It's cope. People desperately want to believe that AI coding is going away so that they can go back to partying like it's 2020.
So there's a huge number of HN posters claiming that the price of tokens will go UP over time rather than down (that's how Moore's Law works, right???) or that code bases that AI contributes to will spontaneously combust, or something.
I don't think it is unreasonable to say both will happen, is it?
In the long term, tokens will fall in price. Obviously. (If "tokens" continues to be the unit)
In the short to medium term, for the IPOs to succeed, people have to start actually paying for what they are using, so the price will go up, and is going up, quite a lot. Once their value is set they will slowly fall from that point (or some point maybe halfway, depending on how much the market is willing to continue to subsidise).
I am an AI cynic, but I am now an informed cynic; I am learning agentic tools so I know where they are useful and I know my enemy.
I think the "fad" here is cloud-based, metered AI being a dominant work mode.
Nothing, so far, has suggested to me that any other outcome is likely than edge- to local-scale, on-device, on-laptop, on-prem models getting good enough to the point where people use them by default and use the cloud models only when they need the extra oomph.
I cannot believe that there is anything other than an enormous incentive for companies like Uber to find local, small model and on-premises solutions to their problems, not least while pricing is so changeable and people are getting nasty surprises.
Betting on OpenAI and Anthropic being around over the long term in the form that they are now, that feels like valley hopium. Utility monopolies essentially always derive from physical/geograpical limitations, don't they?
I mean, there's an "enormous incentive" for people to run their own data centers rather than using AWS. And yet, cloud is growing and on-premise is shrinking.
While I hope local AI continues to exist, I'm skeptical that it will take over, for the same reason running your own servers hasn't taken over. It's just hard, and involves spending huge sums of money up front.
It's also not really clear how much tokens are being subsidized. The discussion reminds me of Uber. For years people on HN claimed that Uber was going to collapse once they ran out of VC money. Then... that never happened, and everyone just moved on to discussing other things.
Infrastructure is massively complex and multi cloud is super hard to do. Switching LLMs is... a drop down.
Now, that doesn't mean running your own LLM will be easy, but this will mean it's a lot more likely that there will be at least regional LLMs, in my opinion. I.e. there will be Google, whichever (if any) is left standing of OpenAI or Anthropic, and then there will be Chinese hosted LLMs, probably Indian hosted LLMs, European hosted LLMs, plus LLMs hosted on managed services (i.e. Bedrock). For sure I see large banks on the like being able to host the best OSS or even licensed LLMs on their own cloud infrastructure accounts (i.e. at AWS, Azure, etc).
And that's on top of the LLMs running on owned server infrastructure plus actual local, on device LLMs.
You're using the future tense, but all of those things already exist. Google exists, Amazon Bedrock exists, DeepSeek's cloud product exists, etc. etc. But this isn't relevant to what the post you are replying to said, which is that "cloud-based, metered AI being a dominant work mode [is a] fad". Since all of those things are cloud-based, metered AI.
I was talking more about on-premises, on private cloud and on-device stuff, as I said.
If you look at what Uber is spending per developer per month, they clearly have some headroom to consider whether more-local, unmetered AI tools on device, on premises, in private cloud, can be cost-effectively used to cut down how much money they are pouring into Anthropic and OpenAI. Not least because a bit of centralised effort might lead them to distilled models that are better for their purposes. Some of that budget could go into simply putting a bit more capacity on a developer's desk.
Can they do it now for everything? Obviously not. But IMO there is no reason at all for planning and scaffolding tasks to be done with cloud models, and there are many reasons why it might be better to do document processing without leaving the premises.
The incentives are there on the technical, operations and particularly on the business levels, and the relative disruption of the switch really small, considering that all the tooling can use different models for different tasks already. They must at least be investigating the possibility; it's irresponsible not to.
Uber's not really a good example because they deliberately incentivized their engineers to spend as many tokens as possible, which was silly. But even assuming that every developer uses the full $1,500 a month of tokens that they are now allowing, that's actually not a lot of money relative to the cost of a single developer for them. It's less than 1/10 of a junior engineer's fully loaded salary.
