An Update On How I Use AI To Help With Techdirt (It’s Still Not Writing Articles)
from the oh-no-techdirt's-gone-slop dept
Over in our Insider Chat, someone recently pointed out a tiered “REAL Rating” five-tier scale designed to replace the fairly blunt and misleading question of “did you use AI on that?” with something a bit more nuanced. As someone who believes the discussion about AI could use a lot more nuance, I’m all for the premise. I ran my own usage through their wizard and came out with a “one” — the tier for automation. While this is technically accurate, it also tells you very little about how I actually use AI.
Which is why I thought it made sense to revisit the public discussion regarding my AI usage.
But going through that thought process reminded me that it’s been a little while since I last spoke about how I’m using these tools here, and given that some readers want more transparency, I thought I’d update you on the latest. I’ve written about it at length a few times before, starting in April of 2024, when I described how I used a tool called Lex to help me edit articles. The key was always that I’d write the articles myself, and then I had trained Lex on my own writing style to ask it how it might improve the article. Sometimes it would give good advice that would make me rewrite, and sometimes… it would give bad advice that I would skip or ignore. But all in all it was a useful forcing function.
In June of 2025, I wrote about how I had built a personal task management tool, using a vibe coding tool called Lovable, which I was using to keep track of various tasks, including what I wanted to write about each day. At the beginning of this year, mostly out of curiosity, I explored whether or not you could, with a ton of scaffolding and detailed instructions and effort, get AI tools to write a passable Techdirt article, and concluded no, you absolutely could not. Every attempt was still a weak facsimile of a Techdirt article with iffy language and clichéd phrases.
In the time since then I’ve run that test for myself a few times, and even as the models have improved, in my experience, my initial analysis stands. It cannot write Techdirt-style articles well, and even if I wanted to use it that way, the rewriting, changing, and fact-checking would take way more time than just… writing it myself. So, yes, I continue to write my own articles.
What’s changed in the last couple of years is that I’ve moved away from Lex (which has not received many updates in the past year or so as the team behind it appears to have moved on to other projects), and I ended up building my own editing tool, which is built directly into the task management tool I mentioned last year (which is now self-hosted and entirely within my control). The benefit now is that I can turn any task immediately into a writing project with a single click:
That was the task that turned into this article about Judge Andrew Oldham’s bizarrely out of context quote on mail-in ballots. From there I can go into the writing project which has a bunch of useful features. Within each writing project I can add sources, which turn out to be really useful during edits. I can upload PDFs, URLs, or just add text myself (and actually, as I was finishing up this article, I had it build a new feature that will automatically add all the sources I link to in a story, so the editor checks them even if I don’t add them manually, which is another example of realizing a potentially useful feature and having it created on the spot). I can also have the tool go out and research if there are other sources I should add, though I haven’t yet found that to be that useful as compared to just searching myself.
The tool is specifically trained on my writing, and has a detailed style guide regarding my voice and style, as well as a detailed system prompt of what I am looking for in editorial help vs. what I’m not looking for (e.g., yes, challenge my assumptions and facts, but no do not think you need to rewrite what I’ve written).
After I’ve written the full piece, there’s a button to open up the “AI review” tool (which lets me use any underlying LLM engine I choose, including local ones), which will give me a critical review and make general suggestions about what works and what doesn’t. I do switch up models from time to time to see how they work, but generally will default to either Claude Opus 5 or Sonnet 5. I am also using Gemma 4 locally and am increasingly trying to use that. It’s good on straightforward pieces, but less so once the argument gets more complicated. I’m hopeful that the next generation of local models will be more reliable.
For example here’s part of the initial review of my first draft on the Oldham ruling:
The only specific line edits tend to be suggestions on typos or confusing sections, such as here where it suggested a way to clean up and tighten a poorly drafted sentence in my initial draft:
You can also see that “Discuss” button in there, which I will frequently use to debate or discuss certain suggested edits if I think the suggestions miss the mark or are unnecessary. Or I’ll just reject them if I totally disagree.
Usually at that point I’ll do a more thorough rewrite based on my own reread of the article with an eye towards what it called out as problematic. For what it’s worth, this is where the inclusion of original sources is often quite helpful, because the AI will press me on things if its review felt that I’m not being completely fair and honest regarding the source material. Sometimes it calls me out for missing stuff and suggest I dig deeper (it’s not always correct, but often is, and pushes me to reread things even more carefully). Here’s one example, where the AI editor tool told me to go back and review the original documents more closely:
It also catches me when I start rambling, which often happens on a first draft when I’m trying to figure out the best way to nail down a certain point:
Again, I don’t always follow its suggestions, but it does make me revisit and look at rewriting sections to make them clearer and tighter.
