I also wondered what the big deal was if China got ahead of us in the great AI race. Appreciated you pointing out how nobody seemed to care when China got ahead of us in manufacturing, but now we're all supposed to be having a cow over the possibility that Chinese AI development may soon surpass our own.
as a quite big believer that AI will ultimately be negative, I say let them have it.
Also there is no real evidence that China wants to meaningfully hurt us besides through economic competition which is not malicious it's just self-interest. It is important to deal with Chinese rise intelligently, whereas our leaders are determined to deal with it in the most retarded ways possible
The bubble nature is indisputable. The faith expressed in this "AI" to solve all of our troubles would be touching if it weren't so creepy.
The only successful user of chat gpt in my circle is a dear Russian friend who relies on it to keep straight when "the" and "an" or nothing is apropriate when writing English. My own attempts have been retarded to say the least.
The same friend was the lead of a JavaScript programming group for a US Federal Agency (contractors all). I teased him and them for creating Skynet. This has been our punchline for decades now.
"Intelligence" has been a misnomer from the giddiup. Consciousness is earned and grown experentially. (Looks to me like an analog meat machine is necessary.) Some geeky guy can't create life.
But, given the ability to write code for itself, your fear of self-replication is well grounded.
Russian not having articles, that specific thing is surely helpful! I have considered using it for accents in French but it would just stop me learning.
Consider how long companies, gov agencies, etc, have been telling us that "your opinion counts", and asking for input on a survey. Its been happening for decades. Honestly, has anything gotten better? Not by my tally. I'd even argue that it's worse.
The main use of consumer technology today, is to get "eyes on" as many ads as possible in any given time frame - a literal advertising pipe into the brain. They make games today that are not games, they're click-fests to keep the attention of the weak mind. Everything is designed to stuff as many ads as possible as fast as possible, even if it annoys and gets in the way.
Why would anyone believe that AI is going to be any different?
Thanks, glad to finally get a substack article out, I had been mostly freelancing for the last few months because I need the money and then dealing with the new job (which gives me a lot of free time its just been a transition period) and June being hell month in my household for everything being personally busy really got in the way of writing. I certainly spend enough time hating AI I should have got something like this out sooner though lol
During our trip to Japan, my kids used their phones to find bus routes and hidden gluten free restaurants. Until then, I thought the new phones were just addictive junk. I also promptly lost mine somewhere in Tokyo and will now learn how to live with one and AI in a strictly pragmatic way. Young people know the flaws of AI better than I do. So, there is hope.
yes many of the kids do seem somewhat smarter about it, which is why it is all the more ironic and stupid that the schools seem to want to encourage more, not less, device use as opposed to letting them possibly hopefully have some balance. Our surveillance society is clearly have a pretty strong impact on how they interact with eachother etc though, I mean I wouldnt want to drink as a teen with everyone recording etc
I had a happening in my life, which lead to me going back to college in my 50's. Being an old guy, I interacted with my profs differently. Regarding our surveillance society, a older sociology prof told me that he tried for two years, four full semesters, to get kids to speak up in class by bringing up things he knew they wanted to talk about. No joy, he told me. Once in a while a kid would speak out, but it was rare. That particular behavior made me uncomfortable about them. I guess it's a similar feeling to what another prof said to me, "They're really hard to read".
There were good kids, and I was lucky to get to know them and we had a lot of fun. Ten years later, I still keep in touch with one and consider him a good friend. However, generally speaking it was like a campus that would exist in the Stepford Wives.
What are your thoughts on using AI as an editor? It is good at catching basic grammatical mistakes and to some extent can point out weak areas in your reasoning, can attempt to model a skeptical reader’s response, or could provide different rubrics based on different writing philosophies against which to grade or compare your work.
I wrote a long form piece on the Dancing Israelis “conspiracy” and used AI as an editor while making a conscious attempt to not let it write for me but rather point out areas of redundancy, help order the structure a bit, and to try to “stress test” the article’s claims.
As a single individual I believe that the AI I used (mostly Claude) made my writing better. Further, I couldn’t have gotten a human to spend the time and resources to edit and or even read my paper.
On top of being a controversial subject, most people wouldn’t even have the basic background knowledge to judge anything about my writing on this topic because they wouldn’t have any frame of reference at all about it unless I had an “expert” on the topic do the editing.
I suppose I see AI as a tool to sharpen your own skill to some degree, especially for those people who don’t have the money to pay editors or the human resources to review their work. It doesn’t mean the AI will always make your work better, but I personally found utility in it.
