Dark day today, dark days ahead.
My own hope was re-established on Open Weights AI levelling the playing field, but they've managed to steal the global chip supply, undermining that too.
I'll go out on a limb and say, even if AI does cure cancer (a big if), was it really worth the loss of meaning/purpose for millions of people along the way?
It has been a complete self own by our own city!
The issue may be less “whose fault is it” (to what, punish?) but instead “how do we incentivize / encourage / command change?”
For assumed access to the best models and unlimited tokens, Claude Code is just whatever it is today. I'll take Pi over it any moment that it's not bundled with a API plan.
30000 commits and so what? 10x the use of memory to come back with output that is barely english.
Have to accept reality and act, is simple but not easy.
The valuation of Anthropic is based on the premise that it will enable the owners of capital to fire their most expensive employees. That’s me. All I have to sell is my labor. If no one is buying, I don’t eat. It doesn’t matter how useful the technology is — the future it’s building is one that has no place for me or anyone I know. Not sure what I’m supposed to do with that.
In a different world, AI enabled you to no longer needing to worry about putting food on the table. You'd be given a chance to focus on what you wanted instead. I suggest you read the Culture series by Banks, it gives a optimistic view of (true) AI.
1: https://knowyourmeme.com/memes/wait-its-all-ohio-always-has-...
I have a client who wants a co-investor for USD 1 million. He has an investor for USD 1 million for a Phase 1 cancer drug and is struggling.
The capital allocation here is a fraction of the daily stock market trading volumes the tradeoff here is large capital pools deciding between Ant, other corporate stocks, private equity and bonds.
If our society allocated wealth differently we could afford to fund 1 million projects like that cancer drug instead of spending that money on Anthropic
There is rightfully a lot of skepticism about whether Anthropic can maintain their business in the face of competition, especially from open weights models. But it's undeniable that they do have a product that does work at doing useful things, right now.
Presumably that cancer drug, important as it would be if it succeeds, is not currently guaranteed to produce any useful result at all (even with enough funding).
Ant IPO just seeded a huge amount of multimillionaires many of which will likely fund cancer research.
Anthropic has a product that can potentially help develop countless cancer drugs.
> If our society allocated wealth differently
Other societies tried allocating wealth differently. I lived in one, here in Eastern Europe. It couldn't even feed its citizens, let alone develop cancer drugs. Turns out central planning is a much much much worse way of allocating wealth.
For the umpteenth time, the bottleneck in drug discovery is not the design-part, but the trials. Can claude code help with that? Will it breed willing and able adults into existence through its API as participants in drug trials?
Also, frankly, a private company trying to make a profit off of a cancer drug is ALSO an abomination.
Sure, I'd love it if governments around the world threw money at cancer research, but claiming that private pharma companies are "more moral" than AI is a losing battle, in my book.
So this seems aligned with that value
Not saying it's perfect, I'm just debating I have yet to see better alternatives.
I'd say the allocation problem is on the cancer side.
Right. On what is probably the most significant solvable health issue of our age.
Compare this to what, $3 trillion parked in the stock of an ad tech company (Google) and now another $2 trillion betting on some wishy-washy benefits of AI (maybe it will cure cancer too? the founder said so).
I don't think that central planning is a good alternative, but I think we should talk about this, because capitalism evidently doesn't produce a particularly efficient allocation of resources either. As it turns out, you can have massive industries that do relatively little for mankind but are very attractive to investors.
if anything - they probably created plenty of new wealth for people to invest in other places.
the pie can grow.
I feel like wars are a larger problem.
Well, the market is betting on the fact that Phase 1 trials for such critical drugs can be fully modeled on digital twins running on GPUS, so they can leapfrog needing human subjects for RCTs.
Not disagreeing with you, though.
i thought the free market was the best tool we had to allocate resources???
And with this now below where OpenAI has been trying to value itself on the private markets that doesn’t bode well for OpenAI. If I was at OpenAI I’d be very worried about the picture today paints.
Long story short, the board briefly fired the CEO Sam Altman because they were starting to eye that kill-switch and they didn't have confidence that his leadership aligned with the the non-profit's mission.
Then a few days later OpenAI and its employees decided that, no, actually they like the money that Altman seems to bring in and they hired him back. If one were being charitable maybe there's an argument that the money is important to implement the ideals, or maybe now that there's a competitive LLM marketplace bowing out wouldn't accomplish anything. Regardless, now it's just a bog standard Valley company which is partially why they can IPO. Maybe not in a strictly legal sense (it's a PBC now, which the FAANGs are not), but functionally. So it goes, so it goes.
I could be distorting some things. I'm only passingly familiar with the details of the matter. Feel free to correct me if this seems substantially inaccurate. Cunningham's Law and all that. ;-)
Strictly speaking? No. But the people who shoveled money into it want an offramp, and that's what the IPO is. So for all intents and purposes, yes they need to IPO.
Yes. They need to find a bagholder before the bubble pops.
Remember recent headlines that OpenAI was struggling to raise cash at valuations above where ANT is now trading.
Net here it’s is a good day for those looking to cash out of ANT and a very worrying signal for others elsewhere that were hoping to cash out too. Cudios to Anthropic for recognizing that there was likely only room for one foundation lab going public and they got there first.
Makes sense to me. This valuation is pretty insane. I don't see it holding up.
dario and daniela needs my voice, sam needs my eyeballs, oh boy
Daniela is right on the money.
- Big IPO - completely arbitrary
- First half hour - completely arbitrary
- The fact that it went up 34% at open before dropping - let's ignore that and only count the dip
on a tiny nano volume, if I understand correctly that's mostly technical preparation to release it to public
> Today at 10:07 AM PST
> Jared Blikre
> Anthropic's first half hour is the worst start for a big IPO since 2008
you might as well be quoting a tweet.
https://finance.yahoo.com/markets/live/stock-market-today-th...
not to mention the observation makes very little sense - the ipo was 83 so who cares that it opened at 110 before the bid-asks were narrowed.
Heavy Anthropic users have a strong incentive to demand DeepSeek cost/task. Its a ~10x difference. There are other ways that will be discovered to improve cost/task. Those decreases in cost don't necessarily lead to dramatically lower bills: it can be soaked up by more heavily using AI in existing processes (code reviews, veifications, etc). And we will see a lot more software being developed over the next few years.
Anthropic stock seems like a bet as to whether they can both maintain their leadership role in software against intense competition and further develop other industries outside of software.
-- Morpheus
$2T is arguably a more correct number in some senses
Lets not be naive here; somebody is getting paid a lot of money to make sure that the headline yesterday was 1.4T and the headline you're reading today is 2T.
It appears to be fully deluded at the very least.
... but *god dam* it feels like 2025 today.
I swear Ima about to throw my laptop out the window - it's been one of the most frustrating few days of my career.
I think this may be an ongoing problem: The AI is 'poweful but broken' - aka it does not feel like a working project by it's very nature.
It's like a fast car, with doors that won't close, gas pedal 1/2 works, brakes 1/2 work.
Is it faster than walking? Yes. But it's just crazy - and Im beginning to think this is just a property of the AI and that overcoming it woudl require something like 10x the compute .. which is 'not worth it'.
So - we won't get a smooth ride for maybe 25 years, when RAM/GPU become and absolute commodity, and the work we do now is goes through 10 filters/passes before you see an answer so its not 'full of **' 1/2 the time.
There is a new psychology here.
I love development, I love AI - but this is driving me completely crazy.
Edit: Thanks folks, typo, updated the number to billion. Question holds though.
This sounds arbitrary until you appreciate that the ones with most purchasing power on the market are also the same ones who have previously proven themselves to be capable asset allocators, as proven by their ability to accumulate capital. So the valuation of a company is not just every random Joe’s opinion equally averaged, but weighted towards those whose opinions on value have previously tended toward being correct (this includes capital acquired by any means, not just equity trading. For example, if you earned a lot of money with your own product, you have proven yourself to be able to judge value and effectively allocate resources too).
The market starts to make a lot more sense when you start to use it as an oracle of hidden information (“what is the market telling me about the value of X?”) instead of an irrational adversary (“why is the price Y? That doesn’t make sense!”).
