1121 points | 2d ago | Discuss on Hacker News | Back to Radar
Especially since this is a tiny percentage of the business on sale — 5%? 10%?
Edit: Another huge buying spike at 1:09 pm at same $95.5 level, but not enough to stop it continuing to slide for it's low so far just below $93.
But then everything about today — the muted reaction, the lack of people losing their minds over a new era — feels more like relief (someone got to IPO) than jubilation.
As a sort of pragmatic cynic, I find it rather reassuring.
Compare that to the $140B owned by the founders, and $200B+ by employees.
I don't know when employee RSUs/options become sellable, but I'm guessing the only employee selling today is part of those 23M shares included as part of the IPO.
They go into the opening with a short position, which they can close out by buying to defend the stock. If the stock goes up rather than down, then they have a "greenshoe" option to buy more shares at IPO price to close the short.
/s
The reality is that the situation is very complex, and saying "it's a bubble" or "insanity" or (Zitron's favorite) "you're the exit liquidity" just misses the forest for the trees. AI is clearly revolutionary, there's also some signs of it being a bubble, the market does seem to be less risk averse than 20 years ago, etc. It's all very muddy, and claiming otherwise with so much bravado is just dishonest.
Costco for example has done quite well this year. How has the AI trade with NVDA worked out this year in comparison?
There's so many ways to make money the default answer of "I don't like this stock" is buying something else.
A: I don't like milk.
B: If you hate milk so much, then borrow someone else's milk and then sell it!!
A: Wtf? No. I'll buy orange juice instead.
So no. None of those 'do this specific thing that should be sane by the science of economy' arguments work. Some powerful institution can just f*ck up your plans just because.
You can say Anthropic as a financial instrument is worth $2T, but then you head into irrational financial bullshit territory where there's a nontrivial chance a lot of people are going to jail.
if Anthropic creates ASI, the economy will probably fall apart. if Anthropic doesn't, its competitors will catch up and the stock will fall below its IPO price. hrm... though I suppose this does argue for trying to short it.
in this IPO, all 7 founders are subject to two lockups for ALL of their shares: 180 days, then they can sell 29%, 365 days, the remainder. only ordinary employees can sell 10% of their vested RSUs / exercise options today, which will be about $400m if everyone sold and everyone got the biggest offer anthropic gives employees, but will be closer to $25m given what the actual rate of hiring (and therefore vesting) and awards were.
Even if the stock drops 95%, they'll still be billionaires.
Sure, Google had the top of the line offering. But you could easily argue that they don't have the moat, that web search is a commodity, and that more established competitors could catch up to Google in no time at all.
Also, unlike with web search, the AI companies are makers of their own destiny - in that they can potentially grow the TAM massively purely by improving their own products. The more capable and price-capable AIs get, the more they cut into various niches, of which there are a great many.
Even in my own field: reverse engineering is hotter than ever, and it's because AI allows RE efforts to scale to the size and complexity of modern software.
Anthropic's share of the total US stock market at IPO is 2%. At the current price, it's 2.74%.
Yet... here we are, google peaked around 5T in October, and GM peaked around 90B in the 2010s. The sheer size and scale of tech companies is hard to wrap our heads around.
That said, your percentage of market capitalization is indeed alarming.
Anthropic is slightly over a third of that with just an AI model. Absolute insanity.
$2T is 87x more.
Is this adjusted for inflation? I find comparisons across different decades to be somewhat useless in establishing a benchmark.
Sergey and Larry hated the idea of mispricing the IPO to generate the "pop". So they did a "Dutch auction" to find the _correct_ price and avoid the unnecessary pop that happens due intentional underpicing. Auctioning (ads) was their bread and butter anyway
Well, it popped anyway and it popped a lot. And no notable company has done Dutch auction since (to the best of my knowledge)
But no worries, pre-IPO investors/employees still hold like 95% of shares, so they didn't leave THAT much on the table.
In different terms, Google IPOd at a market cap ~ 10% smaller than Ford’s at the time.
