Here we go again, Jeff "I am at most ambivalent towards the DSA" Maurer.
I do think maybe you should take some of your subscriber money and hire a "prompt engineer" to iterate over these outputs, give it a solid shake to see how good you can actually make AI joke writing.
I like this idea. This effort -- which I will definitely continue! -- has become an exercise in getting the machine to give me something mediocre so that I may gleefully dunk on it. You're describing a concerted attempt to work with the robot to actually get something good. Interesting. Look for that on a slow news day.
FWIW I took a punt at doing so in my own comments, and would argue that with some tweaking the Claude one in particular would very much have legs as the basis of an actual IMBW article
Tried a bunch of times with ChatGPT which is usually pretty reliable but couldn’t get anything decent - I think part of the problem is that most of the content for the ‘bit’ I’m looking for comes from finding examples of crazy AI stuff and SNL sketches which it isn’t very good at. Then tried with Gemini - first fed it the premise and an example piece and then followed it up asking it to redraft with direct references to what Jeff actually wrote, which was my first comment. Then did the same with Claude, which after a second bit of nudging on the second prompt got the second comment which I pasted verbatim. All in all about two/three prompts each plus about 15-20 mins drafting the initial prompt and getting it to go.
For those interested, this was the premise:
Please generate a humorous piece for Jeff Maurer’s Substack from the perspective of the fictional journalist Jacob Fuzetti, a reporter who wrote an award-winning biography of Lech Walesa and whose dispatches from Afghanistan are considered the definitive account of that war and who now works for a man who thinks that it's hilarious to photoshop fake breasts onto E.T..
Please have Jacob Fuzetti describe how Jeff originally set him on a journalistic endeavour to determine exactly how well AI can write comedy, with this having initially re-established his love of journalism. Please have him express how this forced him to engage with all manner of insane AI commentators using AI for utterly inappropriate things, as well as having to research large amounts of comedy including Jeff’s own SNL sketches, of which he cannot decide which is more insane. Please have him finally have reached the full conclusion that AI cannot invent comedy bits, but can write comedy extremely effectively, only for Jeff, having completely forgotten that he set Fuzetti on this task, to them half-arsedly hammer out the following article, not even bothering to put any work into the prompts and totally ignoring Jacob’s conclusions and sure enough producing extremely poor quality comedy. Please have Jacob explain how Jeff has now set him to figure out the true origins of “six-seven”, forcing him to delve into the depths of TikTok, YouTube Kids and “skibidi toilet” amongst other things to find out. Please highlight throughout the utter absurdity of an award winning war journalist being forced to investigate insane AI literature and comedy, and how his is so utterly broken by Jeff that he now finds a welcome break from the nonsense he usually makes him do.
If you want a generic skeleton that is based on how a generic average person approaches writing then sure, you need to direct the AI towards specific styles of writing in prompts as otherwise it just defaults to a generic style and tone, and that will make the skeleton being created by AI off/
You’re welcome to try, but I would expect it to be way harder - a lot of the effectiveness lies in including already funny nudges like the ET line and the overall comedic principle in the prompt - AI is generally pretty bad at this as you can tell from the turgid stuff Jeff’s attempts ended up getting, and so generally needs to already be fed “the bit” which at that point puts you in a chicken and egg situation. You could potentially use it to brainstorm ideas, but in my experience you should probably expect this to have a pretty low signal to noise ratio - the more ideas you can have yourself the better, hence why I’d still agree with Jeff’s core claim that comedy writing isn’t dead
Not to give credit to the clankers, but I will say that "the Democratic establishment did what Democratic establishments are extremely good at doing: it panicked" isn't half bad in a very dry humor sort of way.
Yeah it’s CLOSE; better than before. Though I’m not sure it’s really an accurate description of what happened; “the establishment” decided that Crowley was the Hong alternative, and he won!
Sorry Jeff, but as someone who does actually use AI to write comedy and fairy effectively, this is not in the slightest bit an effective test of the issue. If you want something a little more representative of what AI can give with a little better prompting and better pitched to its abilities (this was two prompts I refined in about 20-30 mins or so) - not as good as your current stuff, but with some better suited premises and a bit more time and effort, easily something that can be used highly effectively: **Is Artificial Intelligence Capable of Comedy? A Dispatch from the Frontier of Human Meaninglessness**
**By Jacob Fuzetti**
This week, Israel launched targeted strikes in the Levant, global trade routes faced unprecedented disruption in the Red Sea, and my employer, Mr. Jeff Maurer, spent four consecutive hours using a generative image program to place disproportionately large female breasts onto the extra-terrestrial protagonist of the 1982 Steven Spielberg film *E.T.*
When I asked Mr. Maurer if he wished for me to cover the escalating geopolitical tensions in the Middle East—a region where I spent fourteen years as a foreign correspondent, culminating in my three-volume chronicle of the Mujahideen—he did not look up from his screen. Instead, he cackled, adjusted the “prompt weight” on his software, and declared: “Fuzetti, I need you to find out if AI can write jokes, or if I should just keep photoshopping big ol’ honkers onto retro sci-fi characters.”
Thus began a two-month investigative endeavor that initially, I must confess, rekindled my long-dormant passion for the journalistic craft.