Separately, I have a built in set of “sub-agents” that will check certain things, including suggesting better headlines (this is very hit or miss, but usually at least helps me think of something better than either my initial attempt or its suggestions) and one sub-agent whose whole job is to poke holes in the arguments in the piece (this one is fun, but annoying, as it’s basically trolling myself, but in a way that helps me make the arguments stronger):
You may notice that the features I appreciate the most here are the ones that in some sense make my life harder, by telling me to work harder, check my sources, and question my arguments. Considering how often we hear about the problem with these tools being how much smoke they blow up the asses of users, it’s certainly possible to set them up to deliberately challenge and provoke you. The parts I find most useful are the parts built to disagree, because it helps make my work stronger by stress testing it before it goes out in the world.
Anyway, once the piece is then written and edited, it’s a two-click process to import the final version into Techdirt’s WordPress system and hand it off to someone else on the team for a human review before it goes into the queue to get posted.
Compared to how I described the process two years ago, the actual writing process hasn’t changed — which is kind of the whole point. The human part (the thing that is central to what I do) doesn’t get automated away. What’s changed is the structure around it. I have total control over the tool now, it’s better trained and designed to help me write better.
There are a few other more experimental features, including one where after I’ve added all my sources to the writing project, I’ll also write a short paragraph (or two or three) about my focus on the article, and the system will ask me some questions based on the sources and how I want to frame the article. This is sometimes useful in testing some of my assumptions before diving into the writing. This has been most helpful when I’m a bit undecided on what the central backbone of a piece really should be. In the past I would just start writing, realize I went down the wrong path, and then start again. Using the “editorial interview” feature has helped me a few times to think deeper about which way I want to take the article before I get to the actual writing.
I also built in a tool to also write a potential outline of the article based on that “interview,” but I’ve only tried to use that a couple times, and neither time did I follow the outline, so I’ll likely just get rid of that feature. Once again, the benefit of having total control over the tool. For all the (reasonable! understandable!) complaints of big tech companies “shoving AI into everything,” with this tool, I decide what goes where, and if something doesn’t work, I can just remove it.
I’ve also built in a bunch of other features that are simply useful for organizing my day and focus. The same tool creates a morning and afternoon briefing based on my tasks, my email, my calendar, and a growing knowledge base about me and my work. It also texts me news stories it finds that might be interesting for Techdirt (based on search criteria I have provided), which I can turn into tasks/writing projects with one response. Just last week I received my Pebble Index 01 ring that I had mentioned last year, and one of the first things I did with it is hack it so that I can speak into it to create tasks. It can also research things and leave the notes in the task tool (and text me back quick facts if I need them).
In short, as I’ve suggested repeatedly, the power of AI tools is in empowering individuals to be able to accomplish what they want to do, rather than replacing humans or doing the fundamental work. Building very personalized tools that I can constantly update and tweak to meet my needs has made me less and less interested in using software that other people built for me. It’s shockingly empowering to look at a feature that doesn’t quite work the way you want, and to be able to just tell your tools “hey, make this work differently” and have it work properly.
As I mentioned up top, using the wizard on that REAL Ratings site, I came out with a score of “one,” noting that I used AI for automation. You could argue that it maybe should be somewhere between one and two, depending on how much you consider the “brainstorming” part of what the editor feature is doing. But I’d argue this is a long way from the assumption many have of “if you use AI you’re using it for slop.”
Either way, I’m still not writing articles with these tools, and I don’t plan to start. The assumption that any use of AI means you’ve handed your brain over to it misses the part I actually find interesting: building your own tools, on your own terms, and telling them what to do rather than the other way around.
Filed Under: ai, editing, editorial review, llms, writing
Companies: techdirt
Comments on “An Update On How I Use AI To Help With Techdirt (It’s Still Not Writing Articles)”
the ai sounds like a commenter
not me of course
The problem with leaning on LLMs/“AI”/whatever term the HOA-level nitpickers demand I use is that the tech smooths out all the rough edges of your writing. “How is that a problem,” you might ask. Because sure, editing some sections for better understanding and such sounds like a fine idea. But it also robs you of the voice that is uniquely yours. It subtly pushes you towards a version of your writing that’s missing all the different little flourishes and tics that make your writing yours instead of what a machine mind tells you is “better”.