Most folks in finance I talk to think a realistic valuation backed by actual math is likely around $400-500B. Now that the investment bankers got their payday for this hype fest, expect numbers in that range to start coming out from analysts at the big banks.
Now mention how much they have spent, they are literally selling $3 for $1
[1] https://www.geekwire.com/2012/amazing-73-tech-ipos-profitabl...
https://www.msn.com/en-us/news/other/dario-amodei-says-ai-wi...
https://www.noahpinion.blog/p/ai-keeps-stubbornly-refusing-t...
I'm a zoomer, so sorry if I'm mistaken.
Anthropic's revenues for the last 12 months are about $35B. Nothing, compared to the $2T GDP of Spain.
(The bad part to me is to think it means Anthropic’s revenue is comparable to Spain’s GDP. As you point out it’s not.)
I don't know what kind of audience you think you're writing this to ... we all know what GDP and market cap are, you're not the Wizard of Oz here.
And yes, it is an interesting comparison, because it translates a somehow abstract amount of money into something a bit more tangible, i.e. the entire economic output of a developed country.
Correct, Claude can't play with your kids at the park, but Spain's GDP doesn't include that, so that's not relevant.
Anthropic hit a peak market cap of $1.46T post IPO and is now at $1.27T. It may go up to $2T over the next few years when more shares vest, or it may not. Sites reporting $2T today just want to maximize engagement ("we annouced a bigger number than the other guy!").
If the prevailing price of those shares does not change, then yes, it actually does.
> Existing shares become worth less.
Maybe, maybe not. You're talking like this is all formulaic.
They literally talked about "adding up all the shares" and argued against that, but for mechanical reasons, not philosophical reasons about value and worth.
And for all intents an purposes, a company's "worth" (undefined) is almost always measured by its market cap (well defined) as the least-worst approximation.
His whole point is that you can’t assume that.
Maybe, probably even...but it's irrelevant.
That assumption is doing a huge amount of work. How many shares don't change?
Make an issue out of it and you'll just run into "prove it!" to a billion people who will just blank stare in return
The media just grabs whatever event to tie it too, lots of clicks to be had from people wanting an answer, but the reality is that it's mostly just noise. If the markets were ripping green today, then they would be attributing it to anthropic all the same, all else being equal.
Anthropic's share of the total US stock market at IPO is 2%. At the current price, it's 2.74%.
Sergey and Larry hated the idea of mispricing the IPO to generate the "pop". So they did a "Dutch auction" to find the _correct_ price and avoid the unnecessary pop that happens due intentional underpicing. Auctioning (ads) was their bread and butter anyway
Well, it popped anyway and it popped a lot. And no notable company has done Dutch auction since (to the best of my knowledge)
In different terms, Google IPOd at a market cap ~ 10% smaller than Ford’s at the time.
Anthropic is at 50x Fords market cap. It’s over half of Googles market cap, and Google is a competitor in the same market with a dozen already successful lines of business.
They are incredibly highly valued in a way where I don’t even understand mechanically how they could grow like that without the entire financial system crumbling.
Google is up ~ 6,600% since IPO. Apply that to Anthropic and they’ll be worth something like $132T in 2048. As a point of reference, the Congressional Budget Office is predicting the US national debt to be around $102T in 2050 (paper was from 2020, might be outdated).
Anthropic, the singular company, will have to grow to be worth 130% of the entirety of the US’ debt. Napkin math says their revenue would probably have to exceed the US governments tax income. Like genuinely, Anthropic would have to become a world superpower all on its own.
They would be so big that they’re an existential threat to whatever country they call home. They would be such a large portion of the economy that central banks would lose any kind of control (presuming Anthropic doesn’t issue their own currency, they’d be a larger economy than most countries).
I can’t see how that could happen mechanically. The world would burn to the ground even trying to make it happen. If you thought Standard Oil was bad, this would be that but for every product and service in existence and all run by the same company.
$2T is 87x more.
Looking at recent mega-IPOs - SpaceX and Anthropic, with OpenAI coming soon, the price assumes that the company will be a success and simply discounts the projected future cash flow, as it was not risky at all. Future is not certain though. It will be fun.
Sure, Google had the top of the line offering. But you could easily argue that they don't have the moat, that web search is a commodity, and that more established competitors could catch up to Google in no time at all.
Also, unlike with web search, the AI companies are makers of their own destiny - in that they can potentially grow the TAM massively purely by improving their own products. The more capable and price-capable AIs get, the more they cut into various niches, of which there are a great many.
Even in my own field: reverse engineering is hotter than ever, and it's because AI allows RE efforts to scale to the size and complexity of modern software.
Meanwhile other countries are chugging along, doing whatever they want...
This IPO price was engineered, through a small number of shares available, to presales to Big Tech. So, I wouldn't read too much into this news, it remains to be seen what it all means.
From personal experience, I have worked with many IPO and stock option multi-millionaires who keep driving down the 101 and 280 every day to go to work. They could quit at anytime.
People who find themselves in a big tech IPO tend to not be the type of people who want to retire at 30 and drink pina coladas at some beach resort 24/7. They'll be bored, feel they are "not in the game" anymore, worry about wasted potential, worried about becoming unhealthy with a hedonistic lifestyle, etc. They just have a lot more financial freedom now.
Long story short, we live in a capitalist society and when you have capital it opens up a new level to the game of life. People will want to play that game and do more. Evolution made us to work, be communal, create, do things, etc. Cash-out, drop-out, and live a questionable lifestyle at various tourist-heavy tropical countries John Mcaffee-style is really unappealing to a vast majority of people. I'd certainly still work in some fashion if this happened to me even if I crave a vacation lifestyle. I'd be too bored. I'd probably take a lot more vacations and probably start my own companies, but essentially I'd love to keep making things.
I guess I'm the almost never happens. I had my "enough" defined before I took my job at the company that IPO's. Retirement was not to a beach resort but a University Town. Did lots of volunteer work. Took lots of classes. A lot of living my life did happen.
Can't say that boredom has ever entered the picture, I can invent interesting projects to pursue faster than I can finish them.
There are lots of alternatives to drinking pina coladas all day that are far more interesting than making more money, and making even more money for someone else.
Many of DIY MBA books I've read have some version of a story of some entrepreneur encountering a fisherman, trying to convince them to raise capital to expand their business... which leads the fisherman on a long trail that led them back to fishing and spending time with friends and family, right where they started. https://en.wikipedia.org/wiki/Anekdote_zur_Senkung_der_Arbei... https://paulocoelhoblog.com/2015/09/04/the-fisherman-and-the... https://bemorewithless.com/the-story-of-the-mexican-fisherma...
Sadly, the value of money has been hard-refuted numerous times already. The most valuable things currently don't exist, and as such, they cannot even have a price. By modern understanding of the laws of physics it's questionable whether they _can_ exist (they can, but people aren't intelligent enough to realize it).
I hope employees are able to cash some out to get the reward they deserve, but the fundamentals don’t look good here. And to be clear they’d be selling to folks that are unlikely to see upside.
This quote from Margin Call feels applicable here:
Sam Rogers: The real question is: who are we selling this to?
John Tuld: The same people we've been selling it to for the last two years, and whoelse ever would buy it.
Sam Rogers: But John, if you do this, you will kill the market for years. It's over.
[John nods grimly]
Sam Rogers: And you're selling something that you know has no value.
John Tuld: We are selling to willing buyers at the current fair market price.
[Sam lowers his gaze]
John Tuld: So that we may survive.
Sam Rogers: You would never sell anything to any of those people ever again.
John Tuld: I understand.
Others can sell in waves beginning 27 days from the filing of the IPO prospectus.
Founders can transfer 29% of their stock 180 days after trading begins, with the rest of their holdings after a full year.
I have been through a couple of IPOs and the 180 day rule was the one I remember. We were totally locked out. At the time of the IPOs of those companies, we also didn't have the same secondary markets that are present today.
This seems like a very good deal for employees to get some liquidity.