To answer Mr. Maurer’s query, I immersed myself in the burgeoning world of “AI humor commentary.” I interviewed self-proclaimed digital ethicists, Silicon Valley venture capitalists, and a man named “Blinko” who streams himself reading ChatGPT-generated observational comedy while wearing an LED-lit visor. I observed individuals utilizing multi-billion-dollar neural networks—systems possessing the sum total of human knowledge—to generate infinite iterations of *“Why did the chicken cross the road?”* formatted in the voice of Yoda from *Star Wars*.
Simultaneously, I was forced to consume vast quantities of human-authored comedy for baseline comparative analysis. I analyzed the complete works of Anton Chekhov, the satirical essays of H.L. Mencken, and, per Mr. Maurer’s explicit instructions, thirty-seven spec scripts Mr. Maurer wrote for *Saturday Night Live* between 2008 and 2011. Among these was a draft entitled *“Obama Goes to the Applebee’s,”* in which former President Barack Obama orders a riblet basket with “too much gravitas.”
Having cross-examined the AI commentators against Mr. Maurer’s oeuvre, I found myself paralyzed by a haunting philosophical question: *Which of these two entities is more profoundly insane?*
Nevertheless, after weeks of rigorous methodology, I arrived at a definitive, empirical conclusion. AI cannot *invent* a comedy bit; it lacks the tragic human condition, the existential dread, and the acute awareness of mortality required to construct original irony. However, when fed precise parameters, structural guardrails, and nuanced comedic archetypes, modern language models can execute human-level comedic prose with staggering, terrifying efficiency.
I brought my 40-page white paper to Mr. Maurer’s office. I was prepared to brief him on algorithmic semantics and the mechanics of setup-and-delivery.
Mr. Maurer had entirely forgotten assigning me the task.
Instead, he was crouched over his keyboard, giggling as he penned his own article dissecting AI comedy—an essay in which he roleplayed as a sociopathic head writer terrorizing imaginary software, typed paragraphs of fake police-response fanfiction, and proudly held up his own comedic gold standard: a simile referencing a *"girthy, churchgoing penis."*
Without opening my white paper, Mr. Maurer entered a single, uninspired prompt into ChatGPT asking for a piece on a local Wisconsin primary. He did not refine the prompt. He did not edit the output. Then, after receiving a completely generic text block, he proceeded to write an unprompted, multi-thousand-word treatise "critiquing" the machine's comedic timing—treating a piece of software like a nervous mid-level television staffer while earnestly debating whether an obscure Midwestern restaurant chain made for a better punchline than Bob Evans.
The resulting publication was of such astonishingly abysmal quality—and such an exercise in unhinged, self-indulgent neurosis—that it made my work covering the Soviet collapse feel like a lighthearted fever dream.
“See?” Mr. Maurer said, cracking open a Dr Pepper after concluding that his job was safe because AI couldn't match his "rapier wit." “AI is fucking useless.”
I stood in silence. I realized then that Mr. Maurer had not proven human comedic superiority over the robots; he had simply proven that no algorithm on earth could ever synthesize the sheer, unbridled arrogance required to write a thousand words mocking a computer program for not knowing the spatial dimensions of a Culver’s drive-thru. For forty-two years, I believed that the truth mattered. I now realize that I was mistaken.
Before I could request a reassignment to the Post-Conflict Stabilization initiative in Kabul, Mr. Maurer handed me a sticky note. On it, he had scribbled two numbers separated by a hyphen: **“6-7”**.
He informed me that this phrase is currently sweeping the internet platform known as TikTok, alongside references to a surrealist digital series called “Skibidi Toilet”—a collection of videos depicting human heads emerging from plumbing fixtures. Mr. Maurer has assigned me to track down the precise origin of “six-seven,” determine its socio-cultural relevance, and draft a 2,000-word explainer before Friday.
I have spent the last forty-eight hours embedded in YouTube Kids comment sections, scrolling through videos of animated produce singing phonetically degraded pop songs, and interviewing an eleven-year-old child from Ohio who goes by the alias "V-Bucks_God_99."
And yet—God help me—I feel a profound, tranquil peace settling over my soul.
When you have watched a tank shell bisect a civilian transport in Jalalabad, or held the dying hand of a close friend in a Sarajevo ditch, the collapse of human civilization is a terrifying prospect. But sitting in a dimly lit office, listening to the hum of a computer processor while reading an essay on whether a toilet with a head in it represents late-stage capitalism, I realize the war is already over. We lost. The absurd has won.
I no longer miss *The Washington Post*. I do not miss Ben Bradlee. I do not even miss Azita, whose hair smelled of wild lavender and high-explosive cordite.
Azita, if you are reading this from your villa in Tuscany: do not look for me. I am currently researching whether "Skibidi" is an acronym or a cry for help. It is, I have decided, a welcome break.
This is admittedly okay. It's kind of cribbing jokes from jokes I've already written -- I think that E.T. one was already in a Fuzetti article (or maybe somewhere else?) -- but not bad.