The cartoon Home Movies had an episode with a joke where Brendan and his friends ended all their movies with the same line (“it’s time to pay the price”). At the end of the episode, when they try to end a movie with some other line despite that line being perfect for the ending, they stammer and stutter and can’t come up with anything. What does that have to do with your use of “AI”? Simple: Once you see all the little tics and repeated words/phrases pointed out to you, you’ll try to avoid them more often—but in doing so, you’ll end up either running out of things to say or struggling to come up with something that isn’t one of those tics. Your writing will lose its signature and start to sound like everyone else’s.
I do appreciate that you don’t depend on “AI” tools for your writing. But leaning on them even a little is…well, even with all the ethical issues aside, it’s a way to offload your thinking to something that doesn’t actually think. That’s a pretty dangerous road to start going down, no matter how slow you walk.
Re: I don't think this has anything to do with AI
What you describe doesn’t seem to have anything to do with using AI, rather it’s just a side-effect of using an editor. This same effect occurs using a human editor. Many publications end up with a somewhat homogenized “tone” across multiple writers due to editing and feedback within the same group of individuals.
Re: Re:
It’s not so much the editing process that’s the problem, so much as it is what the editing process ultimately strips away. That holds especially true when the “editor” is an LLM designed primarily to lean toward generic speech, because the LLM only “understands” averages and predicts what works “best” to achieve a “mean” result.
I use em-dashes. A lot. (Have been for years, even before AI writing exploded as a thing.) An LLM these days might tell me to use fewer em-dashes or fewer asides because (A) it’s tighter writing and/or (B) it would help dispel the notion that I write using AI. But by giving up the em-dash for the sake of avoiding AI accusations, all I’d be doing is killing a little piece of myself that is part of my style. Hell, I’ll even let you in on a bit of a secret about how I think in regards to my writing comments here: With the longer ones (like this one), as I go back to proofreed, I tend to read them back in my head like I’m doing a wrestling promo. It’s why I often try to end on a strong sentence—I treat that bit like a mic drop moment.
(And since I’m pointing out my dumb bullshit: That “hell” a couple of sentences back, and other similar such one-word exclamatory sentence-starters, are another thing I do often that an LLM would probably tell me to cut in the interests of “better writing”.)
I get that an editor will ultimately aim for helping a writer achieve clarity in their writing. But LLM “editors” aren’t people; they’re not sensitive to writing styles and the actual feelings of actual people. They aim for the middle and try to push people there. In so doing, they make writing that much more generic and unexciting. I’d gladly take a few spelling or grammatical errors in some unique writing over “correct”-but-boring AI-influenced slop every day of the week.
Re: Re: Re:
Please read what I wrote. THAT IS NOT how I use the tools. Why do you keep pretending I’m using it that way?
Yes, if you use generic ChatGPT and ask it to rewrite what you wrote and accept all its suggestions.
I do literally none of those things. I use a tool deliberately trained on my style of writing, with explicit instructions not to adjust or suggest anything having to do with my tone or style, and to give feedback like a regular human editor, with explicit instructions about the audience here, and then it gives me more general suggestions that I use to do my own edit.
The only line edits it gives is in the cases where there are legit typos or run-on sentences or whatever.
Honestly, this feels like you’ve decided in your head how I must be using AI and no amount of “that’s not the tool I use, and it’s not how I use the tool” will move you off of that.
Re: Re: Re:2 “We know by now that technology can be toxic as well as tonic.” — Douglas
Those three respective quotes are from the article and two of your comments. I did see that in the article before I commented on this page.
More from curiosity than anything else, is the tool some sort of neural network or whatnot trained exclusively on your writing, or is it an LLM with its training fine-tuned to your writing? You must have a formidable corpus of your own work, but I doubt it could suffice to train an LLM from scratch.
This whole discussion brings to my mind a quote from the greatest U.S. Supreme Court Justice of all time, William O. Douglas—the First Amendment literalist and civil liberties curmudgeon. Not from the bench, but as much in his private capacity as a Justice ever is, Douglas wrote these words in 1970 (boldface is mine):
In 1970, Douglas’ most urgent concern about technology was that, as he so presciently forewarned, electronic computer databases would destroy privacy and thus impose mass social control as had never before been possible. On those grounds alone, Douglas compared the invention of the computer to that of the atomic bomb. He said so in 1970.
Mike, I presume that everyone in this discussion loves technology. Although of course, I can speak only for myself, I assume that most everyone here has at some point given the “it’s not the tool, it’s how you use it” speech that you make in some of your comments here. This is not exactly an audience of Dr. Kaczynski’s acolytes.
Please consider the reasons why from such technology-lovers, “AI” evokes reactions ranging from sharp criticism, to blazing hatred, to darkest despair. You evidently desire to show everyone what you think are good ways to use “AI”. The quest may be heroic, or it may be quixotic. Either way, please do understand why so many technologists see this technology not merely as the invention of an atomic bomb, but as an active carpet-bombing of virtual atomic warheads upon the fabric of human society.
