"[Linus] ends every session with a lengthy cleanup pass that takes a couple of hours to run, executing tasks like ensuring that the code matches the comments. That amount of work cannot be done with a free LLM account. I asked whether that risked cutting off bright new developers that we would rather hang onto; Torvalds answered that the kernel project only wants serious developers. That includes having access to enough resources to function in this world."
There's this trend in the last couple of years of very successful software engineers heartily endorsing AI usage - I can think of Matz, dhh, Linus, uncle Bob. Probably a lot of other people I don't follow too closely.
I can't pinpoint a common motivation for all of them, other than the obvious (the tools are great and are the future, if you think otherwise you're a loser (paraphrasing dhh) etc).
I think at least at some level, these people are at a point of their careers where they have "been there, done that" (coding) and are happy to go do something else.
All of them have in common not having written code in a long time, instead reviewing patches from others, so the AI thing isn't really too different from what they've been doing for a long time.
Travel agents still exist though; while I couldn't find employment numbers, I wouldn't be surprised if there were more today than 20 years ago (something something's Paradox)
Now, what you're describing is really about the "culture of convenience". People prefer Expedia to human travel agents because it reduces friction (of interacting with people). Same goes for Uber, Waymo and so on. Good piece on it - https://www.cbc.ca/radio/ideas/cult-of-convenience-9.7371392
Totally, I can't quite imagine an AI automating my use of AI tough, how it would infer what prompt I would use at any given time, the problem space is infinite. Am I adding another *arr app to my stack, are we building a SPA, a home automation plugin, whatever. These are things that I want to do and AI helps with that.
In the corporate setting I think though it won't take a lot of time to reduce the layers and number of people that are required to run a software company. Either one becomes a more general work manager or obsolete I guess.
It is remarkable to me how few people seem to understand it is a much smaller step for AI-generated software to kill the need for companies of idea men than it was for them to kill the need for people writing code.
All of your potential customers are just going to vibecode their own software. In many cases they won't even realize that the software they vibecoded even exists as it will just be an invisible tool the LLM built for itself to reproduce a similar solution to a similar prompt over and over.
Once AI flattens out any company enough, the company itself no longer needs to exist, it can trivially be replaced by the model providers. Companies exist to be collections of skilled people with a common goal. When you don't need that, you don't need companies. The agents can work for your customer instead of you and nothing changes except who gets paid (not you).
There's a ton of magical thinking out there where people seem to assume AI will progress right up to the point where it is maximally beneficial to their job without replacing them. Keep huffing that copium, folks.
> There's a ton of magical thinking out there where people seem to assume AI will progress right up to the point where it is maximally beneficial to their job without replacing them.
I've commented previously that there will be much more political interest in AI causing unemployment when the first CEOs are replaced.
That's going to happen. It's part of the "creative destruction" of capitalism.
If AIs maximize shareholder value better than some CEOs, those CEOs are toast.
They either get replaced in the same organization, or a new company arises and crushes them.
I don't think that's ever going to happen when models still hallucinate and prompt-refuse, especially since one of the two SOTA model makers insists on aligning their models toward a progressive ideology.
Just imagine what would happen if JP Morgan did this and the model decided it needs to dump their position in AI data centers. Or maybe it's a health insurance company that fires all of the claim deniers for being unaligned with Dario's ethics.
> There's a ton of magical thinking out there where people seem to assume AI will progress right up to the point where it is maximally beneficial to their job without replacing them. Keep huffing that copium, folks.
Because you are singularly right about the issue? Have some intellectual modesty, brother.
I think it's fair to expect tha AI capabilities will more likely end up far over or under that magic threshold than close to it. Given the way that technology develops, it's probably going to be over.
He started being ambivalent about it maybe two years ago, and reached metastasis in the latest Rails World fiasco, with dhh telling him to his face that his life's work (Matz's terms) is irrelevant now.
>When people claim that it’s no longer sustainable to write code by hand, they’re not entirely wrong, but only in the same way that it’s no longer sustainable to sleep in a bed once someone has taken a shit in it.
This is not representative of what's actually going on, though. It's much more like an artisan carpenter who wants to build houses using bespoke hand-tools and precise joinery. That's great, and there are beautiful works of art created by those skills. But in the last year robots were developed that build homes exactly to spec. The robots work 24/7 and finish entire houses in a few days. That same skilled carpenter is being hired to monitor the robots and check over their work. There is, economically speaking, no place in the process for him to apply his craft. In the time he rips a single board, the robots build the wall. In the time he drives a handful of nails, the drywall is up and the paint is drying. He costs $200/hr, the robots cost $1/hr.
The carpenter is free to exercise his craft as he see's fit, but it's going to take place mostly in his personal workshop, for personal reasons, not to develop homes or run a business in the way he did before.