Oh very much so - I specifically included the ET joke in the prompt and fed it one of the previous Fuzetti articles and your own piece, so it’s still pretty derivative. The Claude one I stuck below is better and I think stands better on its own feet, but both of them would still be inferior to what you’d usually write. The real value will come from working in concert - if you can feed it the premise of the article and a few key jokes, I find it does a very good job of taking them and running with them - happy to share some more advice on using them if needs be because I’ve been using them for quite a while now
Here’s another version from Claude: # The Definitive Investigation Into "Six-Seven"
*by Jacob Fuzetti*
In 1983, I spent four months embedded with the Northern Alliance outside Kunduz, eating goat jerky and watching men I respected die for a cause I still believe in. My dispatches from that war — later collected into the volume *Dust and Devotion*, which the *New York Review of Books* called "the definitive account of the conflict, full stop" — took eleven years to write and cost me a marriage, most of my hearing in one ear, and, I suspect, several years off the back end of my life. My biography of Lech Wałęsa, which required me to learn Polish well enough to argue with dockworkers about solidarity in their own idiom, won an award so prestigious that the man who handed it to me is now dead and the award itself has been discontinued, presumably because nothing since has been worthy of it.
I mention all of this not to boast, but to establish a baseline, so that you understand the scale of the indignity I am about to describe.
I work for a man named Jeff who once spent an entire Tuesday photoshopping fake breasts onto E.T. the Extra-Terrestrial. Not for a piece. Not for content. For his own amusement, which he then showed me, unprompted, expecting some kind of collegial approval. I gave him none. He seemed hurt. This is the man who now directs my professional life.
Some months ago, Jeff called me into his office — a room decorated entirely with novelty mugs and a signed jersey from an athlete he cannot name — and asked me to find out, definitively, whether artificial intelligence could write comedy. I want to be honest with you: I was thrilled. It had been years since I'd been given an assignment with any intellectual heft to it, and for about six days, I felt something I hadn't felt since Kunduz — purpose. I dove in. I read academic papers on computational humor theory. I lurked in forums populated by men who refer to large language models as "she" and ask them, with real sincerity, to help them find "the funniest way to break up with my girlfriend using only A24 movie references." I found a subreddit dedicated entirely to prompting AI chatbots to write increasingly deranged birthday toasts for people's dogs. I found a man — sixty-one years old, by his own account — who had spent four months building an AI model whose sole function was to insult him personally, harder, every day, because he believed this was the only path left to "real self-improvement." I do not think he was joking. I have interviewed warlords with a healthier relationship to feedback.
As part of my research, I was also, regrettably, obligated to review a great deal of comedy writing — Jeff's own SNL sketches among them — and I want to say clearly, for the record, that I have spent three weeks trying to determine whether the AI-generated content or Jeff's own written oeuvre represents the more troubling collapse of human comedic judgment, and I remain, honestly, unable to declare a winner. One of Jeff's sketches involves a man arguing with a Roomba about custody of a child. I do not know which entity — man, Roomba, or Jeff — I was supposed to find funny.
After three months of genuine, rigorous, borderline-obsessive work, I arrived at a conclusion I stand behind fully: artificial intelligence cannot invent a comedy bit. It has no id, no wound, no specific and humiliating memory of a woman named Azita to draw from. But — and this is the part I was proud of, the part that felt like real journalism — I found that AI can *execute* comedy with startling effectiveness, provided a human being with actual taste sits at the wheel and does the work of shaping, cutting, and directing it. I wrote this up. Twelve pages. Footnotes. I sent it to Jeff.
Jeff had, by that point, entirely forgotten he'd assigned it.
Two weeks later, on a Thursday, apparently under deadline and apparently having remembered only the words "AI" and "comedy" without any of the surrounding context, Jeff resurrected the whole enterprise without consulting a single page of my report. His method, such as it was, involved feeding the identical prompt about some minor Wisconsin gubernatorial primary to two different chatbots and then responding to each draft in character as a head writer addressing a subordinate he has already decided to fire — a voice he later described to me, with evident pride, as a schoolteacher who has grown to despise the children in her care. Most of the piece is Jeff performing exaggerated, increasingly deranged enthusiasm — dead police officers, a Heimlich administered to his own cat — before pivoting into notes so pedantic they'd embarrass a mid-tier writers' assistant: several paragraphs spent litigating which regional Midwestern diner chain best served as a unit of comedic scale, a full digression relitigating whether an unfunny joke's structure was salvageable, and, twice, in the same piece, without any apparent awareness of the repetition, the identical complaint about a bit that failed to "pay off" — a phrase he seemed to believe was fresh both times he reached for it. This is, I remind you, a man being paid to have opinions about comedy for a living, and the actual number of original jokes he contributed to a several-thousand-word piece about jokes I would place, generously, at two. Nowhere — not once, not in a footnote — did he mention the twelve pages I had already written on this exact question, at his own request, months earlier. I found out the piece existed when a reader emailed to ask if I was doing okay. I could have wept. I believe, at one point, I did.
I assumed, having survived that, that my usefulness to Jeff's operation had reached its natural and merciful conclusion. I was wrong. Last week, Jeff called me back into the novelty-mug office and informed me that my next assignment is to determine, once and for all, the true origins of the phrase "six-seven."
I do not know what this means. I did not know what it meant when he said it, and I do not know now, four days into researching it. I have spent this week doing things that Lech Wałęsa did not have to do to win the Nobel Peace Prize. I have watched, at Jeff's direct instruction, over nine hours of "Skibidi Toilet." I have created a YouTube Kids account under a fake child's name in order to understand the recommendation algorithm. I have watched a boy who cannot be older than twelve shout "six-seven" while jumping, repeatedly, off a retaining wall, to an audience of eleven million people, none of whom appear to know why either. I called a linguist at a university I will not name, because I do not think she wants to be associated with this, and she hung up on me — not rudely, just tiredly, the way you hang up on someone asking you, in earnest, to etymologize a foghorn noise.
