"These models lack the taste that makes for effective product decision making."
This sentence, whenever I see it, makes me chuckle now.
Surely the leopards won't eat my face. I have taste and you can't get that from a machine.
If the higher-ups don't care about the quality of your code when they compel you to tokenmaxx, what on earth gives you the impression they have a clue about your taste?
If you can direct AI and produce more value per $/time spent then a non software engineer directing AI to develop software then you still can produce a lot of value.
This should be the case now for anything that is not a prototype. The job has changed from "writing code" to "shipping software". A technical lead.
As AI gets more intelligent and the tooling around it develops and organizations figure out how to get it access to the data, the level of involvement needed for software engineers will decrease. Since the cost of developing software is plummeting, the amount of software being used is going to expand greatly- that will keep the technical lead positions available longer and also create more opportunities for consultant engagements.
Although someone can use AI to rewrite every software dependency and every paid piece of software, its still not worthwhile for them for most software, especially if the price of the software is low compared to the complexity it hides.
Users of a paid software product still need a human to be responsible for it.
As you develop your skills with AI, you now have the ability to develop your own software products. So there's an opportunity to become a small business owner. You didn't want to do these non-technical roles, but now you can continue to direct AI to automate all the aspects of the business.
> That first meeting crushed my soul. I heard my ears ring more violently than ever. The color drained from the world. I could barely walk straight.
There was a time that I really cared about my work and about my colleagues livelihoods.
Then I realised, it’s just a job. I’ve lost jobs before. I’ve taken shitty, low paid jobs before. It’s just a job. You turn up, they pay you, then you go home to your actual life, the one that matters.
That's a nice attitude to have when you know you can get another job. That allows you to support your family. Without moving to another city or drastically altering your lifestyle.
Unfortunately not everyone has that luxury, and as AI continues to devalue intellectual labor, fewer will.
Well, I don’t have that luxury ! I’m old, I do have a family to support, and I’ve been out of work for 6 months at a time recently, under financial pressure.
> That's a nice attitude to have when you know you can get another job.
Perspective... there are people crossing the entire world to come and be trashmen or food delivery slaves in our western countries, there are other jobs, you just don't want to do most of them
Don't live like your cushy $$$ desk job is a god given right and you'll be fine
Hot take, but I don't think it actually does, tbh.
Setting aside the temporary delusions of management, the tech itself doesn't devalue anything. It just optimizes away redundant stuff. That was always kinda redundant, but it wasn't "intellectual labor" just because it was done by white-collar workers.
We only have to go back to pre-AI days and look at what kind of mindless nonsense tasks said workers back then complained to see what I mean.
No people ever wanted to do this kind of work anyway, so it is fantastic that they in the future will not have to. The only thing we need to ensure is that the benefits of that will not be harvested by the same rich few. But that was also the mantra before AI, so nothing changed there either.
We can do so, because anything made out of carbon can also be reduced to carbon. Nothing and no one is untouchable, but, ideally, words are the tool. Anything else has unforeseen side-effects.
> No people ever wanted to do this kind of work anyway
No one ever wants to do most of the work they do. That is the literal definition of work, as opposed to a hobby. You do it because otherwise you and your children starve. Job satisfaction is optional (and may cut into your salary).
The author has mix of empathy and fear. They do not necessarily care about the job as a much as much they are probably are not as much stress tolerant and immune to things around them.
Loss of a job/career shouldn’t define a person. Loss of a human being does. For too long we ask people we meet “what do you do for work?” Perhaps the question that lets us better know a person is “who is your family?”
I feel the same as you - but I'm fortunate enough to be young, have a cushion of cash, no mortgage, no mouths to feed. I'm guessing you're similar. We might feel differently without these.
That's really the key - the situation doesn't change, but you can - and how you look at it.
Either you're a helpless waif buffered by the storms of this world, or you're in charge of at least something - picking the first results in reddit-tier apathy and despair.
The second may not cause much immediate change, but at least you'll feel better about it.
Capital and Labor have always had a tense relationship, each one uneasy about the necessity of depending on the other. Unfortunately, Capital’s acquisition of power in the modern Western world has not been accompanied by a commensurate acquisition of confidence. In fact, as Capital’s power has increased, so has its sense of its precarity, becoming so acute as to tip over into paranoia. The moment AI achieved the barest shred of plausibility in its promise to replace Labor, it became an article of faith that Capital feverishly believes in and aggressively acts on.
All of this is to say that we have a case of “the market can stay irrational longer than you can stay solvent” on steroids; Capital is going to fight like its life depends on it to stay irrationally convinced in their belief that AI can replace Labor.
Always interesting when words change their meaning; because in this case the part of Labour that does the actual labour is going to be fine. It'll be the layers of people who do nothing but sit at a desk and press buttons that really feel it.
AI might be the best thing to happen for labour n fact; because they're going to gain a lot more intelligence for the same amount of work.
This is a lesson I learned a long time ago, no matter how good you are, you could be obsolete and replaceable any day and the companies/corporations don't care. Treat your work as a job and don't make it the center of your life.
Hobby woodworkers still build tables themselves, despite the existence of factories, assembly lines, IKEA.
Until the government forbids writing programs by hand, writing programs by hand will remain a satisfying hobby. The only thing that's changing is how much other people appreciate the skill and effort you put into a given program.
>Treat your work as a job and don't make it the center of your life.
Good advice regarding psychological investment.
Not so useful advice when the job provides your actual livelihood, and the changes (e.g. outsourcing or AI or some industry change) mean you might soon hold a "will work for food" sign in some street corner.
I mean these days, it's not usually that people made some particular job their whole personality and "center of their life", the reason that they worry about losing it.
It's that their job is what puts food on the table, and the job pools are drying up.
I understand that aspect but when you treat your work as just a job, you actually live with the mentality that you could get fired tomorrow and that's okay and you're mentally and financially prepared for that.
Also a lot of tech workers earn a higher salary compared to most of other professions, so I'd expect people to put away some buffer for a rainy day.
Being mentally prepared, possibly, and even financially as far as being mentally prepared can go, but then not everyone have be fuelled by FANG perks&benefits.
Remembering when I lost people to offshoring in the mid 00s. I made the case, with metrics, that the quality was nowhere near as good and the spec overhead was non-trivial.
Management's response? It doesn't matter - for the price they could get cheap offshore workers to do it over again 5 times and still save money.
Why would there be a rug pull? Open models do seem to be giving companies like OpenAI and Anthropic some decent-ish competition, so if they try to increase prices too much then people can just leave them.
For the same reason that most businesses are on AWS/GCloud rather than having a server room in the office.
Self hosting is a complex investment. Competitors need an stupidly expensive amount of hardware to enter the scene if they're going to serve SaaS. Chinese backed models can be regulated out of competition. Openai and Anthropic will soon be semi exclusive sources of training data as development happens on private conversations.
Success is not guaranteed, but it is crystal clear that the goal is replacing us and then pricing the replacement as the largest possible percentage of our salaries possible.
It's the oldest trick in the book. Place a heavily discounted supermarket in town, wait for all local businesses to close, raise prices.
All LLM are heavily subsidized by investors. No one is paying the true cost of tokens at the moment.
The rug pull is to get as many people defendant on the tool before they have to massively increase prices.
Sure, open models will definitely be an alternative once the frontier models hike prices to match costs.
But that will be a minority, and people will start to switch, but people use the frontier models for a reason and these AI labs will extort that, once the easy investor money dries up.
My company pay a copilot subscription to everyone, plus my team have a bedrock subscription that cost us 7k last month. In the meantime, some guys created a new internal team and are testing open source models (for our red team) on our own hardware (we didn't sold our 'old', 3 year old servers RAM/CPU, so the only investment was GPUs). My guess is that within the next 6 months our red team will use open source models exclusively, and my team will probably follow within two years. Especially if the GPU prices go down
How long are those models going to be safe to operate with internet access and tool use without ever being updated? Imagine if "do it for my grandma" is still as effective today because the models haven't changed and that every open-source maintainer upset with AI started polluting their libraries' logs with prompts that have been proven to work with no known reliable fix ever coming in this hypothetical future.
I think the crux of the issue is that frontier labs do not make any financial sense. They are spending investor dollars in order to sell tokens at pennies, and then a Chinese lab can build a competing model on those pennies, and the frontier labs can't then start selling the tokens at a price where they'd actually even break even because most of the volume will then shift to those cheaper models.
Republicans were, mostly, the ones responsible for encouraging this in the 80's and 90's with manufacturing.
Now that the unbelievable damage is done to families, towns, and cultures of entire regions of the US, they can swoop in and fix it for white collar workers.
That’s the thing - the FAFO is significantly dampened by the remaining grey beards nudging things back on track over time.
See this over and over on hn and Reddit - CEOs are stupid and don’t understand how much pain offshore/AI causes. No…they’re taking a calculated risk on the organisations resilience and know the people whining about CEOs being stupid will be the ones patching over the cracks
Although obviously I wouldn't like it happening to me, if the offshoring quality was as good as the original -- and I'm sure sometimes it was -- I can't find any valid objection to it. It is, after all, just a different bunch of people being paid to do some work. Why must it be "us" who get that work? Why can't it be "them"?
Being replaced by a machine is different -- but still, not in a way that I can authoritatively say ought not to happen. If you could get your shirt made by a woman, or the same shirt made by a machine that costs 5% of what it would cost to pay the woman, you will pick the machine 99% of the time. Almost anyone would.
Being replaced by other people, or by a machine, is a very sad thing for the person being replaced. But I can't persuade myself that all by itself it is morally wrong in any sense.
The bottleneck in software engineering has always been communication, whether that is engineers communicating amongst themselves or clients communicating what they want, or client's customers communicating to clients who communicate to management who communicate to middle management who communicate to engineers.
Every step introduces misunderstanding and miscommunication, and minds change, and when you go down the wrong path it is hard to alter course or start over while avoiding blame and ill feeling.
Offshoring quality was not as good as the original, because the original was being produced by developers who were close to the clients and steeped in the same assumptions.
With offshoring, all of the weight was borne by the specification - if something was not captured, it did not exist. It doesn't matter how much it was just a basic assumption that something would work that way - if you didn't say precisely what you wanted, you didn't get anything even close.
So someone close to a product could implement something in a week for x money. Management would see that "week" as a fixed unit of work that could be given to someone who would work for less - but that other person would require an order of magnitude more "prompting" in order to produce what you really wanted, because they didn't have the same knowledge and assumptions.
So their calculus was, well, it might actually take 5 weeks to get the same output, but it was still cheaper. But that completely overlooks the local overhead of 5 failed iterations, and the cost to client relations, or the cost to internal staff burning out in frustration at what was once simple becoming multiple needless failures, or the additional cost of documenting the spec to such a degree nothing gets left out. And the latter, you can't do until you've failed enough to know both sides different assumptions and knowledge gaps.
(Edit: also my point wasn't don't do it or that this can never work - it was that in this case management simply didn't care about the demonstrable quality problem because the cost of reiterating was so low they could afford to write off the first couple of failed attempts, and this is exactly the attitude we'll get with AI. If you can get something that's superficially most of the way there in an hour with a prompt all by yourself, there's no cost to just changing your mind, retrying, trying other things, all with no pesky engineers slowing things down)
AI is now management's favorite excuse. But how many software engineers are actually losing their jobs due to the rise of AI and not due to the end of zero interest rates?
In other words, would those people have kept their jobs at current interest rates if there was no AI?
I think most wouldn't and leadership would be saying something else.
This has happened to many times to other careers over the last decades and there wasn't empathy from SWE's then, in fact all you got was "Software is the future, learn to code etc". Now the shoe is on the other foot people are taking suicide, seems a bit self seeking.
How is that even possible? It's a great movie (ah well watched in the period), done by people who were outcast to work on it - similar to the folks sent to the terrace in "Silicon Valley"
There have been six Shrek movies, and it's been 25 years. Shrek hit theaters before the planes hit the towers - and people can be well into their career having been born after both.
We're gettin' old, xxs. Waves cane and yells "Get off my LLM!"
No, for the world! You're thinking way too small, these guys are on a mission to save the planet and in order to do so they are 'just' going to have to tear up the whole stratum that has been built so far. But they have your best interest at heart so you just have to trust them through the slight bit of turbulence that is coming your way.