It will also go like the 2007 crash: once all the layoffs happen, employers will realize that they can do the same work with a fraction of the workforce, thanks to automation. It will be a jobless recovery.
So we may get the crash AND the AI takeover. I just hope governments are able to respond fast enough to radical societal changes.
Are anyone really doubting this part?
$2T is a big number but this Anthropic IPO has been straight down since open, so I wouldnt say the market has gone completely insane.
Thank god.
No, this is cope. The difference with CDOs and llms, is that either you like it or not, they have massive value. They can already automate most of the boring white collar work. Fewer and fewer human decisions remain. They are not perfect, but they are improving massively. The value is there, people are hooked+ brain rotting, they will be dependent forever
I expect this will be the general trend. Not every day, but broadly over time. Anthropic will slowly at first, then quickly just capture most of the US economy. $2T is low for the first company in history on that trajectory.
Make no mistake, more regulations are coming - open models will be regulated to the point where they are useless and Anthropic et al will make sure of that, now that they will have even more pressure since going IPO. We should never forget that consumers deserve choice and open models must be always a valid, usable choice. Not namesake.
He stole all our IP and now wants to make us unemployed. His bankers pumped up the price from 9:30-12:00 in non public trading to $111. Now the thieves sell off slowly during the day.
IP, and copyright in particular, is a scourge upon the land and a violation of our freedom to think. I honestly believe we would all be much happier had the concept never been invented.
Want to keep your intellectual output to yourself? Keep it a trade secret!
* songs - Most artists don't make a living as a professional from royalty either. Most earn it via live gigs, and there are now platforms where fans can support the artists they love, that don't depend on IP.
* books - Most book authors don't make a living by writing books, either. That said, I do believe the same or similar platforms can be used for non-audio works.
* movies - Most people don't make movies, Hollywood makes movies. I don't care how the next Marvel movie is financed, tbh. That said, people still pay for the big screen experience: you can keep the film (or a digital original) a trade secret, ban cams in theatres, all without relying on IP.
Do we all have to endure the insanity of IP so that Taylor Swift can continue to earn gazillions 70 years after she dies (hopefully, far in the future), or to watch seventy-sixth installment of Marvel once-upon-a-time-was-blockbuster?
Here's an excerpt;
Sir, if every good and service in an economy has infinite supply, while demand for each remains finite, and we retain the assumptions of perfect competition, unrestricted access, and zero marginal cost, then the result is profound:
The market ceases to have an economic function as a mechanism for allocating scarce goods.
This is the limiting case of a post-scarcity economy.1. 7 Co-Founders: ~$20+ billion each 2. Several early employees: $1+ billion each 3. 598 employees: $100+ million each 4. 2,800 employees: $1+ million each 5. Total employee equity wealth: $232 billion 6. Employee stock option gains: $119 billion
San Francisco’s median single-family home sale price jumped +29% year over year to $2.39 million in October amid the IPO process.
This marks the largest jump in San Francisco's media home price since February 2018.
Thousands of new millionaires will be minted today.
Great day to be an ANT employee if you are in one of those groups. Life changing and multi generational wealth and early retirement.
in this IPO, all 7 founders are subject to two lockups for ALL of their shares: 180 days, then they can sell 29%, 365 days, the remainder. only ordinary employees can sell 10% of their vested RSUs / exercise options today, which will be about $400m if everyone sold and everyone got the biggest offer anthropic gives employees, but will be closer to $25m given what the actual rate of hiring (and therefore vesting) and awards were.
The reality is that the situation is very complex, and saying "it's a bubble" or "insanity" or (Zitron's favorite) "you're the exit liquidity" just misses the forest for the trees. AI is clearly revolutionary, there's also some signs of it being a bubble, the market does seem to be less risk averse than 20 years ago, etc. It's all very muddy, and claiming otherwise with so much bravado is just dishonest.
So no. None of those 'do this specific thing that should be sane by the science of economy' arguments work. Some powerful institution can just f*ck up your plans just because.
Costco for example has done quite well this year. How has the AI trade with NVDA worked out this year in comparison?
There's so many ways to make money the default answer of "I don't like this stock" is buying something else.
A: I don't like milk.
B: If you hate milk so much, then borrow someone else's milk and then sell it!!
A: Wtf? No. I'll buy orange juice instead.
You can say Anthropic as a financial instrument is worth $2T, but then you head into irrational financial bullshit territory where there's a nontrivial chance a lot of people are going to jail.
There is no global copyright law that governs ownership of digital goods. If it goes online, then you should expect it to be copied, pirated and widely distributed.
Anthropic, Google or OpenAI are not responsible for the mishandling of your copyrighted content that was publicly available on the web.
Just because it is wrong and illegal, does not change the fact that the data has been made available and distilled into open weight models and now every lab likely has used the data in some form directly or indirectly.
Cat's out of the bag and there's no going back. Back to my original point, the data is out there. China would do the same thing and not get in trouble.
We don't know how much this crap costs anyway, my $20 sub is super subsidized. When they crank the dial to the true cost of tokens, we'll see what' what.
I better tune up my compound bow and oil the guns, jic.
update: dow, nasdaq; russell2000; gold, all down ~2% so far today
sorry, what's the good news again?
Like what's a realistic multiplier it could go up at this point?
Not sure why this was downvoted, are we angry at people building transformative products now? Or at startups making it big on their exits?
When these products are built by skirting the law, yes.
When its leadership keeps pushing for regulatory action to stop competition on its track, very much.
When their product allows them to collect the upside but do not give them any responsibility for the downsides, of course.
IPOs these days are not the same as they were 10 years ago. Now, there is less of an upside and the early investors will sell 10% on to retail while you buy at its inflated price.
You are being used as their exit liquidity.
Especially since this is a tiny percentage of the business on sale — 5%? 10%?
Edit: Another huge buying spike at 1:09 pm at same $95.5 level, but not enough to stop it continuing to slide for it's low so far just below $93.
But then everything about today — the muted reaction, the lack of people losing their minds over a new era — feels more like relief (someone got to IPO) than jubilation.
As a sort of pragmatic cynic, I find it rather reassuring.
Compare that to the $140B owned by the founders, and $200B+ by employees.
I don't know when employee RSUs/options become sellable, but I'm guessing the only employee selling today is part of those 23M shares included as part of the IPO.
They go into the opening with a short position, which they can close out by buying to defend the stock. If the stock goes up rather than down, then they have a "greenshoe" option to buy more shares at IPO price to close the short.
I felt like it was all these people came out of the woodwork objecting on very procedural grounds when they obviously didn't actually care about rules relating to stock index inclusion. They just didn't like Musk.
I wonder if people will relitigate these objections now that now that Anthropic will be fast tracked into the indices same as SpaceX. My guess it they won't and my initial instinct of no one really caring was correct.
The business model is in its infancy and whether or not it becomes the google of the next decade or the blackberry of the next decade remains to be seen.
how?
Idk man. I wouldn't buy in now, if anything I'd short it soon. I expect the stock price to be a long slide down until they can figure it out.
I've never ever dealt with a company with such piss poor customer service, and i've dealt with Google.
regardless of the fiduciaries and the bravado about the product, knowing that such a genuinely anti-customer company has found such success irks me the same way as knowing that Musk owns Spacex; it's just a deeply unsatisfying and somewhat troubling truth.
My read is the opposite.
Hn comments have always been heavily cynical of Anthropic’s valuation.
These comments seem way less cynical. Its just my opinion but it feels like that attitude feels a bit defeated or disproven.
Which surprises me because I believe the thesis was always “they aren’t really worth 2 trillion” and not “the market would never value them at 2T.”
It definitely reminds me of early Google Search where normal people didn't viscerally feel or believe the potential of how amazing the product was. So I was content to overpay to buy a very tiny bit of the ANT today.
(I need that reddit /remindme in 10 years bot for this comment...)
Sure, but they did it by stealing all the creative works ever published. Sorry if I don't line up to cheer for their success.