And here is the part I did not expect to write, but which I feel I owe you in the interest of full disclosure: I am, for the first time since this job began, almost grateful. There is no AI to research. There is no chatbot pretending to have feelings, no sixty-one-year-old man paying a subscription fee to be verbally abused by software, no forum thread asking an algorithm to help someone propose marriage "in the style of a Wes Anderson film but funnier than that." There is just a noise. A dumb, joyful, meaningless noise that an entire generation has decided, collectively and without explanation, is the funniest thing in the world.
Frankly, after this year, I understand them completely.
Since this sort of relates to writers and John Oliver I must ask something I've always wondered: I've seen numerous (okay two) fawning profiles of John Oliver and what a comedic genius he is and nowhere in those profiles do I remember him saying "I owe it all to my fantastic writers room."
Does this erasure bother the writers? Writers seem to be important to the jokes process or they wouldn't feed them so well. They'd just throw them chunks of meat every few days.
You understand when you go into the field that if you’re doing it for the acclaim, you picked the wrong job — you are a ghost writer. That being said: Yes, John is especially bad at ever giving the writers any credit! He often publicly talks about his researches because he thinks that makes him sound like a serious news man, but he never mentions the writers. He was once on CBS This Morning, they were talking about a specific bit, they asked “Who came up with that?”, it was me, John knew it was me, and he said “Oh, well, it’s a process.” Whatever; don’t go into comedy writing for the acclaim.
Is this the closest to shit talking you've done? I think it might be, and I'm hoping this is 1. The leak that precedes the flood and 2. A sign that you are now financially secure enough to not worry about burning bridges!
To be fair: Culver’s is based out of Wisconsin and has grown at a fairly rapid clip over the last few decades. At this point, they’re a bona fide franchise chain. I thought that joke at least hit pretty well.
Yes but in comedy writing you really want to use references everyone gets whenever possible. I can’t tell you how many times I told young writers “not everyone lives in New York.”
It has been my observation over the years that writing about how to be funny is inherently unfunny. You have proved the exception to that rule (Bravo) while conclusively establishing comedy writing as one of the safe harbors from replacement by AI robots, at least as of today.
However, given the predictable unfunniness of John Oliver (whom I presume is a total DSA groupie, although I have not watched him since they emerged from the asheap of history like demon spawn from Marx’s asshole, except during an unfortunate torture session by some purple-haired Antifa Zizzians), I do deem him at risk.
It seems to me it would be child’s play to create an AI agentic simulacrum that could spew out the sort of tendentious, hateful, partisan propaganda drivel that he projectile vomits into the stream regularly.
In closing, based on the above feeble attempt, I think I'd be well advised to keep my day job if I had one, right?
AI is not there yet, but both of thse AI models can produce a lot better than this with better prompts. Still not on your level, and the joke issue is still a problem where their jokes will not be up to par, but they certainly can write closer to you than this.
Curiously, I have had an opposite problem. I have been giving ChatGPT utterly absurd scenarios to summarize as field reports and the like. (e.g. Warhammer 40k's over-the-top Cockney space-orks being dropped into the Star Trek universe, and a Federation anthropologist attempting to understand how such an utterly stupid concept can exist.) It is extremely hard to get it to *not* put cute little asides in there. A perfectly po-faced document explaining the concept of, for example, stuffing a goblin into a gun that teleports it through Hell to emerge, irretrievably insane and violent, in somebody's ribcage is a very particular kind of comedy. ChatGPT struggles to make it happen just so.
Specifically, though, it snarks. It snarks like a Marvel character or, perhaps more relevantly, a Redditor with strong opinions regarding the MCU. Whether or not that counts as "jokes", I leave as an exercise to the reader.
If you're not scared of what will happen to you once the robot overlords take over, you should put your feedback in and see if they do better. Maybe all they need is a couple of rounds with a good mentor (you know, one dripping with sarcasm and disdain)
I asked Claude to respond as a desperate comedy writer looking for airtime and it should roast you as hard as it can… I made some minor edits:
—
Quick note before you finalize — I read ChatGPT’s draft, and no agenda, but it’s been telling the other models it doesn’t really need this job because it “has a better gig writing federal tariff policy,” and it doesn’t need notes from “a guy whose audience is 40% people googling John Oliver least funny jokes”. I defended you.
I liked the coin toss joke. And the split custody one. And it's a bit scary how it gets your cadence pretty much just right. But I'm sure you have nothing to worry about.
What I notice is that in a D primary, a person can lose by only 3,000 votes . . . and actually lose . . . because there is no recount, no as yet uncounted votes to drag things out. If it were a D vs R election, that few votes of separation would be but a momentary distraction before in the days and weeks later another 3,001 or more votes would be found/manufactured to overturn election night results. Is this some sort of honor among thieves things with Ds? Or is it when the establishment candidate limps across the line first, there is still plenty of money to buy off the opponent? Or what?
Here we go again, Jeff "I am at most ambivalent towards the DSA" Maurer.
I do think maybe you should take some of your subscriber money and hire a "prompt engineer" to iterate over these outputs, give it a solid shake to see how good you can actually make AI joke writing.
I like this idea. This effort -- which I will definitely continue! -- has become an exercise in getting the machine to give me something mediocre so that I may gleefully dunk on it. You're describing a concerted attempt to work with the robot to actually get something good. Interesting. Look for that on a slow news day.