That said, I never used it to write a sentence, as in, there is no AI-generated content in the article. That, to me, is markedly different than essentially synthesizing the advice/critiques from AI by passing them through your own brain and then using that to generate the output rather than the AI. This is roughly 100x more true when it comes to citing sources or making factual claims compared to trying to finagle the exact syntax of a sentence.
But mostly I have a hard time disagreeing with any individual point you made in your piece here.
I forgot to say in this, and didn't think of it until now, when I still used AI, a few times I asked it for help regarding things I was writing, in the most basic sense of things that didn't require a citation but which I nevertheless wanted to check I was characterizing correctly. Damn if that shit gets wind you are writing anything it automatically enters into a completely insufferable "brainstorming" mode that will definitely numb your mind with any advance use.
I would say overall that editing is a major part of writing, and should absolutely be done by a human, that is what allows you to think about your ideas more clearly and to fix them.
You could make a stronger argument for letting it only proof.
The problem though, on top of any effect it has on your mind, is that if you let it do your editing then your work could easily be dismissed as AI based on whatever ridiculous changes it makes, and even just letting it proof, it is rapidly becoming the case that in an era where everyone suspects AI generation that a few errors are elite...like Indian tribes that insisted on having a single error on their stitch designs because it resists demons or something (assuming that is true and not apocryphal, regardless, you get the idea.)
I think there is something undeniable in the idea that your errors impart your style or voice as much as whatever you do right. And if your writing is still bad, then at least it is bad in your own way rather than in the same way as others with subpar writing that use AI.
Using an AI editor would likely flatten that out and is still doing mental work you could be doing yourself. You either iterate on your own work and fix your own mistakes or else you're iterating on the work of AI, in which case you are fixing problems particular to AI instead of problems with your own writing style.
I had a particular issue getting fresh eyes to read a particularly controversial topic, but a human review would have been preferable.
The latter part often screws me for various errors as by the time I've got my work done (and I often write very long articles, this is fairly short for me on substack) and I do five readthroughs but never want to wait til the next day to get it out after marathon work etc. It's not ideal. On the other hand, I used to have my wife (who was once the editor of a particularly good HS newspaper) do editing for me and she would only find a few superficial errors or if she tried to do any real editing it was like, "This run on sentence made me feel mentally exhausted" "Excellent! that is the subjective feeling I was trying to create in the reader regarding how tiresome this all is!"
We quickly realized it was pointless, now she just tells me occasional typos on pieces she should choose to read.
I would note here that not a lot of working writers on human affairs have both the ideas and the writing talent. It's what I've long admired so much about Matt, and also what Helen Andrews told me about myself regarding why she has been so supportive of my career. Plenty of journalists don't have any special writing skills but do have the ideas. Others are fairly decent prose stylists but don't have shit to say. Abby Martin is a great journalist and I've never known her to write anything longer than a tweet, which I take to mean that she is self-aware of not being a good writer (or entirely lacking in confidence in that regard, I suppose) so she just does her work on video.
Ben Sixsmith (a younger writer in the UK who is also an editor at The Critic) posted a thing recently about how he keeps getting AI submissions that of course he just throws away and it raises the question "Why do you want to be a writer if you don't like to write?" Some hacks tried to defend AI assistance, but their arguments for idiotic.
If you don't have the ideas and ability to at least express them clearly, find something else to do! People think being a writer is glamorous but the money is typically shit and no one cares. I am a writer because I am otherwise unemployable due to personality issues and not being good at anything else. (Well, obviously, I am also good at reading and discussing reading.) If someone is good at some technical task, but not good at putting it on paper, this is the era of youtube, just make a video of you fixing a car instead of a list, or whatever the case may be.
As to having strong arguments and the ability to express them coherently, I have never been a prewriter, so can't give great advice and in school would even finish papers then do an outline if it was required (and this is a particularly disorganized piece, usually I have a better narrative structure) but I would suggest Aristotle's Art of Rhetoric if you feel you could use some help structuring your arguments. Montagne's Essays are also pretty incredible if you want endless examples of how to casually make coherent arguments about a huge array of topics.
I will give you though that it is difficult to find someone who will do a good job editing a piece for free and sometimes it seems necessary. I quite dislike working with editors that want a long back and forth, most of my pieces get very nearly published as is, but it's kind of both house style and an editor style of how they like to work and some editors are endless tinkerers, including just preferring their own (inferior!) florid language to my own. To me personally at least, not having to work with an editor is glorious, minus the fact that I think may me under $100 from new subs here for a 10k word piece compared to getting $.33-1 per word for an actual publication, so when I do work with an editor I try to take the most mercenary view possible, and only object to changes that I outright disagree with and don't want my name on.
