US Job Market Visualizer
karpathy.aiIt's wild that there are as many jobs in the category "Top Executives" as in the category "Retail Sales Worker".
This makes sense given both automation and the US's role in the global economy, but it runs somewhat contrary to standard ideas of class and inequality.
That category has a median pay of $105,350, and includes "general and operations managers" as well as "chief executives". I assume it includes executives of very small enterprises.
Good point. To take it one step further, if they are including 'general managers' and 'operations managers' in this bucket, then that should include the GM and Ops Manager at places like retail stores as well (for example, every Best Buy location has both positions, I'm sure it's similar for Walmart and other big box retailers too).
I took one glance at the chart and decided the results were impossible because of that.
Apparently "top executive" median pay is $105,350 per year: https://www.bls.gov/ooh/management/top-executives.htm
Remember that exec tech salaries are extreme outliers. I worked for an exec in manufacturing. He had full p&l responsibility for a business segment with ~150 employees, $27 million in revenue at 40% gross margins, and a production plant. His total comp was ~$300k.
Now just think of the comp levels in sectors like government, education, etc.
The number of people in the category is simply impossible for any normal person's definition of "top executive".
If you click the link it mentions "general and operations managers". They're tossing a lot of different roles into the category.
> Remember that exec tech salaries are extreme outliers.
It's the combination of tech and big or fast growing companies.
People who operate in FAANG or Silicon Valley bubbles (or who spend too much time on Blind) can lose track of what salaries look like in the rest of the world.
I often share Buffer's open salary page because their compensation is actually pretty normal from all of the data I've seen and hiring I've done: https://buffer.com/salaries
Every time it gets posted there are comments from people aghast that the software engineers "only" make $200K and in disbelief that the CEO's salary is "only" $300K.
Sounds plausible? Even a company with 100 employees and few growth prospects is likely to have a couple of executives, and most companies are small.
If you own a painting company with three employees you are a CEO and fall in the top executives category. You may or may not make 100k a year.
Don't a lot of CEO's famously pay themselves $1 and make their wealth on equity appreciation / capital gains?
Congress people are making about $176k, it's crazy to think top execs "median" make less than a congressman.
This is also base pay
These categories are extremely broad. Top Executive includes general managers, legislators, school superintendents, mayors, city administrators, and a lot of other government jobs. The name is misleading, it's basically non-frontline management.
Chief Executives is actually a specific sub-category of it and is, obviously, much smaller.
The gig economy is ruining government proxy metrics. A good number of ride share drivers are CEOs
> it runs somewhat contrary to standard ideas of class and inequality.
Can you elaborate?
When people think "top executives" they think of a very, very small group of people making tens of millions of dollars a year or much more. The reality is that that's not the case.
If AI produces surplus where does it go? Not talking about investment backed datacenter buildout and AI labs. Talking about the results of AI work...
I think AI outcomes distribute to contexts where it is used, and produce a change in how we work, what work we take on. Competition takes care of taking those surpluses and investing them in new structure, which becomes load bearing and we can't do without it anymore.
In the end it looks like we are treading water, just like it was when computers got 1M times faster in a couple of decades, but we felt very little improvement in earnings or reduction in work.
Surplus becomes structure and the changed structure is something you can't function without. Like the cell and mitochondrion, after they merged they can't be apart, can't pay their costs individually anymore. Surplus is absorbed into the baseline cost.
For my personal projects, any time saved on programming gets used up writing more ambitious programs.
For a business, the question is whether you can make more money by doing more ambitious things.
For a business, the second question is what your competition is going to do. If you have a monopoly over something, you can reap the rewards. But if you're in a space with lots of competition, you might not end up with any better profit margins if everyone's in a Red Queen's Race.
The surplus goes to the owners of the capital. Labor has been losing to capital for sometime now
If existing capital starts to generate excessive profits, more capital will be built, which will require human labor and will make the original capital less valuable.
In theory. In practice, the excessive capital of the incumbent allows them to price out or buy the budding competition, or the legislators, so as to protect their position.
The natural state of a capitalist system is the monopoly.
If AI being a million billion zillion times more productive at doing bullshit jobs nets in very little economic gain, then that lays bare the net economic value of all our bullshit jobs.