Not more than those of farmers who "lost their jobs" to tractors, switchboard operators to switching systems, brokers or postmen (and countless other jobs) to the Internet, etc. And why do we blame the innovators and not the government for not making up useless jobs to keep these people employed? It is just as responsible.
As a product manager my approach and execution of my daily work has massively changed in the last half a year. It never stops making me baffled how little people out of software development understand about what is happening right now.
> It never stops making me baffled how little people out of software development understand about what is happening right now.
What do you expect them to understand?
It amazes me how little people IN software engineering understand what's happening. Go ahead, listen to your bosses and write all your knowledge in AGENTS.md files, you're definitely special and won't be laid off.
Ask anyone in manufacturing how writing things down for your cheaper replacements works out for you.
It's because they've been in FAANG silos with near zero experience in the pre-SDLC (basic research, practical application phases) through to the post-SDLC (tech transfer, commercialization etc) with most having near zero direct experience inside actual QA teams (with QA being beneath a lot of engs). A true AI pipeline hums when this is real end to end cycle. One of the reasons for that is based on how well it manages many-to-many connections inside vector space. One more thing, most soft devs have near zero comprehension of how language models work so leveraging them to manage complexity is very foreign with most not understanding that language models apply far beyond just human language and into any sequence of data e.g. DNA/RNA or just about anything you can chop up. This kind of scale is also a new concept to most.
I would argue that we all know the roles that have been barely justifiable, and now are glaringly obvious came to be as a result of the giddiness of agile and that 'culture'. As I can see now, those who wield a heavy sword are those who are 'allowed' to continue in the LLM-age of product development; I can see managers now developing build projects over teams (which is and can only lead to _less_ productivity and _less_ efficient orgs).
But I think it is important we all emphasise that this LLM-age giddiness sits squarely in those businesses where influence and product have always been their thing. It is unfortunate but, here we are. I don't believe a VP or MD or CXO is ever going to admit they were played by their peers on this one (agentic development).
The positive perhaps is the hardening of personal development, by need or necessity. LLM and agents are terrible at domain expertise still, and only until memory availability becomes abundent again (in end-user equipment), that is not going to change.
My big concern for the industry is when that happens, and anyone can build anything, but with the ability to tune to domain expertise via their own terms; this is the collapse, where no 'moat' exists at all, we are already seeing the collapse of the subscription model, and I believe pay-on-demand / purchase _is_ authentication is a tracking metric to that collapse. I forsee regulation being pushed to ensure the model however and foresee regulation on user equipment.
Larger institutions are perhaps 1-2 years lagging (for regulatory reasons) on uptake and perhaps that is the point in all this giddiness.
Last week "I" have chased down a really sneaky performance bug in a shutdown routine of one of our applications, if it went to prod it would cost us a lot of grief. How did I do it? Well, I didn't, I gave Claude access to kubectl, described the behavior and how to trigger it, and went for a coffee. After 10 minutes of didly-dalling I took a sip of coffe, then copy-pasted 100% spot-on response to a JIRA ticket and sent it to developers for a fix.
Same, I had a very weird issue with tailscale VPN for a very long time, no amount of updates or googling fixed it
I created here something similar to how intranets work, where I can easily define a service as public or internal in k8s through the ingress class
If it's internal, it's accessible only through tailscale, if it's public, it's accessible by anyone, now for some reason internal services have insane delays or very very bad bandwidth, never understood why and couldn't figure out how to even debug this
Ran codex, gave access to the whole node, ah yeah just lower the MTU to this value, lo and behold it fixed everything
The other day I got a cheap thermal printer with the only way to use it is through an app, I thought ok I'll try to reverse engineer it, try to understand the protocol, then write a script to print, few days later and progress was very slow, I thought let's try to let an LLM so it
It reversed the entire protocol, wrote me a web app where I can print whatever I want, I got what I needed, but feel zero accomplishment
The best idea I've come up with is quitting my job, living on my savings and indulging in my hobbies until I run out of money and then hanging myself in the woods.
It's sad to see people lose jobs because executives believe AI could replace them. If it's any consolation, at some point most of those executives could be replaced by AI too.
BMW management believes they can get 30% workforce out and replace the jobs with AI. The executives will not be replaced. Peasants like me will.
ChatGPT can go long way designing printed circuit boards with text based KiCad. Luckily most companies stick to the expensive overengineered (for normal boards) tools like Altium. That part of workforce is safe for a while.
I am moving from electrical engineer to electrician right now. The money is the same (in Germany). I vibecoded my CRM/invoicing/protocol generation/warehouse tool by myself. Over 120 iterations with ChatGPT. And this was scary!!! A decade ago such development would cost 6 figures to develop and 5000-8000€ a year to use. Now it’s on my computer working the way I want. <100€, no need for software team.
My idea is converting to a bicycle mechanic. Very simple and doesn't require a huge amount of very expensive tools as upfront investment, just some experience and practice to do it efficiently.
By the way, my local bike shop has a one month long (!!!) queue for repairs. Just saying.
Where I live many, many older people are now on e-bikes. They are good paying customers, and most don't want to work on their bikes themselves (unlike hobbyists with the pushbikes).
I've worked at a bikeshop when I was a kid. That's long ago so my experience is dated, but here is my takeaway: There are good customers. There are also lots of not so good customers and there are a lot of jerks. People want their bikes ready yesterday, can't imagine why after they neglected their bikes for three years it can't be finished in a few hours.
And today they are taken aback by any reasonable hourly rate because a new bike from China costs less than a few hours of your work. Good luck! Oh, and if you really follow through on this make sure you copy Sheldon's pages just in case they ever go down, you'll have an edge over the rest.
Since long-term foresight beyond the next stock buyback doesn't seem to be materially required anymore for that role anyway, I'd wager they could already be replaced by AI, and with less consequence than any other company roles.
We have a machine that's really good at consuming text and summarizing it for someone else, with the caveat that it never really owns anything or is accountable for it.
Why do we have middle managers still, while we're replacing programmers? Really says a lot about how decisions are made, we're hierarchical monkeys at the core.
> CEO-Bench is an evaluation framework that measures how well AI agents perform executive leadership, long-horizon decision-making, and company management tasks
> If it's any consolation, at some point most of those executives could be replaced by AI too.
I'll believe this the day an AI model can golf with an executive, or wine and dine and call hookers in an overnight business meeting. Sure, that's not every exec, but that's common enough that AI isn't going to work there.
Sadly, it's working so far. Most companies only want a product that works. Best practices, algorithmic efficiency, or taste aren't really on their radar.
> I still do not see how these copy-machines with "super intelligence" inside are going to replace humans
I do.
You don't need any sort of fancy harness as claimed.
All you need to do is type vague ideas into a text box. Then you need some UX taste to pick good concepts.
The building of the software happens trivially now. You do not need a software engineer, sorry guys.
What you need next is someone with time and patience to QA test the software and provide feedback for the AI.
I guess ideally the same person, with the UX taste.
Give it a few weeks and the product is done, without the need for a team of 5-10. All it takes is one competent human.
I know that many here vehemently disagree and raise concerns about the viability and maintainability of the software.
I believe not a single one of them has made that experience, these concerns are based on their imagination of what they think would happen. Usually they tell you that they essentially review every line of code in their projects.
I think this lands in the same area as flying jet liners. A computer can do it. But legally it seems easier if there's a human responsible for the operation of the aircraft. I've conjured my ideas into code via LLM, and it is magical. But the quality is also off, and there are nontrivial issues that I'd like to address. The more a software project is vibed, the less one can manually address things.
Often, the agent will fix a problem without assessing if the fix is worth it, optimal or defective. It's fine to merge fixes like this in codebases occasionally - humans have been managing to do this and to work around this for ages. I'd like to see a software project that is purely LLM driven and doesn't devolve into chaos because of this.
I doubt the need to manually address things. There is little that you cannot discuss and fix with the LLM.
I understand that many doubt this and would like to see evidence. The recent Photoshop clone, PhotoCraft, that gathered 200 contributors in a few days seems like a good example to watch. I understand it's currently around 500k LOC in Rust.
I can, but there often enough changes I need that are easier edited by hand. Not that the skill is in applying a diff from brain to file - the skill I'm paid for is identifying that such a change is required in the first place.
I agree - let's wait and see. I think the outcome of prompting a model will vary wildly depending on who is doing the prompting. Natural language can only get you so far. And if I'm wrong - oh well, I guess I'll need to uproot my life.
What I wrote is based on my personal experience building a commercial app which I am submitting to the AppStore this week.
After >500 hours of work over six months.
I don't think we are screwed. There's risks, but it is also highly likely that living standard will go up for everyone and working time down dramatically.
Yes. If AI doesn't slow down soon, there will no longer be need to build most software. This is not the case yet, and in my opinion we are now at at great time for building new products.
Why would I hire Alice to work for 4 days a week when Bob is willing to work for 5 at the same price (assuming both are equally productive, even if AI has made that productivity extremely high)? So long as there is an oversupply of qualified workers, I doubt working hours will change. To do so would require social contract changes like for the standard 40-hour work week.
>The building of the software happens trivially now. You do not need a software engineer, sorry guys.
This part is quite "load-bearing".
I heard from someone who worked at Lovable that what most customers produce with that magical textbox is software that shouldn't go anywhere near production. Apparently the problem of thinking like an engineer is not solved by these AI tools. Most people who are not software engineers by trade are not capable of verifying what they produced, nor would they even consider needing to.
No offence, I think you might be a classic case of "senior who forgot how it feels to be junior", where after a lot work experience deva start taking their software engineering instincts for granted and forget that they are a skill that takes years to master. You can prompt an AI for a solution, you can even prompt it for ideas, but you cannot prompt it to think for you.
context is important. Depending on what you are doing, you might have to review every line of code two or three or ten times before you sign off on a PR.
> These models lack the taste that makes for effective product decision making. It is cheaper than ever for a product manager to say "Someone complains about this on reddit, let's implement it", while the tasteful leader says 'No, that's not what our product is. Making that means we commit to it for a long time and our customers really want {something else we already do, but not perfectly} instead.'
This is a very good point. It's now easy to just do things, so discernment has become more valuable than ever. If you leave things to Opus 5.5, it'll just say that you need to implement every product-bloating change, and "here let me take care of that for you".
We're heading towards the end of the year and I still do not see how these copy-machines with "super intelligence" inside are going to replace humans. What they have done is convince those with money to spend differently on their fellow humans.
Whether LLMs can successfully replace certain functions is a question of speculation and conjecture. Whether LLMs should replace human roles is both an economic and moral question. When the incentives are what they are, there's little room for objectivity. Both those who stand to—or believe they stand to—benefit most and those who stand to be most harmed by copy-machines replacing humans will be inclined to think of "when it will" happen. Only one of those groups seem to be able to actively participate in asking "whether it should" happen.
> Whether LLMs should replace human roles is both an economic and moral question.
It is also a political question, on which many techies have been maneuvered into abdicating. Stripped of agency, agents they are bamboozled to depend on.
> It is cheaper than ever for a product manager to say "Someone complains about this on reddit, let's implement it", while the tasteful leader says 'No, that's not what our product is. Making that means we commit to it for a long time and our customers really want {something else we already do, but not perfectly} instead.'
This I feel is a strong point. At least in my social-bubble, I notice that many of the vibe-coding CEO's now loose the plot on what their company actually does, and are too focused on solving just any problem any customer presents them. Whether or not that matches their actual company-strength, or their product, is then secondary.
And I miss that. I miss beeing able to say "I work for X, we do Y to solve Z".
However, there's opportunity here. As the cost of building and maintaining software products collapses, taste and carefully choosing what to build becomes a major differentiator.
A lot of us already saw the writing on the wall, how LLMs capabilities could only improve and inevitably lead here. We were called "cultists" and mocked for predicting this exact situation. "It's just parroting the next token." "A SWE will always be needed" "(...)". I always pointed you to the 5 stages of grief and was met with obvious denial and anger.
Seems like we're now crossing into bargaining and depression.