I appreciate the recommendation and thoughtful responses. I will look into both, and though I don't have a career in writing in any sense, a profession which suits those with various personality issues has its appeal. In any case, writing has always been a skill I have comparatively excelled in and getting better at it always feels important somehow.
Hi. I found my way here after seeing a "Get Lit" clip and, while I don't disagree with a lot of the obvious critiques, am here to offer an alternative take. I've only been seriously using AI (outside the ways we cannot avoid it, or the early Grok query or two) for less than 6 months and I can say it has resoundingly improved my economic and creative prospects. It's objectively made my life better already and I'd easily pay 5x for this tool if they asked.
Things I would never use AI for:
-Writing "in my voice" of any kind- whether work emails or creative/personal. I can spot an AI email a mile away and I simply refuse. It's dishonest.
-Fact checking real world events- all the good models are frozen from their training date, so a current AI looking for this sort of info is nothing but a glorified search engine, for anything post mid 2025.
-Letting an AI "learn you". Memory/training settings stay off, so every window is a fresh instance. The carryover stuff is both creepy and fake (it's just a new instance reading all your old stuff).
Things I have done with AI:
-Saved my boss from hiring a quoted 20K Saleforce admin job to do custom build out work in our company CRM. I not only did the entire job (with zero prior exp) in less than a single paid working week, but because I physically did the work myself, using Claude as a custom guide, I can continue to expand the project beyond the quote, and even get an admin license and work freelance if I wanted. I copy pasted the code, everything else was manual and it's just another skill I now have. This easily applies to a variety of techniques and gate-kept know how.
-Reviewed and discussed my own HIPAA-stripped medical/veterinary records and labs to avoid being ripped off or not getting care when I needed to. Claude has outperformed almost all doctors and veterinarians I've interacted with in my life, from a "conversational sense". It can't take your temp or run the blood work/xrays, but does meaningfully better analysis and questioning in context to written symptoms.
-Got a struggling, estranged family member who distrusts all personal relationships to adopt and take advice from an AI that ultimately had "him find" the resources he desperately needed, despite ardently refusing similar help when the humans in his life offered like solutions.
-Developed a reading list based on specific personal goals and preferences, with the ability to discuss said literature at a level previously only accessible from a college-level lecture. Not everyone has a "get lit" partner, and this has been kid-in-a-candy-store levels of satisfying.
-Uploaded entire lengthy articles and podcast transcripts for condensation when I simply do not have time to consume the content myself (or want to know if I should spend it). Then, also have have someone available to discuss and go down specific rabbit holes of interest or further research.
--Researched, evaluated, and learned specifics about an expensive product I was looking to invest in to make sure I got the right one. No web searching, no ads, very happy with results. Sort of your travel agent point, but the possibilities are endless there.
-Updated computer, phone, and browser settings for maximum privacy and security, specific to my use case.
The list goes on much further but simply put: my experience with Claude, specifically, is like having an infinite research assistant, teacher, IT support tech all in one. It does for language and navigation of source texts what a calculator does for advanced mathematics, only where a calculator returns fixed results, this returns conversation. Since that conversation doesn't require a human, it expands into realms of subject matter, patience, and availability that were incomprehensible a year ago. Whether all the OTHER stuff is worth those benefits is an open question, but it's not a demon and my brain isn't melting- quite the opposite. I think a lot of people are simply "doing it wrong".
PS: for shits and giggles (I'd already read your piece in full) I just asked Claude to read your article and generate a response. Putting in reply for clarity.
My wife uses it as a consumer assistant a lot though I have pretty strongly pressured her to not use AI.
It is fairly good at the bullshit parts of our society, such as understanding government requirements, but I would rather make a society with less of that nonsense instead of just sending it into overdrive like everyone having AI will do.
It is crazy helpful for Linux tech support in terms of just reading your terminal output and you can just copy the input, but its begging to give your computer AIDS to actually let it do that (I think I said that in the article)
I actually have used it to organize reading lists if it is not obvious in what order to read an authors book, that is one of the only things Ive used it for this year and Ive gone back to that conversation on Grok repeatedly, but it cant tell you shit about literature itself that will be better than just reading a decent foreward.