But given that the stock market hasn't panicked, this must mean at least one of these premises is false:
1. Economic activity is relatively flat.
2. AI makes us a million billion zillion times more productive than we used to be.
3. The stock market is rooted in reality.
> lays bare the net economic value of all our bullshit jobs.
This was already obvious, the more important question is what are we (collectively, society & our governments) going to do about it?
We (should have) already known most of our jobs were bullshit jobs, especially white collar jobs. The difference is now we might have something coming that will eliminate the bullshit jobs.
But society will always need bullshit jobs or the whole system collapses. Not everyone can go dig ditches, so what do we do?
Clean the bed pans of boomer retirees. And pay into a collapsing social security to finish off the wealth transfer.
Big opportunity for irobot to release a bedpan cleaner.
the market split from reality in 2020 for the last time. This is all just zeroes and ones, which is why they can make the real economy tank.
> In the end it looks like we are treading water, just like it was when computers got 1M times faster in a couple of decades, but we felt very little improvement in earnings or reduction in work.
I think this is a very important point. The hedonic treadmill means real gains are discounted. The novelty information cycle is like an Osborn Effect for improvements, like the semi-annual Popular Mechanic's flying car covers where there is an enticing future perpetually nearly here and at the same time disappointingly never materialized.
It's up to business owners to decide. At the end of the day, in free market economy, goods become more affordable.
Agriculture is a good example of that: http://www.johnhearfield.com/History/Breadt.htm
I think it's gonna mirror how the white collar classes, coastal elites, professional managerial class, whatever you want to call them, sold the countries industrial base to the far east. They got a little bit of money out of it but the biggest gains were the material wealth. $1 widgets instead of $2 widgets. All the people who weren't hurt by it got to live with more material plenty. Of course the nominal values of things didn't go down, but that's just inflation which is somewhat separate of an effect.
This time the jobs most in the crosshairs of AI are the jobs that constituted the paper pushing overhead of modern society, all the paper pushing jobs. Instead of $1 widgets from China replacing $2 domestic widgets it's gonna be $1 AI services replacing $2 services that require a real human.
This is hard to reason about because people tend to consume these kinds of services in big multi hundred or multi thousand dollar increments but in practice what it means is that when you have to engage an accountant, engineer, having something planned out in accordance with some standard, that will be substantially cheaper because of the reduced professional labor component.
And of course, as usual, the string pulling and in investor class will get fabulously wealthy along the way.
Remove AI from the statement and ask again.
Does the work you do provide more or less value to the company than your salary? Where does the difference go? If your killer feature closes a $5M deal, who gets that money?
We live as capitalist serfs. Someone else gets all the value you create, and you should be grateful for the peanuts they toss back to you.
> If AI produces surplus where does it go? Not talking about investment backed datacenter buildout and AI labs. Talking about the results of AI work...
The 1% pockets, this is where the vast majority of the extra productivity computers/internet/automation brought goes to for the last 50 years: https://www.epi.org/productivity-pay-gap/
fix(ui): improve accessibility and screen reader compatibility
Co-authored-by: Claude <claude@anthropic.com>
People who make data visualizations should try to learn the "rules of thumb" to not confuse or exclude readers:
1) Avoid contrasting red/green and blue/yellow, as these are common colorblind pairs.
2) Pick shades that still look different when shown in grayscale.
3) All bar charts should have 0 at one end.
4) Please no 3-D pie charts.
To find good color palettes, check out https://colorbrewer2.org
Everyone is creating visuals, not just data scientists or designers that probably should know these rules.
I generally am against people who have expectations of how they want others to communicate. Be it colors, pronouns, whatever- you’re just setting yourself up for disappointment and it’s not out of malice so just move on or find your own way to deal with what people are putting out there.
Interestingly, it seems from these statistics the median wage for individuals with a Master's is lower than a Bachelor's. I wonder if that's because of immigrants who pursue higher education for visa reasons skewing the data.
Anecdotally, many people get a bachelor's degree to check a box for job applications, whereas many people get a master's degree because they love the field and/or are afraid to leave school.
My friends and I who have a bachelor's degree in CS make more money than my friends who have or are working towards master's degrees in CS, because the former are working in the private sector and the latter are in academia making peanuts.
Other possible reason could be many or most Masters degrees not conferring additional pricing power, and those people’s Bachelors degrees also confer lower pricing power.
Edit: Another possible reason that Masters degrees were less common in the past, so the Bachelors pay statistics skew towards people with more work experience in their higher earning years, whereas the Masters pay statistics skew towards younger people with less work experience.