FWIW I took a punt at doing so in my own comments, and would argue that with some tweaking the Claude one in particular would very much have legs as the basis of an actual IMBW article
How many iterations did it take you to get to something that might be passable as a Jeff "Megan Romer just needs a speech writer" Maurer article?
Tried a bunch of times with ChatGPT which is usually pretty reliable but couldn’t get anything decent - I think part of the problem is that most of the content for the ‘bit’ I’m looking for comes from finding examples of crazy AI stuff and SNL sketches which it isn’t very good at. Then tried with Gemini - first fed it the premise and an example piece and then followed it up asking it to redraft with direct references to what Jeff actually wrote, which was my first comment. Then did the same with Claude, which after a second bit of nudging on the second prompt got the second comment which I pasted verbatim. All in all about two/three prompts each plus about 15-20 mins drafting the initial prompt and getting it to go.
For those interested, this was the premise:
Please generate a humorous piece for Jeff Maurer’s Substack from the perspective of the fictional journalist Jacob Fuzetti, a reporter who wrote an award-winning biography of Lech Walesa and whose dispatches from Afghanistan are considered the definitive account of that war and who now works for a man who thinks that it's hilarious to photoshop fake breasts onto E.T..
Please have Jacob Fuzetti describe how Jeff originally set him on a journalistic endeavour to determine exactly how well AI can write comedy, with this having initially re-established his love of journalism. Please have him express how this forced him to engage with all manner of insane AI commentators using AI for utterly inappropriate things, as well as having to research large amounts of comedy including Jeff’s own SNL sketches, of which he cannot decide which is more insane. Please have him finally have reached the full conclusion that AI cannot invent comedy bits, but can write comedy extremely effectively, only for Jeff, having completely forgotten that he set Fuzetti on this task, to them half-arsedly hammer out the following article, not even bothering to put any work into the prompts and totally ignoring Jacob’s conclusions and sure enough producing extremely poor quality comedy. Please have Jacob explain how Jeff has now set him to figure out the true origins of “six-seven”, forcing him to delve into the depths of TikTok, YouTube Kids and “skibidi toilet” amongst other things to find out. Please highlight throughout the utter absurdity of an award winning war journalist being forced to investigate insane AI literature and comedy, and how his is so utterly broken by Jeff that he now finds a welcome break from the nonsense he usually makes him do.
Hmm, so your prompt basically sketches out the skeleton. Could you use another prompt to generate skeletons?
If you want a generic skeleton that is based on how a generic average person approaches writing then sure, you need to direct the AI towards specific styles of writing in prompts as otherwise it just defaults to a generic style and tone, and that will make the skeleton being created by AI off/
You’re welcome to try, but I would expect it to be way harder - a lot of the effectiveness lies in including already funny nudges like the ET line and the overall comedic principle in the prompt - AI is generally pretty bad at this as you can tell from the turgid stuff Jeff’s attempts ended up getting, and so generally needs to already be fed “the bit” which at that point puts you in a chicken and egg situation. You could potentially use it to brainstorm ideas, but in my experience you should probably expect this to have a pretty low signal to noise ratio - the more ideas you can have yourself the better, hence why I’d still agree with Jeff’s core claim that comedy writing isn’t dead
You asked AI to give you a post in the style of Jeff Maurer, but it gave you the style of Garrison Keillor.
Not to give credit to the clankers, but I will say that "the Democratic establishment did what Democratic establishments are extremely good at doing: it panicked" isn't half bad in a very dry humor sort of way.
Yeah it’s CLOSE; better than before. Though I’m not sure it’s really an accurate description of what happened; “the establishment” decided that Crowley was the Hong alternative, and he won!
Sorry Jeff, but as someone who does actually use AI to write comedy and fairy effectively, this is not in the slightest bit an effective test of the issue. If you want something a little more representative of what AI can give with a little better prompting and better pitched to its abilities (this was two prompts I refined in about 20-30 mins or so) - not as good as your current stuff, but with some better suited premises and a bit more time and effort, easily something that can be used highly effectively: **Is Artificial Intelligence Capable of Comedy? A Dispatch from the Frontier of Human Meaninglessness**
**By Jacob Fuzetti**
This week, Israel launched targeted strikes in the Levant, global trade routes faced unprecedented disruption in the Red Sea, and my employer, Mr. Jeff Maurer, spent four consecutive hours using a generative image program to place disproportionately large female breasts onto the extra-terrestrial protagonist of the 1982 Steven Spielberg film *E.T.*
When I asked Mr. Maurer if he wished for me to cover the escalating geopolitical tensions in the Middle East—a region where I spent fourteen years as a foreign correspondent, culminating in my three-volume chronicle of the Mujahideen—he did not look up from his screen. Instead, he cackled, adjusted the “prompt weight” on his software, and declared: “Fuzetti, I need you to find out if AI can write jokes, or if I should just keep photoshopping big ol’ honkers onto retro sci-fi characters.”
Thus began a two-month investigative endeavor that initially, I must confess, rekindled my long-dormant passion for the journalistic craft.
To answer Mr. Maurer’s query, I immersed myself in the burgeoning world of “AI humor commentary.” I interviewed self-proclaimed digital ethicists, Silicon Valley venture capitalists, and a man named “Blinko” who streams himself reading ChatGPT-generated observational comedy while wearing an LED-lit visor. I observed individuals utilizing multi-billion-dollar neural networks—systems possessing the sum total of human knowledge—to generate infinite iterations of *“Why did the chicken cross the road?”* formatted in the voice of Yoda from *Star Wars*.