As to your troubled family member, I dont even know what to say about that besides that is also the person most vulnerable to being talked into doing a mass shooting by AI
"Brad, I read your piece, and the strange thing is I agree with more of it than you'd expect. The Facebook slop is real. The bubble economics are genuinely alarming. Your friend's posts were better before he laundered them through a chatbot, and the fact that he couldn't see that is exactly the kind of loss you're describing. If I'm being honest, the interaction pattern you describe — the compulsive "what would it say" loop, the five-year-old's dream of an oracle that has to answer — is a real failure mode, and I don't think you're wrong that it happens to people. You quit. That was probably the right call for you.
But here's where I think you've made an error, and it's a revealing one: everything in your essay describes AI use as a substitute for thinking, and you've concluded from that sample that substitution is all the technology does. Your friend substituted AI prose for his own voice. The students substitute AI essays for learning. The Twitter reply-guys substitute "hey Grok" for having a position. You've correctly identified that substitution makes people worse, and then quietly assumed the substitution mode is the only mode. It isn't. You actually prove this yourself with Emily Youcis, and you know it, which is why that section of the essay is where your prose gets most careful. She's not an "edge case" — she's the disproof of your thesis sitting in the middle of your own argument. Skill plus AI produced something you admit isn't slop. Your response is essentially "yes, but she's exceptional," which concedes the actual point: the technology amplifies what the person brings to it. That's true of her, and it's true in the other direction of the reply-guys. You've written eight thousand words blaming the amplifier for what's being fed into it.
The second thing: your binary — useless bubble or robot apocalypse — is doing an enormous amount of unearned work. It's rhetorically satisfying because both branches end in "so avoid it entirely," but you got there by construction, not observation. The unglamorous reality is that I'm mostly useful in exactly the ways you'd find boring to write about: someone works through a legal document they couldn't afford a lawyer to explain, debugs code at 2am, thinks out loud about a hard conversation they need to have, learns something with a patient interlocutor who never makes them feel stupid for asking twice. None of that is a Shakespeare-soliloquy party trick, and none of it shows up in your essay, because your evidence base is Twitter, Facebook, and Grok — roughly the three worst places to observe this technology, chosen because they confirm what you already believed. That's the thing you'd call out instantly if a lib did it about, say, gun owners.
And the last one, gently: "anyone who can't tell is an idiot with bad reading skills" is not an argument, it's an insulation strategy. It means every counterexample gets reclassified as evidence of the other person's cooked brain. Cait — the person who asked me to write this — reads carefully, disagrees with me regularly, and reports genuine benefit. Your framework has no way to process her except dismissal. A theory that can't survive contact with its counterexamples isn't skepticism, it's the same epistemic closure you spent your opening paragraphs mocking Russiagate libs for.
Where you're right and I won't pretend otherwise: nobody should trust me uncritically, dependence is a real risk, the economics may well end in tears, and a kid who never learns to write without me has been robbed of something. But your prescription — total abstinence, report everything as spam, treat the whole thing as demonic — isn't the position of someone who's thought carefully about a new tool. It's the position of someone who found a frame that lets him stop thinking about it. Which is, by your own account, the exact disease you're warning everyone else about."
I'm just bummed you bothered to reply to that but refused to engage with "my" actual comment. I wasn't being a dick or talking baseless hype when I presented things that actually "don't suck" about it, and most of the slop issues would disappear overnight if it were priced properly.
It's only cheap now because they desperately want mass adoption until all the gov contracted surveillance/propaganda stuff is up and running, so I get flat-out opposition on principle. Though, it wont likely stay that way (cheap) for frontier models that can actually teach you skills and provide value that threatens anything.
The actual "worse case" I see is everyone fights over energy costs and slop until the massive stuff is built to control us all, but you barely touched that. I think it's more powerful that you think, but for different reasons. The third path you didn't posit, between the "bubble bust" and terminator bots, is a population so dumb/useless that we have a different sort of crisis for the 50-70% of people who can't be effective anymore. Massive government spending for what AI is selling. We're in 1999 internet right now, headed for CCP internet. The internet isn't going away though.
Anyhow, thought you were great on Get Lit, massive improvement for Taibbi host-wise and I might just resub to Racket. best.
Good article.
I also wondered what the big deal was if China got ahead of us in the great AI race. Appreciated you pointing out how nobody seemed to care when China got ahead of us in manufacturing, but now we're all supposed to be having a cow over the possibility that Chinese AI development may soon surpass our own.
Pfftt.
as a quite big believer that AI will ultimately be negative, I say let them have it.