Anthropic is at 50x Fords market cap. It’s over half of Googles market cap, and Google is a competitor in the same market with a dozen already successful lines of business.
They are incredibly highly valued in a way where I don’t even understand mechanically how they could grow like that without the entire financial system crumbling.
Google is up ~ 6,600% since IPO. Apply that to Anthropic and they’ll be worth something like $132T in 2048. As a point of reference, the Congressional Budget Office is predicting the US national debt to be around $102T in 2050 (paper was from 2020, might be outdated).
Anthropic, the singular company, will have to grow to be worth 130% of the entirety of the US’ debt. Napkin math says their revenue would probably have to exceed the US governments tax income. Like genuinely, Anthropic would have to become a world superpower all on its own.
They would be so big that they’re an existential threat to whatever country they call home. They would be such a large portion of the economy that central banks would lose any kind of control (presuming Anthropic doesn’t issue their own currency, they’d be a larger economy than most countries).
I can’t see how that could happen mechanically. The world would burn to the ground even trying to make it happen. If you thought Standard Oil was bad, this would be that but for every product and service in existence and all run by the same company.
Looking at recent mega-IPOs - SpaceX and Anthropic, with OpenAI coming soon, the price assumes that the company will be a success and simply discounts the projected future cash flow, as it was not risky at all. Future is not certain though. It will be fun.
Not sure why this was downvoted, are we angry at people building transformative products now? Or at startups making it big on their exits?
When these products are built by skirting the law, yes.
When its leadership keeps pushing for regulatory action to stop competition on its track, very much.
When their product allows them to collect the upside but do not give them any responsibility for the downsides, of course.
Also the fake moral outrage. The west has stolen 100s if not 1000s of trillions in wealth from other nations over centuries and causing people in those countries to leave their own homes to come to the west because of mass fuckery. Oh please spare me crocodile tears.
I hope employees are able to cash some out to get the reward they deserve, but the fundamentals don’t look good here. And to be clear they’d be selling to folks that are unlikely to see upside.
This quote from Margin Call feels applicable here:
Sam Rogers: The real question is: who are we selling this to?
John Tuld: The same people we've been selling it to for the last two years, and whoelse ever would buy it.
Sam Rogers: But John, if you do this, you will kill the market for years. It's over.
[John nods grimly]
Sam Rogers: And you're selling something that you know has no value.
John Tuld: We are selling to willing buyers at the current fair market price.
[Sam lowers his gaze]
John Tuld: So that we may survive.
Sam Rogers: You would never sell anything to any of those people ever again.
John Tuld: I understand.
Thank god.
Others can sell in waves beginning 27 days from the filing of the IPO prospectus.
Founders can transfer 29% of their stock 180 days after trading begins, with the rest of their holdings after a full year.
I have been through a couple of IPOs and the 180 day rule was the one I remember. We were totally locked out. At the time of the IPOs of those companies, we also didn't have the same secondary markets that are present today.
This seems like a very good deal for employees to get some liquidity.
No, this is cope. The difference with CDOs and llms, is that either you like it or not, they have massive value. They can already automate most of the boring white collar work. Fewer and fewer human decisions remain. They are not perfect, but they are improving massively. The value is there, people are hooked+ brain rotting, they will be dependent forever
The only money to be made is in selling compute and Anthropic rents all that.
$2T is a big number but this Anthropic IPO has been straight down since open, so I wouldnt say the market has gone completely insane.
I expect this will be the general trend. Not every day, but broadly over time. Anthropic will slowly at first, then quickly just capture most of the US economy. $2T is low for the first company in history on that trajectory.
It will also go like the 2007 crash: once all the layoffs happen, employers will realize that they can do the same work with a fraction of the workforce, thanks to automation. It will be a jobless recovery.
So we may get the crash AND the AI takeover. I just hope governments are able to respond fast enough to radical societal changes.
Anthropic prices initial public offering
From personal experience, I have worked with many IPO and stock option multi-millionaires who keep driving down the 101 and 280 every day to go to work. They could quit at anytime.