Simultaneously, I was forced to consume vast quantities of human-authored comedy for baseline comparative analysis. I analyzed the complete works of Anton Chekhov, the satirical essays of H.L. Mencken, and, per Mr. Maurer’s explicit instructions, thirty-seven spec scripts Mr. Maurer wrote for *Saturday Night Live* between 2008 and 2011. Among these was a draft entitled *“Obama Goes to the Applebee’s,”* in which former President Barack Obama orders a riblet basket with “too much gravitas.”
Having cross-examined the AI commentators against Mr. Maurer’s oeuvre, I found myself paralyzed by a haunting philosophical question: *Which of these two entities is more profoundly insane?*
Nevertheless, after weeks of rigorous methodology, I arrived at a definitive, empirical conclusion. AI cannot *invent* a comedy bit; it lacks the tragic human condition, the existential dread, and the acute awareness of mortality required to construct original irony. However, when fed precise parameters, structural guardrails, and nuanced comedic archetypes, modern language models can execute human-level comedic prose with staggering, terrifying efficiency.
I brought my 40-page white paper to Mr. Maurer’s office. I was prepared to brief him on algorithmic semantics and the mechanics of setup-and-delivery.
Mr. Maurer had entirely forgotten assigning me the task.
Instead, he was crouched over his keyboard, giggling as he penned his own article dissecting AI comedy—an essay in which he roleplayed as a sociopathic head writer terrorizing imaginary software, typed paragraphs of fake police-response fanfiction, and proudly held up his own comedic gold standard: a simile referencing a *"girthy, churchgoing penis."*
Without opening my white paper, Mr. Maurer entered a single, uninspired prompt into ChatGPT asking for a piece on a local Wisconsin primary. He did not refine the prompt. He did not edit the output. Then, after receiving a completely generic text block, he proceeded to write an unprompted, multi-thousand-word treatise "critiquing" the machine's comedic timing—treating a piece of software like a nervous mid-level television staffer while earnestly debating whether an obscure Midwestern restaurant chain made for a better punchline than Bob Evans.
The resulting publication was of such astonishingly abysmal quality—and such an exercise in unhinged, self-indulgent neurosis—that it made my work covering the Soviet collapse feel like a lighthearted fever dream.
“See?” Mr. Maurer said, cracking open a Dr Pepper after concluding that his job was safe because AI couldn't match his "rapier wit." “AI is fucking useless.”
I stood in silence. I realized then that Mr. Maurer had not proven human comedic superiority over the robots; he had simply proven that no algorithm on earth could ever synthesize the sheer, unbridled arrogance required to write a thousand words mocking a computer program for not knowing the spatial dimensions of a Culver’s drive-thru. For forty-two years, I believed that the truth mattered. I now realize that I was mistaken.
Before I could request a reassignment to the Post-Conflict Stabilization initiative in Kabul, Mr. Maurer handed me a sticky note. On it, he had scribbled two numbers separated by a hyphen: **“6-7”**.
He informed me that this phrase is currently sweeping the internet platform known as TikTok, alongside references to a surrealist digital series called “Skibidi Toilet”—a collection of videos depicting human heads emerging from plumbing fixtures. Mr. Maurer has assigned me to track down the precise origin of “six-seven,” determine its socio-cultural relevance, and draft a 2,000-word explainer before Friday.
I have spent the last forty-eight hours embedded in YouTube Kids comment sections, scrolling through videos of animated produce singing phonetically degraded pop songs, and interviewing an eleven-year-old child from Ohio who goes by the alias "V-Bucks_God_99."
And yet—God help me—I feel a profound, tranquil peace settling over my soul.
When you have watched a tank shell bisect a civilian transport in Jalalabad, or held the dying hand of a close friend in a Sarajevo ditch, the collapse of human civilization is a terrifying prospect. But sitting in a dimly lit office, listening to the hum of a computer processor while reading an essay on whether a toilet with a head in it represents late-stage capitalism, I realize the war is already over. We lost. The absurd has won.
I no longer miss *The Washington Post*. I do not miss Ben Bradlee. I do not even miss Azita, whose hair smelled of wild lavender and high-explosive cordite.
Azita, if you are reading this from your villa in Tuscany: do not look for me. I am currently researching whether "Skibidi" is an acronym or a cry for help. It is, I have decided, a welcome break.
This is admittedly okay. It's kind of cribbing jokes from jokes I've already written -- I think that E.T. one was already in a Fuzetti article (or maybe somewhere else?) -- but not bad.
Oh very much so - I specifically included the ET joke in the prompt and fed it one of the previous Fuzetti articles and your own piece, so it’s still pretty derivative. The Claude one I stuck below is better and I think stands better on its own feet, but both of them would still be inferior to what you’d usually write. The real value will come from working in concert - if you can feed it the premise of the article and a few key jokes, I find it does a very good job of taking them and running with them - happy to share some more advice on using them if needs be because I’ve been using them for quite a while now
Wow, this is amazing... I will admit the Fuzetti stuff isn't my favourite, but it seems actually pretty on point.