Masters seems to be a common theme in a few lower paying expansive fields like social work and education. I don't think that someone with a masters is typically making less in the same field all else equal.
Wow, I had no idea the reason my peers and I can't find another position in less than 12 months is because the market for software developers is growing faster than average!
Every year US absorbs 120k+ H1B+L1+OPT new visa holders. Considering there are 1.9M software engineers, market has to grow by 5% every year just to stand still. Add US graduates and you are talking about 10% growth required just to maintain employment. It's not realistic long term.
Congress/president should pause H1B visas or hike up fee to 200-500K so that only truly exceptional talent are allowed in. Right now it's just give away to corporations that are laying off people by tens of thousands.
you're not factoring in a few specific things:
1) how many of these people leave the country in this analysis.
2) OPTs likely will get h1b/l1s/leave the country and are being counted distinctly.
3) not all h1b/l1/OPTs are for tech. majority for sure, but there's a conversation factor.
specially in the current situation that green cards are much harder to obtain and many OPTs don't find a job, I expect 1 to be much larger than in the past.
as a more general observation, this line of reasoning does fit lump of labor fallacies: https://en.wikipedia.org/wiki/Lump_of_labour_fallacy
Oh, there's a name for it! I've sometimes been struggling to verbalize in the past the logical issue I perceived with the "immigrants steal are jobs" absolutists, and this is a useful reference.
Are new H1Bs a thing anymore?
Since the fee went up to $100k, I’m not aware of any companies still sponsoring hires who need a new H1B
As far as I understand the $100k fee applies only to consulate issued H1Bs. L1 -> H1B path (via AOS) is possible without fee. (Recent) US university graduates can also use similar path from what I understand.
We will see how much the $100k fee affects things during this H1B lottery round in few weeks.
Exactly https://news.bloomberglaw.com/daily-labor-report/few-us-busi...
> Only about 70 employers have paid a $100,000 Trump fee on H-1B workers from outside the US since it was imposed through a September White House proclamation, a government attorney said Thursday.
I think a lot of people have just moved to L1/O1/etc visas to get around it as OP pointed out, although a lot of people are still hiring H1B's. Amazon has applied for over 2000 H1B's so far this year, which puts them on track for ~7000 for the year https://www.uscis.gov/tools/reports-and-studies/h-1b-employe...
We have hit the cap for H1B's every year and we will always do so until we get rid of the program. Cheap labor will always be in demand.
A 100k one-time fee is nothing for big employers. That's 25k/year for 4 years, and if you realize that H1B's can't easily leave their job it's obviously worth it.
Compare hiring an H1B that is stuck at their job, to an American who can leave at any time. You can pay the H1B a lower wage to compensate for the fee you paid to get them into the role. 25k/year for 4 years is worth it for not only the reduced churn that comes with training a new person, but also you don't have to pay any of the incentives that come with getting a new employee into the role like sign-on bonuses, wage bumps, benefits etc.
There's an X account which just posts universities hiring H1B's for ~half of what it would normally cost to hire people. An 80k/yr senior software developer will always be in demand, especially if the team is already predominantly non-american
Universities typically are in the public sector side of the equation... and the public sector doesn't pay any non-administrative role the Big Tech rate.
Pulling up my alma mater... https://www.openthebooks.com/wisconsin-state-employees/?Year...
The various roles that you'll find for software developers: Sr Is Specialist, Is Tech Srv Cons/Adm, Sr Inform Proc Conslt, Sr Systems Programmer
And you can pull up the pay scale at https://hr.wisc.edu/standard-job-descriptions/?job_group=Inf...
$80k/y isn't "we're paying H1-B half of what the going rate is" but rather "the state legislature has set this pay scale and we're paying everyone that amount" ... And many times, H-1B visas aren't eligible to work in those roles.
> Universities typically are in the public sector side of the equation... and the public sector doesn't pay any non-administrative role the Big Tech rate.
There's absolutely no reason government couldn't pay competitive rates for software engineers. They do it for doctors and administrators of state-owned medical centers. Not to mention football coaches
It's only a little bit lower than salaries for non-Big Tech that are in the area. Again, for Madison compare it to https://www.levels.fyi/companies/american-family-insurance/s...