People who find themselves in a big tech IPO tend to not be the type of people who want to retire at 30 and drink pina coladas at some beach resort 24/7. They'll be bored, feel they are "not in the game" anymore, worry about wasted potential, worried about becoming unhealthy with a hedonistic lifestyle, etc. They just have a lot more financial freedom now.
Long story short, we live in a capitalist society and when you have capital it opens up a new level to the game of life. People will want to play that game and do more. Evolution made us to work, be communal, create, do things, etc. Cash-out, drop-out, and live a questionable lifestyle at various tourist-heavy tropical countries John Mcaffee-style is really unappealing to a vast majority of people. I'd certainly still work in some fashion if this happened to me even if I crave a vacation lifestyle. I'd be too bored. I'd probably take a lot more vacations and probably start my own companies, but essentially I'd love to keep making things.
I guess I'm the almost never happens. I had my "enough" defined before I took my job at the company that IPO's. Retirement was not to a beach resort but a University Town. Did lots of volunteer work. Took lots of classes. A lot of living my life did happen.
Can't say that boredom has ever entered the picture, I can invent interesting projects to pursue faster than I can finish them.
There are lots of alternatives to drinking pina coladas all day that are far more interesting than making more money, and making even more money for someone else.
Many of DIY MBA books I've read have some version of a story of some entrepreneur encountering a fisherman, trying to convince them to raise capital to expand their business... which leads the fisherman on a long trail that led them back to fishing and spending time with friends and family, right where they started. https://en.wikipedia.org/wiki/Anekdote_zur_Senkung_der_Arbei... https://paulocoelhoblog.com/2015/09/04/the-fisherman-and-the... https://bemorewithless.com/the-story-of-the-mexican-fisherma...
Sadly, the value of money has been hard-refuted numerous times already. The most valuable things currently don't exist, and as such, they cannot even have a price. By modern understanding of the laws of physics it's questionable whether they _can_ exist (they can, but people aren't intelligent enough to realize it).
Of which they only paid a $1.5 billion penalty. A pittance compared to the $80 billion they are raising and the $2T price that the market is valuing them at. Certainly a lesson to learn for future entrepreneurs ...
There is no global copyright law that governs ownership of digital goods. If it goes online, then you should expect it to be copied, pirated and widely distributed.
Anthropic, Google or OpenAI are not responsible for the mishandling of your copyrighted content that was publicly available on the web.
Just because it is wrong and illegal, does not change the fact that the data has been made available and distilled into open weight models and now every lab likely has used the data in some form directly or indirectly.
Cat's out of the bag and there's no going back. Back to my original point, the data is out there. China would do the same thing and not get in trouble.
Anthropic's revenues for the last 12 months are about $35B. Nothing, compared to the $2T GDP of Spain.
I don't know what kind of audience you think you're writing this to ... we all know what GDP and market cap are, you're not the Wizard of Oz here.
And yes, it is an interesting comparison, because it translates a somehow abstract amount of money into something a bit more tangible, i.e. the entire economic output of a developed country.
(The bad part to me is to think it means Anthropic’s revenue is comparable to Spain’s GDP. As you point out it’s not.)
GDP is value per time (year) and market cap is value.
(Value/time)*Time=value
Spain GDP for one year=Anthropic market cap
How long should it be before a country's GDP matches the value of that country? Should GDP integrated over any timeframe represent the value of a country?
From a strictly rational market perspective, the value of a company is the net present value of all it's future profits; assuming that's true, how do you make an equivalent comparison to GDP? GDP isn't measuring profit or even revenue.
These are all nonsensical questions.
It seems what he said is correct.
If you're going to make that comparison,
How long will it take for Anthropic to buy all of Spain's asset (land, real estate, companies, ...) with its revenues?
I don't know the answer to your question, because the value of a country's assets is a very difficult number to find.
Fortunately, that's not the case for company market caps or country GDP numbers.