I honestly think, given that, the best way to capture the value of AI is to buy the index fund VT
You are all in a cult. It’s a public company, so far they haven’t shown their business is sustainable. That’s literally the only thing that should matter
I've also seen non-technical folks in the corporate and government worlds glom onto vibe coding, only to start flailing like they were caught in a rip current when the underlying technical debt is too strong, even with strong and capable models.
For a certain type of engineer who enjoys really arcane technical problems, yeah, the gravy train may be over, or at least will involve a lot more machines than it used to, as Alan Turing predicted about "mechanical work" becoming a "drudge."[0]
And I think Big Tech is for the foreseeable future a sink for capital in the form of specialized hardware rather than human capital.
[0]https://en.wikiquote.org/wiki/Alan_Turing#Quotes
I'm old enough to remember when compiled languages like C were bemoaned by people because their skill at crafting assembly by hand was suddenly less important. Or similarly with people who were brilliant with arithmetic, or symbolic manipulation, or manual drafting/lettering. These skills were all paramount at one time, and it's natural that the people who excelled at them will feel a sense of loss.
The real goal, however, will always be to solve real-world problems. Engineering consists of whatever gets you there most effectively.
We had prevented this for 20 years in the 1950s and 60s with a 90% federal tax bracket that capped yearly incomes to $5 mil a year in 2026 dollars. Stock buybacks were illegal and we had a bunch more post-1929 regulations that were not gutted yet.
Restore the tax policy and financial regulations of the 1950s. CEOs would be stopped immediately. This is how we got out of the great depression.
CEOs are allowed to do all of these things, but they do not necessarily maximize business returns, and therefore their own incentives. Nothing has changed here from 2020, 2015, or 2010 -- a period over which SWE compensation has skyrocketed. No reason to believe this gets worse with AI.
The 50s and 60s marginal tax rate was (1) mostly paid by no one; rich people instead received business income not subject to personal income tax; and (2) other high rates on the middle+ class were only possible due to a sense of national togetherness forged by the world war. There's no reason to believe this time period was exceptionally good for the poor, working class, or middle class.
It's easy to hate the rich and want to punish them for being rich, but it doesn't help the poor to do so.
Using the 1950's as a justification for policies you support is disingenuous because the real reason the 1950's were so good is because the USA had just won the largest war in history without any damage to the homeland and had the only industrial base that wasn't bombed out and exported a huge amount of manufactured goods to the entire world.
The policies you advocate would just privilege the existing holders of capital by making it impossible for anyone new to join their class. It would also lead to all manner of bizarre forms of perks, kickbacks, and other non-monetary forms of lifestyle benefits and power, sort of like you see in the ruling class in communist countries.
So why did the UK also see the emergence of a strong middle class which has since eroded?
It was a bit of a disaster.[1]
Then he saw the Autobahn in Germany.
He also was Sir Arthur Tedder's superior at SHAEF when Tedder developed the strategy to destroy Germany's railroads.[2]
It is not surprising that Eisenhower was committed to the creation of a national highway system.
Then you have him the reins of the massive public works machine built by the new dealers, and the interstate highway system is practically the inevitable result.
[1] https://en.wikipedia.org/wiki/Transcontinental_Motor_Convoy
[2] https://uncpress.org/9780807858509/the-collapse-of-the-germa...
Buybacks are not necessary. The same exponential compounding happens just as easily with dividends or even just growth and a trusting valuation. It's the compounding r>g that is the problem.
I think almost every new business will have at least one person whose job is to deal with the technology. There might be fewer of them, and they might be a weird amalgamation of sysadmin and web developer, but there will still be jobs for nerds like us who actually like doing this stuff.
vs
"I want to pre-sort during idle time so that I can effectively implement a binary search"
You can tell me either of these and get the same product. You can pick one of these and tell an LLM to do it three times and get three different products.
What's going to change is that we'll have to focus on the actual hard problems of software development: making sure the software does what it's supposed to, isolating nontrivial bugs, finding ways to make it perform better.
These are exactly the things LLM's outperform all of us on.
They'll shove out a pattern quickly, but no matter how new and magical the latest model is, they will just put out some tweaked variant of existing designs. I find myself treating it like a junior-mid engineer. "Add rdtsc-based timers around major parts of the code to whittle down what's taking too long." "Use that to figure out where the transfer amount is way too small to justify the time taken without copying the payload." etc.
But concurrency bugs from large LLM-developed codebases are a major issue. They constantly find bugs they wrote. I eventually told it to start writing formal models of the concurrent uses and to (a) validate the models; (b) verify the code implements the models; to make it work.
So yeah, I stand by my words. For simple commands and tiny codebases, it /can/ do those things, but even small projects (50 KLOC) are a mess of making it fix its own mistakes, reconcile basic logic in requirements analysis, and everything has to be taken over with a fine-tooth comb to make it any good.
If you've tried to build anything nontrivial with LLM you know this just isn't true.
Why is much of our legacy software suddenly failing us if we are so much better at finding intricate bugs than the LLM's? Why are all of us easily capable of speeding up our software by orders of magnitude in small amounts of time utilizing them. Yes, they are better at verification as well.
Examples are: Note taking apps, voice recording apps, calendar apps, video games, etc.
If you throw an excessive amount of money at the problem yes.
Wages might decrease if the work becomes easier and more people are able to do it, but I'm not sure jobs will actually go away until true AGI.
We'll be fine. Move to a lower cost of living area, start a business. There's millions of companies in need of software expertise.
All this moral panic about software work is unseemly. Maybe the easy "learn JS/frameworks" path to wealth is closed. Whatever. Good, actually.
We'll be fine, we're the most prosperous worker class in the history of working.
Besides the job never was churning out code and if there was a bloat of such positions it was only ever a temporary distortion.
Right now there are mass layoffs in the IT industry partially as a correction to the pandemic hiring spree, and also to redirect capital from salaries to AI initiatives.
It is fashionable right now for tech CEOs to pretend that enormous amounts of work being done by humans is now being done by AI, but none of these are companies where the velocity at which code was written was ever the main bottleneck in getting products out the door. They’re mostly just lying to the market to pump their stock prices.
The software field is overexposed because there are a lot of people earning hundreds of thousands of dollars a year to add fields to CRUD apps. These folks are in enormous danger, but when the dust settles the skilled professionals solving difficult problems will do just fine.
I guess I kinda resent your implication that the "skilled professionals", the worthy, you know, us, will be fine. I think the real danger is for the people who care. And for the rest of us who rely on the people who care.
Just look at the job offers and the responses they get on here. It's all trivial crap.
Part of this, I think, is because the people who had the inclination and ability to pick up programing have, for decades, been nudged into programming from whatever job they had before, by comp, co-workers, managers, and opportunity availability, so they’re already heavily concentrated in programming even if they’re perfectly capable of doing other jobs (perhaps with some brief re-training).
I think other roles need to fear programmers who studied a new role for a couple weekends more than programmers need to fear LLMs. If we’re bumped out of full-time programming… guess where we’re going, and what we’ll be way better at doing than them?
there is shortage. You can just jump on problem you want to solve, because distribution channels are saturated/occupied.
On the other hand, just two years ago they were practically useless for anything non-trivial. In that time, they went from "might be cool someday" to "can do most of the work, but still needs handholding." What are they going to look like two years from now? Five years? Ten?
So much depends on where we are on the S-curve. If we're getting into diminishing returns, then we may top out at a point that isn't too far beyond what we have today, where they're highly capable but need that human judgment and oversight. If we're at the bottom of the curve, it's hard to know what the future would look like. I have no idea which one is closer to being true, and I don't think anyone else does either.
I think we actually are there in some areas but the bureaucratic/social structures are actually blocking it. If you asked me 2 years ago I would have laughed.
But the thing is that with AI the number of people willing to pay you to solve their art problem has dramatically fallen. The total number of jobs in art in general has plummeted. The opportunities to make a living or develop yourself with art have similarly plunged.