I have empathy for those who were open minded enough not to mock those of us who were warning against this or swiftly dismissed our opinions. I hope you can now be open minded to the fact this will continue, and will extend to ALL jobs, that our society WILL dramatically change to the point of becoming unrecognizable in the not so distant future. We tend to be consistent in making the same errors when predicting the future. You just need to look into past predictions about our present time. Most people still make those errors. But once you see them you can try to avoid them, though unfortunately not making those mistakes usually leads you to seem crazy to others, which is also part of the reason those mistakes are so common.
AI did not take your job; a person decided to use the productivity improvement from a new technology to realize increased short term profitability at the cost of losing you. That is not so much of a value judgement on the choice than a plea to the people making the choices to realize how short term that thinking is and to the author to not let them off the hook for making it.
I just had a realization that the people who still have resources will need to invest their money to have any income. But it seems that Investment funds have been quite good in doing wealth transference, which basically means that overtime, any investment you do might have negative.
Which all sound a bit alarmist, but it needs to break at some point where investing money will not work because there will be no payoffs.
Before I went to college, I worked as an electrician apprentice. One of the guys I worked with was 50+, and had worked many jobs over his life. He told me that from 18 to to his late 20s he worked a factory job. Then, one year, they stared rolling down production. The company had decided to move production to China.
Most were just let go. No severance, or comp packages. Just a good luck with your job search, it was nice while it lasted.
For his next gig, he worked in fisheries. His 10 years of experience at the factory didn't mater, and he started with entry level pay. This lasted some years, until some bigger company bought his employer, and merged. Lots of people had to go, him included.
Then he landed work delivering mail. Worked there for 15 years, until the postal service started downsizing due to ever decreasing volume in mail being delivered.
So his last attempt was to just go to trade school, and try some trade. 50+ years old working on a apprentice salary, once again starting at the bottom rung of the ladder.
My point is, there are many people out there that have lived this life. For every software developer that have made bank for the past 20 years, and are now faced with the reality of being made obsolete, there are probably 10 people who've gone through this multiple times, but gone from one minimum wage job to another. Going back to start every time.
It sucks for everyone. But you just have to adapt. I guess one silver lining is that when the "white collar professional jobs"-apocalypse arrives, that's when we'll finally look at things like UBI.
Time and time again, we've seen that when working class jobs disappear, it is met with some sort of aw-shucks / but-they'll-find-something-else-to-do reaction. Whereas when the middle/professional class is affected, it is seemingly a end of times sign.
Historically I agree with you, but considering how many people reacted negatively to the mathematicians' concerns about development in ai, and people calling them hobbyists, it seems to me someone is trying to create infighting as always. Seems like the reaction may yet again be the mentioned aw-shucks, but they'll find something else to do
I have, too, observed some sense of schadenfreude in the comments on recent headlines about layoffs in the white collar sector.
It is hard to know if it is just bot traffic to sow discord, or if plenty of people are genuinely delighted that office workers are finding themselves unemployed.
Certain sectors and jobs will last much longer. Like healthcare.
But, I think the cold hard truth is that if you're a professional worker, with a six figure salary, the idea of going back to school, and working with something completely unrelated to your interest - all while taking a steep pay cut...it is unthinkable for many.
Yes, but then there is a new industry to being in. This time it is different: there won't be any place where just having a brain that works is going to be a source of income.
>Do you know how hard it is to sleep when all you can hear is your own heartbeat? Do you know how hard it is to function with four effective hours of sleep when you offer your body ten? It is rough.
This hits home. It's both a shame and infuriating what's happening in the market right now. One (unsolicited) piece of advice I would give to anyone impacted is to build their own shit. If these companies stop paying for your lunch, then eat their fucking lunch with the same gimmicks they convinced themselves are adequate to stop paying for your lunch.
I'm not just talking about the enterprise giants - build competing products , lower their margins build similar products and offer them for less. Cut out the middle-men. They might have the contract now but their operating costs are higher. In the long run they are uncompetitive.
For now, those with Capital will probably be ok. Those without Capital? In the EU, they'll probably be ok. In the US? Who knows. In China, India, LATAM, Africa - who knows?
Ultimately in the future (near? middle? far?), Capital will also be useless because the marginal value of everything will be driven towards zero, and who will be the buyers? Certainly not those with Capital, because the marginal cost of anything being close to zero would mean that they would not need to buy from others - they can just make their own. And those without Capital won't have the means to buy anything, even when the cost is close to zero.
Interesting times ahead?
Anyone care to speculate what that future will look like? Two different eras: post-employment, and post-capitalism?
> Capital will also be useless because the marginal value of everything will be driven towards zero
Capital is that fraction of produce set aside for future production. OpenAI is capital but not all wealth is capital. And capital isn't going anywhere - the machines will have gigafactories etc. The value of labor will plumet but present-day capital that is usable by the machines or can be scrapped and repurposed will retain value.
The value of square meter of land + mineral/sunlight rights (not capital) is likely to rise - the amount of wealth that somebody would give you for that asset would probably increase, as more wealth can be generated from it.
Individuals without wealth now will likely end up with a much lower allocation of those things which retain value (land, materials, etc) while having a much higher quantity (relative to now) of those things whose value will drop - AI tokens, electronics, entertainment, etc.
Essentially our current trajectory taken to the limit. Land expensive, food expensive, interest low, smartphone cheap.
> who will be the buyers
Imagine an impoverished and technologically backward country which has been colonized/occupied, and where the most valuable assets are exported well below market price or stolen. The people still trade their own labor for subsistence. They still produce some goods which have value in outside countries, they receive some charity, there is still tourism, their labor has some value (servitude, trafficking). AI would be the same (assuming laws hold up) - e.g. you could rent out your solar panels to OpenAI in exchange for a loaned domestic robot and a token subscription.
Energy, land and row materials will be the main limiting factors.
Mega corp will advance technology exponentially, but for the rest of us we will live in a mix of medieval life style assisted by robotic and low tech. People will exode city and go back to the countryside.
For what concern my self, I'm buying arable land, maybe in couple of years home food production will be fully automated.
Hopping that space colonization will become a thing for the top 0.0001% and that earth will not be their main concern, but yes I’m an optimistic person by nature.
I do wonder what will happen to the outsourcing industry in India. In the UK at least it seems that most corporate development has been sent to Infosys, Wipro, etc.
Not just India. Millions of young, able bodied, angry people, all of a sudden finding themselves not having anything else better to do. Sharing borders with countries who have their own millions of young, able bodied, angry people.
> Those without Capital? In the EU, they'll probably be ok.
They probably won't. EU is not the same economic bloc that it was up to 2004 anymore. I am not sure that there is big country in EU that is having consistent high growth. I expect in the next 15 years the western eu countries to experience the joy of living standards collapse that the eastern bloc enjoyed during the 90s. But without some of the factors that made it bearable for the east.
If the aim of each country is just job creation, leaving aside "comparative advantage", job quality, salary, etc:
1. There are plenty of things being outsourced, They can be done locally.
2. There are plenty of things currently not being done by locals - like elderly care.
3. We can find new things to do.
4. We can work less.
And even if those things raise living standard somewhat, the efficiencies from AI/robots should be more than enough to compensate.
But, in a capitalistic globalized economy this isn't how things work. So we'll see some shift from that. And it will work better in countries with a more democratic socialistic leaning rather than capitalistic.
We're already starting to see that, with the recent EU trade deal with china, limiting the amount of cars china can send to Europe.
Automation kills some jobs and creates more jobs. Nothing about AI (other than the statements of people with lots of shares in AI companies) looks to be any different.
> Nothing about AI (other than the statements of people with lots of shares in AI companies) looks to be any different.
Everything looks to be a lot different. Once AI is better than humans at knowledge work, why would you even keep human in the loop? (That's assuming AI reaches the level of being better than humans - I'm not arguing for that assumption, but about it's consequences and that in this scenario we won't get much new jobs).
You might not always, just as we don't always keep humans in the loop when picking groceries for delivery[0]. Great. But that doesn't mean we can do that en masse.
You can give a workman a power tool and he shouldn't instantly assume that his housebuilding job is under threat, because corporations will build a house-building power tool.
Why won't he think that? Two reasons: 1) any tool that becomes more powerful tends to become more specialised and expensive to operate, so an electric drill doesn't scale to a house-building drill without incredible tradeoffs, and 2) he isn't reading a million breathless class warfare-igniting articles whose writers are paid through attention.
Smaller and/or heavily quantized models have a tendency to do this. The reason is because the auto-regressive nature of the llm dictates that its output should be its own input. You effectively purposefully create a feedback loop. That loop will manifest like this once the possible number of output states starts to converge on a smaller subset of the available tokens. A really interesting side effect of using speculative decoding is that you can detect that the llm is going to loop. You'll see this monotonic increase in the efficiency of the predictor and once that cross a certain threshold the loop is either already underway or it becomes inevitable. I've seen such loops range across 10K token and if not for the speculative decoding signal I would have never realized a loop was underway, but once you know it you can write a simple piece of software to detect the loop. The lower the 'temperature' at which you run the LLM the higher the chance of a loop (because you are constraining the number of branches that lead out from every token). The smaller the model the bigger the chance that it will loop. The more quantized a model is the bigger the chance that it will loop.
There are interesting parallels between LLMs and biological structures as well, for instance spontaneous oscillation in nervous systems lacking inputs but with a feedback loop in the topology.
Any kind of feedback system is riding a fine line between amplification/generation and runaway behavior resulting in a loop. Electronics, mechanical control systems, biological systems and LLMs all share the same property: some fraction of the output is coupled back to the input. With LLMs the fraction is 100%. It's a small miracle they don't loop more than they do!
What makes it so hard to read for you?
The contrast is fine, it even has a narrow column layout to limit the distance your eyes have to travel, the writing style reads quite human and clearly laid out.
Or are the 5 pictures giving you trouble?
Recently, an agency that used to give me work handed me an app they had built using AI.[1]
But even though they had used only the best AI of the time, like Astra and Fable, in reality it had many bugs and many wrong points.
If you're not a programmer, there's a lot of room to make mistakes even if you use AI.
So honestly, I don't believe the claim that Fable can do everything.
It's smart, and it writes code blocks far better than I do, and it writes algorithms far better than I do.
But that doesn't mean it has the global consistency a program should have.
The irony looms large. Most of us have worked on projects that replaced people. I remember one from 2005; automating returns at an online retail company. The returns department was 15 people. We did a project that removed steps and automated others. Ten people were let go.
More so was agile having developers replace the heavy old product and project management organizations a lot of enterprises had. And then came devops were developers replaced sys admins? Remember those folks?
The billionaire class tell us the utopia is coming. Why should we believe them? Given that they’ve failed to share the benefits of the massive productivity gains of the last 50 years (stagnant pay, 30 hour work weeks, health care and social safety net?) why should we trust them to share this? We shouldn’t imo.
The world where programmers and frankly most intellectual workers enjoyed privileged position and high salaries is ending. Some will survive most will have to adjust, start new careers, possibly never come close to income levels they enjoyed.
I was part of it in online poker world. You needed 2 functioning braincells to make 6 figures while playing a few hours a day back in 2005/2006. It got more and more difficult from there. First it got guys "living their life" while hustling on the side who thought it's an easy mode forever and didn't even bother to learn new tools. Then it turned previously good players into marginal grinders and then only a small group of winners remained. Many went to different/easier pastures - private games, heavy game selection at (super) high stakes. Most though lost their income and with it years of life that now seemed wasted in front of the computer.
Among my friends I've seen suicides (two), lost families (more than two), despair and poverty (too many to count). There are also positive stories - people who started doing something else, adjusted their expectation and are now living happy life even though their income will never be close to what their experienced in their early twenties.
In online poker - instead of Fable - we had ever developing tools. Databases, HUDs, HUDs connected to databases, auto-seating scripts (where a script looks for "donks" and auto seats you at the table when they seat) and then bots and solvers. I was on the frontier lab side I guess as the author of the most popular solver when they first came out. Now Fable is coming after my edge.
I am still a tech optimist. I think it will be a better and more abundant world at the end of it. There is still a lot of jobs we are not doing now because we don't have enough people. The transition is going to hurt though, especially those who don't have energy to start from scratch.
Gambling for living taught me no income stream is forever. Especially lucrative ones are going to be competed for and will dry out. Live below your means, try to save money, don't compete for status. It was always a good advice but it never hits until it's you who it applies to.
i sympathise with the individual of course, and i wish they did not have the experience they wrote about. it makes me feel quite bad to read it actually.
but it also seems like tech workers are getting a taste of their own automation medicine they've been happily developing for the last 50 years.