Well sometimes you need to do that, eg if you want to figure out if it’s better to buy or rent. I don’t see why that’s a problem. You just have to remember what you’re comparing.
Correct, Claude can't play with your kids at the park, but Spain's GDP doesn't include that, so that's not relevant.
I'm a zoomer, so sorry if I'm mistaken.
IPOs these days are not the same as they were 10 years ago. Now, there is less of an upside and the early investors will sell 10% on to retail while you buy at its inflated price.
You are being used as their exit liquidity.
Edit: Thanks folks, typo, updated the number to billion. Question holds though.
[1] https://www.geekwire.com/2012/amazing-73-tech-ipos-profitabl...
Their operating loss is one third of last years with revenue going 12x. Let it grow for a bit more, they'll be profitable but at this growth rate it would be illegal financially not to invest in that growth momentum.
Now mention how much they have spent, they are literally selling $3 for $1
Their S-1 filing says that they've recently reached $2.08 earned for every $1 spent on compute and infrastructure, so nearly the opposite.
https://www.noahpinion.blog/p/ai-keeps-stubbornly-refusing-t...
Firing current employees is not easy in many jurisdictions, plus companies sometimes at least try to be gentle to them.
https://www.msn.com/en-us/news/other/dario-amodei-says-ai-wi...
But overall I agree that the claim is not very realistic: keep looking Claudson.
This sounds arbitrary until you appreciate that the ones with most purchasing power on the market are also the same ones who have previously proven themselves to be capable asset allocators, as proven by their ability to accumulate capital. So the valuation of a company is not just every random Joe’s opinion equally averaged, but weighted towards those whose opinions on value have previously tended toward being correct (this includes capital acquired by any means, not just equity trading. For example, if you earned a lot of money with your own product, you have proven yourself to be able to judge value and effectively allocate resources too).
The market starts to make a lot more sense when you start to use it as an oracle of hidden information (“what is the market telling me about the value of X?”) instead of an irrational adversary (“why is the price Y? That doesn’t make sense!”).
Nit: "Agreed to pay" should instead be "paid".
Buyers agree to pay more than they actually end up paying. Sometimes much more.
Most folks in finance I talk to think a realistic valuation backed by actual math is likely around $400-500B. Now that the investment bankers got their payday for this hype fest, expect numbers in that range to start coming out from analysts at the big banks.
Obviously you are going to hit growth ceilings, because there are only so many humans on earth, and open weight models could become "good enough" but when you are projecting out growth like this, you can really come up with just about any number you want and justify it.
Seems like you need to start with how a valuation is generated, first, before asking about one of the most complex valuations we've seen so far lol
Revenue from a single quarter is so far from enough data to even ask this question in jest, especially for a new tech company.
If they become the dominant AI company it could make sense. That's pretty uncertain though.
Meanwhile other countries are chugging along, doing whatever they want...
It's like that meme from about a decade back about the millionaire saying that the only reason millennials can't afford houses is that they eat too many avocados. That advice makes sense in a long-gone era where houses are cheap and avocados are expensive, but in the 21st century the price of personal avocado consumption would never add up to a house in a reasonable amount of time no matter how excessive your avocado lifestyle may be. jedbergh is going to be an "avocado guy" from now on (again not that I actually know anything about him).
What is falling away is that we just won't need as many SWEs as we used to. The way to make sure you don't "fall off the ladder" is to make sure you're the best at the new way of doing things.
It doesn't matter how good you are at anything when you can't even get a reply from a person. And absolutely forget about learning anything on the job. The ladder is gone. Combined with the war on the middle class, am looking at a retirement in poverty. Even social security is scheduled to be empty in a few years when available.
Sorry, I just checked and I don't see anything from you. Where did you ask? I try to respond to everything I get, but sometimes gmail doesn't surface random messages.
edit: Body contains Fry's Canoga Park etc.
As mentioned, the ladder is gone. Also, judging by the number of news articles, people care about 5x more about young people not able to work over older.