We can talk for ages about the quality of the robot's work, the maintainability of the resulting home, etc. But it's clear that, at this time, the market and ecosystem are embracing the robots whole-heartedly so those of us used to the old ways are having to rapidly adjust.
This analogy hinges on the economic argument (that a robot costs 200x less than a human), but far from being an established fact, that notion is more or less entirely based on vague, misleading and exaggerated claims from the token peddlers.
The truth is no one knows how much the robot costs, and we probably won't know until its too late. Its extremely naive, though, to simply buy into the PR from the peddlers and redistribute their propaganda verbatim in the way you are doing. Although none of us know for sure, considering R&D, infrastructure capex and externalities, the robots are most definitively _not_ cheap.
> Its extremely naive, though, to simply buy into the PR from the peddlers
OK, so don’t buy their PR, and try it yourself.
But if you do this, you will find there is a lot of evidence for the economic argument.
I tested out Google Argon 4 release by having it build a Web app to replace a binder I have for reconciling payments with reimbursable expenses.
It did it, as well as I would have and in much less time. I just set it to high and supervised it, it cost about $2. Even if you throw in a “supervision cost” to account for the time I had to spend interacting with it, it was substantially cheaper than had I done it at normal salary rates. Probably 2 orders of magnitude cheaper, even.
I think the OP was talking about robotics specifically and not tasks like the one you described. Nobody can argue that for tasks like your binder replacement cost savings are enormous.
No, they were carrying on the analogy about artisanal home builders being replaced by mythical robots that do it perfectly.
In this analogy, we human software devs are the artisanal homebuilders and the fancy LLMs are the robots.
And yes, there's quite a cost to build all the datacenters for all the AI that Big Tech believes that people will want. But as you mention, the cost savings are enormous for even despite that, and will make a much larger set of problems accessible to real software solutions, that previously either stayed on paper, in peoples' heads, or maybe an Excel sheet somewhere.
Almost anyone will accept a 50% quality decrease for a 99% cost decrease. The only reasons I'm not living in a stack of cardboard boxes for $1 rent are that it's illegal, and the police won't protect it from break-ins.
With the money you wouldn't be paying in rent, you could build a heck of a lot of stuff. Cardboard box is a metaphorical starting point. You could build some kind of little wooden box vehicle and travel the country. If you stayed in town you could have takeout for every meal too and still have money left over. The point was not paying rent to the corrupt system, the point wasn't the exact living specifications.
You can go and live in cardboard boxes right now for free. Why aren't you? Being technically legal or not is besides the point - there are places where it is tolerated.
>You can go and live in cardboard boxes right now for free. Why aren't you?
this is why i try to avoid analogies whenever i can. we've cycled through like 3 different iterations of the analogy in 5 comments, just to land on this question.
For corporations, sales, probability and feature release cycles take priority over maintainability or code quality. AI is allowing businesses to build features faster than ever, so management is happy.
By the time maintainability issues hit, they've either pivoted, raised more capital or can throw money at the problem.
Not that tired old parable again. It doesn't hold.
If someone complains they take pride in understanding and designing deterministic processes, you can't reply that this is just like when robots took over manual labor.
No, because it is not manual labor, and robots hasn't taken over. Robots are computers. Robots are deterministic.
A better comparison would be an artisan architect, a blueprint designer being replaced by a special paint splatter, which when thrown against a piece of paper looks almost exactly like a blueprint of an almost completely liveable house. Great! What's not to like? It will allow anyone to design a house.
> But in the last year robots were developed that build homes exactly to spec.
That is not what's happened at all. That's the entire point of the discussion. The robots aren't as good as humans, they can't follow specs or instuctions as well, and their output is markedly worse than human output in a lot of ways.
The only difference is that the robots are faster and don't need rest. And we're discovering now that that is not good enough.
Unless they manage regulatory capture and with this kind of money at stake I would not bet against that. It's going to get very ugly. Read up on transhumanism to see what the stakes are, they're playing for keeps here.
Regulatory capture - I tried to bring this up in another comment of mine, but not much resonance.
I wondered if many a press article seen in media re security breaches and LLMs based hacks are not in part at least campaigns to legitimize regulation.
How would regulatory capture of LLM realistically look like? Do you have any ideas here?
> How would regulatory capture of LLM realistically look like? Do you have any ideas here?
Let's flip this around. What do you imagine the most likely form that regulatory capture crafted by the major LLM Manufacturers would be?
When rolling this question around in your head, remember that the US Congress gave the telephone companies that blatantly and knowingly violated FISA retroactive immunity for their crimes in the 2008 FISA Amendments Act. Congress will move mountains if the majority of its members are scared enough and/or see enough profit in something.
