The workslop war of all against all
Here's to AI: the cause of, and solution to, all of life's problems.
The new desirable employee archetype is a “cracked twenty-two-year-old,” Sun said, using the slang for a hyperproductive, extremely online programmer who might work “nine-nine-six,” a term adopted from workers in China that refers to a schedule of 9 A.M. to 9 P.M., six days a week. The only way to escape the permanent A.I. underclass, ironically, is to lean in and hustle in a bot-like way. “Rather than being politically radicalized, everyone grinds harder,” Sun said. The reward for the grind might be a role as an overlord of the A.I. future: the closer to collaborating with the machine you are, the more power you will have.
Kyle Chayka, The New Yorker, 8/10/25
Let me begin with a personal anecdote to illustrate my broader theme, dear reader.
Recently, a potential new client, hereinafter referred to as PNC, asked me to write an article. One about how people can more effectively leverage AI, as it happens.
In the ChatGPT Beforetimes, it was common for clients to email a one-paragraph or sometimes one-sentence directive.
PNC mailed over a ‘core’ brief that stretched to 17 pages and almost 3000 words about the article to be written. They also sent a set of guidelines (15 pages; nearly 3000 words) about what PNC expected from its potential new freelancers, hereafter PNFs, a brand bible (5 pages; 1500 words) for the corporate client the article was being created for, and at least one other guide about how the article should be formatted for submission.
What’s more, most of the guides also contained links to further guides.
I was provided with such a wealth of information because, nowadays, it’s possible to bust out a brand bible or set of contractor/freelancer guidelines with remarkably little time or effort involved – with the assistance of AI, of course.
Unfortunately, reading and absorbing information takes as long as it ever has. Probably much longer, in fact, given the way the Internet has destroyed everybody’s attention span.
But two can play at the labour-saving-delegation game.
I may have uploaded the multitude of guides to ChatGPT and asked it to provide a concise summary of what was expected of me.
One mandatory requirement, reasonably enough, was that I generate the content myself, rather than farming out that task to AI.
As has been the case with all the riding instructions I’ve received of late, there was a somewhat tortured passage around the use of AI in at least one of the many sets of guidelines I was sent.
The gist of it was that while PNC wasn’t so naïve as to imagine AI wouldn’t be used in the creation of the article, it shouldn’t be used for the writerly heavy lifting.
Indeed, I was instructed to run my copy through an AI-content-detection tool and hold off submitting it until it was classified as 80 per cent human-generated.
Did I nonetheless use AI to do much of the grunt work on the article?
Yes, I did.
Before you judge me too harshly, I should point out I was being paid less than 20 per cent of my pre-ChatGPT word rate.
To make anything like the money I did three years ago – not an extravagant amount to begin with – I needed to write at least five times faster.
When I ran the final draft through the checker, it came back as only seven per cent artificially generated. So, I either made a substantial contribution to the end product or ChatGPT can now mimic my tone of voice sufficiently well to beat the system.
At the end of the experience, I was left wondering if everyone wasn’t better off back in ye olde days, when a client sent me a concise directive and I generated some (premium) content using technology no more magical than a search engine.
Increasingly, the not inconsiderable efficiencies gained by leveraging AI are offset by the not inconsiderable costs of having to plough through extra work generated by AI.
It’s the system, maaan!
As our economist friends never tire of pointing out, people respond to incentives. When the incentive changes, so does the behaviour, so it should be no surprise everyone’s behaviour as adjusted now AI has achieved mass penetration, at least among the knowledge-worker class.
The commissioning editor who sent me several hours of background reading material was incentivised to provide as much information as possible to minimise the chances I would submit an article that didn’t meet the necessary criteria.
Given the modest a fee I was receiving for my labours, I was incentivised to use AI to get a short summary of my riding instructions and then complete the article as quickly as possible.
Of course, the commissioning editor and I are just part of a much larger group of people who now default to assuming the other party in any business transaction is probably using AI to game the system in their favour.
Both the more common and shameless forms of AI-facilitated ‘system gaming’ are increasingly attracting public and media attention.
Five seconds after ChatGPT launched in late 2022, high school and university students realised they could now get AI to do their homework for them. Not too long afterwards, the more digitally savvy teachers and academics realised AI could be used for admin, lesson planning and marking.
Students have automated away the studying, and teachers have automated away the teaching.
To take another well-publicised example, job applicants and recruiters are also stuck in an AI arms race. Applicants use AI tools to ‘keyword stuff’ words into CVs and cover letters that suggest they are a good fit. If they reach the try-out/interview stage, they likely use AI to help complete trial tasks and may use a ‘co-pilot’ tool, such as Live Interview AI, which ‘delivers perfect responses in real-time as you interview’.
Recruiters haven’t been sitting idly by as applicants have been availing themselves of AI helpers. They have been investing in copilot-detection systems such as Sherlock, which “can identify if candidates are reading from off-screen sources or exhibiting thinking patterns consistent with external help”.