Where I would expect to see people invest in local models is in cases where a company has regulatory requirements to keep data local, or where they're doing some specialized kind of work. Neither of those really apply to Uber. In 2026, it would absolutely be irresponsible for a taxi company to try to build a better Claude than Claude.
Now what this looks like 5 or 10 years from now, it's hard to say. A lot will depend on whether China keeps releasing open weight models and whether people can still run those open weight models on commercially available hardware.
> So there's a huge number of HN posters claiming that the price of tokens will go UP over time rather than down (that's how Moore's Law works, right???)
I mean, Github Copilot's pricing just went up considerably, so I guess they were right?
Token costs do go down over time for sure due to software optimizations (i.e. better attention kernals) but acting like hardware INFLATION isn't happening for at least a few more years is just nonsense. Objectively an A100 is more expensive to rent today than it was in 2024 (a 7 year old GPU - Big short guy is a turbo idiot) and rising. As such, over short time horizons, it's possible to see limited amounts of "price per token goes up" for the same model.
It's a mix. If the current wave of LLM businesses crater, demand for LLM specific hardware (and related hardware) will crater. GPUs were propped up by crypto currencies and now by LLMs. They're still great at doing fundamental math operations, but for their value to stay up another massive business opportunity involving matrix multiplication and the like would need to rise as soon as the current business cycle winds down.
Because if I want a low-cost, low-quality dev, and Claude fills that role better than an outsourced dev, such people will "decrease in numbers" due to supply and demand.
Perhaps my point wasn't clear. The article suggests humans are needed. So does the person here I responded to. Someone has to tell the LLM what to do and verify it.
Outsourced can be anywhere and when you're referring to US, anywhere else is cheaper. Commonwealth devs for example are at least 20% the cost of a US dev. That 20% more than covers the cost of access to a frontier model that this dev could drive.
Thinking of LLMs in human units is super strange and if you wish to consider the economics you have to also account for the lowering of barriers to entry.
My point is that current-year LLMs are a better replacement for low-skilled developers than high-skilled developers. For example, if your mental model in 2016 was that you would have a senior engineer write a specification, and have an outsourced team in India implement the specification, in 2026, an LLM replaces the outsourced team in India, not the senior engineer. As a bonus, everyone is now in the same time zone, and there are no language or cultural barriers.
Frontier models are pretty good at executing on a spec when all the decisions have been made already. The author reached the same conclusion. It's the agency that the humans bring that ties it all together, which is still required. Doesn't sound like you disagree.
I think where were missing each other is that the human agent can also be outsourced. Not all outsourcing involves rote tasks and the lowest skilled workers. The difference between the mean US senior dev salary and everywhere else is large enough to cover a manageable LLM cost. So there's no reason why a company looking to cut costs need stop at the lowest rung to makeup for LLM expenditure.
Outsourcing comes with significant costs. One of them is that it takes time to communicate with the outsourced team. For example, if you want to talk to someone in India, you might have to send an email that they read the next day, because they're almost 12 hours separated from you. So exchanges that would have just been 15 minute chats with a US-based developer turn into multi-day back-and-forths. If you are hiring someone to be the architect of your system, don't you want to be able to talk to that person? That's why even the most outsourcing-obsessed companies usually did not outsource architects.
Another issue is that there are cultural barriers. People in India or elsewhere may say "yes" when they really mean "hmm, probably not" because saying no to a superior could be considered rude.
If you replace a big team with a small team and LLMs, you are actually saving money overall because LLMs are much cheaper than humans. But you may actually need more skilled humans than previously, not less skilled ones, because they need to be able to manage a large volume of code being generated. LLMs are not good news for outsourced developers. They are the opposite: a cheaper substitute for the grunt work that they had been providing.
You don't have to convince me, I've experienced what you're describing. I'm just being realistic with the options decision makers have.