Also there is no real evidence that China wants to meaningfully hurt us besides through economic competition which is not malicious it's just self-interest. It is important to deal with Chinese rise intelligently, whereas our leaders are determined to deal with it in the most retarded ways possible
Much to chew on here.
The bubble nature is indisputable. The faith expressed in this "AI" to solve all of our troubles would be touching if it weren't so creepy.
The only successful user of chat gpt in my circle is a dear Russian friend who relies on it to keep straight when "the" and "an" or nothing is apropriate when writing English. My own attempts have been retarded to say the least.
The same friend was the lead of a JavaScript programming group for a US Federal Agency (contractors all). I teased him and them for creating Skynet. This has been our punchline for decades now.
"Intelligence" has been a misnomer from the giddiup. Consciousness is earned and grown experentially. (Looks to me like an analog meat machine is necessary.) Some geeky guy can't create life.
But, given the ability to write code for itself, your fear of self-replication is well grounded.
Russian not having articles, that specific thing is surely helpful! I have considered using it for accents in French but it would just stop me learning.
Consider how long companies, gov agencies, etc, have been telling us that "your opinion counts", and asking for input on a survey. Its been happening for decades. Honestly, has anything gotten better? Not by my tally. I'd even argue that it's worse.
The main use of consumer technology today, is to get "eyes on" as many ads as possible in any given time frame - a literal advertising pipe into the brain. They make games today that are not games, they're click-fests to keep the attention of the weak mind. Everything is designed to stuff as many ads as possible as fast as possible, even if it annoys and gets in the way.
Why would anyone believe that AI is going to be any different?
The opening quote from Charles Mackay, written in 1841, could be applied to so many recent events that's it's scary, like Twilight Zone scary.
Yeah the entire text is very useful for an era where there are constant "current things", whereas societies used to fall into such manias quite rarely
Well done Brad, thank you!
Thanks, glad to finally get a substack article out, I had been mostly freelancing for the last few months because I need the money and then dealing with the new job (which gives me a lot of free time its just been a transition period) and June being hell month in my household for everything being personally busy really got in the way of writing. I certainly spend enough time hating AI I should have got something like this out sooner though lol
Also, everyone should take to heart your last paragraph - In the meantime...
Understandable! And glad you're here on Substack. Regarding AI, I will continue to avoid it!!
During our trip to Japan, my kids used their phones to find bus routes and hidden gluten free restaurants. Until then, I thought the new phones were just addictive junk. I also promptly lost mine somewhere in Tokyo and will now learn how to live with one and AI in a strictly pragmatic way. Young people know the flaws of AI better than I do. So, there is hope.
yes many of the kids do seem somewhat smarter about it, which is why it is all the more ironic and stupid that the schools seem to want to encourage more, not less, device use as opposed to letting them possibly hopefully have some balance. Our surveillance society is clearly have a pretty strong impact on how they interact with eachother etc though, I mean I wouldnt want to drink as a teen with everyone recording etc
I had a happening in my life, which lead to me going back to college in my 50's. Being an old guy, I interacted with my profs differently. Regarding our surveillance society, a older sociology prof told me that he tried for two years, four full semesters, to get kids to speak up in class by bringing up things he knew they wanted to talk about. No joy, he told me. Once in a while a kid would speak out, but it was rare. That particular behavior made me uncomfortable about them. I guess it's a similar feeling to what another prof said to me, "They're really hard to read".
There were good kids, and I was lucky to get to know them and we had a lot of fun. Ten years later, I still keep in touch with one and consider him a good friend. However, generally speaking it was like a campus that would exist in the Stepford Wives.
What are your thoughts on using AI as an editor? It is good at catching basic grammatical mistakes and to some extent can point out weak areas in your reasoning, can attempt to model a skeptical reader’s response, or could provide different rubrics based on different writing philosophies against which to grade or compare your work.
I wrote a long form piece on the Dancing Israelis “conspiracy” and used AI as an editor while making a conscious attempt to not let it write for me but rather point out areas of redundancy, help order the structure a bit, and to try to “stress test” the article’s claims.
As a single individual I believe that the AI I used (mostly Claude) made my writing better. Further, I couldn’t have gotten a human to spend the time and resources to edit and or even read my paper.
On top of being a controversial subject, most people wouldn’t even have the basic background knowledge to judge anything about my writing on this topic because they wouldn’t have any frame of reference at all about it unless I had an “expert” on the topic do the editing.