I can see this going a few different ways. We could be approaching a future where the machines can drive themselves and the computer goes from being a tool to a partner (ie Data from Star Trek). Or maybe we're approaching the future where computers still require humans to tell them what to do but they're good enough at figuring things out that nobody even needs to understand how a computer works to build the next enterprise-grade large-scale thingamabob.
The future you describe where software engineers still exist but they've carved out a niche managing AI software factories only makes sense if we've just reached the end-state of generative AI and there will only be incremental improvements from here on out. Otherwise I don't think you can safely assume that the AI software factory actually needs anybody with an advanced skillset to tell it what to do.
Success in a prior time does not guarantee success in the present or future as too many variables are different
Demographics have shifted wildly and most users are not a soft spoken grama needing help setting up pop3 in Outlook Express
Look around you; everyone uses their phone at whatever level of expertise they need now
Goodhart's Law has come for how we used to measure need and todos in tech. We hit the top of the S-curve relevant to this form of computing devices and there is nowhere else to go with it
I would understand this comment if I was telling you how to live frugally because it's not my lived experience. But in this case, it is my lived experience. I literally work at a job where I have to build a software factory to be successful.
Perhaps. But I work right now, and I actually care about people other than myself. I would like to see the industry that made me successful keep making others successful.
We'll have to wait a month or two I suppose to find out if you're right, but I think all companies will be moving to the model of software factories. And It's not something I thought of recently -- I've been building a software factory with AI for a year now, and working with others who are building entire businesses on it.
And FWIW, pretty much every tech company is now moving towards the same model.
This is my first reading about your work but I like your attitude. Databricks is also a good service.
From what I am seeing, those who embrace AI are embracing the 'computer' from the 80's. The landscape and reality is still being defined.
To the AI haters and doomers I say buy the largest house with cash you can to enable your retirement. The stress that people experience is the thought of their crushing debt or obligations that can be met with a 90% reduction in spend. Of course, that requires a lifestyle change that most are unwilling to make.
I don't really see a future where human labour at all can still be competitive beyond like 2050 in any field. Anyone who's ridden in a Tesla in FSD knows transport jobs are defacto over, the tech works and it will be copied and deployed en masse sooner or later. Humans don't scale and we haven't meaningfully mentally improved in 100k years, we're limited to 20W of compute, take 20+ years to train and go EoL in 40 after that. It's inevitable that we'll be outclassed and it only needs to happen once. Once a model exists that is slightly better at just one specific task, it only gets cheaper over time and it can be used for that forever.
large groups of people dont have good taste. Its always going to be reversion to the mean. which is just mediocre software
I assume the premise of a software factory is that it can churn out any software idea you have. You won't need to build a new distinct factory for each new distinct piece of software you dream up. Otherwise it wouldn't be called building a software factory, it would just be called building software.
So if somebody gets this right (this remains to be seen, the various Gas Towns of the world have not proven their value), won't one good software factory be enough for everyone to download and use? Wouldn't we all then be done with the labor of making both software and software factories?
For example, I could clearly see having a "Static web page with HTML and CSS only software factory" that just churned out that kind of software. Or a "Postgresql optimized SQL" software factory. And so on.
> end up with a factory factory
java nightmare ensues...every company is making a totally different product so its hard to see how every company is its own 'software factory' outside of such components (which they were already doing even before llms)
And while many companies may have differences, there are certain things most companies need, like a marketing website.
For example I pointed an agent at the bulletproof react website and had it generate an AGENTS.MD with the top level best practices and links to specific docs ("when creating hooks, read docs/hooks.md"), and a note to consult bulletproof react when in doubt. My react code was pretty solid, even though I haven't coded javascript in 8 years or so and was out of the loop on current techniques.
Unless we organize to prevent it, technology will continue be commoditized like all other technology into the hands of a few players.
And especially with AI it’s going to look more and more like nuclear or biological technology and that’s why you see so much push for regulatory capture.