(An historical footnote: Then-Congressperson Gerald Ford tried to get Justice Douglas impeached for the above-quoted book. Partly for its content, which frankly reminded the young people of 1970 that they have a right to rebel, and even to overthrow the government—and partly because excerpts thereof were published in Playboy and in the Evergreen Review. On the latter account, Grove Press later lost a First Amendment case on a 4-4 tie because Douglas ethically recused himself to avoid a conflict of interest.)
Re: Re: Re:3 do re mi
Addressing the first point about training an LLM on Masnick’s writing, Techdirt has been running for, what, 20 years or so? I think he can find a couple of writing samples kicking around here somewhere.
Re: Re: Re:4
29 years as of this week. Next year will be 30 years.
And the site has 86,000 posts, 53,000 of which were written by me. So, yeah, there’s a pretty decent corpus of material for training.
Re: Re: Re:5 Is that the only corpus?
I knew off the top of my head that Techdirt was about thirty years old, for I saw it relatively early. Though this isn’t one of those “I’ve been reading you for so many years!!” things. My life has been convoluted; it’s not you, it’s me. I remembered this site in early 2025, when I was trying to rediscover Web 1.0 sites—oh, yay, it’s now an anti-Trump site.
A “formidable corpus”, as I put it in my great-grandparent comment.
But that neatly leaves unanswered a question which at least two different people (including me) have politely asked: Is your “AI” writing assistant tool trained exclusively on your writing, or is its existing model weighted towards your writing?
53,000 articles seems small. Perhaps my inferences may be mistaken about your tool, but the full text of Wikipedia would seem small. LLM trainers have ingested shadow libraries with hundreds of millions of books and academic papers, they’ve relentlessly scraped untold billions of pages from the Web, and they’ve even crashed the Internet Archive scraping old websites that no longer exist. A corpus of 53,000 articles sounds like a training set for the types of ML-nerd projects that people did before ChatGPT made LLMs famous.
I don’t like to feel like I’m publicly interrogating you about your computer gizmos. But in the article and in at least two of your comments on this page (as I quoted in GGP), you’ve stood explicitly on the assertion that your tool is trained on your own writing. You didn’t say if it’s trained only and exclusively on your own writing.
It’s a relevant question. In a context where you saw a need to append to your article’s headline, “(It’s Still Not Writing Articles)”, and you’ve thrice emphasized that it’s trained on your own writing—all to defend and positively to advocate for something which, for well-considered reasons, many people frankly hate—it’s unavoidably relevant.
Re: Re: Re:2
Fair point; I assume responsibility for the errors in my thinking.
That said: Are you sure it’s trained only on your style, or is it an existing model that’s weighted toward the local data you overlaid on it? Because if it’s the second option, you might be putting a bit too much trust into a model that isn’t fully “yours”.
Re: Re: Re:2 Ceci n’est pas un humain.
Mike, by the way, in case you may have wondered: My email address is real. I’ve even used it publicly; in the archives of a mailing list, you can find a PGP key for me as an downloadable attachment at the bottom. It’s an unusual vanity address; it’s splendid for my handle. Try it if you want, preferably E2EE.
So-called “AI” has sucked the social discourse into a vortex of inhumanity, amidst which I’m left invisibly waving my arms and shouting, “I’m human! I’m real!” My text is not I—it is my representation of me.
But Stephen T. Stone is a obviously an AI summarizer: He repeated my question in a concise form. 😉
Re: Re: Re:3
Hey now, I take great offense to being called a summarizer! I’ve worked too hard at being long-winded to take that kind of shit. 🙃
Re: Re: Re:4 “and the devil hath power”
Oh, no, you’re so nice and concise. If I don’t watch myself, I could accidentally write a ten thousand word commentary on the Shakespearean aphorism, “brevity is the soul of wit”.
Re: Re: Re: I so love em dashes, I’ve got one—on—my—Xorg—keyboard—layout.
A “good” writer obeys such rules. A great writer knows when and how to break them.
Kids who grow up dependent on this shit will never explore the artistic—or sometimes ironical—or even self-referentially, recursively ironical effect that one can attain (or, I ought say as I’m here unsure of hitting my mark, to which one may aspire) when one dares disobey mind-numbing stylistic “rules”.
—Whoopsie. Did I just do something that an LLM cannot fucking do because it’s a dumb text-extruder?
LOL. It’s such a very self-aware cheating tool, it instructs you on how not to get caught with it. (To be clear, that’s multiply-layered sarcasm; I am not suggesting that the LLM is actually intelligent or self-aware!)