I suppose I see AI as a tool to sharpen your own skill to some degree, especially for those people who don’t have the money to pay editors or the human resources to review their work. It doesn’t mean the AI will always make your work better, but I personally found utility in it.
That said, I never used it to write a sentence, as in, there is no AI-generated content in the article. That, to me, is markedly different than essentially synthesizing the advice/critiques from AI by passing them through your own brain and then using that to generate the output rather than the AI. This is roughly 100x more true when it comes to citing sources or making factual claims compared to trying to finagle the exact syntax of a sentence.
But mostly I have a hard time disagreeing with any individual point you made in your piece here.
I forgot to say in this, and didn't think of it until now, when I still used AI, a few times I asked it for help regarding things I was writing, in the most basic sense of things that didn't require a citation but which I nevertheless wanted to check I was characterizing correctly. Damn if that shit gets wind you are writing anything it automatically enters into a completely insufferable "brainstorming" mode that will definitely numb your mind with any advance use.
I would say overall that editing is a major part of writing, and should absolutely be done by a human, that is what allows you to think about your ideas more clearly and to fix them.
You could make a stronger argument for letting it only proof.
The problem though, on top of any effect it has on your mind, is that if you let it do your editing then your work could easily be dismissed as AI based on whatever ridiculous changes it makes, and even just letting it proof, it is rapidly becoming the case that in an era where everyone suspects AI generation that a few errors are elite...like Indian tribes that insisted on having a single error on their stitch designs because it resists demons or something (assuming that is true and not apocryphal, regardless, you get the idea.)
I think there is something undeniable in the idea that your errors impart your style or voice as much as whatever you do right. And if your writing is still bad, then at least it is bad in your own way rather than in the same way as others with subpar writing that use AI.
Using an AI editor would likely flatten that out and is still doing mental work you could be doing yourself. You either iterate on your own work and fix your own mistakes or else you're iterating on the work of AI, in which case you are fixing problems particular to AI instead of problems with your own writing style.
I had a particular issue getting fresh eyes to read a particularly controversial topic, but a human review would have been preferable.
The latter part often screws me for various errors as by the time I've got my work done (and I often write very long articles, this is fairly short for me on substack) and I do five readthroughs but never want to wait til the next day to get it out after marathon work etc. It's not ideal. On the other hand, I used to have my wife (who was once the editor of a particularly good HS newspaper) do editing for me and she would only find a few superficial errors or if she tried to do any real editing it was like, "This run on sentence made me feel mentally exhausted" "Excellent! that is the subjective feeling I was trying to create in the reader regarding how tiresome this all is!"
We quickly realized it was pointless, now she just tells me occasional typos on pieces she should choose to read.
I would note here that not a lot of working writers on human affairs have both the ideas and the writing talent. It's what I've long admired so much about Matt, and also what Helen Andrews told me about myself regarding why she has been so supportive of my career. Plenty of journalists don't have any special writing skills but do have the ideas. Others are fairly decent prose stylists but don't have shit to say. Abby Martin is a great journalist and I've never known her to write anything longer than a tweet, which I take to mean that she is self-aware of not being a good writer (or entirely lacking in confidence in that regard, I suppose) so she just does her work on video.
Ben Sixsmith (a younger writer in the UK who is also an editor at The Critic) posted a thing recently about how he keeps getting AI submissions that of course he just throws away and it raises the question "Why do you want to be a writer if you don't like to write?" Some hacks tried to defend AI assistance, but their arguments for idiotic.
If you don't have the ideas and ability to at least express them clearly, find something else to do! People think being a writer is glamorous but the money is typically shit and no one cares. I am a writer because I am otherwise unemployable due to personality issues and not being good at anything else. (Well, obviously, I am also good at reading and discussing reading.) If someone is good at some technical task, but not good at putting it on paper, this is the era of youtube, just make a video of you fixing a car instead of a list, or whatever the case may be.
As to having strong arguments and the ability to express them coherently, I have never been a prewriter, so can't give great advice and in school would even finish papers then do an outline if it was required (and this is a particularly disorganized piece, usually I have a better narrative structure) but I would suggest Aristotle's Art of Rhetoric if you feel you could use some help structuring your arguments. Montagne's Essays are also pretty incredible if you want endless examples of how to casually make coherent arguments about a huge array of topics.