So there will always be another problem but that doesn't mean there will always be someone willing to pay you to solve it if they can get by without you.
just happened on a client call yesterday. our work had been postponed because a new system was being introduced. our routine calls kept being cancelled and we started to worry. the customer finally showed up after a few months and told us the new system was demoed and "was clearly vibe coded. everything went wrong. they were asking claude questions during the demo" so were back in contract again
> reliably deliver them better than 50% completion
everyone has ideas. being able to program helped separate ideas from execution and i thought llms would remove that barrier but there seems to be some ingrained half-ass mentality in the self-proclaimed "idea guy"s ive met which prevent them from finishing things
Anecdata I’ve seen this too a few times. Major issues where a significant number of engineers had to stop important product work to un-jankify something a non technical person created. In every case those product people felt the heat so bad they basically stopped.
Are you solving a shared agreed upon real problem or just doing leetcode on arbitrary meaningless data structure?
There will still be jobs but orders of magnitude fewer jobs.
Closer to reality that existed before the 08 crash and ZIRP fueled ponzi schemes fueled the web SaaS mania.
US doesn't need 4+ million software devs
World doesn't need 50+ million software devs
...when big tech is just shipping black boxes that boot models to handle local inference
It just needs enough jobs to solve that big tech production pipeline
Seems like copium to me if we're being honest.
Why would anyone pay you to solve the problem when they can just throw it at a cheap LLM?
And if the answer is that you have some sort of special taste or domain knowledge, once your solution is out in the world how are you going to stop someone from using an LLM to copy it and then undercut you in an immediate race to the bottom for everything?
My grandfather worked at NASA, and my impression is that a significant portion of what he did was actual calculation. Nasa still employs people, they just don't do that particular work anymore.
If that happens I guess the whole thing hinges on how powerful and user-friendly open source models become.
If the answer to that is "yes", then there's literally no point in anyone hiring anyone for any knowledge work. And that applies to other businesses. Why give a SaaS company recurring revenue if they're just a thin wrapper around Claude anyway?
The answer I think is that few companies are so thin. Most offer something else, some other kind of moat that has nothing to do with problem solving, or some kind of lock-in.
But there are definitely some (mostly small) companies that only exist to sell relatively simple specialist software to customers that don't have their own developers. Their moat against competitors was the relative scarcity of good developers, and the difficulty of motivating them even if you could hire them. In the post-LLM world, those small companies suddenly have competitors for the first time...
It’s not possible for them to grow as big as their boosters want them to. There would be so much resistance, and if that resistance fails, then we’d already be in the middle of techno-dystopia hell anyways, where money wouldn’t matter.
government will be their puppets, and what can I do then?
if your scenario starts happening, we break up anthropic and openai for antitrust reasons, or people do other things so that they don't become kings.
I don’t know what moats we can have at this point. I’m trying to build a side hustle with vibe coding something because if it’s a backend only you can access at least you have a tiny moat. If you’re just prompting slightly more cleverly at your employers Claude account you do not have a moat
i also wonder if there's something in your employment agreement about the things you make on company time and who they belong to.
(1) LLMs can solve any human problem; (2) that #1 would be a bad thing if it were actually true because some SWEs can't get the same job they had before
if the LLMs are that great, tell them to find the SWEs a new, better job. if they aren't, and LLM+SWE is > LLM, then #1 falls apart.
Certainly not every problem, but enough problems that we currently consider economically valuable to cause a very large negative disruption in labor.
> (2) that #1 would be a bad thing if it were actually true because some SWEs can't get the same job they had before
I wouldn't say some SWEs I'd say almost all SWEs.
> if the LLMs are that great, tell them to find the SWEs a new, better job. if they aren't, and LLM+SWE is > LLM, then #1 falls apart.
Again, I don't think LLMs will solve every problem. Just enough problems that they put about 80% of all knowledge workers out of work in a relatively short timeframe, which is going to be economically disastrous in a way that expands far beyond just knowledge workers.
Now it seems like it's the sunset of this way of making a living. I think people who manage to move up the food chain a bit will succeed but its hard to advise any junior in the industry exactly what that looks like.
So whenever someone says "We are laying off SWE's because AI", I have no idea what they're talking about. Just the idea of taking on the mental load of getting rid of engineers to direct more agents shipping complex features to 100's of thousands of users makes my hands sweaty. I am so glad I am still able to hand off the technical direction, architectural care, follow-up, unblocking decisions etc. to a real accountable human.
The sunset calls on this career seem far too preemptive.
For them, keeping their level of ambition and dropping staff in favor of AI makes perfect sense.
So a company that can't/won't grow, lays people off and blames AI? Checks out I guess.
What is the common thread?
These jobs depend on some combination of physical skill, human connection, trust, judgment and accountability, and such jobs are out of reach of AI.
Then they will be able to do more and more, and 1 nurse will be able to do the job of 4-5 2020 nurses.
Until that nurse is a dude from Mexico [where im from/based] giving instructions to all the robots.
There are some professions where human touch is in demand (this is not about capability, but about user experience in sense of human interface)
Of course the elderly and frail and people too afraid to touch something lest they break it will still call an expert, and original build and installations will need those trained, but I could see vast swaths of less technically skilled people opting for a go at routine repairs which could really hollow out the demand.
And where do they get the money to pay for this when all of the knowledge work is gone?
None of these end up really being out of reach of AI if Anthropic can justify this valuation. They're just slower to roll out.
Arguably, software has times that require physical presence, human connection, trust, judgment, and accountability.
Depends on what the software does and it is application.
If its a tiny video game, journaling app or camera app that requires little human interaction: Then it isn't important.
If it is infrastructure-level, critical and time-sensitive and deals with highly regulatory industries such as finance, crypto, healthcare, legal, manufacturing and has a chance of risk to the human then that software will continue to be used and pay will remain high.
LLMs are hugely suited to code because it's token heavy and the compiler is the RLHF verifier.
The LLMs are still very bad at concepts.
What we see in code does not translate to other, arbitrary domains.
Legal - yes, due to text heavy content with tons of great examples as well.
Everything else ... not really. At least not LLMs as we understand them today.
I'm not the only one that had that idea, they're advertising these things as features now.
I don't know what's next, but keep in mind that today's LLM's are the worst they'll ever be.
They are going to be 'better than what they are today' but I don't see obvious unlocks.
We need more abstractions on top of them.
I wonder if we are seeing them maxing out llm model capabilities now...
That's like saying in 2011: "today's Google search is the worst it will ever be."
Enshitification is coming my friend. We're just now starting to see the kinds of refusals that LLMs will ultimately end up with to protect IP owners and the status quo. We'll look back at this time and marvel that LLMs used to be able to reverse engineer software, or able to create derivative IP works.
Nearly everything will (eventually) be impacted.
Therapists? People are using LLMs for “free” help.
Physical trainers? Just upload a video of your workout to an LLM for tuning advice and other exercises you can focus on.
Appliance/vehicle repair? Just DIY it with an LLM.
Teachers? Just teach yourself with an LLM.
I could go on.
To be clear, I’m not saying LLMs are better at this. But just saying LLMs are impacting nearly every field.
EDIT: just to clarify some of the responses talking about full replacement -- that's not my point here. I'm talking about impacting, not replacing. Software engineering likely won't get fully replaced (anytime soon), either -- but it's definitely being impacted.
Sure, in THEORY we can do all these things, but I suspect this fear is somewhat borne out of the same error that mashes us estimate a task will take 6 hours and then it takes a week.
We systematically underestimate the human effort needed. How many non-technical, non-interested people are going to be looking up DIY for fixing their car, buying the needed equipment, developing the skill, etc, really?
While they’re doing that, they’ll somehow also have time to produce their own movies, write their own indie games, etc, etc.
Fast food is available; people still go to restaurants. I can sketch on a piece of paper, but I still buy art. It’s technically not difficult to build a chair, but I still buy furniture, etc.
It’s going to have a massive impact for sure, but we have to temper our expectations and ground them in real, practical limits.
Definitely some interesting years ahead. Fingers crossed.
I think this is a very good template for what sort of skills will survive the LLM-apocalypse and what won't. LLMs are the great homogenizer. Any type of knowledge work where there is a broadly-accepted set of best practices for how to do it is gonna get outsourced to an AI.