I’m sometimes shockingly naïve. Since I am not on social media, I am so oft not attuned to current trends.
Do you suppose that the reason why very few people have replied to my comments is an assumption that I’m copypasting LLM output? Am I being paranoid? The thought first occurred to me when I was writing my first comment on this page. It is distinctly horrifying.
Thereto, I’d assumed that I was being TL;DRed—which I should expect on a Net brain-damaged by Twitter and by smartphones with tiny screens—or folks just didn’t like me—which would be fine, on grounds that not everyone in the world needs to like me.
As for you, holy hell. Never change your human writing style to conform to machine-biased denials of your humanity! When a human is so dehumanized, we are living in a cliché sci-fi dystopia at the plot point right before the humans rise up to smash the machines.
Of course, what I just said shows that humans are in a double-bind. It is necessary to out frauds who pass off LLM output as their own writing—and preferably, if possible, to electrocute them. I don’t know what to do.
And so it came to pass that as humans felt a need to avoid natural writing that was thought to smell LLMish, the humans homogenized their own writing into generic slop of what was thought to be “humanlike”—upon which the LLMs did then train…
Re: On Writing and Reading
Well said. Albeit not always practicable, it’s preferable to offer a constructive solution when critiquing a problem; thus, I will offer to all and sundry some unsolicited advice about writing.
The best way to improve your own writing is twofold: (a) Read long-form prose by authors of great eloquence, and (b) avoid reading and writing Twitter-tier “txt speak” garbage.
Whereas reading quality text exercises your mind, and it thus helps you to be the best writer you can be. Beyond that, habitually reading and writing trains your inner monologue into an inner dialogue. The process of writing then becomes a medium for exploring, organizing, and developing your own thoughts; you thus discover new ideas that never occurred to you before you wrote them down. I need not remark that an LLM can’t do that for you.
For my part, I have never used an LLM writing tool of any kind whatsoever. The only automated writing tool that I use is spell-check to assist reading proof—just as I’ve done since my childhood word processor experiences. Incidentally, I have never taken a writing class in my life.
I read. I read quality prose. I love to read. When I write, this is an example of a comment that I dashed off today while I had a splitting headache. I did substantially edit the first draft, and I read proof; I spent much more time editing than I did on initial writing. In the process, I refined my ideas such that I understood my own thinking better than when I started.
I’ll let my prose tell the reader whether or not I’m a competent writer…
…and I’ll try to keep such things to myself, to avoid being plagiarized by LLMs. It is not about money, which I’m not being paid. It is not about copyright; in principle, at least, I yearn to abolish copyright laws! I do not consent to be used to train mass-plagiarism machines.
Moreover, I’m deeply disturbed to realize that if I don’t disclaim using an LLM, people will probably assume that I do. Worst of all, some may not even believe my claim to write all by myself. So-called “artificial intelligence” has turned the Web into a classroom where everyone who’s not a drooling idiot is assumed, or even conclusively presumed to cheat.
The predictable outcome is the ultimate reduction of society to the lowest common denominator, the final commodification of art, the death of the intellect, and the end of human discourse as a meeting of human minds.
Re:
I don’t know what to tell you other than this is just not even remotely true. It might be true if I fed all my stuff through a generic AI tool and told it to rewrite everything, but I’ve been pretty clear I don’t do that.
My tool is heavily trained on my existing writing and knows what to expect, and does very, very, very few specific line edit suggestions, and never to “remove the edge” from what I write. (It did once complain that I hadn’t been edgy enough in explaining how “fucking crazy” a story really was…)
None of the suggestions it makes are regarding “voice.”
I’m not sure why you assume otherwise when I’m pretty clear about this. Also, I don’t magically turn my brain off and accept every suggestion it makes.
So I really don’t understand your complaint other than that you insist I must be using AI in a way I directly have explained I’m not.
Honestly, your complaint applies WAY more to editors of mainstream publications that I’ve written for than anything I’m using AI for.
Re: Re:
As a former 4chan veteran: Yeah, nah, you don’t get nearly as fuckin’ vulgar in your articles as you probably could. 🙃
In the couple of times I fiddled with an LLM in this regard, I remember it suggesting language changes that felt like it was aiming less for mere clarity of communication and more for the kind of middling, try-to-please-everyone language that I was talking about. And that’s on top of the usual structures of AI writing that it likely suggested. It’s been so long since I tried it that I forget the exact suggestions it made.
And considering most of them are likely in the bag for AI… 🤣
Re: Did you not read the article
Or just not understand it?
What a crock.
Good for you?