Literally all I do is spend my time trying to figure out how to prevent this, and still make progress towards AGI for everyone, but it doesn’t seem like people care
Everyone just wants their bag everything else be damned, and that’s what we’re seeing
See web dev jobs tanking. Backend is next.
Backend dev opinions/arbitrary strings are being compressed out of the stack.
Devs that don't understand hardware development are naive about the endgame; black boxes that boot to a Google Opal like GUI.
Boot to Blender-like/WYSIWG environment loaded from a model itself. All the context we put in files added as vectors on the presentation layer.
Users will conjure a word processor or email UI and then a video game.
Network gear will boot a model to handle the specific job the model was trained on. That's what all the Stargate stuff is about but obfuscated behind marketing to hide implementation goals details from knowledgeable labor to avoid public scrutiny. Vague scrutiny without concrete proof is fine but if we could point to the details too they'd be naked.
50 million devs globally among 8 billion that do not want a subscription service to an expert when they have no such leverage over the expert themselves.
This isn't empty speculation. Hardware industry executives are responding to social pressure at scale.
America was able to leverage its clout to empower IT to pretend it's more than just math relative to memory and display matrices but America no longer has the global clout to assure that. To Chinese engineers it's just math.
American software devs were engaged in sophistry and social moat building. Ya got disrupted by the world. Believed they would ignore their eyes for the rhetoric and hype forever; rather Orwellian really.
The work of software engineers has enabled a (temporary?) profession of mostly busy-hands sort of work to thrive, mostly achieved by automation. For any software to presently work entails a substantial amount of things working in an automatic or "factorized" fashion. Thus, the engineering aspects will now move onto enabling a different sort of work, before it factorizes it, also.
This is multi-generational change happening in a decade. This is disruption across nearly all industries simultaneously. We can't just waive that away. There are going to be consequences.
But... I've also spent an incredible amount on poorer quality shoes and regretted it often. Many £50-150 shoes that just didn't last. The obvious planned obsolescence of eg trainer (sneaker) soles is just so obvious.
Damn it, why can't I learn the lesson!
£300 is equiv to £485 today too
Similar Loakes go for ~£320 and Cheaneys ~£500 today! https://herringshoes.co.uk/collections/oxford-shoes?filter.p...
We have to balance "a lot of people can get the product" with "product is high quality" somehow.
Eventually we just accept the new status quo, and the QOL hit that comes with it.
Anyone who hasn't FIRE'd yet should aggressively be saving as much as possible before shit hits the fan.
Good luck everyone! It's going to get ugly.
(And I'm getting lots of job offers. But I'm in Singapore, and that's presumably not where you are.)
I'm 40 now. Not sure if that counts as older.
"if its used correctly and broadly"
It's not enough to just have a claude code subscription for every engineer.
The powerband of AI comes through automation, integration, and tooling. When literally everything is touched by AI, and when most tasks get passed through AI before seeing a human, only then are you actually at an AI-first shop.
Only then will you actually realize how little you do in this AI-first world.
"You are holding it wrong."
Engineers are now more productive than ever. Increasing productivity drives salaries up. Secretaries are paid more today than they were when they didn't have computers to use. Why wouldn't this be the same?
It was trivial for 1 Rails dev in 2010 to do 10-20x more than a Java dev back then.
Today’s world is an interesting dynamic but I’m pretty excited about it.
I tend to be more on the "Jevons" side of things, where I do think that demand will more or less expand. There are lots of projects I've been putting off for years because there would be an aspect about it that I hate doing and don't have the money to hire people to do for me (e.g. frontend web development). Those are projects that otherwise wouldn't exist that do now exist and potentially might make money or something (though with my business acumen that's debatable).
Java itself was a much slower standard workflow. Dynamic languages in general, like Ruby, Python, Perl or PHP were much faster to work with than the burdensome recompile workflows. Think instant feedback vs waiting for your CI to finish after every single save. The speed of iteration was critical.
Beyond that, when the speed of development is slower, more people get hired to do the work. When more people get hired to do the work, there's more communication relay as the team size grows. Teams coordinating between each other. Decisions going to committees. Middle management layers.
