trust my first drafts and will set aside a piece for a few hours or overnight, before reading it with an editor's eye, and then clarifying or reorganizing as needed. As my (old-fashioned) journalism professor used to tell us: Writing is re-writing.
“The Economist has done an amazing job of editing all of its writers to sound like they’re the same person — a wry British know-it-all.”
I like listening to The Economist weekly audio edition as a NPR substitute. It’s not literally a single voice; they employ 3 or 4 readers for each addition but they all come across (enjoyably) as British know-it alls
I encountered the Economist when I was managing a small company in England for the parent company I worked for. I was also looking for the best magazines to keep up with world affairs. The Economist came out at the top of the list, and I found the style delightful - higher-quality writing than that for the average American in US news weeklies, and some classic phrases which remain embedded in my mind, like "form swallows function" (architecture) and "towering edifices of supposition" (theories of universe structure).
The problem was that almost everything in the magazine was worth reading. It came out weekly, but took all week to read it! I just couldn't keep up.
That’s why I listen to the audio edition. I make use of the time when I can’t read. The level of research seems higher than any other general news publication.
Unfortunately, I am constantly building a reference library of links, quotes, and illustrations on my computer. Audio doesn't do it for me. Even the print magazines I get are a problem - when I've found a particularly good article, I have to go to the computer and try to find it to bookmark or quote it.
"for each edition"
Editing needed.
Great article, Noah. I love the insider viewpoint, and I agree with you. As a consumer of this stuff, I am looking forward to the day where all "volume" op-ed writers are independent. The ones that suck would fall off the edge of the cliff while the good ones would prosper like you have. It would also be better for the reader. I like reading Paul Krugman, David Brooks, and Russ Duhat in the Times, but I would like reading them much better on Substack, for the reasons you mention. I'm sure others like their other writers, too, and they would do fine independently. That would leave the Op-Ed section to guest writers like elected officials, Oprah, CEO's, candidates, disgraced former RNC committee chairs or whomever that is involved with a story to provide a unique viewpoint that a journalist may not be able to provide. I'd rather the Times focus on providing me all the news fit to print, accurately and objectively. It may still be the battering ram of the national liberal conversation, but it would be less in-your-face about it. Thanks for what you do, Noah. I read you and Matt regularly over my morning coffee. If you run into Professor Krugman, please encourage him to jump ship.
Krugman actually has a Substack, which he's published to very occasionally, as well as his freeform longer newsletter posts from the Times.
“Facts are pretty much the same thing no matter who’s reporting them…”. Sweet and old-fashioned, and rightfully how it should be. Sadly rare, though.
Facts are snippets we take out of context to support our narrative. Usually this involves discarding or ignoring a greater number snippets that might weaken the narrative.
The universe of facts is very large while the selection of narratives on offer is very limited. Substack helps expand the selection of narratives.
Love your first paragraph.
Are you thinking about "Alternative facts?"
The only "facts" that most all MSM outfits will print these days are ones that fit their narrative. And, of course, just throw in some "sources" and "experts" is if this establishes their cherry picked "facts" as credible. "Facts" that do not support the narrative are discounted or outright ignored. as if a fact in and of itself is dangerous if it is a "bad" fact.
When you mentioned the (delete-block) error I thought you were going to talk about what I call a "wordo" instead of a typo.
When I first started using the word star word processor in about 1988, I loved being able to just let it roll, knowing that I could go back and make Corrections without scribbling all over a yellow pad. But one drawback was that I would make a small change that would change the tense of a sentence and and then I'd miss another word that needed to be changed to match that change.
Your points on the (lack of) value of LLMs to writing are true in my experience, and especially about voice, that most important and subtle aspect of writing. So far I've seen nothing that suggests LLMs can create maintain a consistent writing (or chatting) voice. Historical or literary character AIs like the ones from Khanmigo or deepAI all sound like AI generated art portraits look. They seem like eager fifth grade history teachers putting on a costume: pedantic, eager to please, and unable to get any deeper than an encyclopedia-level answer. I'm not sure with a generative book like Cowan's GOAT, but it seems like he wrote the content, and it is the form that is new. Anyone have examples of LLMs successfully generating a consistent voice?
Nice arguments. I (a tech worker) am persuaded by the analogy. Getting the producer (of content, of an application) closer to the end user almost always improves performance. That was the whole point of Agile. You could say Substack is Agile applied to opinion.
I disagree with your point on the usefulness of LLMs in the writing process however. I think LLMs are most useful in discovery, or pre-writing. They accelerate the ability to organize thoughts and information, especially for novices in certain topic areas or infrequent writers. I agree that LLMs probably subtract voice, but I don’t think that’s the step in the process in which they are optimally employed. Yan Lecunn himself called them “mere writing aids” and I think this “0 to 1” step is what he had in mind.
Writing aids seems correct, especially as you say, for novice writers. Voice is what separates writing produced mostly through LLMs and good human writers. It also why I prefer to read good human writers over novice human writers.