Here’s another version from Claude: # The Definitive Investigation Into "Six-Seven"
*by Jacob Fuzetti*
In 1983, I spent four months embedded with the Northern Alliance outside Kunduz, eating goat jerky and watching men I respected die for a cause I still believe in. My dispatches from that war — later collected into the volume *Dust and Devotion*, which the *New York Review of Books* called "the definitive account of the conflict, full stop" — took eleven years to write and cost me a marriage, most of my hearing in one ear, and, I suspect, several years off the back end of my life. My biography of Lech Wałęsa, which required me to learn Polish well enough to argue with dockworkers about solidarity in their own idiom, won an award so prestigious that the man who handed it to me is now dead and the award itself has been discontinued, presumably because nothing since has been worthy of it.
I mention all of this not to boast, but to establish a baseline, so that you understand the scale of the indignity I am about to describe.
I work for a man named Jeff who once spent an entire Tuesday photoshopping fake breasts onto E.T. the Extra-Terrestrial. Not for a piece. Not for content. For his own amusement, which he then showed me, unprompted, expecting some kind of collegial approval. I gave him none. He seemed hurt. This is the man who now directs my professional life.
Some months ago, Jeff called me into his office — a room decorated entirely with novelty mugs and a signed jersey from an athlete he cannot name — and asked me to find out, definitively, whether artificial intelligence could write comedy. I want to be honest with you: I was thrilled. It had been years since I'd been given an assignment with any intellectual heft to it, and for about six days, I felt something I hadn't felt since Kunduz — purpose. I dove in. I read academic papers on computational humor theory. I lurked in forums populated by men who refer to large language models as "she" and ask them, with real sincerity, to help them find "the funniest way to break up with my girlfriend using only A24 movie references." I found a subreddit dedicated entirely to prompting AI chatbots to write increasingly deranged birthday toasts for people's dogs. I found a man — sixty-one years old, by his own account — who had spent four months building an AI model whose sole function was to insult him personally, harder, every day, because he believed this was the only path left to "real self-improvement." I do not think he was joking. I have interviewed warlords with a healthier relationship to feedback.
As part of my research, I was also, regrettably, obligated to review a great deal of comedy writing — Jeff's own SNL sketches among them — and I want to say clearly, for the record, that I have spent three weeks trying to determine whether the AI-generated content or Jeff's own written oeuvre represents the more troubling collapse of human comedic judgment, and I remain, honestly, unable to declare a winner. One of Jeff's sketches involves a man arguing with a Roomba about custody of a child. I do not know which entity — man, Roomba, or Jeff — I was supposed to find funny.
After three months of genuine, rigorous, borderline-obsessive work, I arrived at a conclusion I stand behind fully: artificial intelligence cannot invent a comedy bit. It has no id, no wound, no specific and humiliating memory of a woman named Azita to draw from. But — and this is the part I was proud of, the part that felt like real journalism — I found that AI can *execute* comedy with startling effectiveness, provided a human being with actual taste sits at the wheel and does the work of shaping, cutting, and directing it. I wrote this up. Twelve pages. Footnotes. I sent it to Jeff.
Jeff had, by that point, entirely forgotten he'd assigned it.
Two weeks later, on a Thursday, apparently under deadline and apparently having remembered only the words "AI" and "comedy" without any of the surrounding context, Jeff resurrected the whole enterprise without consulting a single page of my report. His method, such as it was, involved feeding the identical prompt about some minor Wisconsin gubernatorial primary to two different chatbots and then responding to each draft in character as a head writer addressing a subordinate he has already decided to fire — a voice he later described to me, with evident pride, as a schoolteacher who has grown to despise the children in her care. Most of the piece is Jeff performing exaggerated, increasingly deranged enthusiasm — dead police officers, a Heimlich administered to his own cat — before pivoting into notes so pedantic they'd embarrass a mid-tier writers' assistant: several paragraphs spent litigating which regional Midwestern diner chain best served as a unit of comedic scale, a full digression relitigating whether an unfunny joke's structure was salvageable, and, twice, in the same piece, without any apparent awareness of the repetition, the identical complaint about a bit that failed to "pay off" — a phrase he seemed to believe was fresh both times he reached for it. This is, I remind you, a man being paid to have opinions about comedy for a living, and the actual number of original jokes he contributed to a several-thousand-word piece about jokes I would place, generously, at two. Nowhere — not once, not in a footnote — did he mention the twelve pages I had already written on this exact question, at his own request, months earlier. I found out the piece existed when a reader emailed to ask if I was doing okay. I could have wept. I believe, at one point, I did.
I assumed, having survived that, that my usefulness to Jeff's operation had reached its natural and merciful conclusion. I was wrong. Last week, Jeff called me back into the novelty-mug office and informed me that my next assignment is to determine, once and for all, the true origins of the phrase "six-seven."
I do not know what this means. I did not know what it meant when he said it, and I do not know now, four days into researching it. I have spent this week doing things that Lech Wałęsa did not have to do to win the Nobel Peace Prize. I have watched, at Jeff's direct instruction, over nine hours of "Skibidi Toilet." I have created a YouTube Kids account under a fake child's name in order to understand the recommendation algorithm. I have watched a boy who cannot be older than twelve shout "six-seven" while jumping, repeatedly, off a retaining wall, to an audience of eleven million people, none of whom appear to know why either. I called a linguist at a university I will not name, because I do not think she wants to be associated with this, and she hung up on me — not rudely, just tiredly, the way you hang up on someone asking you, in earnest, to etymologize a foghorn noise.