AI is a slopening of everything that makes us humans. If we cant see it that is only because the last 30-40 years have been a constant assault on human intelligence. we became dumb and sloppy as well. peak western society ended back then maybe around the time the first Concorde airplane flew for the first time. Everything went downhill as technology progressed and people thoughts could be controlled and manipulated in ways where intelligence and curiosity were not seen as relevant skills to acquire and nurture.
peak western society ended back then maybe around the time the first Concorde airplane flew for the first time.
This seems a bit arbitrary to me. Why 1969, exactly?
Personally I'd go for the "democratization of Internet" as the start of the slope.
So perhaps 1989 ("1989: Quantum Computer Services and Apple end their partnership. Quantum Computer Services renames itself America Online. Two years later, America Online for DOS is released, and a year after that, America Online for Windows.")?
Or arguably even a bit later...
You and the GP are effectively in agreement. It's not so much about when the exact peak was but about whether there was a peak and whether or not it was somewhere around that time.
Thing is that you cannot do 24/7 oncall with two person and Opus 5.5 . Also bus factor start to be pretty big if there is only two persons etc. There is still reasons why FAANGS are so big.
Sure you can, as long as nobody complains too much about problems during night if those 2 people can work overtime every now and then.
You can probably even promise 99% uptime (sounds good enough to most buyers probably?) even if people sleep 8 hours a day as long as the failures comes no earlier than than 15 minutes into their sleep and only happens at most once a month (1% of 30 days is 7,8 hours).
That's dumb and just an excuse. If it were true, why not keep the whole team and empower them and deliver 10x the existing value?
The dark truth is that tech executives have been extremely pissed off by tech workers in the 2010s leveraging their specialist knowledge into exorbitant salaries, autonomous working conditions and even political demands (remember the Google in China backlash?), and they are now using AI as an excuse to do mass layoffs in order to scare the uppity proles-craftsmen into falling in line. As Kalecki said, they value discipline in the factories more than raw profits
Even if they had, doesn't mean it's a good idea to flood the product with changes, nor it's necessary that the market has the capacity to asborb and expand your sales, or that the existing customer base will appreciate it.
Well I did say "good", which excludes things that don't fit in the context of the wider business.
I think startups have an easier time for 2 reasons:
They are innovating on the product side already, and probably were already capacity constrained for work on the top ideas.
They also have a high proportion of tech debt which never bubbled to the top of effort vs reward (competing against product work) before but would still be valuable to fix. E.g. https://posthog.com/blog/sql-parser was on the HN frontpage a few months ago, and this work would never have happened in the before times
Even if the idea fits the context of the wider business, there's such a thing as too much change "just because you can", which alienates users and can make the end product an unfocused mess.
If it's easier for startups, it's for the same reason doing a "pivot" is easier: they usually don't have much of a product or revenue or customer base to begin with, they throw things out there and see what sticks.
>That's dumb and just an excuse. If it were true, why not keep the whole team and empower them and deliver 10x the existing value?
Because they don't need 10x the existing value, nor is the extra value they can actually use and benefit from that expandable.
And despite being quick to call this "dumb", you missed the most obvious reason: by not keeping the whole team they save money. Which makes their expenses lower and their VCs/shareholders happy. Keeping the team and "delivering 10x the existing value" doesn't leverage AI for reducing costs, and is hardly guaranteed to bring in much in the way of returns either.
Usable new value goes through a "pipe" so to speak, which constrains how much you can use or need. And that's the business and market and regulation realities.
If you're making Instagram for example you aren't going to change everything "just because you can", nor is there any reason to make 10x intrusive changes to the backend "to make it faster" or something with the danger of disrupting something, if it goes predictably well enough already.
Your marketing team, your engineering team (what's left of it), your business team, also need to understand and pace any changes, and have a specific vision, not just "anything possible, even 10x the things".
If you think AI can replace experienced engineers you're not working on anything complicated. If your product is a web app, sure, maybe. If you're working on actually difficult problems, then people who know what they're doing are irreplaceable. AI can supplement them and make them more productive, sure. AI does not know what NOT to build.
i'm tired man. AI this, AI that, here's how we productmaxxed our deliveries, here's how many people we could remove, etc etc. Everyone is going crazy about ai, but no one is remembering the people affected.
Corpo culture is disgusting around layoffs. Zero notice terminations should not be legal, anywhere. Have some fucking heart guys.
To be fair Zero notice terminations are illegal in many Countries. Furthermore I am yet to see solid evidence that replacing humans with AI is beneficial. Actual crazy times.
I don’t think the blame belongs with corpo culture. It belongs with engineering culture. Engineers have always been at the forefront of technological disruption, and for the most part they’ve been spared its consequences. The people displaced by those efficiency and productivity gains stayed invisible to them, or were written off as a necessary sacrifice for progress. Those people lost the old way of doing things and were told to keep up or get left behind. The irony is that now it’s engineers getting hit hard, facing the threat of being made obsolete in the name of efficiency and productivity. They’re finally going through what so many others went through before them. And they can’t expect any compassion or help from their own kind.
We will lay you off in 3 months - and having for a 3 months dead man walking in the company is probably worse for everyone. Since said employee probably won't break records for productivity during this period - it is easier for everyone just to give the one let go some lump sum.
Anyway I think that in this post the context is that they will be let go during good times - without the informal - we know the company is in trouble and layoffs as coming common knowledge that big companies have.
"both of them would lose their job next week with zero notice."
What kind of 3rd world country do you live in where this is legal? I just want to point out that businesses needing to balance their workforce is normal. Living in a society where letting go of employees without notice is not. Maybe that's the part that needs to be addressed as a matter of urgency, not the management's delusions of AI capabilities - those will verify themselves very quickly in practice - employment law takes time and effort to change
My last job I was notified via text message on a Saturday that I didn't need to come back into the office Monday. The USA has basically zero protections for employees.
Yep. My company, based in the US, did that to most of our US team. Being somewhere with decent employment regulation meant that they kept us and not the US guys.
The good US workplaces will give you a few months of severance, but you can still expect to receive an email late Friday night that says "don't come in on Monday". They typically don't want fired/laid-off employees in the office at all (in theory for "security", but mostly to prevent labour organisation)
That's fine - being paid for your time in lieu of notice is practiced even in countries with very strong employee protections. That's still "notice" as far as I'm concerned.
Yeah well in the US employees can be fired at will. However in the post it seems the author didn't tell his reports they risked/were about to getting fired. I don't understand that kind of loyalty to your employer, esp. when, as is the case here, the employer has zero obligations towards their staff.
I think it's pretty normal to have like one/two week notice period in most of the work except for maybe in Europe where the notice period can be 3 months
To be fair, what is often called "zero notice" actually means "you are fired, this is your notice, and for the next N weeks until your employment ends, you don't need to come in anymore as we have already revoked your access, but of course we pay you the salary that you are entitled to." That us just much longer to say and less catchy too.
"we can't ship your stuff because of some material restriction or let you pick up a box with some dignity, so we destroyed your stuff"
Really? never heard of anything like this even when I was in the states. If anyone tried this anywhere in Europe the case would almost certainly end up in a parliamentary questions session.
If you want something outrageous, I was fired as my dad was on his deathbed, he died before my 2 weeks notice finished and that created a lot of heat since I did not send them their laptop in my last day as I was busy preparing his funeral instead. They did not care, it was surreal, and I completly lost faith in humanity as a result
SWE as a job is over, this should have been obvious at start of the year, its is simple as knowing the capability and watching the progress of the models. Sitting waiting for it to happen and crying about it is a bad coping mechanism, accept it and have a plan B/C/D in the works while you are one of the lucky ones where it company has not made the move yet.
"These models lack the taste that makes for effective product decision making."
This sentence, whenever I see it, makes me chuckle now.
Surely the leopards won't eat my face. I have taste and you can't get that from a machine.
If the higher-ups don't care about the quality of your code when they compel you to tokenmaxx, what on earth gives you the impression they have a clue about your taste?
If you can direct AI and produce more value per $/time spent then a non software engineer directing AI to develop software then you still can produce a lot of value.
This should be the case now for anything that is not a prototype. The job has changed from "writing code" to "shipping software". A technical lead.
As AI gets more intelligent and the tooling around it develops and organizations figure out how to get it access to the data, the level of involvement needed for software engineers will decrease. Since the cost of developing software is plummeting, the amount of software being used is going to expand greatly- that will keep the technical lead positions available longer and also create more opportunities for consultant engagements.
Although someone can use AI to rewrite every software dependency and every paid piece of software, its still not worthwhile for them for most software, especially if the price of the software is low compared to the complexity it hides.
Users of a paid software product still need a human to be responsible for it.
As you develop your skills with AI, you now have the ability to develop your own software products. So there's an opportunity to become a small business owner. You didn't want to do these non-technical roles, but now you can continue to direct AI to automate all the aspects of the business.
> That first meeting crushed my soul. I heard my ears ring more violently than ever. The color drained from the world. I could barely walk straight.
There was a time that I really cared about my work and about my colleagues livelihoods.
Then I realised, it’s just a job. I’ve lost jobs before. I’ve taken shitty, low paid jobs before. It’s just a job. You turn up, they pay you, then you go home to your actual life, the one that matters.
People take their jobs too seriously.
That's a nice attitude to have when you know you can get another job. That allows you to support your family. Without moving to another city or drastically altering your lifestyle.
Unfortunately not everyone has that luxury, and as AI continues to devalue intellectual labor, fewer will.
Well, I don’t have that luxury ! I’m old, I do have a family to support, and I’ve been out of work for 6 months at a time recently, under financial pressure.
I still feel the same way.
> That's a nice attitude to have when you know you can get another job.
Perspective... there are people crossing the entire world to come and be trashmen or food delivery slaves in our western countries, there are other jobs, you just don't want to do most of them
Don't live like your cushy $$$ desk job is a god given right and you'll be fine
> AI continues to devalue intellectual labor
Hot take, but I don't think it actually does, tbh.
Setting aside the temporary delusions of management, the tech itself doesn't devalue anything. It just optimizes away redundant stuff. That was always kinda redundant, but it wasn't "intellectual labor" just because it was done by white-collar workers.
We only have to go back to pre-AI days and look at what kind of mindless nonsense tasks said workers back then complained to see what I mean.
No people ever wanted to do this kind of work anyway, so it is fantastic that they in the future will not have to. The only thing we need to ensure is that the benefits of that will not be harvested by the same rich few. But that was also the mantra before AI, so nothing changed there either.
We can do so, because anything made out of carbon can also be reduced to carbon. Nothing and no one is untouchable, but, ideally, words are the tool. Anything else has unforeseen side-effects.
> No people ever wanted to do this kind of work anyway
No one ever wants to do most of the work they do. That is the literal definition of work, as opposed to a hobby. You do it because otherwise you and your children starve. Job satisfaction is optional (and may cut into your salary).
The author has mix of empathy and fear. They do not necessarily care about the job as a much as much they are probably are not as much stress tolerant and immune to things around them.
I agree that we take our job/careers too seriously. Especially when it can lead to serious consequences such as https://en.wikipedia.org/wiki/Takotsubo_cardiomyopathy Broken heart syndrome.
Loss of a job/career shouldn’t define a person. Loss of a human being does. For too long we ask people we meet “what do you do for work?” Perhaps the question that lets us better know a person is “who is your family?”
I couldn’t agree more !
I feel the same as you - but I'm fortunate enough to be young, have a cushion of cash, no mortgage, no mouths to feed. I'm guessing you're similar. We might feel differently without these.
Nope, I’m old, with family, been jobless recently.
Experience has taught me that investing the OP’s amount of emotional capital into your job leads to unhappiness.
That's really the key - the situation doesn't change, but you can - and how you look at it.
Either you're a helpless waif buffered by the storms of this world, or you're in charge of at least something - picking the first results in reddit-tier apathy and despair.
The second may not cause much immediate change, but at least you'll feel better about it.