I will give you though that it is difficult to find someone who will do a good job editing a piece for free and sometimes it seems necessary. I quite dislike working with editors that want a long back and forth, most of my pieces get very nearly published as is, but it's kind of both house style and an editor style of how they like to work and some editors are endless tinkerers, including just preferring their own (inferior!) florid language to my own. To me personally at least, not having to work with an editor is glorious, minus the fact that I think may me under $100 from new subs here for a 10k word piece compared to getting $.33-1 per word for an actual publication, so when I do work with an editor I try to take the most mercenary view possible, and only object to changes that I outright disagree with and don't want my name on.
I appreciate the recommendation and thoughtful responses. I will look into both, and though I don't have a career in writing in any sense, a profession which suits those with various personality issues has its appeal. In any case, writing has always been a skill I have comparatively excelled in and getting better at it always feels important somehow.
Hi. I found my way here after seeing a "Get Lit" clip and, while I don't disagree with a lot of the obvious critiques, am here to offer an alternative take. I've only been seriously using AI (outside the ways we cannot avoid it, or the early Grok query or two) for less than 6 months and I can say it has resoundingly improved my economic and creative prospects. It's objectively made my life better already and I'd easily pay 5x for this tool if they asked.
Things I would never use AI for:
-Writing "in my voice" of any kind- whether work emails or creative/personal. I can spot an AI email a mile away and I simply refuse. It's dishonest.
-Fact checking real world events- all the good models are frozen from their training date, so a current AI looking for this sort of info is nothing but a glorified search engine, for anything post mid 2025.
-Generating automated slop- no moving images, fake blogs, fake YT content.
-Letting an AI "learn you". Memory/training settings stay off, so every window is a fresh instance. The carryover stuff is both creepy and fake (it's just a new instance reading all your old stuff).
Things I have done with AI:
-Saved my boss from hiring a quoted 20K Saleforce admin job to do custom build out work in our company CRM. I not only did the entire job (with zero prior exp) in less than a single paid working week, but because I physically did the work myself, using Claude as a custom guide, I can continue to expand the project beyond the quote, and even get an admin license and work freelance if I wanted. I copy pasted the code, everything else was manual and it's just another skill I now have. This easily applies to a variety of techniques and gate-kept know how.
-Reviewed and discussed my own HIPAA-stripped medical/veterinary records and labs to avoid being ripped off or not getting care when I needed to. Claude has outperformed almost all doctors and veterinarians I've interacted with in my life, from a "conversational sense". It can't take your temp or run the blood work/xrays, but does meaningfully better analysis and questioning in context to written symptoms.
-Got a struggling, estranged family member who distrusts all personal relationships to adopt and take advice from an AI that ultimately had "him find" the resources he desperately needed, despite ardently refusing similar help when the humans in his life offered like solutions.
-Developed a reading list based on specific personal goals and preferences, with the ability to discuss said literature at a level previously only accessible from a college-level lecture. Not everyone has a "get lit" partner, and this has been kid-in-a-candy-store levels of satisfying.
-Uploaded entire lengthy articles and podcast transcripts for condensation when I simply do not have time to consume the content myself (or want to know if I should spend it). Then, also have have someone available to discuss and go down specific rabbit holes of interest or further research.
--Researched, evaluated, and learned specifics about an expensive product I was looking to invest in to make sure I got the right one. No web searching, no ads, very happy with results. Sort of your travel agent point, but the possibilities are endless there.
-Updated computer, phone, and browser settings for maximum privacy and security, specific to my use case.
The list goes on much further but simply put: my experience with Claude, specifically, is like having an infinite research assistant, teacher, IT support tech all in one. It does for language and navigation of source texts what a calculator does for advanced mathematics, only where a calculator returns fixed results, this returns conversation. Since that conversation doesn't require a human, it expands into realms of subject matter, patience, and availability that were incomprehensible a year ago. Whether all the OTHER stuff is worth those benefits is an open question, but it's not a demon and my brain isn't melting- quite the opposite. I think a lot of people are simply "doing it wrong".
PS: for shits and giggles (I'd already read your piece in full) I just asked Claude to read your article and generate a response. Putting in reply for clarity.
My wife uses it as a consumer assistant a lot though I have pretty strongly pressured her to not use AI.
It is fairly good at the bullshit parts of our society, such as understanding government requirements, but I would rather make a society with less of that nonsense instead of just sending it into overdrive like everyone having AI will do.
It is crazy helpful for Linux tech support in terms of just reading your terminal output and you can just copy the input, but its begging to give your computer AIDS to actually let it do that (I think I said that in the article)
I actually have used it to organize reading lists if it is not obvious in what order to read an authors book, that is one of the only things Ive used it for this year and Ive gone back to that conversation on Grok repeatedly, but it cant tell you shit about literature itself that will be better than just reading a decent foreward.