Trying to make state government competitive with Big Tech salaries (especially in states that aren't California) would not go over well with voters.
While private sector deals with layoffs and uncertainty, the public sector has things like "budget not good this year? Two weeks unpaid vacation for everyone" - https://docs.legis.wisconsin.gov/code/executive_orders/2003_... ... 401k matching? How about a fully funded pension instead. https://reason.org/commentary/the-wisconsin-retirement-syste...
Football coaches are revenue generating for universities... software developers at universities not so much. Doctors are licensed professionals that have a decade of schooling... software developers frequently reject licensure and celebrate their lack of a formal education.
But there's no reason they couldn't just pay them more. The way to make them pay more is to force them to hire applicants at market rate, and when that's impossible, they'll go to the state legislature. Allowing for the H1B loophole is the problem universities are too eager to abuse
Maybe it’s because you are really a computer programmer. Computer programmer employment is expected to decline by 6%.
Data is from 2024.
Are childcare and kindergarten teachers really exposed to AI? In theory, we could put a class of 30 children in front of chatbots with one supervisor. But I doubt we would chose to do this as a society. If office work becomes more automated, early childhood education is actually one area I'd expect to take up the Slack. I can't imagine a situation where we have millions of unemployed former office workers but we leave them idle and let our children waste away in front of screens.
Childcare and education requires a specific tolerance, mindset and passion to be effective though. I'd be curious how many previously-PMs or HR drones or email jockeys would be adequate (let alone thrive) in an environment where there are next-to-nonexistent budgets, and you're servicing literal babies and tiny children lol
On second thought, client service folks might do extremely well here!
>specific tolerance, mindset and passion
What you mention here is the exact thing why my earlier relationship went bust, because I didnt have any of these, then the children arrived :-X
Con: kids
As a current parent, I assumed this was due to people having fewer kids, not AI. Additionally, with childcare centers becoming more expensive, many more families are looking to be stay at home parents or using grandparents / relatives to watch their kids during work hours.
There are a lot of education and curriculum companies pitching basically this- replace those 'expensive' teachers with aides making minimum wage as all they need to do is recite curriculum and help them log in to be evaluated.
I'd say yes. One teacher can use AI and be able to cover for more children. Thus less teachers are required.
That could work in ideal world where children behave nicely, and are eager to learn. But in reality that's not the case. Especially in high school big part of teacher's job is keeping order and being the authority figure. Good luck replacing that with LLM.
Bet on high job-growth market: Security guard for data centers.
Stand in front with a gun while mobs come to burn down the data center that took their jobs.
(I think I'm half joking).
> Stand in front with a gun while mobs come to burn down the data center that took their jobs.
I doubt it'd be old hit-and-run, more like small scale Ukraine with drones filled with explosives.
Still need rocket defense though.
Oh please that will be done by drone and killer robots.
Drones took the jobs of angry mobs and the world will never be the same.
My takeaway here: 3.XT $ of US salaries are the TAM for AI companies.
Apple, a very successful company, makes 300B/y revenue? (ish)
~10% is all you need to be Apple.
And, it can work by taking all of 10% of the jobs and collecting the whole salary (the AI employee -- dubious proposition),
or by taking 10% of everyone's salary and automating part of everyone's job (the AI "tool" -- much more plausible).
If "part" being automated is >10%, we all win in the long run, every company gets productivity growth without cost growth, etc etc.
If you add in data center costs, and multiple competing AI companies, and then expand the TAM to all white collar work worldwide, you can make everyone successful beyond their wildest dreams with a "20% of work for 20% of the cost" model. Again, how you distribute that 20% remains to be seen (20% new unemployment, or new 0% unemployment with "tools".
I formalized my thoughts here: https://jodavaho.io/posts/ai-jobpocolypse.html
Your math is missing the fact that Apple products are differentiated from their competitors. If AI becomes a ubiquitous commodity, it's not worth 300B/y.
Potable water is far more important than AI or iPads ever will be, but the world's most valuable water company only does about 5B/year in revenue: https://en.wikipedia.org/wiki/American_Water_Works
He originally vibe coded this as AI Exposure, but deleted that one because people were misinterpreting it. Someone mirrored that one here https://joshkale.github.io/jobs/ . EDIT: D'oh, I now see Karpathy didn't delete it, he just made it into a "Digital AI Exposure" button.