Capital and Labor have always had a tense relationship, each one uneasy about the necessity of depending on the other. Unfortunately, Capital’s acquisition of power in the modern Western world has not been accompanied by a commensurate acquisition of confidence. In fact, as Capital’s power has increased, so has its sense of its precarity, becoming so acute as to tip over into paranoia. The moment AI achieved the barest shred of plausibility in its promise to replace Labor, it became an article of faith that Capital feverishly believes in and aggressively acts on.
All of this is to say that we have a case of “the market can stay irrational longer than you can stay solvent” on steroids; Capital is going to fight like its life depends on it to stay irrationally convinced in their belief that AI can replace Labor.
Always interesting when words change their meaning; because in this case the part of Labour that does the actual labour is going to be fine. It'll be the layers of people who do nothing but sit at a desk and press buttons that really feel it.
AI might be the best thing to happen for labour n fact; because they're going to gain a lot more intelligence for the same amount of work.
This is a lesson I learned a long time ago, no matter how good you are, you could be obsolete and replaceable any day and the companies/corporations don't care. Treat your work as a job and don't make it the center of your life.
but my job is also my hobby, it is the center of my life i cant pretend i dont care
I sincerely hope you are a founder of your company.
You can continue to be a hobbyist mountain climber even when the job market stop asking for any.
Hobby woodworkers still build tables themselves, despite the existence of factories, assembly lines, IKEA.
Until the government forbids writing programs by hand, writing programs by hand will remain a satisfying hobby. The only thing that's changing is how much other people appreciate the skill and effort you put into a given program.
My flight instructor said it succinctly - "The graveyard is full of people nobody could do without. Don't press the weather!"
That was in regards of not trying to fly when conditions were marginal, but it applies to much more.
Though, originally that quote is from Ronald McDonald.
>Treat your work as a job and don't make it the center of your life.
Good advice regarding psychological investment.
Not so useful advice when the job provides your actual livelihood, and the changes (e.g. outsourcing or AI or some industry change) mean you might soon hold a "will work for food" sign in some street corner.
I mean these days, it's not usually that people made some particular job their whole personality and "center of their life", the reason that they worry about losing it.
It's that their job is what puts food on the table, and the job pools are drying up.
I understand that aspect but when you treat your work as just a job, you actually live with the mentality that you could get fired tomorrow and that's okay and you're mentally and financially prepared for that.
Also a lot of tech workers earn a higher salary compared to most of other professions, so I'd expect people to put away some buffer for a rainy day.
Being mentally prepared, possibly, and even financially as far as being mentally prepared can go, but then not everyone have be fuelled by FANG perks&benefits.
Remembering when I lost people to offshoring in the mid 00s. I made the case, with metrics, that the quality was nowhere near as good and the spec overhead was non-trivial.
Management's response? It doesn't matter - for the price they could get cheap offshore workers to do it over again 5 times and still save money.
It's better to work for companies that understand time/opportunity cost, and so won't penny pinch on something which will make them lots of money.
The "it saves money" mindset is bad for everyone, distorts everything away from valuable work and investments that generate the most revenue.
At least offshoring was seemingly stable. Here we're all waiting for the rug pull.
Why would there be a rug pull? Open models do seem to be giving companies like OpenAI and Anthropic some decent-ish competition, so if they try to increase prices too much then people can just leave them.
For the same reason that most businesses are on AWS/GCloud rather than having a server room in the office.
Self hosting is a complex investment. Competitors need an stupidly expensive amount of hardware to enter the scene if they're going to serve SaaS. Chinese backed models can be regulated out of competition. Openai and Anthropic will soon be semi exclusive sources of training data as development happens on private conversations.
Success is not guaranteed, but it is crystal clear that the goal is replacing us and then pricing the replacement as the largest possible percentage of our salaries possible.
It's the oldest trick in the book. Place a heavily discounted supermarket in town, wait for all local businesses to close, raise prices.
All LLM are heavily subsidized by investors. No one is paying the true cost of tokens at the moment. The rug pull is to get as many people defendant on the tool before they have to massively increase prices.
What about people who are running open models on their own GPUs? They are being subsidised too?
That's the point gp was making. If the big labs all increase prices 10x, the cost of running your own open models stays the same
Sure, open models will definitely be an alternative once the frontier models hike prices to match costs. But that will be a minority, and people will start to switch, but people use the frontier models for a reason and these AI labs will extort that, once the easy investor money dries up.
My company pay a copilot subscription to everyone, plus my team have a bedrock subscription that cost us 7k last month. In the meantime, some guys created a new internal team and are testing open source models (for our red team) on our own hardware (we didn't sold our 'old', 3 year old servers RAM/CPU, so the only investment was GPUs). My guess is that within the next 6 months our red team will use open source models exclusively, and my team will probably follow within two years. Especially if the GPU prices go down
> Especially if the GPU prices go down
I rather expect them to go up. I bought a bunch three years ago and a lot more recently. All of them I could sell for 2x what I paid.
How long are those models going to be safe to operate with internet access and tool use without ever being updated? Imagine if "do it for my grandma" is still as effective today because the models haven't changed and that every open-source maintainer upset with AI started polluting their libraries' logs with prompts that have been proven to work with no known reliable fix ever coming in this hypothetical future.
Evidence ?
Who are these investors subsidizing random inference providers on https://openrouter.ai/ ?
I think the crux of the issue is that frontier labs do not make any financial sense. They are spending investor dollars in order to sell tokens at pennies, and then a Chinese lab can build a competing model on those pennies, and the frontier labs can't then start selling the tokens at a price where they'd actually even break even because most of the volume will then shift to those cheaper models.
That's not the rug pull.
The rug pull is every software engineer losing their job.
I could see a republican administration banning this entirely, though they would just find cheaper workers in the U.S.
I wonder when this comes from other white class jobs— process engineering, autocad drafter, electronics design.
Republicans were, mostly, the ones responsible for encouraging this in the 80's and 90's with manufacturing.
Now that the unbelievable damage is done to families, towns, and cultures of entire regions of the US, they can swoop in and fix it for white collar workers.
Amazing.
Serious question, what sort of FAFO timelines did you see?
(And I assume that there's a sliver of people who never went this route in the first place, and another sliver who never stopped).
That’s the thing - the FAFO is significantly dampened by the remaining grey beards nudging things back on track over time.
See this over and over on hn and Reddit - CEOs are stupid and don’t understand how much pain offshore/AI causes. No…they’re taking a calculated risk on the organisations resilience and know the people whining about CEOs being stupid will be the ones patching over the cracks
Although obviously I wouldn't like it happening to me, if the offshoring quality was as good as the original -- and I'm sure sometimes it was -- I can't find any valid objection to it. It is, after all, just a different bunch of people being paid to do some work. Why must it be "us" who get that work? Why can't it be "them"?
Being replaced by a machine is different -- but still, not in a way that I can authoritatively say ought not to happen. If you could get your shirt made by a woman, or the same shirt made by a machine that costs 5% of what it would cost to pay the woman, you will pick the machine 99% of the time. Almost anyone would.
Being replaced by other people, or by a machine, is a very sad thing for the person being replaced. But I can't persuade myself that all by itself it is morally wrong in any sense.
The bottleneck in software engineering has always been communication, whether that is engineers communicating amongst themselves or clients communicating what they want, or client's customers communicating to clients who communicate to management who communicate to middle management who communicate to engineers.
Every step introduces misunderstanding and miscommunication, and minds change, and when you go down the wrong path it is hard to alter course or start over while avoiding blame and ill feeling.
Offshoring quality was not as good as the original, because the original was being produced by developers who were close to the clients and steeped in the same assumptions.
With offshoring, all of the weight was borne by the specification - if something was not captured, it did not exist. It doesn't matter how much it was just a basic assumption that something would work that way - if you didn't say precisely what you wanted, you didn't get anything even close.
So someone close to a product could implement something in a week for x money. Management would see that "week" as a fixed unit of work that could be given to someone who would work for less - but that other person would require an order of magnitude more "prompting" in order to produce what you really wanted, because they didn't have the same knowledge and assumptions.
So their calculus was, well, it might actually take 5 weeks to get the same output, but it was still cheaper. But that completely overlooks the local overhead of 5 failed iterations, and the cost to client relations, or the cost to internal staff burning out in frustration at what was once simple becoming multiple needless failures, or the additional cost of documenting the spec to such a degree nothing gets left out. And the latter, you can't do until you've failed enough to know both sides different assumptions and knowledge gaps.
(Edit: also my point wasn't don't do it or that this can never work - it was that in this case management simply didn't care about the demonstrable quality problem because the cost of reiterating was so low they could afford to write off the first couple of failed attempts, and this is exactly the attitude we'll get with AI. If you can get something that's superficially most of the way there in an hour with a prompt all by yourself, there's no cost to just changing your mind, retrying, trying other things, all with no pesky engineers slowing things down)
AI is now management's favorite excuse. But how many software engineers are actually losing their jobs due to the rise of AI and not due to the end of zero interest rates?
In other words, would those people have kept their jobs at current interest rates if there was no AI?
I think most wouldn't and leadership would be saying something else.
This post is so sad it makes me wonder how many suicide notes Amodei and Altman had on their doormats, or will in the future.
This has happened to many times to other careers over the last decades and there wasn't empathy from SWE's then, in fact all you got was "Software is the future, learn to code etc". Now the shoe is on the other foot people are taking suicide, seems a bit self seeking.
It's a price they're willing to pay.
Or they tell themselves the usual stuff - if they didn't do it then someone else would, technological progress is inevitable, etc.
>> "It is difficult to get a man to understand something, when his salary depends upon his not understanding it."
Well, it's the rest of us paying the price, they're reaping the rewards.
To quote the wise Lord Farquaad: "some of you may die, but that's a sacrifice I'm willing to make."
I think it was the brave captain Zapp Brannigan who said it first, but might be mistaken here.
Pretty sure they're not the ones paying the price.
It is a Shrek meme: https://www.youtube.com/watch?v=roNlgIY9QKw
Thanks. I didn't know this one.
How is that even possible? It's a great movie (ah well watched in the period), done by people who were outcast to work on it - similar to the folks sent to the terrace in "Silicon Valley"
There have been six Shrek movies, and it's been 25 years. Shrek hit theaters before the planes hit the towers - and people can be well into their career having been born after both.
We're gettin' old, xxs. Waves cane and yells "Get off my LLM!"
It's à joke. Shrek reference: https://youtu.be/hiKuxfcSrEU?si=anKoBVyhRIegW_X4
That's the joke.
And they're doing it for your own good!
No, for the world! You're thinking way too small, these guys are on a mission to save the planet and in order to do so they are 'just' going to have to tear up the whole stratum that has been built so far. But they have your best interest at heart so you just have to trust them through the slight bit of turbulence that is coming your way.
"just accept that bad things will happen to you peasant"
Not more than those of farmers who "lost their jobs" to tractors, switchboard operators to switching systems, brokers or postmen (and countless other jobs) to the Internet, etc. And why do we blame the innovators and not the government for not making up useless jobs to keep these people employed? It is just as responsible.
It would have happened even without them specifically. There's no shortage of greedy bastards.
All the while Karpathy talking about how only a fraction of people understand the capability of (frontier) AI models: https://news.ycombinator.com/item?id=50047422.
As a product manager my approach and execution of my daily work has massively changed in the last half a year. It never stops making me baffled how little people out of software development understand about what is happening right now.
> It never stops making me baffled how little people out of software development understand about what is happening right now.
What do you expect them to understand?
It amazes me how little people IN software engineering understand what's happening. Go ahead, listen to your bosses and write all your knowledge in AGENTS.md files, you're definitely special and won't be laid off.
Ask anyone in manufacturing how writing things down for your cheaper replacements works out for you.
It's because they've been in FAANG silos with near zero experience in the pre-SDLC (basic research, practical application phases) through to the post-SDLC (tech transfer, commercialization etc) with most having near zero direct experience inside actual QA teams (with QA being beneath a lot of engs). A true AI pipeline hums when this is real end to end cycle. One of the reasons for that is based on how well it manages many-to-many connections inside vector space. One more thing, most soft devs have near zero comprehension of how language models work so leveraging them to manage complexity is very foreign with most not understanding that language models apply far beyond just human language and into any sequence of data e.g. DNA/RNA or just about anything you can chop up. This kind of scale is also a new concept to most.
They are only capable of Technical reasoning, other perspectives are often beyond them.