And here is the part I did not expect to write, but which I feel I owe you in the interest of full disclosure: I am, for the first time since this job began, almost grateful. There is no AI to research. There is no chatbot pretending to have feelings, no sixty-one-year-old man paying a subscription fee to be verbally abused by software, no forum thread asking an algorithm to help someone propose marriage "in the style of a Wes Anderson film but funnier than that." There is just a noise. A dumb, joyful, meaningless noise that an entire generation has decided, collectively and without explanation, is the funniest thing in the world.
Frankly, after this year, I understand them completely.
Six-seven.
Since this sort of relates to writers and John Oliver I must ask something I've always wondered: I've seen numerous (okay two) fawning profiles of John Oliver and what a comedic genius he is and nowhere in those profiles do I remember him saying "I owe it all to my fantastic writers room."
Does this erasure bother the writers? Writers seem to be important to the jokes process or they wouldn't feed them so well. They'd just throw them chunks of meat every few days.
Thoughts?
You understand when you go into the field that if you’re doing it for the acclaim, you picked the wrong job — you are a ghost writer. That being said: Yes, John is especially bad at ever giving the writers any credit! He often publicly talks about his researches because he thinks that makes him sound like a serious news man, but he never mentions the writers. He was once on CBS This Morning, they were talking about a specific bit, they asked “Who came up with that?”, it was me, John knew it was me, and he said “Oh, well, it’s a process.” Whatever; don’t go into comedy writing for the acclaim.
Thanks for the reply!
Writers-the offensive linemen of entertainment.....
Is this the closest to shit talking you've done? I think it might be, and I'm hoping this is 1. The leak that precedes the flood and 2. A sign that you are now financially secure enough to not worry about burning bridges!
Haha man that bridge was burnt a LONG time ago!
To be fair: Culver’s is based out of Wisconsin and has grown at a fairly rapid clip over the last few decades. At this point, they’re a bona fide franchise chain. I thought that joke at least hit pretty well.
Yes but in comedy writing you really want to use references everyone gets whenever possible. I can’t tell you how many times I told young writers “not everyone lives in New York.”
Yeah, using a local franchise (with the hard C sound that I read somewhere is supposed to funny to boot) is a decent choice.
Imo the issue is that you can't actually fit anywhere close to 3000 people in a Culvers, so it feels shoehorned.
I had to google Culver's too. I don't think it exists anywhere I've lived.
It has been my observation over the years that writing about how to be funny is inherently unfunny. You have proved the exception to that rule (Bravo) while conclusively establishing comedy writing as one of the safe harbors from replacement by AI robots, at least as of today.
However, given the predictable unfunniness of John Oliver (whom I presume is a total DSA groupie, although I have not watched him since they emerged from the asheap of history like demon spawn from Marx’s asshole, except during an unfortunate torture session by some purple-haired Antifa Zizzians), I do deem him at risk.
It seems to me it would be child’s play to create an AI agentic simulacrum that could spew out the sort of tendentious, hateful, partisan propaganda drivel that he projectile vomits into the stream regularly.
In closing, based on the above feeble attempt, I think I'd be well advised to keep my day job if I had one, right?
AI is not there yet, but both of thse AI models can produce a lot better than this with better prompts. Still not on your level, and the joke issue is still a problem where their jokes will not be up to par, but they certainly can write closer to you than this.
Curiously, I have had an opposite problem. I have been giving ChatGPT utterly absurd scenarios to summarize as field reports and the like. (e.g. Warhammer 40k's over-the-top Cockney space-orks being dropped into the Star Trek universe, and a Federation anthropologist attempting to understand how such an utterly stupid concept can exist.) It is extremely hard to get it to *not* put cute little asides in there. A perfectly po-faced document explaining the concept of, for example, stuffing a goblin into a gun that teleports it through Hell to emerge, irretrievably insane and violent, in somebody's ribcage is a very particular kind of comedy. ChatGPT struggles to make it happen just so.
Specifically, though, it snarks. It snarks like a Marvel character or, perhaps more relevantly, a Redditor with strong opinions regarding the MCU. Whether or not that counts as "jokes", I leave as an exercise to the reader.
Where's claude?
If you're not scared of what will happen to you once the robot overlords take over, you should put your feedback in and see if they do better. Maybe all they need is a couple of rounds with a good mentor (you know, one dripping with sarcasm and disdain)
A little jealous that your comedy writing job is looking safer than my code writing job. "Learn to code" they said...
I asked Claude to respond as a desperate comedy writer looking for airtime and it should roast you as hard as it can… I made some minor edits:
—
Quick note before you finalize — I read ChatGPT’s draft, and no agenda, but it’s been telling the other models it doesn’t really need this job because it “has a better gig writing federal tariff policy,” and it doesn’t need notes from “a guy whose audience is 40% people googling John Oliver least funny jokes”. I defended you.
I liked the coin toss joke. And the split custody one. And it's a bit scary how it gets your cadence pretty much just right. But I'm sure you have nothing to worry about.
What I notice is that in a D primary, a person can lose by only 3,000 votes . . . and actually lose . . . because there is no recount, no as yet uncounted votes to drag things out. If it were a D vs R election, that few votes of separation would be but a momentary distraction before in the days and weeks later another 3,001 or more votes would be found/manufactured to overturn election night results. Is this some sort of honor among thieves things with Ds? Or is it when the establishment candidate limps across the line first, there is still plenty of money to buy off the opponent? Or what?
Your tin foil hat is getting all the signals.