Work slop

noreply@beansprout.observer (B.S. Observer) — Sun, 01 Nov 2026 22:05:02 +0200

Today I’m reporting from muddy trenches in the fields of mediocrity where work slop is not just accepted, it’s endorsed.

The first lines of trenches were drawn when your colleagues suddenly discovered that the synthetic output of a text-extruding machine could pass for a job done. Everyone was a bit excited at the time, and generating slop was a bit of a silly take on the job at hand. The trench digging started right after your colleague asked you to review more lines of code or text than usual. It continued with you running to the trenches after you had your first argument with your coworker about the chatbot being wrong and about the need to double-check its output. Now we’re here, covering in deep trenches under a bombardment of thousands of lines of code and text that a human would never have produced in the past, yet a human is expected to read and vet.

There is a limit to your patience when covering in the trenches. At some point you’re likely to go on attack. The poor sod bearing the brunt of the heat may find it unexpected but the fault ultimately lies with your existing biases amplified by their work slop. Being just annoyed or having an argument with your peer is one thing; however if you are in a position of power, your amplified biases may affect your coworker’s performance evaluation and eventually their salary.

As Alex Hanna puts it in her video¹ explaining the myth that “AI use will raise wages”:

… generative AI can reduce wages in other ways, namely by punishing workers who use the tools and pass on work slop to their colleagues. … researchers report that receiving generative AI text or images from colleagues results in lower perceptions of the competence of their colleagues. Getting slop from your colleague makes you think worse of them. And if those workers are women or older or people of color, then it confirms their co-workers biases against them.

There are ways to deal with biases, and some work slop can be eliminated by existing processes at your disposal, or just by communication. The problem arises when the pressure to generate work slop is exerted by external forces. If your boss expects you to produce results faster using a word-guessing program, are you going to lower your quality bar or slowly slide towards karōshi² - a death by overwork?

Your health is important, but so is the sense of a job well done. If you don’t trust the machine, you spend a lot of time reviewing its output. If it outputs a large amount of data, you become a bottleneck when reviewing it all. At some point you find yourself at odds with the clock. When you reach that fork in the road, you will likely opt for lower-quality output because time is the scarcest resource, and spending it on slop is not worth it.

The low quality, however, will come back to haunt you sooner or later. It can be your coworkers thinking less of you, the hours of maintenance later, or unhappy customers trashing your brand. Outsourcing quality to a word-guessing program, will leave you with the fallout.

The trenches in the fields of mediocrity are filling up with slop. Climbing out means joining a herd of reverse centaurs³ who are there not to oversee the AI’s work, but to take the blame for its mistakes. Staying in means keeping control but also working more hours to keep up with the slop while being called a loser or “the professional fax machine operator in the internet age”⁴.

The future may seem bleak, but don’t worry. The AI bubble will soon burst, and you’ll be in a better position being assisted by a word generator than assisting one.

References

Comments

@beansprout_observer@mastodon.social

Deskilled by word-guessing program

noreply@beansprout.observer (B.S. Observer) — Sat, 10 Oct 2026 20:02:23 +0200

This is my first report from the muddy trenches of the 21st century. We’re being overrun by a technology which goes by many names, a word-guessing program (by Cory Doctorow), stochastic parrot² (by Timnit Gebru & Emily M. Bender), large language model, chatbot, but many people call it just “AI”. Pronounced like “pain” - synonym for anguish and your first thought whenever someone hands you a slop generated by a theory-free inference³ machine.

Many of us felt awestruck the first time we saw an output generated by a word-guessing program. Some may have even experienced revelation about statistical regularity of routine sentences⁴. You no longer need an infinite amount of time for key-hitting monkey⁵ to produce any given text. You just need a large enough typewriter where keys are bananas of varying appeal to the monkey.

The technology had potential to become normal technology⁶, a code completion plugin, a word processor extension, a text summarizer add-on, a better language translator, etc. Instead, thanks to AI boosters and doomers, it’s pushed as an anything-machine to all aspects of our lives while at the same time being presented as a danger to our lives. Go figure!

You can still shut it off from your personal life by making decisions about where and how you shop, search, study, write, code, etc., but if your boss is trying to replace you with a word-guessing machine it’s something you can’t simply ignore.

Your first instinct may be to find ways to incorporate the machine into your work and hope for the best, but what if the wannabe anything-machine produces non-stop error output and you’re forced to correct its errors? Or what if the machine produces rubbish but you can’t recognize it because you didn’t have time to learn the needed skill?

Instead of you producing good output, your work is reduced to identifying bad output and in the worst case, fixing the bad output by applying more of the same.

Rather than learning new skills, you’re forced to produce synthetic output and hope that no one will notice or judge you by the output.

We’re deskilling by being just an appendage of the machine or by not learning the skills at all.

Software engineers are at the forefront of the struggle between token-printing machine overlords and the rest of us. They labor in trenches when their boss asks them to add an asshole ✦ button to every web page and app they can think of. They smile & wave when their manager shows them an app created using slop-bot. They fend off requests to move faster and silently curse all the useless code and text they need to review to keep the software running well. But they also see how everyone around them is slowly deskilling.

Juniors are replaced by token-generators that will never learn, seniors are pushed to uncomfortable roles where they are forced to use slop-machine without learning necessary skills, leaders are pushed to let seniors cut corners for the sake of showing the product can be built faster. No one learns, no one keeps knowledge, everyone cuts corners, everyone becomes dependent on overlord’s token-printing machine and crappy software is being built just as a side effect.

Don’t get me wrong, I think it’s great people can automate their daily lives writing small scripts or personal apps they would not be able to make without learning the skill of writing code. But let’s not pretend that building a shack in your backyard is the same as building a bridge or that an app built by a coding agent has the same quality as one built by an experienced team of software developers. You cannot contain human knowledge and experience in an agent prompt.

The problem doesn’t lie so much in the quality of machine’s output. It lies in its forced adoption, in disregard for engineering experience and especially in deepening the inequality between those who can afford the price of output and those who cannot.

How long is it going to take for anything-machine skills to be a requirement for a wide array of jobs? If you can’t afford paying for machine tokens, how can you qualify for such jobs?

Some of us have already lost the ability to read long texts, and I commend you if you are reading these lines. Some of us don’t know the answer to 6x7; some don’t remember their parents’ or spouse’s phone number. We’re relying more and more on machines. But there’s a major difference between owning a calculator or a phone book and Silicon Valley overlords owning your means of getting a job or means of doing your job.

References

Comments

@beansprout_observer@mastodon.social

Bean Sprout Obeserver

Work slop

References

Comments

Deskilled by word-guessing program

References

Comments