I wish there would be a color blind friendly version of this. I have deuteranopia and can’t distinguish red from green in the page.
I'm colorblind as well and what's fascinating to me is that this is the second AI created chart in a week I've seen that I can't read. Surprisingly I've found such agressively colorblind-unfriendly charts to be far less common when created by humans.
Out of curiosity, what colors (or text treatments) do you personally prefer to confer "gain" and "loss"?
It's not about preference, but about being able to see the differences in the first place, so any sufficiently contrast combination should work.
If you turn on the color filters in accessibility settings in macOS you can see what the contrast could look like to a colorblind person.
You're not gonna believe it but NONE colors, my brain doesn't process colors because I know I cannot trust them to be correct. So I use shapes, numbers, arrows.
Ask a colorblind person to explain how to win at candy crush and you'll be surprised (hint: we do not use colors, we use the shapes).
I don't have any color discrimination deficiencies, but it is my understanding that for various types of signage, the move has been towards RED=bad/danger/etc, and BLUE (instead of green)=good/safe/etc.
For color deficiencies, different lightnesses are safe e.g. dark for loss and light for gain (could be dark reds for loss and light greens for gain, but don't mix the lightnesses). Other options are icons/shapes (like up/down arrows) or pattern fills (like stripes for loss).
The general trick is you can rely on differences in color lightness, patterns, text and icons, but not differences in color hue. The page should be usable in grayscale.
> You are an expert analyst evaluating how exposed different occupations are to AI. You will be given a detailed description of an occupation from the Bureau of Labor Statistics.
> Rate the occupation's overall AI Exposure on a scale from 0 to 10.
The sad part isn't that this is low-effort AI slop, but that intelligent people and policy makers are going to see it and probably make important decisions impacting themselves and others based on these numbers.
This is 99.44% slop! You are completely correct. The "exposure" is based entirely on vibes and does not correspond to observable reality. Down here in the real world the very first sector that is being disrupted is manual farm labor. They are out here with machine vision and quadcopters picking fruit. But according to the prompt that produces the treemap, manual labor has an exposure rank of zero.
This is a surprisingly bad treemap, which is surprising given how easy it is to build a good one these days.
"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."
Looks average to me. Maybe I don't know how the good ones look? Can you give some examples?
Treemaps are used to show hierarchical data. But here he doesn't even bother to show more than one-level of hierarchy. If I want to find, e.g. Police, it's near impossible, since I have to scan with my eyes (when, again, it's trivial to add another rectangle to show Law Enforcement or other).
In addition, little work is done to separate the classes. He has probation officers in the same node as teachers, completely separate from law enforcement.
Here's some much better examples:
- https://www.washingtonpost.com/nation/2022/05/04/abortion-nu...
- https://flowingdata.com/2015/04/02/how-we-spend-our-money-a-...
It's kinda cool to see a whole lot of otherwise intelligent people who are so dogmatically and ideologically opposed to anything AI that they're going to willfully dismiss anything that AI produces regardless of utility.
It's not great for them, but it's a definite advantage for people who are already in the mindset of distinguishing and discriminating information and sources on merit, instead of running an "AI bad" rubric as part of their filter.
AI has already won. It's taking over. It might be a year or two, or five, or ten, but AI isn't slowing down, nobody is going to pause, and there's a whole shit ton of work people do that won't be meaningful or economically relevant in the very near term. Jevons paradox isn't relevant to cognitive surplus - you need a very different model to capture what's going to happen.
It's time to surf or drown, because it doesn't look like any of the people in charge have the slightest clue about how to handle what's coming.
- 6 months ago
i wonder what people in restaurants use and why
From my experience, restaurants hand-wash some stuff (anything that needs scrubbing such as cookware) and use dishwashers for light-soil service items (plates, glasses, cutlery). But these aren't dishwashers like you have at home. They run very hot water and complete a wash/rinse in just minutes.
I get where you're coming from but dishwasher is definitely a "could live just fine without."
Fridge OTOH, not so much.
This is a great analogy, because just like AI, microwaves are good for quick fixes, tasks where you don't really care about the quality and would rather minimise the effort.
I think the analogy is a bit inaccurate here when people are talking about automation.