> Anyone have examples of LLMs successfully generating a consistent voice
I’ve seen it done successfully- but it requires fine-tuned training in a reasonably large corpus of prior works (hundreds or even thousands of articles). You can’t (consistently) do it with prompt engineering. Anthropic’s are much better for this than OpenAI’s models.
That's interesting that you describe Matthew Yglesias's voice as "dry and snarky". One of the things I find that I like about his writing on Substack is that he is actually quite earnest, and doesn't seem very snarky at all. When I see his tweets, I'm always much less happy, because they do seem to be on the snarky side (which I think is a feature of the medium).
I think Yglesias enjoys being snarky and knows it drives engagement from Twitter, but I agree, his posts are more earnest exhortations for better political and policy thinking.
Another great feature here is that readers get to have a bit of back and forth with you in the comments. And also you get to write the title of your post instead of someone writing a clickbait title that the original author would not have chosen.
The first thing that comes to mind to describe your voice is "cheerful", which maybe seems dismissive, but, like, your posts almost always make my day better. I am a tech optimist but a bit of a social/political pessimist. (I'm still pretty worried about the core argument of MattY's "American Democracy is Doomed" essay from a while back, based on Juan Linz's work.) Focusing on the bright side helps one to get through the day, and maintain hope that "doing the work" will actually lead to a better future. Politics is a tedious slog. "The slow boring of hard boards." It's easier to keep trudging along when you can lift your eyes and see the goal, now and again.
Oh, and regarding the block delete problem -- dude. Get yourself a desk with a keyboard tray, so you're not typing with your wrists elevated, and/or leaning way down to look at the screen. A laptop is fine for a little while, but if you really are going to work for an extended period, the relationship between the position of the screen and the position of the keyboard is terrible. Your future self will thank you for avoiding carpal tunnel syndrome and pinched nerves in your neck.
(Given I work 8-10 hrs a day in front of a computer, I have a decent desk with tray, and I have a big screen that is positioned above my laptop, and I can drag windows between the small and large screen. Usually I'm keeping my email and teamchat messaging on the small screen, and the big screen has various windows up on it for data visualization and command consoles for our onsite control devices.)
Re: Noah's voice, I'd say "enthusiastic and excited" is what comes to mind for me
Of course, really my opinion on American Democracy is, "It would be a great idea, we should try it some time."
I’m not sure if this is the same topic, but the gap between the article title (which the editor or whoever chooses) and the article itself can often do a disservice to the writer.
This is a massively important point. I can't tell you how many times I've discovered that the headline doesn't properly represent the contents. This has gotten worse in the digital age, where the headlines are competing for eyeballs and clicks. Having the author write the headline is a little-appreciated bonus.
I wonder if it would have been difficult for you to learn on the fly without an editor early in your career. Is it possible that the utility of an editor declines as the writer becomes more knowledgeable but that it's important early on?
As an editor of opinion pieces (though primarily a copyeditor, which gets me something of a pass), I'm going to disagree. If the reading public saw some of the stuff that doesn't get to them, or that doesn't get to them in the form it eventually does, there'd be more appreciation for editors. Granted, I'm thinking about this in aggregate terms; it's certainly possible for an editor to have a negative impact on an article. Some more thoughts on copyediting: https://www.splicetoday.com/writing/meditations-of-a-copy-editor
It may well be that the main impact of an editor is slightly negative on 90% of submissions but hugely positive on 10% of them.
That is probably the main benefit Noah overlooked: curation. As a writer, slowing down the ideation to publish process is must be annoying- but as a reader, I want both interesting pieces from authors who have a perspective I already value as well as novel works I wouldn’t otherwise encounter. One of the things I value from the Substack writers I subscribe to is when they recommend specific pieces. Most writers are themselves voracious consumer so getting their filter is valuable. On the editorial side- we no longer have limited print pages- so more outlets could publish aggressively and then have editors curate the best stuff to the main page. How else will we fine the next Noah’s?
I wouldn't guess at numbers, but I think that is a key thing. Without editors, there'd be opinion pages with some truly terrible stuff.
While I also think your remark about editors often removing voice is spot on, I do think this is going a bit far. The main problem I see is not that we have op-ed editors, but that we have plenty of bad editors, who rarely add value to what goes through them and unnecessarily stifle the voice of op-ed writers.
Thus, I think this post is shifting attention away from the real issue. What we should be doing is not removing op-ed editors, but rather look into why do we have so many poor examples, and what we can do to remedy that.
I strongly believe that having a second opinion on a piece is important, as one can easily fall into a pattern, not seeing the big picture. I am not saying this necessarily applies to you, as you don't take yourself too seriously, and you are capable of looking at your pieces critically. My point is that this is a skill that the majority of people do not have. Of course, the ideal situation would be to have an editor who works with op-ed writers, not over their head, which looks like is not the case at present.
Re "having a second opinion on a piece is important"
The thing is, readers provide that second opinion in much more rich fashion.
fantastic as always
We haven’t even talked about headlines. Editors are notorious for bad headlines.
So glad your opinion is all yours! And so varied in substance.
I often forward your commentary to my friends and my Fiduciary in hopes they sign up. Cheers!