I would argue that we all know the roles that have been barely justifiable, and now are glaringly obvious came to be as a result of the giddiness of agile and that 'culture'. As I can see now, those who wield a heavy sword are those who are 'allowed' to continue in the LLM-age of product development; I can see managers now developing build projects over teams (which is and can only lead to _less_ productivity and _less_ efficient orgs).
But I think it is important we all emphasise that this LLM-age giddiness sits squarely in those businesses where influence and product have always been their thing. It is unfortunate but, here we are. I don't believe a VP or MD or CXO is ever going to admit they were played by their peers on this one (agentic development).
The positive perhaps is the hardening of personal development, by need or necessity. LLM and agents are terrible at domain expertise still, and only until memory availability becomes abundent again (in end-user equipment), that is not going to change.
My big concern for the industry is when that happens, and anyone can build anything, but with the ability to tune to domain expertise via their own terms; this is the collapse, where no 'moat' exists at all, we are already seeing the collapse of the subscription model, and I believe pay-on-demand / purchase _is_ authentication is a tracking metric to that collapse. I forsee regulation being pushed to ensure the model however and foresee regulation on user equipment.
Larger institutions are perhaps 1-2 years lagging (for regulatory reasons) on uptake and perhaps that is the point in all this giddiness.
> The joy of programming is dying.
Ever since LLMs took over everything I'm working or have worked on, I no longer have fun, no longer feel joy, or challenged to solve issues.
My question is, what now? What's the backup skill or job to develop next and still feel joy with it?
It might be shocking but there is like a whole world outside of computers and even outside of your job.
As a counterpoint, I have enjoyed that last several months greatly. I have been able to solve real, long standing, issues quickly and efficiently.
Not sure what to recommend for you though.
> I have been able to solve
Is it you though?
Last week "I" have chased down a really sneaky performance bug in a shutdown routine of one of our applications, if it went to prod it would cost us a lot of grief. How did I do it? Well, I didn't, I gave Claude access to kubectl, described the behavior and how to trigger it, and went for a coffee. After 10 minutes of didly-dalling I took a sip of coffe, then copy-pasted 100% spot-on response to a JIRA ticket and sent it to developers for a fix.
I did maybe 2% of work there.
Same, I had a very weird issue with tailscale VPN for a very long time, no amount of updates or googling fixed it
I created here something similar to how intranets work, where I can easily define a service as public or internal in k8s through the ingress class
If it's internal, it's accessible only through tailscale, if it's public, it's accessible by anyone, now for some reason internal services have insane delays or very very bad bandwidth, never understood why and couldn't figure out how to even debug this
Ran codex, gave access to the whole node, ah yeah just lower the MTU to this value, lo and behold it fixed everything
The other day I got a cheap thermal printer with the only way to use it is through an app, I thought ok I'll try to reverse engineer it, try to understand the protocol, then write a script to print, few days later and progress was very slow, I thought let's try to let an LLM so it
It reversed the entire protocol, wrote me a web app where I can print whatever I want, I got what I needed, but feel zero accomplishment
> then copy-pasted 100% spot-on response to a JIRA ticket and sent it to developers for a fix.
I get these and the fix ai suggests is often wrong or hallucinated!
The best idea I've come up with is quitting my job, living on my savings and indulging in my hobbies until I run out of money and then hanging myself in the woods.
It's sad to see people lose jobs because executives believe AI could replace them. If it's any consolation, at some point most of those executives could be replaced by AI too.
BMW management believes they can get 30% workforce out and replace the jobs with AI. The executives will not be replaced. Peasants like me will.
ChatGPT can go long way designing printed circuit boards with text based KiCad. Luckily most companies stick to the expensive overengineered (for normal boards) tools like Altium. That part of workforce is safe for a while.
I am moving from electrical engineer to electrician right now. The money is the same (in Germany). I vibecoded my CRM/invoicing/protocol generation/warehouse tool by myself. Over 120 iterations with ChatGPT. And this was scary!!! A decade ago such development would cost 6 figures to develop and 5000-8000€ a year to use. Now it’s on my computer working the way I want. <100€, no need for software team.
Good job adjusting quickly.
I used to repair cars. I guess if nothing else works, I can do that until EVs take over that.
My idea is converting to a bicycle mechanic. Very simple and doesn't require a huge amount of very expensive tools as upfront investment, just some experience and practice to do it efficiently.
By the way, my local bike shop has a one month long (!!!) queue for repairs. Just saying.
Where I live many, many older people are now on e-bikes. They are good paying customers, and most don't want to work on their bikes themselves (unlike hobbyists with the pushbikes).
I've worked at a bikeshop when I was a kid. That's long ago so my experience is dated, but here is my takeaway: There are good customers. There are also lots of not so good customers and there are a lot of jerks. People want their bikes ready yesterday, can't imagine why after they neglected their bikes for three years it can't be finished in a few hours.
And today they are taken aback by any reasonable hourly rate because a new bike from China costs less than a few hours of your work. Good luck! Oh, and if you really follow through on this make sure you copy Sheldon's pages just in case they ever go down, you'll have an edge over the rest.
That's quite ingenious I would say, when life gave you lemons you actually made lemonade.
Since long-term foresight beyond the next stock buyback doesn't seem to be materially required anymore for that role anyway, I'd wager they could already be replaced by AI, and with less consequence than any other company roles.
We have a machine that's really good at consuming text and summarizing it for someone else, with the caveat that it never really owns anything or is accountable for it.
Why do we have middle managers still, while we're replacing programmers? Really says a lot about how decisions are made, we're hierarchical monkeys at the core.
> we're hierarchical monkeys at the core
How could you ever think otherwise?
Given the level of compensation of US CEOs, it's obvious they should be first in line to be replaced by machines.
> CEO-Bench is an evaluation framework that measures how well AI agents perform executive leadership, long-horizon decision-making, and company management tasks
https://ceobench.com/
> If it's any consolation, at some point most of those executives could be replaced by AI too.
I'll believe this the day an AI model can golf with an executive, or wine and dine and call hookers in an overnight business meeting. Sure, that's not every exec, but that's common enough that AI isn't going to work there.
Sadly, it's working so far. Most companies only want a product that works. Best practices, algorithmic efficiency, or taste aren't really on their radar.
Not only CEOs. Companies themselves can be replaced by AI. What do I need Adobe for, if I can build me a Photoshop application using AI?
I don't want anybody to be replaced by AI :/
I want everyone replaced with AI and the end of human labor
> I still do not see how these copy-machines with "super intelligence" inside are going to replace humans
I do.
You don't need any sort of fancy harness as claimed.
All you need to do is type vague ideas into a text box. Then you need some UX taste to pick good concepts.
The building of the software happens trivially now. You do not need a software engineer, sorry guys.
What you need next is someone with time and patience to QA test the software and provide feedback for the AI.
I guess ideally the same person, with the UX taste.
Give it a few weeks and the product is done, without the need for a team of 5-10. All it takes is one competent human.
I know that many here vehemently disagree and raise concerns about the viability and maintainability of the software.
I believe not a single one of them has made that experience, these concerns are based on their imagination of what they think would happen. Usually they tell you that they essentially review every line of code in their projects.
I think this lands in the same area as flying jet liners. A computer can do it. But legally it seems easier if there's a human responsible for the operation of the aircraft. I've conjured my ideas into code via LLM, and it is magical. But the quality is also off, and there are nontrivial issues that I'd like to address. The more a software project is vibed, the less one can manually address things.
Often, the agent will fix a problem without assessing if the fix is worth it, optimal or defective. It's fine to merge fixes like this in codebases occasionally - humans have been managing to do this and to work around this for ages. I'd like to see a software project that is purely LLM driven and doesn't devolve into chaos because of this.
I doubt the need to manually address things. There is little that you cannot discuss and fix with the LLM.
I understand that many doubt this and would like to see evidence. The recent Photoshop clone, PhotoCraft, that gathered 200 contributors in a few days seems like a good example to watch. I understand it's currently around 500k LOC in Rust.
Let's see if their app is going to beat eg. GIMP.
I can, but there often enough changes I need that are easier edited by hand. Not that the skill is in applying a diff from brain to file - the skill I'm paid for is identifying that such a change is required in the first place.
I agree - let's wait and see. I think the outcome of prompting a model will vary wildly depending on who is doing the prompting. Natural language can only get you so far. And if I'm wrong - oh well, I guess I'll need to uproot my life.
The problem with this is that it is not written by a human. It’s one of the versions of: „Since AI I finished all my spare time projects.“ Etc.
The actual reality is, that it is too early to actually say where this is going to end. Either AI is going to fix what it is causing right now or not.
We are screwed either way. So is the person or prompter behind the message above.
If Ai is universal there is no need for UX, because a real personal assistant, that takes personal care of you is the only UX necessary in the end.
What I wrote is based on my personal experience building a commercial app which I am submitting to the AppStore this week.
After >500 hours of work over six months.
I don't think we are screwed. There's risks, but it is also highly likely that living standard will go up for everyone and working time down dramatically.
Yes. If AI doesn't slow down soon, there will no longer be need to build most software. This is not the case yet, and in my opinion we are now at at great time for building new products.
Why would I hire Alice to work for 4 days a week when Bob is willing to work for 5 at the same price (assuming both are equally productive, even if AI has made that productivity extremely high)? So long as there is an oversupply of qualified workers, I doubt working hours will change. To do so would require social contract changes like for the standard 40-hour work week.
>The building of the software happens trivially now. You do not need a software engineer, sorry guys.
This part is quite "load-bearing".
I heard from someone who worked at Lovable that what most customers produce with that magical textbox is software that shouldn't go anywhere near production. Apparently the problem of thinking like an engineer is not solved by these AI tools. Most people who are not software engineers by trade are not capable of verifying what they produced, nor would they even consider needing to.
No offence, I think you might be a classic case of "senior who forgot how it feels to be junior", where after a lot work experience deva start taking their software engineering instincts for granted and forget that they are a skill that takes years to master. You can prompt an AI for a solution, you can even prompt it for ideas, but you cannot prompt it to think for you.
I believe you kind of can prompt it to think for you.
Is it really so hard to select Opus 5.5 in Claude Code and type into the box that you'd like an audit for production readiness?
I think you need a competent person, perhaps a product owner, in front of it to make sense of the output and direct it.
Maybe you are right and for a non-technical person this is harder than I imagine. But it is getting easier every month now, isn't it?
context is important. Depending on what you are doing, you might have to review every line of code two or three or ten times before you sign off on a PR.
> These models lack the taste that makes for effective product decision making. It is cheaper than ever for a product manager to say "Someone complains about this on reddit, let's implement it", while the tasteful leader says 'No, that's not what our product is. Making that means we commit to it for a long time and our customers really want {something else we already do, but not perfectly} instead.'
This is a very good point. It's now easy to just do things, so discernment has become more valuable than ever. If you leave things to Opus 5.5, it'll just say that you need to implement every product-bloating change, and "here let me take care of that for you".
> The joy of programming is dying. The joy of delivering results lives on.
This sums it up well and will ring true no matter which side of the argument you're on.
> Whether LLMs should replace human roles is both an economic and moral question.
It is also a political question, on which many techies have been maneuvered into abdicating. Stripped of agency, agents they are bamboozled to depend on.
> It is cheaper than ever for a product manager to say "Someone complains about this on reddit, let's implement it", while the tasteful leader says 'No, that's not what our product is. Making that means we commit to it for a long time and our customers really want {something else we already do, but not perfectly} instead.'
This I feel is a strong point. At least in my social-bubble, I notice that many of the vibe-coding CEO's now loose the plot on what their company actually does, and are too focused on solving just any problem any customer presents them. Whether or not that matches their actual company-strength, or their product, is then secondary.
And I miss that. I miss beeing able to say "I work for X, we do Y to solve Z".
However, there's opportunity here. As the cost of building and maintaining software products collapses, taste and carefully choosing what to build becomes a major differentiator.
A lot of us already saw the writing on the wall, how LLMs capabilities could only improve and inevitably lead here. We were called "cultists" and mocked for predicting this exact situation. "It's just parroting the next token." "A SWE will always be needed" "(...)". I always pointed you to the 5 stages of grief and was met with obvious denial and anger.
Seems like we're now crossing into bargaining and depression.