Microwaves do one thing, but they do it reliably. Microwaves didn't affect the culinary industry because cooking is far more than just heating food, and many tasks are very difficult to automate. LLMs are more general-purpose - the average Joe is now relying on them as a source of truth, advice and mental work across the board. However, LLMs can't be guaranteed to always be reliable, it's all probabilistic. The threat of automation here is in taking away a lot of the less important or less complex work. Low impact + high precision (microwave) vs. high impact + low precision (AI)
how is it a great analogy? do microwaves improve as fast as AI has been?
Yes, they did, back in their day.
> and cooking with an oven is more of a special occasion thing than the default cooking method that it was before.
That really only makes sense if for households with a toaster oven, single adults, childless couples, and retired people. A toaster oven makes a lot more sense for small meals, in part because it can heat up much faster than a full oven.
Otherwise, a daily family meal isn't a special occasion.
They won at automating a task and becoming indispensable in the larger ecosystem of related tasks.
This was a real, unironic mindset for a while: https://a.co/d/0iYb8mlz
It looks like a template with placeholders. Before it was "big data", then "Crypto", now they used "AI"
Microwaves are the trend of the past! It sounds like you don't own an air fryer.
What is the argument here? Someone had a wrong take on something completely unrelated, so it somehow applies to this?
You think "there's a whole shit ton of work people do that won't be meaningful or economically relevant in the very near term" is wrong?
AI being bad isn't in conflict with AI winning or taking over. I think all of those things are true. I think what we currently call social media is bad. And it's won. No conflict there either.
> AI has already won. It's taking over. It might be a year or two, or five, or ten, but AI isn't slowing down, nobody is going to pause, and there's a whole shit ton of work people do that won't be meaningful or economically relevant in the very near term. Jevons paradox isn't relevant to cognitive surplus - you need a very different model to capture what's going to happen.
No, AI has not "already" won. And phrasing it as you do, "It's taking over. It might be a year or two, or five, or ten" is an admission of that.
People may indeed not pause, but there's never any guarantee that the next step of progress is possible; whatever we reach may be all we can do, and we'll only find out when we get there. Or it might go hyperbolic and give us everything.
I'm not certain, but I suspect Jevons paradox is probably the wrong thing to bring up here, that's about cheaper stuff revealing more latent demand, and sure, that's possible and it may reveal a latent demand for everyone to build their own 1:1 scale model of the USS Enterprise (any of them) as a personal home, but we may also find that AI ends the economic incentives for consumerism which in turn remove a big driver to constantly have more stuff and demand goes down to something closer to a home being a living yurt made out of genetically modified photovoltaic vines that also give us unlimited free food.
(I mean, if we're talking about the AI future, why not push it?)
What I do think is worth bringing up is comparative advantage: Again, this is just an "I think", I'm absolutely not certain here, but if AI can supply all demand at unlimited volumes*, I think the assumptions behind comparative advantage, break.
> It's time to surf or drown, because it doesn't look like any of the people in charge have the slightest clue about how to handle what's coming.
Yes, and I think they've also not even managed to figure out the internet yet.
* and AI may well be able to, even if all models collectively "only" reach the equivalent of a fully-rounded human of IQ 115; and yes I know IQ tests are dodgy, but we all know what they approximate, by "fully rounded" I mean that thing their steel-man form tries to approach, not test passing itself which would have the AI already beat that IQ score despite struggling with handling plates in a dishwasher.
Asking an LLM to analyze data directly doesn’t work. But they’re great at writing scripts to analyze (and visualize) data. Anthropic just figured this out last week and gave Claude a mode that does that for you.
This. I only ask LLMs to summarize non-critical stuff, i.e. just give me a general summary of all the work done over the past week.
If I were in need of hard analytics you can be damn sure I'd have it build a tool with a solid suite of tests following a rigorous process to ensure the outputs are sound. That's the difference between engineering and vibing.
Yes, you have to calibrate the effort to the task, you can't just blindly vibecode it. But if you treat it like a new college hire who still remembers their stats course, rather than a senior analyst who will just come back with the right answer, you can do some pretty high-level stuff that's trustworthy. It's so fast that it's no problem to double/triple check everything and even do it with multiple methods.
I'm wondering if you're confusing "AI" with "LLMs" here.
I think LLMs are the equivalent of someone with a PhD in English literature and a few other things, and can be very intelligent and literate without being particularly good with numbers.
On the other hand you have plenty of machine learning numbers that are absolute beasts at everything number-related. I'm assuming you wouldn't put George RR Martin in charge of building your datasets.