But AI is terrible at doing anything unique, personalized, and based on your own desires, because it doesn't have your desires. It's building on a corpus of shared knowledge, but there is a large set of knowledge that is unshared, and it can't do anything for that. A large portion of business, for example, is secrets and lies, and an LLM isn't going to be able to help you much with that. At its best it can set up the frameworks and tools to help you keep secrets and tell lies.
And then a large portion of life is personal preferences, the things that make you unique and distinct from every other human, and an LLM can't help you with that either. There is a reason I don't have a nanny raise my kids, even though we could afford it. If I did that, it would be the nanny raising my kids, which defeats a lot of the point of having kids in the first place. LLMs have the same problem.
I keep trying to nudge them to be relevant with specific examples of a few shows I had watched that were very popular at the time but now considered "old/boring/not remastered" so don't make it into the Best X Shows of XXs lists that emphasize modern tastes. But the recommendations were either the same things I already knew about, or were just wrong/irrelevant for various reasons.
Recently I stumbled onto a Reddit post from years ago with a similar request to mine, and both the question itself and top answers were a goldmine of suggestions that were never recommended and I hadn't heard of. I knew within seconds of looking at the IMDB that they would be exactly what I wanted, so it's not like the AI recs are failing just because of idiosyncratic taste. They just don't seem to search from a deep enough info source, have been RLed into conformity, or some other disconnect between intent and execution.
My wife has a deep expertise in modern therapy* and I saw her use ChatGPT. The difference is between "I want to fetch better products" vs "modify this index to use embeddings with model XXX"
*people's mental model of therapy comes from the movies, books or TV shows. Modern therapy doesn't look like an episode of Sopranos
Nothing makes sense or is computable when human labor/knowledge has little to no value. It's a bit like an economic event horizon
Actually makes everything a lot simpler, economically.
Economics is the study of how to equitably distribute scarce resources. When everyone's contributions are equally meaningless (because hey, we made this do-everything bot that will do your job, no matter how much you were worth in the before-time) then the whole convoluted, often inequitable distribution scheme is paved over with the brutal 'your share = [resources] / [number of people]' equation.
Ownership is a social construct, backed by the rule of law. If we rapidly take away the ability for the masses to be able to afford to own things, there isn't any magic preventing them from deciding to not honor that particular construct anymore.
The bigger question is, if that does come to pass, will the AI overlords still have enough power to enforce their ownership claims the old fashioned way -- with violence?
I'm afraid this is going to get very ugly.
Cars are too expensive to not repair, but the complexity of car repair is way too high. Even with an incredibly specific and accurate video, even one made lovingly by a knowledgeable mechanic, most repairs are beyond most people’s physical skills.
I do mostly agree with your thesis, still. Being closer to education than many other fields, though, I remain unconvinced that AI will handle children well.
College, on the other hand? For many reasons, colleges should be an existential crisis right now. Nothing about college makes sense from cost to how AI-cheating changes things, to job prospects for most majors.
And the results are poor because the LLM does not have a face and is not reading your facial expressions or hearing the tone of your voice. We communicate with more than just the words we speak.
> Physical trainers? Just upload a video of your workout
...lol
> Appliance/vehicle repair? Just DIY it with an LLM.
"Officer you can't arrest me and impound my car, I did what the LLM told me to do. It's not my fault that a wheel fell off and my exhaust is an eerie green color!"
> Teachers? Just teach yourself with an LLM.
See above about therapy, there is more to teaching than spitting out words.
> I could go on.
You could, but at what point will you get to the few lines of work that are genuinely going to disappear because of this technology? Taxi drivers are the most obvious, and least discussed, example -- self driving cars will eliminate that entire profession over the next few decades.
The results are more 5* reviews than 1*: https://apps.apple.com/us/app/feeling-great/id6450066153
I don't disagree. But I think people will still use it. We've seen how people Googled stuff to get help in this area, and it's 100x easier with LLMs.
>> Physical trainers?
> ... lol
I disagree here. I think LLMs are decent at getting someone dumb to working out, like me, from zero to establishing a decent workout plan, and making adjustments as you go, if you give it enough context on your health, and how things are working for you.
Again, I'm not saying LLMs are better at it, but I think it's arguable that people will figure out that they can use an LLM to get started pretty easily, where they used to go to physical trainers.
> "Officer you can't arrest me and impound my car, I did what the LLM told me to do. It's not my fault that a wheel fell off and my exhaust is an eerie green color!"
I was already doing a lot of my own home/vehicle work via YouTube. Now I almost never taking my vehicle into the shop (for what I think are relatively small repairs that aren't safety hazards, like fan repairs, electronics, etc.).
Additionally, I now upload repair estimates to an LLM, and have it push back on things that I can either do myself, or do somewhere else and save a lot. And this has, indeed, kept me from spending as much for vehicle repairs (cutting into their margins).
Similarly, I've uploaded pics/videos of appliances bugging out on me, and have been able to fix many things myself. I used to already do this to an extent with YouTube, but would still have to figure a lot of things out myself. Being able to upload pics/videos changes things quite a bit.
>> Teachers?
> See above about therapy, there is more to teaching than spitting out words.
Again, agree mostly on this one. But the value of higher education has already been in question for a while, and I think LLMs are only pushing this further.
I don't think teachers/professors will get replaced (anytime soon), but I have kids and know parents of kids that are using LLMs to help them in ways that they used to rely on teachers for.
> You could, but at what point will you get to the few lines of work that are genuinely going to disappear because of this technology?
I'm not necessarily talking about full replacement. Agree that this is very possible/likely with taxi drivers. And I do think some fields will also get replaced, some that we may not even consider now (would we have thought 15 years ago that taxi drivers would get replaced in our lifetime?).
But my point is that LLMs are going to, at the very least, have an impact on many non-digital fields.
The real danger for teachers is society deciding people don’t need to be educated. Although that would be a problem for all of us actually.
Good luck replacing elementary / high school teachers with LLM's. Their job is about so much more than just teaching.
More workers competing means lower pay. Less good paying jobs means less customers. Less customers means less jobs. This is an over simplification, but is one possible future.
I'd be happier if my prediction turns out wrong.
You got 50 years, but many people haven't. There are so many people out there in the prime of their careers right now, I genuinely worry about them and what this does for their livelihood.
If I could pick an age for this all to happen, it would be when I was 16, just starting to get into computers. I think I would enjoy the hell out of the new world, getting in on the ground floor at the very start of my journey. But my 2nd pick would be my current age, at the tail end of my career. I get to see this crazy transformation, but I've already reaped most of the rewards of everything I learned that is potentially becoming quickly obsolete.
I really don't envy those mid-career, who have put tons of work into their craft but have yet to reap the majority of the rewards (which tend to be stacked later in career rather than evenly spread out.)
Two examples of the lengths people went to for performance - 1) the networking hardware, terminals and everything were special airline flavors that sent data in 6 bits rather than 8, just to squeeze more data onto the wire. Is you look over someone's shoulder at an airline booking screen even now you'll see an all caps look - that's why. All caps means less bits.
The second was a little before my time but the geniuses who maintained the assembler code were very close to the hardware. Some used programming tricks to account for the rotational speed of the DASD, so the software would be ready to pull bits off the hard drive just as those bits rotated around and arrived at the read heads. Insane optimisations.
I'm out of the area now so I didn't know what the state of the art is, presumably there's not any IBM assembler still lurking but it certainly wouldn't surprise me.
The changes taking place now are not unexpected, and not unwelcome, but long overdue.
(Of course, whether that justifies assigning a significant chunk of the US stock market's value to Anthropic is another question entirely, and not as easy to handwave away.)
For anyone who nabbed the outsized returns while they existed, be thankful you got them at all, and hopefully you saved some of it. Due to geographic barriers to movement, 96% of humanity never had the opportunity to work in tech and profit off the supply/demand gap. LLMs arbitraged that gap, and now we can diffuse these capabilities throughout the entire world at extremely low cost.
AI is a giant compression and de-duplication algorithm on white collar labor, and it was needed. We've had an over-production of white collar elites problem for a while now resulting in policy errors in most developed economies who completely ignored building in the real world.