What you don’t seem to get is that in the real world outside of your bubble the usage demanded by corporate shit heads is one where the AI does all the work and you are the reviewer.
Re:
No, I absolutely do get that, which is why I have called out repeatedly why that’s a stupid way to use the tool, and put up posts like this to show that there are better ways.
What do you run the local stuff on? Mostly curious if you need something like a DGX Spark/Mac Mini for the ram for larger models, or what GPU to keep the output snappy.
Would love to see more stuff on local models, despite all the problems I have with the industry, that is one aspect that is kind of exciting
Re:
I have a Linux desktop machine with an AMD Ryzen chip. The output is… not snappy if I’m using the larger models. For simpler tasks I use the 4B model which is reasonably quick. For bigger tasks I have the 31B model. And it’s slow. But I’m willing to deal with the slowness to keep things local. I did have to change some of my tools so they didn’t time out, but it’s fine. I can set it to do the editorial review while I go work on something else (like responding to comments).
Re: Re:
I’m currently hovering over the button to buy some cast off enterprise gear so I can run dual Tesla P40s at home. Surprisingly I can get it all for about 1500 bucks. Includes a 24 disc JBOD shelf populated with 900 GB drives. should allow me to run a 70B model comfortably.
I ended up ditching AI altogether. Having to go through several paid products or at least one paid product to train an LLM on writing styles and my own creative twist thrown in and spending hours babysitting AI to try and come to a conclusion that could have easily happened by me just writing an article and being done with it was simply not worth it. It was certainly an experiment, but the insane complexity of trying to wrestle AI into submission to do something half decent just removed me from the equation altogether. Much happier just loading up the back end, writing out the article with raw by-hand HTML, doing the usual research to bolster the points I’m making, and publishing. I think my money is better spent hiring a human if I ever get to that point.
I know several software engineers on Mastodon who’ve found this same thing: the big LLMs are useless, but if you build a small model trained on your own code and documentation it can produce usable results. By small I mean something you can train and run on a single Threadripper box. That makes sense, an LLM produces the most likely next text based on it’s training data and if it was trained on your code it’s going to produce the kind of code you’d produce rather than what the average coder on the Internet would produce.
I’ve found them quite helpful for improving my writing as a sort of beta reader, but I only ask for criticism and tell it not to give me examples or specific suggestions. This way my voice doesn’t get fucked with at all but I still get good feedback on what needs attention. I think I would find your workflow much too intrusive, and I do think it has altered your voice somewhat.
Useful but is all this worth trillions in debt?
Transparency
Thanks for this, I do appreciate the transparency. I wish more writers followed your example here, even though we can’t expect most of them to be as tech-savvy as you in building custom tooling.
I think it’s also a good example of LLMs being especially useful for everything but prose generation, something I also believe about the use of LLMs for Wikimedia projects.
This is very disappointing
Unlike most readers of this site, I’ve actually studied AI (and related technologies) and I had the singular privilege of doing so under some of the people who wrote textbooks on the subject(s). So — also unlike most readers of this site — I understand how it works…or doesn’t. With that in mind:
This comment has been flagged by the community. Click here to show it.
Always knew the trump derangement syndrome garbage on this site was AI generated
Why do you have a button just for arguing with the computer?
Choosing to draft your work with this process is of course your own prerogative, but I find it difficult to understand what benefit lies in posing questions to something which does not actually understand any points you make. Sure, you get a conversation-shaped exchange for what that’s worth (rubber-duck programming comes to mind), but how can you truly grow and advance your craft when this dialogue is shackled to the realm of text? A human editor (in ideal circumstances) is a colleague, not just someone who looks at a document and makes suggestions…
Re:
As mentioned in the article, a human editor reviews everything I write. But what I’m delivering to them is way better.
I’ve had many human editors over the years, and I don’t know what to tell you other than that you’re wrong. A well crafted AI tool with a detailed prompt (i.e., not generic Claude or ChatGPT or whatever), gives way more thoughtful, detailed editorial notes than any human editor I’ve ever had, including when I’ve written for major publications.
Also, as I’ve said in the past, one of the benefits of an AI tool is that when it’s wrong, it’s much easier to ignore than when a human editor is wrong and I have to fight over a change they want.
So I’m very supportive of having human editors, but this is a situation where both have their place, and my lived experience is that having the well-trained AI editor do the first pass is very, very helpful in making sure that when I do get the final draft to my colleague, the human editor, it’s in a much, much stronger position, and we can have a more thoughtful conversation about where to take the piece.
I feel like much of the criticism about things like “conversation-shaped exchange” is based on people who haven’t used the latest models or haven’t used tools that are deliberately trained in a certain way.