I have empathy for those who were open minded enough not to mock those of us who were warning against this or swiftly dismissed our opinions. I hope you can now be open minded to the fact this will continue, and will extend to ALL jobs, that our society WILL dramatically change to the point of becoming unrecognizable in the not so distant future. We tend to be consistent in making the same errors when predicting the future. You just need to look into past predictions about our present time. Most people still make those errors. But once you see them you can try to avoid them, though unfortunately not making those mistakes usually leads you to seem crazy to others, which is also part of the reason those mistakes are so common.
AI did not take your job; a person decided to use the productivity improvement from a new technology to realize increased short term profitability at the cost of losing you. That is not so much of a value judgement on the choice than a plea to the people making the choices to realize how short term that thinking is and to the author to not let them off the hook for making it.
This is covered in the post.
I just had a realization that the people who still have resources will need to invest their money to have any income. But it seems that Investment funds have been quite good in doing wealth transference, which basically means that overtime, any investment you do might have negative.
Which all sound a bit alarmist, but it needs to break at some point where investing money will not work because there will be no payoffs.
What will break first? The people or the system.
> The joy of programming is dying. The joy of delivering results lives on.
Before I went to college, I worked as an electrician apprentice. One of the guys I worked with was 50+, and had worked many jobs over his life. He told me that from 18 to to his late 20s he worked a factory job. Then, one year, they stared rolling down production. The company had decided to move production to China.
Most were just let go. No severance, or comp packages. Just a good luck with your job search, it was nice while it lasted.
For his next gig, he worked in fisheries. His 10 years of experience at the factory didn't mater, and he started with entry level pay. This lasted some years, until some bigger company bought his employer, and merged. Lots of people had to go, him included.
Then he landed work delivering mail. Worked there for 15 years, until the postal service started downsizing due to ever decreasing volume in mail being delivered.
So his last attempt was to just go to trade school, and try some trade. 50+ years old working on a apprentice salary, once again starting at the bottom rung of the ladder.
My point is, there are many people out there that have lived this life. For every software developer that have made bank for the past 20 years, and are now faced with the reality of being made obsolete, there are probably 10 people who've gone through this multiple times, but gone from one minimum wage job to another. Going back to start every time.
It sucks for everyone. But you just have to adapt. I guess one silver lining is that when the "white collar professional jobs"-apocalypse arrives, that's when we'll finally look at things like UBI.
Time and time again, we've seen that when working class jobs disappear, it is met with some sort of aw-shucks / but-they'll-find-something-else-to-do reaction. Whereas when the middle/professional class is affected, it is seemingly a end of times sign.
Historically I agree with you, but considering how many people reacted negatively to the mathematicians' concerns about development in ai, and people calling them hobbyists, it seems to me someone is trying to create infighting as always. Seems like the reaction may yet again be the mentioned aw-shucks, but they'll find something else to do
I have, too, observed some sense of schadenfreude in the comments on recent headlines about layoffs in the white collar sector.
It is hard to know if it is just bot traffic to sow discord, or if plenty of people are genuinely delighted that office workers are finding themselves unemployed.
> But you just have to adapt.
Yes, but this time around there may be nothing left to adapt to.
Certain sectors and jobs will last much longer. Like healthcare.
But, I think the cold hard truth is that if you're a professional worker, with a six figure salary, the idea of going back to school, and working with something completely unrelated to your interest - all while taking a steep pay cut...it is unthinkable for many.
The alternative is of course unemployment.
If someone is willing to settle on salary and job comfort levels, there are plenty of jobs, at least currently.
Lots of people have their lives set up in a way that an instantaneous significant reduction in income is going to have severe effects.
Sure, that's bad.
But that's true for anybody that need to start as a beginner in a new industry.
Yes, but then there is a new industry to being in. This time it is different: there won't be any place where just having a brain that works is going to be a source of income.
>Do you know how hard it is to sleep when all you can hear is your own heartbeat? Do you know how hard it is to function with four effective hours of sleep when you offer your body ten? It is rough.
This hits home. It's both a shame and infuriating what's happening in the market right now. One (unsolicited) piece of advice I would give to anyone impacted is to build their own shit. If these companies stop paying for your lunch, then eat their fucking lunch with the same gimmicks they convinced themselves are adequate to stop paying for your lunch.
I'm not just talking about the enterprise giants - build competing products , lower their margins build similar products and offer them for less. Cut out the middle-men. They might have the contract now but their operating costs are higher. In the long run they are uncompetitive.
This is actually a great idea. It's easy to forget that if AI lives up to its promise, it cuts both ways.
For now, those with Capital will probably be ok. Those without Capital? In the EU, they'll probably be ok. In the US? Who knows. In China, India, LATAM, Africa - who knows?
Ultimately in the future (near? middle? far?), Capital will also be useless because the marginal value of everything will be driven towards zero, and who will be the buyers? Certainly not those with Capital, because the marginal cost of anything being close to zero would mean that they would not need to buy from others - they can just make their own. And those without Capital won't have the means to buy anything, even when the cost is close to zero.
Interesting times ahead?
Anyone care to speculate what that future will look like? Two different eras: post-employment, and post-capitalism?
> Capital will also be useless because the marginal value of everything will be driven towards zero
Capital is that fraction of produce set aside for future production. OpenAI is capital but not all wealth is capital. And capital isn't going anywhere - the machines will have gigafactories etc. The value of labor will plumet but present-day capital that is usable by the machines or can be scrapped and repurposed will retain value.
The value of square meter of land + mineral/sunlight rights (not capital) is likely to rise - the amount of wealth that somebody would give you for that asset would probably increase, as more wealth can be generated from it.
Individuals without wealth now will likely end up with a much lower allocation of those things which retain value (land, materials, etc) while having a much higher quantity (relative to now) of those things whose value will drop - AI tokens, electronics, entertainment, etc.
Essentially our current trajectory taken to the limit. Land expensive, food expensive, interest low, smartphone cheap.
> who will be the buyers
Imagine an impoverished and technologically backward country which has been colonized/occupied, and where the most valuable assets are exported well below market price or stolen. The people still trade their own labor for subsistence. They still produce some goods which have value in outside countries, they receive some charity, there is still tourism, their labor has some value (servitude, trafficking). AI would be the same (assuming laws hold up) - e.g. you could rent out your solar panels to OpenAI in exchange for a loaned domestic robot and a token subscription.
Energy, land and row materials will be the main limiting factors.
Mega corp will advance technology exponentially, but for the rest of us we will live in a mix of medieval life style assisted by robotic and low tech. People will exode city and go back to the countryside.
For what concern my self, I'm buying arable land, maybe in couple of years home food production will be fully automated.
>but for the rest of us we will live in a mix of medieval life style assisted by robotic and low tech
Awfully optimistic. Why would they let you do that?
Hopping that space colonization will become a thing for the top 0.0001% and that earth will not be their main concern, but yes I’m an optimistic person by nature.
I do wonder what will happen to the outsourcing industry in India. In the UK at least it seems that most corporate development has been sent to Infosys, Wipro, etc.
Not just India. Millions of young, able bodied, angry people, all of a sudden finding themselves not having anything else better to do. Sharing borders with countries who have their own millions of young, able bodied, angry people.
> Those without Capital? In the EU, they'll probably be ok.
They probably won't. EU is not the same economic bloc that it was up to 2004 anymore. I am not sure that there is big country in EU that is having consistent high growth. I expect in the next 15 years the western eu countries to experience the joy of living standards collapse that the eastern bloc enjoyed during the 90s. But without some of the factors that made it bearable for the east.
If the aim of each country is just job creation, leaving aside "comparative advantage", job quality, salary, etc:
1. There are plenty of things being outsourced, They can be done locally.
2. There are plenty of things currently not being done by locals - like elderly care.
3. We can find new things to do.
4. We can work less.
And even if those things raise living standard somewhat, the efficiencies from AI/robots should be more than enough to compensate.
But, in a capitalistic globalized economy this isn't how things work. So we'll see some shift from that. And it will work better in countries with a more democratic socialistic leaning rather than capitalistic.
We're already starting to see that, with the recent EU trade deal with china, limiting the amount of cars china can send to Europe.
Reminds me of The Zero Marginal Cost Society by Jeremy Rifkin.
> Anyone care to speculate what that future will look like? Two different eras: post-employment, and post-capitalism?
In a word, bleak.
Automation kills some jobs and creates more jobs. Nothing about AI (other than the statements of people with lots of shares in AI companies) looks to be any different.
> Nothing about AI (other than the statements of people with lots of shares in AI companies) looks to be any different.
Everything looks to be a lot different. Once AI is better than humans at knowledge work, why would you even keep human in the loop? (That's assuming AI reaches the level of being better than humans - I'm not arguing for that assumption, but about it's consequences and that in this scenario we won't get much new jobs).
You might not always, just as we don't always keep humans in the loop when picking groceries for delivery[0]. Great. But that doesn't mean we can do that en masse.
You can give a workman a power tool and he shouldn't instantly assume that his housebuilding job is under threat, because corporations will build a house-building power tool.
Why won't he think that? Two reasons: 1) any tool that becomes more powerful tends to become more specialised and expensive to operate, so an electric drill doesn't scale to a house-building drill without incredible tradeoffs, and 2) he isn't reading a million breathless class warfare-igniting articles whose writers are paid through attention.
[0] https://www.youtube.com/watch?v=4DKrcpa8Z_E
What's the hmmm-hmmm part about? I didn't get that reference.
maybe it is about staring at thinking traces?
llm loop
TIL.
I guess I'm not using this tech enough :-)
Smaller and/or heavily quantized models have a tendency to do this. The reason is because the auto-regressive nature of the llm dictates that its output should be its own input. You effectively purposefully create a feedback loop. That loop will manifest like this once the possible number of output states starts to converge on a smaller subset of the available tokens. A really interesting side effect of using speculative decoding is that you can detect that the llm is going to loop. You'll see this monotonic increase in the efficiency of the predictor and once that cross a certain threshold the loop is either already underway or it becomes inevitable. I've seen such loops range across 10K token and if not for the speculative decoding signal I would have never realized a loop was underway, but once you know it you can write a simple piece of software to detect the loop. The lower the 'temperature' at which you run the LLM the higher the chance of a loop (because you are constraining the number of branches that lead out from every token). The smaller the model the bigger the chance that it will loop. The more quantized a model is the bigger the chance that it will loop.
There are interesting parallels between LLMs and biological structures as well, for instance spontaneous oscillation in nervous systems lacking inputs but with a feedback loop in the topology.
Any kind of feedback system is riding a fine line between amplification/generation and runaway behavior resulting in a loop. Electronics, mechanical control systems, biological systems and LLMs all share the same property: some fraction of the output is coupled back to the input. With LLMs the fraction is 100%. It's a small miracle they don't loop more than they do!
Am I the only one who's having a hard time reading white text on a brown background?
It makes me think about how many people just stop reading because it's not comfortable, and they don't bother turning on reading mode.
Just enable reader view in Firefox. Makes reading half of peoples personal blog post viable, since often they have weird font size choices etc.
The animal cartoons are also incredibly off-putting. I don't get it.
Any chance you have astigmatism? Some people's eyes don't respond well to dark mode.
This post is illegible.
What makes it so hard to read for you? The contrast is fine, it even has a narrow column layout to limit the distance your eyes have to travel, the writing style reads quite human and clearly laid out. Or are the 5 pictures giving you trouble?
Tf;dr (Too Furry; Didn't Read)
Recently, an agency that used to give me work handed me an app they had built using AI.[1] But even though they had used only the best AI of the time, like Astra and Fable, in reality it had many bugs and many wrong points. If you're not a programmer, there's a lot of room to make mistakes even if you use AI. So honestly, I don't believe the claim that Fable can do everything. It's smart, and it writes code blocks far better than I do, and it writes algorithms far better than I do. But that doesn't mean it has the global consistency a program should have.
[1]https://www.makonea.com/en-US/casual/fixing-a-vibe-coded-app
The irony looms large. Most of us have worked on projects that replaced people. I remember one from 2005; automating returns at an online retail company. The returns department was 15 people. We did a project that removed steps and automated others. Ten people were let go.
More so was agile having developers replace the heavy old product and project management organizations a lot of enterprises had. And then came devops were developers replaced sys admins? Remember those folks?