You'd routinely see companies with 3-5 Rails devs outperforming shops with 50 Java devs. Communication and iteration speed was huge. The other piece of it was the convention over configuration aspect because it removed so many things you'd have to discuss and make decisions about as a group which further accelerated communications.
10-20x multiplier isn't an exaggeration at all. If anything I could be underselling it.
On one extreme, a solo dev could prompt a frontier SOTA agent with not much more than "Build Uber", give it the necessary hardware/software/permissions to build itself a useful feedback loop and act autonomously, put it in goal mode, and let it cook, and as long as the tokens kept flowing they'd eventually end up with something reasonable in actual orders of magnitude less time than it would have taken that dev to build the same thing — especially if you're counting calendar days rather than hours worked, and extra especially if they wouldn't have had the luxury of investing full-time hours into that particular thing. At the opposite extreme, BigCo full-time SWE #5000 tasked with an extremely specific ticket for a critical revenue-generating brownfield system might experience even "20%" as liberal, and correctly observe that writing code was never the bottleneck for their job. And then there's the massive spectrum of circumstances in between those two ends.
None of that means that the gains aren't real, or that anyone in particular is wildly exaggerating in either direction. It just means the gains aren't evenly distributed.
If you're allowed to use the newer libraries and tools, it has actually been a perfectly fine language to write software in, even relatively quickly, since ~Java 8 or 11.
I have had elaborate fantasies of what I would do to the designers of JavaEE if I ever found them, and I had assumed that because writing JavaEE is so unbelievably horrible that the entirety of the platform was like that. Years later I had a different job that was much more amenable to me using "modern" Java, so I used Vert.x and Disruptor and properly learned how to use NIO and and I actually had fun.
I think Java has been the "regression to the mean" language primarily because it's taught to frequently in schools and it allows you to write in a very verbose and irritating style, meaning that it was a great fit for mediocre-to-incompetent engineers who want to look like they're doing a lot.
sigh
Not some greenfield dinky todo list startup project with 0 users
Why do HNers always talk like these are the same as real products
I don't believe this.
in aggregate there is not as much need as we can create now
Thinking human skill will remain necessary for this is a huge bet against the AI companies.
If productivity increases lead to wage increases, being a weaver would be one of the best paid jobs today.
If it wasn't absolutely essential that a person was there, it would have been automated away long before AI.
My point is that MS Office has really automated everything about the job that can be automated. I don't think LLMs will really dent secretarial/PA work much further. Those people have their jobs not because the work hasn't been automated, but because their bosses are mediocre and relatively disorganised, like most people.
Silly Jarvis fantasies aside, an LLM cannot make your boss less shit at returning calls or managing their work life. The people who need support humans will always need support humans, until they themselves are replaced.
Post computers there are now 50 who see 20% more income
LLM might get rid of a couple more but it's not going to make a major difference
There are now 10,000 SWEs
LLMs will drop that to 50 who see 20% more income
I'm sure you are one of the 0.5%
No it doesn't. Salaries have been decreasing in real terms over the last decade despite productivity improving. There was a period of time in which productivity increased alongside salaries but that does not prove causation.
Do you have a source for that?
https://www.marketplace.org/story/2026/09/04/wage-growth-can...
Obviously, the above is talking about labor generally, not just SWEs. But speaking as a professional computer scientist at a place with decent salaries and excellent benefits, my raises over the last 5 years have barely outpaced inflation. It’s pretty annoying to realize that I probably would have earned a lot more money had I NOT gone back to school for a PhD. Not that I would make a different decision given the choice again, but money is much more abstract in your 20s than it is in your 40s.
Likewise with garbage collection, OOP, and SQL. Each has greatly enhanced programmer productivity, and the result is an entire ecosystem of software that would have been too big and complex for programmers to develop using the previous generations of development tools. It has gotten to the point where we can sit here looking down on all those "CRUD web app crap" projects that programmers can throw together -- try to imagine building one using the languages and tools that were available in 1970.
