The irony is that AI is best at replacing the work of the upper classes. Synthesizing different opinions, summarizing them, and producing outputs based on statistics are things AI does well.

But AI is actually not very good at replacing an entire lower-level worker’s job as a whole. It works well only when that work is broken down into smaller and smaller tasks.

The core problem is this: the coercive force of AI use is felt by the lower classes, while the upper classes still have the freedom not to use it. AI may be able to make decisions based on more data than executives do, and perhaps even make better decisions than management. Yet the people being replaced are the lower-level workers.

This is the problem. The upper classes, who claim that AI is an essential tool, still have the freedom not to use it. But the lower classes cannot survive unless they use it. It becomes a tool required for survival, while at the same time being treated as something wrong, inferior, or low-status if you use it.

To get a job, AI becomes an essential survival tool. But culturally, it is also treated as a tool that damages creativity. I see this in open-source communities as well, in the class discourse around open source.

The same culture appears on Hacker News. Among the upper layer of open-source communities, there is often hostility toward AI-generated code, based on ideas of human purity: AI code is said to have no meaning, no responsibility, no real authorship. So even within open source, this takes on a class character.

But as a freelance developer, I have to trade against my own code-writing ability in order to survive and deliver. Because of AI, the floor price of software delivery has collapsed. If I do not use AI, I cannot meet the new requirements.

In the past, a job that would have given me two months and paid $5,000 is now expected to be completed in two weeks for the same $5,000. Without AI, that volume of work is impossible to handle.

This kind of discourse always makes me uncomfortable. I dislike it, but I have to use it.

AI lowers the barrier to creation and learning, but the way it lowers that barrier can also bypass the training of thought itself. It turns young people into both beneficiaries and damaged subjects at the same time.

And we live under this loop of coercion. Sometimes I think I do not want to use AI.

But if I want to survive, I have to use it. I feel the abilities I once took pride in beginning to decay, and I feel myself becoming increasingly bound to AI companies. At the same time, I also feel another kind of ability beginning to emerge.

Perhaps growing older means learning how to live inside irony.

>The irony is that AI is best at replacing the work of the upper classes. Synthesizing different opinions, summarizing them, and producing outputs based on statistics are things AI does well.

AI just repeats whatever the prevailing opinion is at that time. I am a very heavy AI user (Claude, Gemini, ChatGPT) and have queried it on a variety of topics. AI is not thinking, it is repeating.

> In the past, a job that would have given me two months and paid $5,000 is now expected to be completed in two weeks for the same $5,000.

So you have quadrupled your income? That seems like the opposite of a collapse.

> The irony is that AI is best at replacing the work of the upper classes. Synthesizing different opinions, summarizing them, and producing outputs based on statistics are things AI does well.

It’s really not though. A lot of sort of incidental communication that people don’t really read carefully it’s fine at. But the actual hard stuff it’s just not that great. It’s basically average by design so it almost definitionally incapable of being great.

I’ve been trying to use AI for putting together job applications, tailor my resume and cover letter and stuff. But it’s just not good. It’s decent if I want a sanity-check analysis as in “how does this come across generically.” But if I ask it to write anything it sound like LinkedIn slop. I expect if everyone starts using this to write their basic communication you’ll be in a world where nobody is reading anything anyone says because it’s so BORING. Everyone sounds the same, everything is phrased in this sort of mushy and generic way. At that point the intended communication isn’t happening! We will have sanded away all the rough edges to the point where you can’t actually get a grip on anything anymore.

At another point I tried to test it out and asked it to wireframe a very basic application for me. Something that should have been a very lightweight thing that runs on device to do a basic background process it architected as some crazy overcomplicated enterprise-scaled thing as if I’m trying to build a unicorn startup out of this rather than just a toy app to organize my shopping list. If I wasn’t technically savvy enough to recognize it was way overcomplicating things I’d have just run with this. And what’s worse is, it burns through all your token budget to figure out a bunch of problems you don’t need to have!

Obviously I’m using it and find it useful, but I’ve started to develop serious doubts about how useful it will ever be without an informed and accountable attendant overseeing it.

> The irony is that AI is best at replacing the work of the upper classes

This is why the harshest critics of AI tend to be white collar workers of this social class. The same kinds that told coal miners and autoworkers to "learn to code" and called them deplorables for voting nativist in 2016.

Any chance to build mutual trust was squandered. The jobs worst impacted by AI are jobs where most of the workers are Democrats and live in blue states that don't swing.

Meanwhile, those manufacturing, construction, and healthcare jobs that are becoming a bigger part of the economy tend to be in the purple part of the country so their needs are heard.