> Read up on transhumanism to see what the stakes are...
Eh. Transhumanism is very good. Every medical intervention that improves or consensually prolongs the life of a human is transhumanist. Our natural state is incredibly godawful. [0] The best future for humanity is one where one need not work, need not get sick, need not die... where all of the awful parts of life are optional.
What the TESCREAL cult is peddling is incredibly bad. They have folks believing that not only are they ushering in the arrival of Robot Jesus, that nearly any terrible thing done now to speed up the project is justifiable because it hastens Robot Jesus's arrival. In actuality, the only thing that's hastened is the accumulation of wealth of awful humans like Altman, Amodei, Huang, and many others.
The core tenant of Rationalism [1] is great. It's just a huge shame that so many of the folks who were attracted to the philosophy ended up being so bad at learning how to do the one thing it tries to get you to do.
[0] Remember that it wasn't that long ago that humans had a 50/50 chance of dying before the age of five!
[1] Your brain is really bad at thinking carefully. You have to train yourself to think carefully despite how shitty your brain is at it. You will inevitably fail at this, but that's no reason to stop trying to get better at it.
Nah. All three are covered by my original statement:
The best future for humanity is one where one need not work, need not get sick, need not die... where all of the awful parts of life are optional.
Work, sickness, and death are -for the most part- all things that are awful. As it turns out, there are many more than three awful aspects to being a present-day human. For brevity's sake, I picked three that seemed to me to be most likely to be widely understood and widely considered to be awful.
> What is "you" in this scenario if not your brain?
That's an excellent question. Are we the blob of fat that hangs out in our skull, our entire body (because of how changes in "distant" parts of it affect what our brain does), the parts of the patterns in our brain that seems to correspond to our conscious mind, some other thing or things?
I am utterly unequipped to answer that question, but hope that we collectively get around to trying in earnest to answer it. Maybe we might even do that before we get around to collectively deciding that it's a catastrophically bad thing for people to go without food and shelter? [0]
[0] Given what I see here in SF and elsewhere, it looks like we currently collectively consider it to be somewhere between "Not at all a problem" and "Yeah, maybe we should get around to fixing that some day. I _guess_.".
The difference with those is that they're systems of governance... which means that they must scale if they're to be applied to large groups of people. And -well- they scale very poorly.
Except the robots actually cost far more than $1/hour, it’s just that the owners have enough capital to operate at a massive loss until the carpenters have largely been replaced. Only then will we learn what the robots actually cost.
Plus robots (software and hardware) will increasingly be used to invade even the most basic privacy, to control people etc. They are not just productivity tools anymore, they are also mass surveillance tools and every bad thing that comes with it.
Self-hosted and/or open models are fast approaching (or have reached) the quality bar necessary to handle a great deal of software development work. The models needed to cause the OP's problem are becoming commoditized.
Forget robots, what you're describing already happened in the trades. Modern construction techniques are all about lowering the skill required to build a house and standardizing the process so it can be done with minimal training. E.g. drywall and caulk replacing plaster and precise joinery.
You can find example of this in all of the trades. Often, these are also talked about as "safety improvements" - which they sort of are. Given the same baseline intelligence, it's harder to dangerously fuck up plumbing or electrical in 2026 than it was in 1926.
I have been told it is the future of sleeping, and if I don‘t sleep with a plastic cover directing the pooping agent I will fall behind and eventually become obsolete. And that sleeping like this will eventually cure cancer and build swarms of mining rigs in space.
I think the author is trying to say that a code base once infected with vibes, becomes indecipherable to the average human being. It's still just code, but full of Claudisms.
100%. Although large handwritten software code bases are often indecipherable to the average human being (actually, almost always indecipherable to the average human being, but often indecipherable even to skilled software engineers too); sometimes due to inherent complexities, sometimes due to Enterprise-isms.
sounds like bullshit. I'm actually glad that the people that hold this view are getting out of the way, they are the people that gatekept open source and made it annoying to work with. Do I miss having to wait for months to fix a trivial bug because I first need to convince 15 people that the fix is even necessary?
Next time someone complains about software do me a favor and compare it to a nail through the foot or grabbing a hot dish and burning yourself or something else negative in the human experience. There is plenty to choose from.
We don't need 300 variations of enshitification, the folks at Pompeii and Herculaneum have that well covered.
I was thinking about this recently. HN is of course much older than the current AI hype cycle, but maybe its time to update HN policies to forbid bots, and start thinking about ways to enforce that. There might not be any 100% tamper-proof way nowadays, but reducing the bots by some percentage will still be a win.
> Please don't put generated text in HN posts. Write your text yourself—HN is for sharing between humans. For the same reason, please don't automate posting.
reply