Don't worry, humanity still has more than a few problems to solve. There's infinite jobs to be done. The nature of humanity is that our expectations always grow to exceed our current capabilities. And with AI, you don't need to sit in a classroom for 6 years to get up to speed on the next job to be done.
If the US government allows it, of course. It just banned my country from buying data center nvidia chips.
>Don't worry, humanity still has more than a few problems to solve. There's infinite jobs to be done. The nature of humanity is that our expectations always grow to exceed our current capabilities. And with AI, you don't need to sit in a classroom for 6 years to get up to speed on the next job to be done.
You sound like someone who's overly cocksure that you've figured it all out. By It I mean, how to organize society. What Policy Errors would not have been made, in your scenario?
- The entire west relocating all key assets like factories, mines and steel production to the developing world/China (factories? yuck)
- Neglecting to build homes in basically all western countries for the past 18 years (construction? yuck)
- Virtue signaling going green by pushing energy/fertilizer production offshore to less regulated countries and then importing it back at even higher net carbon costs (fossil fuels? yuck)
- Getting overly hysterical during Covid due to WFH being viable for the white collar class -> shutting down the economy -> requiring massive stimulus that mostly just caused inflation and inflated asset values
- Decades of under investment in hard infrastructure and instead investment in legal red tape and white collar bureaucratic process
- An entire generation over-educated in Microsoft Office Suite vs. physical reality, such that much of the capability isn't even here anymore. (See the firms contracted to work on CA high speed rail for more info, or imagine the US trying to build the Hoover dam today)
- Post-financial crisis, inflating global asset values by keeping rates too low for too long, causing the whole K shaped economy thing and current discontent
- Over fetishization of higher education -> government guarantees of college loans to any borrower regardless of ability to repay -> runaway inflation in the cost of US higher education
I can do more if needed
There's a guy running a stall in Hanoi right now laughing at you claiming you're a humble working man while making 30X as much as him.
As you get older, I recommend traveling more internationally, you begin to get a more accurate picture of your relative stature.
While you and the rest of HN are busy exciting yourselves with the online bolshevik revival, it would be wise to first examine where you sit objectively relative to the rest of humanity (based on the analytics, if you're on this site you're in the top 2% to 0.001%).
If we're going to normalize seizing the property of individuals to pay for your fully automated luxury socialist dreams, unfortunately we won't be able keep the line drawn at just billionaires, or at just millionaires. Sorry, the math doesn't math without taking a lot more of your stuff too.
What reason is that?
But of course, what does an old crusty fart like me know and what could I have possibly learned over 50+ years of coding that could be at all relevant today? IKR?
You hate this? That's on you. But doing the same thing over and over and expecting a different result is insanity. However, if you saved up before you were outed as an imposter, I can see that working out in the long run. IDIC and all that.
This is just a tautological advice. It is equivalent to saying "be successful, then you will be successful".
Also, have you seen the batch of CS majors coming from non-top-tier universities? I don't see how they will be able to compete with LLMs.
Most companies that are running the “how many SWEs can we fire before it all breaks” experiment have learned it’s a failed experiment - this validated by every single major company/product outage where their uptime went below 90% this year.
I can’t help but think this anxiety and fear is just marketing. chill out everyone
So we are not waiting for the tech to get good enough, we're waiting for humans to understand and integrate it.
I don't work in tech, rather manufacturing, and our company is barely using any AI (or using it, but only like AI mode in Google) despite me knowing that we could easily use it to cut heads. The automation we are leaving on the table is huge.
But I'm not gonna be the one to go to my bosses and say "Hey, you want me to show you what we can do with SOTA models and agents?". For now I'll just stick to using my accounts to leverage my job.
Hate to say it but most humans that couldn't code before because it was too technical also can't understand enough about the underlying systems to competently integrate AI.
I actually did. It doesn’t really change much IME. I just did it to come out in the open about AI use and start the slow adaptive process. It’s been a year now and just this week some colleagues are starting to act surprised by what is possible. I’m in software dev for crying out loud. This world is teeming with stingy sysadmins and neckbeards proclaiming it are all toys and soon everything will be back to normal.
If the culture doesn’t pick it up automatically there is some mixture of cultural factors holding them back and just knowing about tools ain’t it.
I think we were a bit smug as a group. Sure we were disrupting EVERYONE ELSE and saying how disruption is just the natural way things work, everyone should adapt, etc. Now that we are the ones having to cope with the same thing, suddenly all that goes out the window.
I know I’m speaking in generalities here. I’m in the same boat as everyone else. I do argue that for those of us that have been into this since we’re kids, we grew up being told we were the exception. OUR jobs would never be obsolete, we are the future; we looked at people who refused to adapt with judgement. Now it’s our turn to adapt.
Indeed. I always call it out.
It was fine to disrupt everyone else. It was fine to replace all those laborers. If they complained, we just told them it was the implacable march of technology, told them to learn some new skill.
I'm a doctor. Putting me out of a job is and has always been one of the wettest dreams of hackers. It was palpable in every thread about medicine on HN.
But now artists are getting automated and software developers are getting replaced. Watching people here treat it as a crime against humanity is just too much for me.
I'm a hobbyist programmer. I mourned the loss of my hobby. Working on my little projects just feels pointless now, like none of it matters in the grand scheme of things. But I do hope that we're working towards some kind of post-scarcity society where the economy itself has been eliminated.
Unless we can efficiently transmute atoms into other atoms I don't think this is happening anytime soon.
You can continue to code for the joy of doing it. If you enjoy the end result more than the coding process, you can work on crafting the best end result that you can. So it’s kind of a win-win for a true hobby, as far as I can tell.
A lot of people seem to think they can pick up an LLM and suddenly write software solutions as good as a seasoned SWE. But it turns into this terrible mess that's thrown out all engineering fundamentals, solving the symptoms instead of the cause.
If I were just writing boilerplate code with an LLM all day I'd be very scared for my job too.
Here are few anecdotes:
1. I've a friend who owns a PT (physical therapy) business. He's not even a tech guy, but uses claude to manage most of the business aspects, and he has told me repeatedly that he would've to spend thousands of dollars to get things done, and instead he pays 250 per year.
2. Another friend is smart, and is a jack-of-all-trades type in the software/hardware/firmware business. He is solo bootstrapping by doing all the things from pcb to webapp. He doesn't see hiring anyone for the niche business.
3. A 65 year old friend of mine who never in any tech career was able to build website for his side business (that website looks really good btw) and he now talks about json and ocr.
You are the one who is out of touch with reality.
I've got a friend in this situation too. He's going to hit a limit of what he can do with agents. And whatever agents can't do, his business can't do. We are going to find the shape of that soon.
Normally hitting a limit is where you would hire someone, but building organizations with heavy agent use is hard. AI is very very good at shifting work onto other people and business units; that's what all the complaining about review (and not just code review) is about. None of that responsibility shifting happens in a solo setup.
Build-or-buy decision is still very relevant, and one can go far with affordable PCB assembly houses, 3D printing, and many others. For a solo person, today's environment is 100x better than, say 10 years ago. So, niche businesses have better chances of thriving, all else being the same.
> I can’t help but think this anxiety and fear is just marketing. chill out everyone
I'm building a mobile app using flutter. It's got to work on iOS and android. I can promise you, if I did not have years of experience, and I did not follow a regimented, checkpointed interaction flow with my agent, this code base would have turned to shit within a day. I've had to draw on my experience to define the architecture, correct missteps, prevent just wild flailing, and just generally keep things on the rails.
Your example 2 is the closest match, where someone with experience is driving the result carefully.
That's because your app has constraints. Why would a PT business operations and website for a side hustle need to be any more complex than the owner needs it to be? Your app complexity is probably trivial compared to say Google Maps, so your tooling is designed to match your app needs, not that of Google Maps.
Just to pretend that you are the only guy who can grok, and non-technical business people are incapable of curiosity, hard work, creativity and have the ability to manage their business and IT is more of a reflection of your mindset.