The billionaire class tell us the utopia is coming. Why should we believe them? Given that they’ve failed to share the benefits of the massive productivity gains of the last 50 years (stagnant pay, 30 hour work weeks, health care and social safety net?) why should we trust them to share this? We shouldn’t imo.
If you think AI can't get so good that it is a superset of Human capabilities, there's a good news.
We will expand the job market and create new jobs that weren't possible before.
How?
Specific mechanism:
1. jobs being automated implies companies can do way more with fewer workers
2. this means they have extra profit that they can push somewhere
3. they will push it again to capital by finding new ventures to scale out
4. for achieving (3) they need labour pool, which is where we come in
Few open questions
- What kind of new ventures? Colonising Mars is an example, but ambition is unlimited
- Will companies choose to scale out or just eat profits? I think they will scale out due to ideology and market dynamics
- Biggest question: can humans still provide value in some dimension where AI can't? I don't know..
“AI makes features not software.” - MM
>MM
Marylin Monroe? Mickey Mouse? Michael Myers?
The world where programmers and frankly most intellectual workers enjoyed privileged position and high salaries is ending. Some will survive most will have to adjust, start new careers, possibly never come close to income levels they enjoyed.
I was part of it in online poker world. You needed 2 functioning braincells to make 6 figures while playing a few hours a day back in 2005/2006. It got more and more difficult from there. First it got guys "living their life" while hustling on the side who thought it's an easy mode forever and didn't even bother to learn new tools. Then it turned previously good players into marginal grinders and then only a small group of winners remained. Many went to different/easier pastures - private games, heavy game selection at (super) high stakes. Most though lost their income and with it years of life that now seemed wasted in front of the computer.
Among my friends I've seen suicides (two), lost families (more than two), despair and poverty (too many to count). There are also positive stories - people who started doing something else, adjusted their expectation and are now living happy life even though their income will never be close to what their experienced in their early twenties.
In online poker - instead of Fable - we had ever developing tools. Databases, HUDs, HUDs connected to databases, auto-seating scripts (where a script looks for "donks" and auto seats you at the table when they seat) and then bots and solvers. I was on the frontier lab side I guess as the author of the most popular solver when they first came out. Now Fable is coming after my edge.
I am still a tech optimist. I think it will be a better and more abundant world at the end of it. There is still a lot of jobs we are not doing now because we don't have enough people. The transition is going to hurt though, especially those who don't have energy to start from scratch.
Gambling for living taught me no income stream is forever. Especially lucrative ones are going to be competed for and will dry out. Live below your means, try to save money, don't compete for status. It was always a good advice but it never hits until it's you who it applies to.
i sympathise with the individual of course, and i wish they did not have the experience they wrote about. it makes me feel quite bad to read it actually.
but it also seems like tech workers are getting a taste of their own automation medicine they've been happily developing for the last 50 years.
Wait until the ceo finds out all his business secrets are sold to his competitors.
AI is a slopening of everything that makes us humans. If we cant see it that is only because the last 30-40 years have been a constant assault on human intelligence. we became dumb and sloppy as well. peak western society ended back then maybe around the time the first Concorde airplane flew for the first time. Everything went downhill as technology progressed and people thoughts could be controlled and manipulated in ways where intelligence and curiosity were not seen as relevant skills to acquire and nurture.
peak western society ended back then maybe around the time the first Concorde airplane flew for the first time.
This seems a bit arbitrary to me. Why 1969, exactly? Personally I'd go for the "democratization of Internet" as the start of the slope. So perhaps 1989 ("1989: Quantum Computer Services and Apple end their partnership. Quantum Computer Services renames itself America Online. Two years later, America Online for DOS is released, and a year after that, America Online for Windows.")? Or arguably even a bit later...
You and the GP are effectively in agreement. It's not so much about when the exact peak was but about whether there was a peak and whether or not it was somewhere around that time.
> The frontier models are not good enough to replace an experienced engineer.
Yeah they are.
Two pizza team = 2 person team + Opus 5.5
Or for the really enterprising, solopreneur scaling to > $2.5 ARR (more than Goog's revenue per employee).
This is a fundamentally different game now.
I don't know why anyone assumes it just stops here.
It's a game where you are obsolete, unless you are independently rich, so not much to cheer about
Thing is that you cannot do 24/7 oncall with two person and Opus 5.5 . Also bus factor start to be pretty big if there is only two persons etc. There is still reasons why FAANGS are so big.
Sure you can, as long as nobody complains too much about problems during night if those 2 people can work overtime every now and then.
You can probably even promise 99% uptime (sounds good enough to most buyers probably?) even if people sleep 8 hours a day as long as the failures comes no earlier than than 15 minutes into their sleep and only happens at most once a month (1% of 30 days is 7,8 hours).
>Thing is that you cannot do 24/7 oncall with two person and Opus 5.5
Almost all 24/7 on call is support theater anyway.
>Also bus factor start to be pretty big if there is only two persons etc.
Not if a newcomer can be assisted by the other person or even just by AI in understanding the codebase.
depends where you work. on a tiny team the 3am page is usually real and its usually you
Ok, three persons.
That's dumb and just an excuse. If it were true, why not keep the whole team and empower them and deliver 10x the existing value?
The dark truth is that tech executives have been extremely pissed off by tech workers in the 2010s leveraging their specialist knowledge into exorbitant salaries, autonomous working conditions and even political demands (remember the Google in China backlash?), and they are now using AI as an excuse to do mass layoffs in order to scare the uppity proles-craftsmen into falling in line. As Kalecki said, they value discipline in the factories more than raw profits
> That's dumb and just an excuse. If it were true, why not keep the whole team and empower them and deliver 10x the existing value?
That requires having enough good new ideas for everyone to work on, and sadly a lot of companies don't have those
Even if they had, doesn't mean it's a good idea to flood the product with changes, nor it's necessary that the market has the capacity to asborb and expand your sales, or that the existing customer base will appreciate it.
Well I did say "good", which excludes things that don't fit in the context of the wider business.
I think startups have an easier time for 2 reasons: They are innovating on the product side already, and probably were already capacity constrained for work on the top ideas. They also have a high proportion of tech debt which never bubbled to the top of effort vs reward (competing against product work) before but would still be valuable to fix. E.g. https://posthog.com/blog/sql-parser was on the HN frontpage a few months ago, and this work would never have happened in the before times
Even if the idea fits the context of the wider business, there's such a thing as too much change "just because you can", which alienates users and can make the end product an unfocused mess.
If it's easier for startups, it's for the same reason doing a "pivot" is easier: they usually don't have much of a product or revenue or customer base to begin with, they throw things out there and see what sticks.
>That's dumb and just an excuse. If it were true, why not keep the whole team and empower them and deliver 10x the existing value?
Because they don't need 10x the existing value, nor is the extra value they can actually use and benefit from that expandable.
And despite being quick to call this "dumb", you missed the most obvious reason: by not keeping the whole team they save money. Which makes their expenses lower and their VCs/shareholders happy. Keeping the team and "delivering 10x the existing value" doesn't leverage AI for reducing costs, and is hardly guaranteed to bring in much in the way of returns either.
Usable new value goes through a "pipe" so to speak, which constrains how much you can use or need. And that's the business and market and regulation realities.
If you're making Instagram for example you aren't going to change everything "just because you can", nor is there any reason to make 10x intrusive changes to the backend "to make it faster" or something with the danger of disrupting something, if it goes predictably well enough already.
Your marketing team, your engineering team (what's left of it), your business team, also need to understand and pace any changes, and have a specific vision, not just "anything possible, even 10x the things".
Because too many people = too many opinions, communication loss, vision conflict.
10 people don't 10x. They barely 3x.
If you think AI can replace experienced engineers you're not working on anything complicated. If your product is a web app, sure, maybe. If you're working on actually difficult problems, then people who know what they're doing are irreplaceable. AI can supplement them and make them more productive, sure. AI does not know what NOT to build.
> ackchually difficult problems
That's likes rocket science, av, automata and things like that.
Less than 1% engineers work on that and even they are using ai.
For other 99%, one person knowing what to build is enough.
Fwiw the guy you're replying to is trying to replicate 1:1 the entire Adobe creative suite (and now Microsoft Office?) with the help of AI.
Had to Google "two pizza team". On what planet does 2 large pizzas feed 5-10 people? One pizza barely feeds me.
i'm tired man. AI this, AI that, here's how we productmaxxed our deliveries, here's how many people we could remove, etc etc. Everyone is going crazy about ai, but no one is remembering the people affected.
Corpo culture is disgusting around layoffs. Zero notice terminations should not be legal, anywhere. Have some fucking heart guys.
To be fair Zero notice terminations are illegal in many Countries. Furthermore I am yet to see solid evidence that replacing humans with AI is beneficial. Actual crazy times.
I don’t think the blame belongs with corpo culture. It belongs with engineering culture. Engineers have always been at the forefront of technological disruption, and for the most part they’ve been spared its consequences. The people displaced by those efficiency and productivity gains stayed invisible to them, or were written off as a necessary sacrifice for progress. Those people lost the old way of doing things and were told to keep up or get left behind. The irony is that now it’s engineers getting hit hard, facing the threat of being made obsolete in the name of efficiency and productivity. They’re finally going through what so many others went through before them. And they can’t expect any compassion or help from their own kind.
The severance is your notice.
We will lay you off in 3 months - and having for a 3 months dead man walking in the company is probably worse for everyone. Since said employee probably won't break records for productivity during this period - it is easier for everyone just to give the one let go some lump sum.
Anyway I think that in this post the context is that they will be let go during good times - without the informal - we know the company is in trouble and layoffs as coming common knowledge that big companies have.
I'm not just talking about the post though. In many places it IS legal to remove someone zero notice with or without a severance package.
"both of them would lose their job next week with zero notice."
What kind of 3rd world country do you live in where this is legal? I just want to point out that businesses needing to balance their workforce is normal. Living in a society where letting go of employees without notice is not. Maybe that's the part that needs to be addressed as a matter of urgency, not the management's delusions of AI capabilities - those will verify themselves very quickly in practice - employment law takes time and effort to change
My last job I was notified via text message on a Saturday that I didn't need to come back into the office Monday. The USA has basically zero protections for employees.
Yep. My company, based in the US, did that to most of our US team. Being somewhere with decent employment regulation meant that they kept us and not the US guys.
https://www.dir.ca.gov/dlse/TerminationOfEmployment.pdf
So, USA is the answer?
The good US workplaces will give you a few months of severance, but you can still expect to receive an email late Friday night that says "don't come in on Monday". They typically don't want fired/laid-off employees in the office at all (in theory for "security", but mostly to prevent labour organisation)
That's fine - being paid for your time in lieu of notice is practiced even in countries with very strong employee protections. That's still "notice" as far as I'm concerned.
Yeah well in the US employees can be fired at will. However in the post it seems the author didn't tell his reports they risked/were about to getting fired. I don't understand that kind of loyalty to your employer, esp. when, as is the case here, the employer has zero obligations towards their staff.
I think it's pretty normal to have like one/two week notice period in most of the work except for maybe in Europe where the notice period can be 3 months
Would it make much difference if you got two weeks or a months notice. Out of the bad things corporations do this seems a long way down the list.
To be fair, what is often called "zero notice" actually means "you are fired, this is your notice, and for the next N weeks until your employment ends, you don't need to come in anymore as we have already revoked your access, but of course we pay you the salary that you are entitled to." That us just much longer to say and less catchy too.
The most outrageous sentence frankly is
"we can't ship your stuff because of some material restriction or let you pick up a box with some dignity, so we destroyed your stuff"
Really? never heard of anything like this even when I was in the states. If anyone tried this anywhere in Europe the case would almost certainly end up in a parliamentary questions session.
If you want something outrageous, I was fired as my dad was on his deathbed, he died before my 2 weeks notice finished and that created a lot of heat since I did not send them their laptop in my last day as I was busy preparing his funeral instead. They did not care, it was surreal, and I completly lost faith in humanity as a result
SWE as a job is over, this should have been obvious at start of the year, its is simple as knowing the capability and watching the progress of the models. Sitting waiting for it to happen and crying about it is a bad coping mechanism, accept it and have a plan B/C/D in the works while you are one of the lucky ones where it company has not made the move yet.