> AI has already won. It's taking over. It might be a year or two, or five, or ten, but AI isn't slowing down, nobody is going to pause, and there's a whole shit ton of work people do that won't be meaningful or economically relevant in the very near term.
I think AI is not going anywhere.
I also don't think the future will play out as you envision. AI is a very poor replacement for humans.
And I say this as a misanthrope who doesn't have a particular beef against AI.
I have found:
Published AI generated code is a mild negative signal for quality, but certainly not a fatal one.
Published AI generated English writing is worthless and should be automatically ignored.
What doesn’t make sense to me about the AI Inevitabilism Embrace Or Die trope is how there’s going to be a sudden trap door which will eliminate all the naysayers which can be avoided by Embrace. Because that doesn’t cohere well with how autonomuous AI is or will be.
I could understand if all the naysayers doing old fashioned stuff like work all of a sudden have no more work to do. But the AI Embracers will have what, in comparison? Five years of experience manipulating large language models that are smarter than them by a thousand fold?
> Jevons paradox isn't relevant to cognitive surplus
Could you elaborate on this? Is it just a claim, or is there some consensus out there based on something that it doesn't/shouldn't apply?
The best predictor I can find of a segment being green in median wage is if it's red in digital AI exposure.
So... What exactly are you talking about?
He is talking about the same thing as you, no? As you point out, the more AI exposure (red), the more likely to have higher wages (green). Which suggests that those who are embracing AI are those who are thriving the most. Same as what he suggested.
No, those numbers don't mean that.
Whether people are adopting AI or not, everybody doing the same kind of job gets the same number for exposure to AI.
You can claim that AI is creating a Jevons paradox situation and making companies hire as crazy the people it nominally replaces. But then you would have to point any instance of that happening, because it's clearly not there either.
They must, else you'd have elaborated what they do mean already.
Assume I want to believe exactly what you're saying. What is that, though?
a. "Has already won"
b. "Might be a year or two, or five, or ten"
You’re playing chess. You see that you have a forced mate in several moves. You’ve already won, but it will happen in 2, or 5, or 10 moves.
(This isn't how chess mate-finding works.)
Yes it is.
I think a lot of the pushback comes down to your attitude. The way you're talking about AI is like how the crypto bros talked about bitcoin. Just being very insistent on your point of view is a red flag. Either you can present new data to convince people, or your insistence will just look like it's emotional rather than rational.
I use AI every day as part of my work, it's very unclear to me where it's going and we have no idea if we're on an exponent or S-curve. Now, normally people talk with conviction because they have more data. But one of the breakthroughs of crypto was this social convention of just have very strong opinions based on nothing. A lot of that culture has come over to AI.
Your comment typifies this, it's all about I need to get on board, AI has already won, you've got an advantage over me because you realise this.
Go back, look at the actual article you're commenting on. Did the AI analysis of job exposure provide anything of value. I'm not totally convinced it did, and you didn't even think about it. What critical thinking did you do about the data that came out of this dashboard.
100% on point, the similarity to crypto slogans and posts is very suspicious. But crypto made a lot of people rich so I guess it works...
Well what do you mean by "works" the guys on twitter screaming about "Have fun being poor", were by and large just trying to scam you. Like, I could phone up your grandmother and convince her she's got a virus and she needs to transfer her life savings to me before the hackers get it, that could make me rich, does that "work". I don't know what Crypto actually worked for beyond - creating a target rich environment for scammers, a neat way to buy drugs, and good way for criminals and rogue nations to launder money.
Uh huh.. but the data in Andrej's visualizer is showing software development growth outlook is at 15% (much faster than average)
Over the past year (where Opus has supposedly changed the game), we're seeing ~10% more job postings for software developers compared to this time last year [1,2]
A huge amount of our work is not easily verifiable, therefore it's extremely hard to actually train an LLM to be better at it. It doesn't magically get better across the board.
AI HAS WON. SURF OR DROWN. YOU DONT KNOW WHATS COMING!!!?!?!
Stop with this doomer drivel. It's sick. It's not based in reality and all it does is stress innocent people out for no reason.
>opposed to anything AI
AI is great for searching. I ll give you that. And that itself is a big deal. In software development, there is also real value provided by AI if you use it for code reviews. But I am not sure how much worth it would be if you have to retrain a model with new information just to give better search results and for code reviews..