In 20~ years we will probably be talking about all the generic "Agentic ABC/ZYX apps" that an entire generation of programmers builds their careers on and what it would have taken to build such things using the tools that were available in the CRUD web app crap era before coding agents. There will be more programmers, because people who are not programmers today will become programmers with these AI systems (much like Lotus 123 turned accountants and actuaries into programmers -- yes I know about the London whale, I am not saying there will not be problems that come up).
Then inflation eats up the salary, but doesn't lower the productivity requirement.
Also, there are fewer secretaries.
From https://www.bls.gov/ooh/office-and-administrative-support/se...:
"Overall employment of secretaries and administrative assistants is projected to decline 2 percent from 2025 to 2035.
Despite declining employment, about 314,400 openings for secretaries and administrative assistants are projected each year, on average, over the decade. All of those openings are expected to result from the need to replace workers who transfer to other occupations or exit the labor force, such as to retire."
Inflation is about the real price for a dollar, ie how many goods a dollar can buy. It's not about the relative price between, say, your labour and the goods and services you want to afford. Inflation is a red herring in this discussion.
The last 40+ years of data on wage growth vs productivity growth says otherwise.
Much better than the bad old days of the 1940s to 1960s: if you had told someone back then that in 2026 obesity will be a bigger problem for the global poor than starvation, they would have called you a crazy dreamer.
The gravy train for shoemakers was over. What remained was a small number of shoe designers and (initially) a large number of unskilled factory workers. The same thing happened with many other skilled jobs.
But much will depend on how much professionally made software the world really needs. Demand for software could be a lot more flexible than the demand for shoes.
the market for SWEs will continue to be K shaped where the top of the K are SWEs that can solve ambiguous business problems and take accountability
The gatekeeping of making software is insane. Most people are building bullshit products for scum of the Earth corporations but they act like they're saving lives and changing the world.
We should all want more people to be able to make comfortable livings. If you don't want that, then you're pretty much the enemy of the people.
And more software does not necessarily mean more apps.
It does not.
Productivity increases and salary increases have been uncoupled for half a century
We’re talking about corporate compensation to labor vs shareholders and admin.
Why is there a steep decoupling on this graph?
Your argument is: automation making work easier is an appropriate reason to stop paying people comfortable wages, all at once?
I’m not convinced.
https://fred.stlouisfed.org/blog/2018/08/corporate-profits-v...
Because productivity measures all productivity, while salaries only track human labor? You do understand that in and around the 1970s we started meaningfully adding non-human labor into the mix, right?
"Increasing productivity drives salaries up"
To which I said: No, that's not the case; productivity and salary have been uncoupled for a long time. That's a different claim from "salaries did not increase", which is what you are countering.
The wage-productivity gap is well known and I won't attempt to school anyone on it. Just look it up. Its really disingenuous to claim that the gap is going "to the robots". Robots don't give a shit. The gap goes to the _robot owners_.
It is the case, though. Salaries have been driven up with increased productivity.
> productivity and salary have been uncoupled for a long time.
This is also the case. Once we added non-human laborers then obviously total productivity had to start to rise beyond salaries, which only track human labor. However, that doesn't mean salaries haven't also gained within increased productivity.
For a fairer comparison you should be looking at the human labor productivity vs salary chart, not total productivity vs salary chart, but I know that isn't as fun for the emotions.
Are you posting from 1990?
If they’re needed still, they are really needed!
My bank account disagrees with that statement. We are just expected to do more for the same pay.
I can name 5 other pieces of software I have open right now, more likely than not there are 100x supercharged engineers in this thread who work on them, and they're also the exact same piece of shit they were 2 years ago.
As a software user, when can I expect to see results of this unprecedented productivity boost? Are you people completely detached from reality, do you not see the output of the industry hasn't moved in the slightest?
Showing 400 of 2254. Read the rest on Hacker News
Comments are loaded live from Hacker News and are not stored by Mid or Real.