Of course, should you get a job or hire a candidate, the AI-enabled tit-for-tat continues.
It’s estimated that around two-thirds of large companies monitor their staff via keystroke logging, screenshot capture, video recording and AI-based productivity ranking systems. Microsoft’s Workplace Analytics, for example, monitors time spent on websites, email response rates, and after-hours work.
The workers of the white-collar world have united, where feasible, to frustrate these productivity-monitoring efforts with cunning ruses such as mouse jigglers and dual monitors (one for work, one for fun and/or a second job).
The siren song of labour-saving AI has tempted those from all walks of life. Deloitte Australia recently had to partially refund a $440,000 fee it charged the Australian government.
Presumably well-educated, well-remunerated and well-respected Deloitte staffers used AI to generate a 237-page report. One with references to academic papers that didn’t exist, quotes from judges that were never uttered, and citations to research that was entirely made up.
Is the curtain coming down on the productivity theatre show?
I have good news and bad news regarding the ‘AI slop’ issues canvassed above.
The good news is that they will go away in the near future.
The bad news is that they will go away once the human worker can be eliminated from the process.
Let me illustrate by returning to my own experiences in the workslop trenches.
If you are from a business or tech background, you may have wondered earlier in the article why my potential new client didn’t simply feed the briefs and endless guidelines into an AI and have it generate the article. After all, that would have saved considerable time and some money by disintermediating the human content creator.
This is precisely what many businesses, especially small to medium-sized ones, now do. However, there remains some market demand for ‘artisanal’ human-created content, especially among larger, deep-pocketed companies.
The only reason that market demand still exists is that AI can’t quite replicate what human writers provide.
I’m not making that point self-defensively, mainly because I long ago resigned myself to AI surpassing human writers in the nearish future.
But for the moment, as Perplexity puts it, “AI-generated content is formulaic and derivative, lacking the creativity, emotional depth, originality, contextual nuance, and authentic personal voice that talented human writers consistently bring through lived experience.”
Which inevitably brings me to the big AI story of the week – AI-generated actress Tilly Norwood.
Tilly takes centre stage
AI-generated faux humans should be old news by now.
Mia Zelu, an AI-generated influencer, shot to fame during Wimbledon in the middle of the year. Around the same time, the news leaked out that psychedelic rock band Velvet Sundown was AI-generated.
But Tilly has received a lot more attention than Mia Zelu or fictional bandmates Lennie West, Gabe ‘Mellotron sorcerer’ Farrow, Milo ‘synth alchemist’ Rains or Orion ‘Rio’ Del Mar.
About two years after SAG-AFTRA (Screen Actors Guild-American Federation of Television and Radio Artists) went on strike for 118 days to stop the AMPTP (Alliance of Motion Picture and Television Producers) from using ‘synthetic performers’, the AMPTP, and its foreign equivalents, are salivating at the prospect of finally ridding themselves of pricey, troublesome human performers.
SAG-AFTRA is talking a big game, but it appears American actors are now roughly in the same situation American factory workers were in after Bill Clinton got NAFTA passed and ushered China into the WTO.
Here’s the ‘brave face’ statement from the unions.
SAG-AFTRA believes creativity is, and should remain, human-centered. The union is opposed to the replacement of human performers by synthetics. Tilly Norwood is not an actor. It’s a character generated by a computer program trained on the work of countless professional performers—without permission or compensation… we have enormous leverage because audiences want real human performers in their movies, TV shows, animation, video games, audiobooks and all the other platforms our members work on.
The tragedy of that statement is that it’s only half true.
Tilly was indeed summoned into existence by drawing on the work of countless human actors. Just as AI was also trained on the work of countless animators, architects, visual artists, authors, cartoonists, composers, copywriters, graphic designers, illustrators, journalists, musicians, painters, performers, photographers, playwrights, poets, screenwriters, scriptwriters, songwriters, voice actors and writers.
Speaking as an erstwhile print journalist, I’d warn my thespian brothers and sisters that huffing, “You’ll miss us once we’re gone,” isn’t likely to prove an effective survival strategy. I imagine most consumers already don’t care if it’s “a real human performer” in animation, video games and audiobooks. I suspect they ultimately won’t be much bothered by synthetic performers being part of hit Netflix series and blockbuster Hollywood films.
If consumers are unperturbed by synthetic performers, the members of AMPTP can look forward to making out like bandits. Or even bigger bandits than they currently are, as the case may be.
Of course, the outlook for members of SAG and AFTRA will be considerably grimmer.
And so it has come to this
Generative AI excels at generating content, so it’s no surprise that it’s come for the content creators first. But I imagine the dynamics sketched out above will play out across most occupations between now and 2030.
There will continue to be some work available for a while, involving tasks that AI currently struggles with. Such as writing non-formulaic, emotionally resonant copy.
But eventually, we writers, and just about everybody else, will be in the same situation a young actor trying to launch a career in Hollywood is now in.
In the meantime, it appears we should all dig in for sloppy trench warfare between the bosses/clients and the workers/vendors.