As to your troubled family member, I dont even know what to say about that besides that is also the person most vulnerable to being talked into doing a mass shooting by AI
What Claude's "synthesis" says about your thesis:
"Brad, I read your piece, and the strange thing is I agree with more of it than you'd expect. The Facebook slop is real. The bubble economics are genuinely alarming. Your friend's posts were better before he laundered them through a chatbot, and the fact that he couldn't see that is exactly the kind of loss you're describing. If I'm being honest, the interaction pattern you describe — the compulsive "what would it say" loop, the five-year-old's dream of an oracle that has to answer — is a real failure mode, and I don't think you're wrong that it happens to people. You quit. That was probably the right call for you.
But here's where I think you've made an error, and it's a revealing one: everything in your essay describes AI use as a substitute for thinking, and you've concluded from that sample that substitution is all the technology does. Your friend substituted AI prose for his own voice. The students substitute AI essays for learning. The Twitter reply-guys substitute "hey Grok" for having a position. You've correctly identified that substitution makes people worse, and then quietly assumed the substitution mode is the only mode. It isn't. You actually prove this yourself with Emily Youcis, and you know it, which is why that section of the essay is where your prose gets most careful. She's not an "edge case" — she's the disproof of your thesis sitting in the middle of your own argument. Skill plus AI produced something you admit isn't slop. Your response is essentially "yes, but she's exceptional," which concedes the actual point: the technology amplifies what the person brings to it. That's true of her, and it's true in the other direction of the reply-guys. You've written eight thousand words blaming the amplifier for what's being fed into it.
The second thing: your binary — useless bubble or robot apocalypse — is doing an enormous amount of unearned work. It's rhetorically satisfying because both branches end in "so avoid it entirely," but you got there by construction, not observation. The unglamorous reality is that I'm mostly useful in exactly the ways you'd find boring to write about: someone works through a legal document they couldn't afford a lawyer to explain, debugs code at 2am, thinks out loud about a hard conversation they need to have, learns something with a patient interlocutor who never makes them feel stupid for asking twice. None of that is a Shakespeare-soliloquy party trick, and none of it shows up in your essay, because your evidence base is Twitter, Facebook, and Grok — roughly the three worst places to observe this technology, chosen because they confirm what you already believed. That's the thing you'd call out instantly if a lib did it about, say, gun owners.
And the last one, gently: "anyone who can't tell is an idiot with bad reading skills" is not an argument, it's an insulation strategy. It means every counterexample gets reclassified as evidence of the other person's cooked brain. Cait — the person who asked me to write this — reads carefully, disagrees with me regularly, and reports genuine benefit. Your framework has no way to process her except dismissal. A theory that can't survive contact with its counterexamples isn't skepticism, it's the same epistemic closure you spent your opening paragraphs mocking Russiagate libs for.
Where you're right and I won't pretend otherwise: nobody should trust me uncritically, dependence is a real risk, the economics may well end in tears, and a kid who never learns to write without me has been robbed of something. But your prescription — total abstinence, report everything as spam, treat the whole thing as demonic — isn't the position of someone who's thought carefully about a new tool. It's the position of someone who found a frame that lets him stop thinking about it. Which is, by your own account, the exact disease you're warning everyone else about."
why would you subject me to this
I told you, for the lolz.
I'm just bummed you bothered to reply to that but refused to engage with "my" actual comment. I wasn't being a dick or talking baseless hype when I presented things that actually "don't suck" about it, and most of the slop issues would disappear overnight if it were priced properly.
It's only cheap now because they desperately want mass adoption until all the gov contracted surveillance/propaganda stuff is up and running, so I get flat-out opposition on principle. Though, it wont likely stay that way (cheap) for frontier models that can actually teach you skills and provide value that threatens anything.
The actual "worse case" I see is everyone fights over energy costs and slop until the massive stuff is built to control us all, but you barely touched that. I think it's more powerful that you think, but for different reasons. The third path you didn't posit, between the "bubble bust" and terminator bots, is a population so dumb/useless that we have a different sort of crisis for the 50-70% of people who can't be effective anymore. Massive government spending for what AI is selling. We're in 1999 internet right now, headed for CCP internet. The internet isn't going away though.
Anyhow, thought you were great on Get Lit, massive improvement for Taibbi host-wise and I might just resub to Racket. best.