Re: Re:
That probably comes from people hearing about AI chatbots and the like, then thinking “holy fuck, why would I talk to a machine like it’s a person”. Which…yeah, I get that—I wouldn’t want to treat an LLM like a person, either.
Re: Re: Re:
Yeah. To be honest, I think the “chatbot” setup is understandable as to why it took off, but I personally don’t get it at all. It just feels weird.
Though, also, I found this interview with Bridget Todd quite interesting today:
https://www.nytimes.com/2026/08/19/podcasts/bridget-todd-ai-chatgpt.html
Bridget is very thoughtful, and a smart critic of tech where it deserves criticism. And my position has been similar to where her’s was, that using AI as any sort of human shaped companion is a mistake and not good. But she found one (limited) use case where it was apparently incredibly, deeply, helpful to her.
Re: Re: Re:2
I’m kind of lucky in that regard, in that the sycophantic yes-man nature of LLM-powered chatbots can’t really affect me because I’m too fucked up to accept even random mild compliments normally, never mind a constant stream of ass-kissery.
I do believe the technologies we colloquially refer to as “AI” can and will have legitimate uses in some fields. Science and health will be especially important fields for such technology.
Where I draw the line are creative fields. The very things that make us human, that connect and bind us in a shared experience—art, music, movies, stories, poetry—shouldn’t be offloaded to a fucking machine. It can never think or feel; it can never have an “a-ha!” moment of breaking through writer’s block or an “oh, that’s interesting” feeling when you experiment with an art style or a character design and find something that works perfectly with what you were drawing.
Offloading cognitive function to LLMs is bad enough. Offloading the creative drive goes beyond laziness and speaks to a wrongly entitled nature. Generative AI is meant to allow wealth to access talent without letting talent access wealth; it’s a shortcut to a result that misses how the work is the truly important part of creativity.
Re: Re: Re:3
So for a project I’m working on, I recently went and looked at some of the discussion about when film cameras were introduced in the 19th century, and people made very similar claims about that, being a lazy tool for failed painters and whatnot.
I agree that it will lead to lazy art, but I’m not entirely convinced that’s such a big problem. Such art will suck and will disappear. At the same time I keep coming across other examples, like the exiled Iranian filmmaker who wanted to make a documentary about the protests from earlier this year but (for obvious reasons) was unable to return to Iran to do so, and used AI tools to help make the film, which has won praise on the film circuit.
So, yes, lazy artists beget lazy art, but I’m not sure the tools automatically lead to lazy art. For some artists, it can be empowering.
Re: Re: Re:4
I can’t draw for shit. Would it be “empowering” for me to go to a site with an AI generator, grab a model that lets me make art in a style that looks like the work of my favorite artists, and call the end result “my art” even though I didn’t draw it and I had to steal other people’s work to make it?
Part of the issue with AI-generated slop is the ease with which it can create passable-looking work. Another part of it is the major ethical issue with how the models for such generators are trained—specifically, the ethical issue of taking a shitload of people’s works without their knowledge or consent and training a model on said works even (and sometimes especially!) when an artist says “don’t use my work for AI, you fucking weirdos”. And that’s not even getting into all the other issues with generative AI and the “AI” industry in general, especially the “environmental damage” issues.
I’m reminded of three “AI”-related stories in the news within the past couple of weeks:
Generative AI is supposedly “empowering” and the future of creativity, if’n you believe the AI evangelists. So why do people and corporations seem to be terrified of being caught using it?
Ask AI for me about choices
Which families drank polluted water, inhaled diesel exhaust, lived under light pollution, and paid more expensive electric rates so you could do this?
Thanks for helping to raise the cost of memory chips for me. That’s useful!
The point of writing is to write
Not to get someone to write for you!
I can write but do not really edit. Also the words that I write down are not the ones I am thinking of, I have been shocked when I go back later. Stupid copilot trying to suggest words when I type them!
Actually writing stuff down on paper has been my most successful writing exercise to date. 2000 words in 2 hours, took 6hrs to type up!
Who cares what you use it for?
The first question we should be asking is, “Is the LLM you’re using ethical to use?” Since it’s Claude, the answer is: No, absolutely not. Those tools were built on mass copyright infringement, and are explicitly intended by their creators to further concentrate power into an unaccountable fascist technical elite class. “The purpose of an LLM is to give wealth access to skill without giving skill access to wealth.”
As a judge might say, we don’t even need to reach the question of whether the way you use the tool is effective or not, or what negative effects it might have: Using the tool itself is already so hilariously unethical that the actual task at hand is irrelevant.
Re: “AI” has corrupted free culture.