Maybe that will be subsidized by all the people like you who want everything to be done by AI, for the rest of us to use it as a better search tool and use it for quick reviews..who knows!
> AI has already won. [...] It might be a year or two, or five, or ten
brainbroken by chatbots lmao
Wishful thinking. AI is useful, but it’s far more niche than militantly pro-AI people like you want to believe. It’s a useful tool, nothing more.
I immediately turn off when I read extreme points of view. It tells me they are people that lack traits such as nuance - essentially time wasters.
Eh, idk about this. One nice thing about humans is that they still feed themselves when the economy collapses.
"AI has already won. It's taking over. "
Man.. I suggest you touch some grass. You are living in a bubble.
It is free for you to say this, because if you're wrong, there will be no consequences. Words are cheap. No different than various CEOs saying "AI will replace these workers" and now having to hire back those they laid off. Klarna, Salesforce, etc. Will be a great comment to reference in the future to capture the exuberance of the times.
Companies Are Laying Off Workers Because of AI’s Potential - Not Its Performance - https://news.ycombinator.com/item?id=47401368 - March 2026
> Some companies that announced large headcount reductions because of AI have since revised their talent strategies or have faced public criticism. Klarna, for example, the Swedish fintech that offers “buy now, pay later” e-commerce loans, reduced its human workforce by 40% between December 2022 and December 2024 as it invested in AI. (The company used a hiring freeze and natural attrition, not layoffs to achieve this cut.) But in 2025 the company’s CEO told Bloomberg that Klarna was reinvesting in human support, explaining that prioritizing lower costs had also led to “lower quality.” A spokesman told HBR that the company has hired about 20 people to deal with customer service cases the AI assistant can’t handle, and that the use of AI “changes the profile of the human agents you need in the customer support role.” The language-learning company Duolingo announced that AI would be used to replace many human contractors, and it faced considerable criticism on social media.
> For one, AI typically performs specific tasks and not entire jobs. As an example, Nobel laureate Geoffrey Hinton stated in 2016 that it was “completely obvious” that AI would outperform human radiologists within five years. A decade later, there is no evidence that a single radiologist has lost a job to AI—in part because radiologists perform many tasks other than reading scan images. Indeed, there is a substantial shortage of them.
The 'AI-Washing' of Job Cuts Is Corrosive and Confusing - https://news.ycombinator.com/item?id=47401499 - March 2026
* Companies are "AI washing" layoffs, blaming artificial intelligence for workforce reductions they would have made anyway, according to OpenAI CEO Sam Altman.
* A Resume.org survey found that 59% of hiring managers say they emphasize AI's role in layoffs because it "is viewed more favorably by stakeholders than saying layoffs or hiring freezes are driven by financial constraints".
* The stated reason for the layoff matters more than the fact of the layoff, and framing cuts as proactive restructuring around AI can result in a valuation boost, even if the technology doesn't actually work.
> The AI premium isn’t even reliable. By late 2025, Goldman Sachs group Inc. found that investors were actually punishing AI-attributed layoffs, with shares falling an average of 2%. The analysts concluded that investors simply didn’t believe the companies. But Block’s surge shows the incentive hasn’t vanished. It’s just a lottery instead of a sure thing. And executives keep buying tickets.
> The broader data confirms the gap between narrative and reality. A National Bureau of Economic Research study published in February surveyed thousands of C-suite executives across the US, UK, Germany and Australia. Almost 90% said AI had zero impact on employment over the past three years. Challenger, Gray & Christmas tracked 1.2 million layoffs in 2025, and AI was cited in fewer than 55,000 of them. That’s 4.5%. Plain old “market and economic conditions” accounted for four times as many.
So! Sophisticated capital market participants don't believe this; why do people here?
AI is making CEOs delusional [video] - https://www.youtube.com/watch?v=Q6nem-F8AG8
I'm very confused how you can put up such an obvious strawman, say all these wildly unsubstantiated things, and yet still get engagement. Who are you even talking to?
It's been several years and nothing has changed except the AI grift is crumbling as we get out of the post-covid slump.
I'll tell you what AI apparently can't replace and that is information graphics designers who are familiar with the exotic detail known as "contrast".
Keyboard Shortcuts
- j
- Next item
- k
- Previous item
- o / Enter
- Open selected item
- ?
- Show this help
- Esc
- Close modal / clear selection