Simply put, if you want frontier models at API prices, you can make up for that expense by hiring non-US talent. Many who have a good command of English and are willing to overlap US hours. There's plenty in LatAm alone. Whether or not that's a good choice for a business isn't relevant to the point I was making.
Where you appear to be stuck is that you think outsourcing is only workers with poor English who do grunt work. It's a rather myopic view of the situation to be polite.
You appear to be assuming that frontier LLM models will be more expensive than outsourced developers. That seems like a crazy assumption to me. Even with current prices of about $1000 a month, that's still just $12,000 a year, which you cannot live on, even in Eastern Europe. Maybe some people in India could live on that, but probably not the people you want writing your code.
and this is my developer. She consumes one hundred fifty gigabytes and runs two hundred thousand dollar, custom-tooled GPUs at ten thousand tokens per minute. It costs four hundred thousand dollars to develop…for twelve seconds.”
[Laughs]
“Oh my Claude, who touched settings.json? Alright…Who touched my LLM!?”
To me it's dishonest to write an article like this without mentioning the reason why Spirit's employees and customers are in this situation: because the owners realized that it wasn't a viable business but were blocked from selling their company by the government.
It's not like the article doesn't go into the history of the company either. There's a paragraph that starts with "Spirit has been unprofitable since the pandemic..." To deliberately leave out one of the most important parts of that history is extremely deceptive.
Companies often expect employees to give advance
notice of termination of employment, but employers
do not feel beholden to their employees to
reciprocate. This is unreasonable.
Workers can leave their jobs with no notice period at all and suffer no legal consequences. The traditional two weeks notice period is a courtesy and not legally required. Meanwhile, employers often have extensive legal limitations on how many workers they can fire at a time (for example, because of the WARN act). It may be impossible to fire some workers (for example, those on disability leave). It's traditional for employers to pay severance to employees that are fired or laid off, and this may sometimes be legally required.
Nobody can force a worker to work, but an employer is often forced to employ someone they'd rather not. Maybe sometimes this makes sense, but it's actually the exact opposite of what you're trying to claim.
Many companies showed improved productivity when
people were given flexibility. Claims to the contrary
are often not backed up by actual data outside of
cherry-picked results chosen by those in managerial
positions.
If remove work is so productive, then management should choose remote work without any coercion. If you want more benefits, then just be honest and don't make claims that you know better than management about what is productive.
MTG is a creative endeavor, and its output is
protected by copyright. Copyright is, so far
(thankfully), only enforceable when the media
is produced by humans. Protections against the
use of generative AI are not only good for
workers' long-term productivity, but also
for legal purposes
Copyright applies to works that are partially produced by AI and partially produced by humans. Refusing to use the tools that everyone else is using to be more productive is just going to make companies into dinosaurs.
I think maybe you should stop licking boots and
start understanding why any of our society works
today at all.
The union is also a corporation, which makes profit for its officers. Maybe you should stop licking its boots?
As a hint, every protection you have
ever enjoyed (40-hour work weeks, pensions or retirement
funds, health insurance guarantees, mandatory paid time
off, health and safety regulations, and so on) was paid
for in blood by someone a company decided was not more
important than profit. Unions are a fundamental capability
by workers to ensure their own safety. Shockingly, I think
it more reasonable to rely on the interpretations of those
working in these jobs who want to unionize than some random
person online who thinks "no wait, but think of the poor
endangered company! :("
I'm not a member of any union. The protections I enjoy are because of the law, which applies to everyone, and because of the economic situation. If the economic situation changes, I will be fired no matter what unions or governments do, the same way saddle-makers were after the rise of the automobile.
Since you think Cuba and China are such nice places, perhaps try living there. You'll quickly find out about their "merits" (such as the fact that they execute dissidents).
So there's a huge number of HN posters claiming that the price of tokens will go UP over time rather than down (that's how Moore's Law works, right???) or that code bases that AI contributes to will spontaneously combust, or something.
reply