Well said, except that copyright is not the issue. I’m anticopyright in principle. I used to recommend Anna’s Archive to people, before they became “AI” cheerleaders selling special access to the trainers of mass-plagiarism machines.
People who are pro-copyright are obviously against the “AI” industry. People who are anti-copyright should also be—but as I said in another comment that’s not yet showing up, “AI” is like Trump: It corrupts all what it touches.
I’m against LLM training on public domain books, which have no copyright. How about you? I’m against copyright licensing deals with LLM trainers, and I would never agree to one no matter what money I was offered. You?
It’s idealistic to free culture from the corporate copyright industry, to give unlimited access to books to underprivileged humans, and to encourage humans to read. It’s disgusting to support cheating tools that not only strip-mine culture for the purpose of polluting the world with slop, but also poison the well of human discourse with the assumptions that everything is fake, everyone who’s too smart must be cheating, and maybe cheating is okay because everyone does it.
“AI” is the first technology ever deployed to connect exploitation to pollution, with nothing in between and nothing beyond. LLMs are created by ridiculously evil companies who corrupt the financial markets so they can buy up all the world’s computational capacity and spend unlimited energy, in a quest to arrogate to themselves unlimited power. Regardless of copyright, I agree that Claude is created by unethical people by unethical means to unethical ends, and it’s thus unethical to use.
Re: Mass-Plagiarism of Free and Open-Source Software
You express a concern about “copyright infringement”. What is your position on the “AI” strip-mining of free and open-source software?
justine.lol(personal website).The author of that piece is a scary-smart, ingeniously creative programmer—who has worked on “AI”, who contributes to open-source “AI”. She’s horrified to discover that LLMs trained on code she’s licensed as open-source do not give her any credit. All she wants is credit for her work!
Boldface is mine:
Well, good luck enforcing your copyright license against the mass-plagiarism of your work that you benevolently gave to the world as a gift. I say that what the LLMs are doing is wrong even with PD/Unlicense/CC-0 code. As I remarked last week, the complete works of Shakespeare are in the public domain; that doesn’t make Shakespeare fair game for plagiary.
I’m not nearly as talented a programmer as Justine is—but I’m good enough to understand and to appreciate Cosmopolitan, which is arguably the most brilliant low-level portability hack ever invented. For my part, as I remarked in another thread here last week, the existence of “AI” deters me from contributing to open-source software.
“AI” : Technology :: Trump : Politics
A confession of a dirty secret: I agree with some of the positions that Donald Trump purports to hold. It’s unsurprising. I’m an avowed independent, off the left-right spectrum—and as for Trump, as the saying goes, even a stopped clock is right twice a day.
I don’t support Trump on the issues where he allegedly agrees with me. That’s not mere partisanship, but to the contrary: I am protecting my pet issues.
Trump is the corruptor. All what Trump touches turns to shit. Trump first corrupts, then discredits.
If I have spent the past thirty years futilely advocating Foobarian Quxification, and Trump suddenly throws his power behind a policy of “Foobarian Quxification”, I feel sick to my stomach. I know that my beloved Quxification project is toast. Trump will somehow twist Foobarian Quxification into a way to hurt people while enriching himself and his cronies, most likely without even delivering any of its actual benefits. It’s just what he does—and he always does this. And as a second-order result, Foobarian Quxification will be tarred with such infamy that reasonable people will treat me as batshit crazy for even mentioning it.
If I believed in God, I would pray for Him to change Donald Trump’s heart to disagree with me about everything.
I am not a domain expert on “AI”—far from it. But I’m enough of a technophile that years before ChatGPT, I was basically aware of generative networks and evolutionary algorithms and other ML goodies. It seemed so exciting!
Reviewing the discussion on this page, revising my opinions, I now feel for “AI” such a hatred that I know how Dr. Kaczynski must have felt when he sent his mailbombs.
“AI” is the Donald Trump of technology. It corrupts all what it touches. It is poison for human society. It turns good things bad, and bad things worse. I’m now ashamed that I used to say that some of the various technologies in the “AI” grab-bag were really cool—and I know that if I’m still a sort of closeted admirer of some of those technologies, I’d better work overtime to burn “AI” to the ground and salt the ashes.
I most certainly will not write essays on why Trump is allegedly useful for a few good things. Nor will I minimize, nor downplay, nor equivocate about Trump’s badness on similar grounds. Indeed, by doing the latter, some intelligent authors have diminished, or even altogether discredited themselves in my eyes. It’s particularly been a problem with libertarians—including at least one person whom I’d admired for many years, whom I’ve seen associate in the past on an issue I care about (Internet free speech) with Mike Masnick.