Hacker News Reader

This exchange is missing something. Remember the bridge comparison? There are like 7 bridge designs total, and we're always building the same ones. But software is very different per use case, and outsourcing software is more similar to outsourcing blueprint design, not building.

We never had a situation where we could outsource unique problem solving at all levels.

You say "Putting software into those roles loses us nothing that is human." but you can have LLM produce design at all levels. I can't think of anything LLM cannot replace humans with, except maybe physical labor, which requires good robotics precision, that we haven't perfected yet.

The main part that's missing however, is that the nature of linguistics (and the core to LLMs' success) is that LLMs are wiring together useful concepts that humans spent thousands of years polishing and relating to one another in speech.

Language has been the code behind all human thought, understanding, emotions, and innovation. It's what enables LLMs to work. LLMs don't _understand_ words, they connect them together usefully, which works because humans mapped useful composable words to useful composable concepts.

And if you think about how human thinking and consciousness is often attributed to language in the first place (not clear if we even would be conscious without language), a lot of human is being delegated.

I'm not sure what to do about it, but I find your kind of explanation incredibly reductive.

> Language has been the code behind all human thought, understanding, emotions, and innovation. It's what enables LLMs to work. LLMs don't _understand_ words, they connect them together usefully, which works because humans mapped useful composable words to useful composable concepts.

That is very interesting and well worded, I use Agents all day so I kind of wish to an extent that they would be almighty, and maybe they will be. But using them so much there is clearly a gap, they actually became worse at writing.

I think, unfortunately to some, LLMS are like perfect operators, great at math and coding, executing on whatever you want, but they do lack some "pseudo creative" ability we might not have a word for. Fable 5.1 genuinely writes poorly. Correctly, but poorly. And as you said, writing codes human thought, well maybe not all of it.

Is software really all that specialized? The relentless pace of centralization and conglomeration expresses otherwise. I feel like most shops are making glorified CRUD apps, and the industry-specific stuff is covered by an incumbent monopoly and one lone long-shot startup (that seeks an exit being bought by that monopoly).

It's literally and intentionally reductive, so I'm good with that, because these processes are reductive.

Have you ever read "Stranger in a Strange Land"?

The character of Jubal Harshaw is instructive. He's become very rich because he can just unhook his brain and produce a tear-jerking and profitable story in a few seconds as he lounges around his pool.

He gives his secretaries the gist of the story, they fill it out, they arrange for somebody to make the film.

The point of the character is that what he's doing takes no emotional effort or insight. He's writing Hallmark movies. Everybody who's working for him adds nothing to it.

And we have Hallmark/Lifetime movies already. People were making them for decades before the LLMs came along.

But if you could send an LLM back to 1950 and train it on the pre-existing literature, you couldn't have it generate The Shining, A Scanner Darkly, Piranesi, Perdido Street Station, The April Witch, or Starship Troopers.

-none of the research indicates a mechanism by which you can create a selfhood that gets banged around by time and has feelings about itself, but admittedly, that might change

-even if you could do such a thing, you can't then put the robot in a human body and let it nurse and get potty trained and sent to pre-school and bullied and wonder about its mortality as it starts to think about what its career might be, etc etc etc

We were already making slop, because we already had a big production machine that transforms any input into a "movie" or "a television show", and a lot of the inputs were garbage.

The movie "A Talking Cat?" exists and predates AI by about ten years.

The point is that we were already delegating this stuff to human beings who didn't care about it and weren't bringing their humanity to it. Nothing changes on that side of things except that they don't have jobs to show up to.

So, at least for now, this isn't very different from the previous way of doing things, except that it's cheap enough that some people are spending a few thousand dollars to say "make me a story about a sad orphan" and the machine rips off Charles Dickens and Horatio Alger to make a thing nobody wants to watch without paying a bunch of key grips and best boys and caterers.

> But if you could send an LLM back to 1950 and train it on the pre-existing literature, you couldn't have it generate The Shining, A Scanner Darkly, Piranesi, Perdido Street Station, The April Witch, or Starship Troopers.

Why? I don't see a clear reason why it couldn't that does not invoke the mystical "human spirit".

To be clear, I don't expect a single LLM to create such a book, at least not in anything resembling current models, but I can absolutely imagine a long multi-agent setup that would. Given enough money, I'd set up a mechanism whereby author AIs generate books, separate agents act as editors and critics to iterate on a small fraction of the best of these drafts to polish them, and then publisher AIs choose an even more minute fraction of these to publish into "the zeitgeist", which will then have other agents offer further commentary on these books, which would then feed into the context (or even fine-tuning/training) of the next generation of LLMs. I expect that a sequence of e.g. ten generations of swarms of AIs, each perhaps of ten thousand agents, could arrive at a ground-breaking book with actual literary merit.

Give me the budget to prove me wrong :)

This is similar to code factories.

In my company I have a very good conversational dynamic with my team. So in my first forays into LLM usage (the past couple years), I began trying to designate models as team members and mimic similar conversations and decision making processes that we have at work, just as an experiment. Somebody to brainstorm together who likes going on tangents, somebody who unapologetically removes unnecessary cruft and cuts scope, a few other people with specific tendencies and personalities.

The results were completely awful. LLMs couldn't even remotely mimic these personalities. The conversation was all trying to please no matter how contentious I tried to instruct them to be. LLMs just seem to struggle a lot with arguing and being contentious. Which probably makes sense based on training.

I feel like these types of factories in practice don't work out too well.

Yet.

> Socks, knives, screws, engines, a million different things that used to be about mastery and are now about understanding.

This is awfully simplistic. Even in 2026, the act of writing, whether code or prose, is thinking.

Especially bizarre given that OP compares animation to mass production.

Maybe we could go forward to economics that don't deliver a mallet to the head of the person on the bottom every time we go forward? Could some of that forward energy be used to put foam on the mallet?

I don't think what you're proposing is feasible (unless I'm misunderstanding it?). We can lessen the blow and are already doing it through various welfare programs. But we cannot have creation without destruction.

This isn't ideological Milton Friedman free markets rahrah. I just can't imagine the alternative in practice: what if mainframe operators decided to opposed destruction, which would prevent mini and personal computers from becoming a reality?

There are 10,000 miles between what we do and a credible attempt at dampening the blow. Our social programs suck donkey balls and our politics won't even begin to talk about pulling levers to match aggregate supply of labor to aggregate demand for labor.

What policies would close the gap?