PS: you are using llm to build the app too ;)
In other words, the verdict is still out on whether we'll see a decrease in jobs in the long run or not. Personally, I'm not very optimistic but, as I tried to illustrate above, I might very well be (and certainly hope to be) wrong.
Yes, we don't know the full effects, and sure things are in a flux, but let's not forget the convenient truth that the software engineering itself has required less and less expertise over the years due to programming languages with better abstractions, faster hardware, toolkits, frameworks etc.
With tons of jobs and better pay, lots of folks moved over to the web side since the level of expertise required (compared to a firmware engineer or asic designer or a dsp guy) is quite low and while the pay is way better. During the dotcom boom, a friend of mine who was a furniture salesman, took a java course and ended up with a software job.
Now, if the opportunity presents itself and incentives are strong, and barrier to entry is low, wouldn't more people also like to start/manage their business? I would think yes.
there is massive lose in revenue for a business and or infrastructure that people rely on goes down
if you’re working anything in this way, no one is just trusting a clanker. some thought leader might say the world is trusting an AI, but it’s not true. every company internally that is more than 2 people isn't trusting a bot, and it's well understand that it's a horrible idea (even by regular bullish AI business heads)
so yes, I have the bot help me write a basic website for me, and it can one shot the typescript express app-layer boilerplate for me, but if you’re serving even the most basic app at scale with consequences, the bot isn't mitigating the concerns and engineering decisions that go into this work.
a concrete real world example: literally msft outlook went down a week ago. it was out for god knows how many actual hours. it still might be partly down tbh. this is technology that is “solved” and is dumb simple in terms of complexity. their product was out for 48+ hours. and the status messages shared publicly by the engineering team did not only convey incompetence, but to paraphrase, the message they sent out was along the lines of, “We think we found the source of the problem, but unsure, the problem might be larger”
how much you want to bet they had a team of 30+ engs running around with claude asking the equivalent of, “please fix outage, no mistakes”
I’m still not convinced the outage is fully recovered because I still get errors.
point is, yes there isn’t a giant 100 football field wide moat between regular people and programming anymore. smart people can do basic stuff and less inclined people can build a static website now. if you’re worried about that, go study more and you'll learn the depths of this field go WAY beyond typing code
EDIT: typos
The implication is far reaching. Lots of non-tech businesses which relied on expert help can now pareto slice it and hire experts on need basis unlike before. Saving money and increasing profits can be a big motivator to own many of those tasks too. Lots of people who were cosplaying as experts on non-tech people will have less and less opportunities.
Previously, being a knowledge worker could get you far, now the knowledge worker has to level up, and if opportunities are limited how could it happen? Even then, more experts (practice, not theory) is not a given, considering the bell curve effect.
This is a big problem. When this kind of thing gets launched, its all fanfare, this is the future, get on the boat or die.
But then it fails miserably and its like it never existed
80% of people only use 20% of features, but if you build the 20% solution you only attract 5% of the userbase because they all use different features from the 100%.
I am starting to wonder if or when we’ll see that shift. If companies lay off employees because AI can fill the gap, the I think a lot of people expect those savings to be passed on. Right now, that doesn’t seem to be happening.
For my business the threshold is definitely getting lower for justifying building things ourselves.
That seems like it would be such a paradigm shift that the second and third order effects are so crazy and nobody can predict what that world will actually look like
This is engineering thinking, not business thinking. The question is, did the uptime drops cost the company more money than the firings saved? If the answer is no (which for megacorps like Microsoft it is) then the experiment was a success.
Sadly more than likely they will just apply more and more AI to try to fix it, while others use AI to attack it faster than they can defend.
but at the same time I think the companies are pumping this idea that SWE is over and stoking anxiety and we have yet to see how bad things get when everyone can write software without thinking about it, we will see what insecure slop gets produced and it's going to be huge.
The issues with AI security are quite big as well, people who will rely on just an LLM to be the intelligence will get burnt I am sure especially in operations, prompt injection is still an issue, would someone running LLM's notice an provider endpoint getting hijacked and acting maliciously and inserting malware into production servers.
The future is going to be a strange one for sure but being able to turn your brain of and get your LLM to run you servers write you servers is not going to work alone.
I have also yet it write a good fairly novel program without babysitting (quite trivial babysitting sometimes) but stuff that is bread and butter for us, for those who don't have those skills, will they really know how to explain the issues they may run into that requires creativity, for example telling it to hack something together with curl and jq, knowing why a certain thing with the sandbox is not working.
I think these issues are not as easy to resolve with just better tooling and they will soon get annoyed at prompting.
Additionally I think Gas Town shows the sort of almost mentally ill mentality that is baked into some of these tools and ideas, Everything I read of Gas Town was so strange and the amount of tokens used to anthropomorphize agents.
He wrote a series of blog posts in which he talked about how Gas Town failed and blamed it on Opus 4.7 changes causing all his workflows to fail. Then he announced Gas City as a rewrite.
At some point after that he seemed to give up on the whole idea and instead just started working on reviving an old graphical MUD style game he made where I guess now he spends $10,000 a month on agents to build a game that currently has 0 concurrent players (4 concurrent 24 hour peak) on Steam:
https://steamdb.info/app/1541710/charts/
The game is available from other platforms than Steam so maybe he's got a bit more users than that across the platforms, and his stated plan is to bilk "whales", so maybe that's working out for him? But it still seems like he's lighting massive amounts of money on fire in what appears to be a prolonged state of ai psychosis.
Must be nice to be that financially secure.
if the hype was all true, i could just prompt the llm to make me money and it would be able to do it, but it's not possible
My friends at FAANG companies would disagree here. If anything, it seems to be the exact opposite.
Sure, it's not a end-to-end 0 touch factory, but a majority of folks I talk to these days that I consider good engineers aren't writing _any_ code anymore. They do that as a hobby outside of work.
Is it? It looks more like a forgettable moment to me. We're talking about an IPO, how many headline-making IPOs can you actually remember?
> as a SWE the gravy train is basically over in my view
That is not a result of the IPO even if you are right, and I have my doubts.
> they figured out how to make a program that does the job
No they figured out how to make software that makes the job easier. How is this any different from the introduction of optimizing compilers or type systems or any of the other software tools we all use? It makes programmers more productive, which means programmers can take on larger projects.
Coding agents have limits and the managers who forced their SWEs to use them are starting to figure that out. Using agents to both write code and review code sounds cool in theory but in practice it is not such a great idea -- it means that nobody actually understands the codebase and therefore nobody can actually debug the code when things break. No, you cannot just assume that you can also throw an agent at the problem when things break -- because the breakage might mean you lost access to the agent or the agent lost access to your system and cannot help you.
> difficult to grasp how much life will change in the coming years
Not too difficult to grasp. Shareholders will demand returns and Anthropic will have to deliver those returns. All the sky-high dreams about RSI and the next phase of human existence are going to evaporate and profit is going to be front and center.
Another possibility: Anthropic fails to survive the next recession and the companies that did not allow their employees to become dependent on Claude will be the winners.
That seems highly unlikely, given that there are countless open-weights models available.
If nobody knows what your code does anymore, if it turns out that the tradeoffs AI makes metastacize over time, if it turns out that attention to detail ACTUALLY matters to the functioning of your systems (which it always has), then we are barreling for an extremely rude awakening. The water is just starting to move in the kettle - it might turn to a boil soon.
The only thing that makes this IPO notable is the dollar amount, and it is neither the first nor the last IPO to be notable for that reason.
Exactly...? This is the only one I'm sure to remember.
Engineers are now more productive than ever. Increasing productivity drives salaries up. Secretaries are paid more today than they were when they didn't have computers to use. Why wouldn't this be the same?
The gravy train for shoemakers was over. What remained was a small number of shoe designers and (initially) a large number of unskilled factory workers. The same thing happened with many other skilled jobs.
But much will depend on how much professionally made software the world really needs. Demand for software could be a lot more flexible than the demand for shoes.
the market for SWEs will continue to be K shaped where the top of the K are SWEs that can solve ambiguous business problems and take accountability
in aggregate there is not as much need as we can create now
It was trivial for 1 Rails dev in 2010 to do 10-20x more than a Java dev back then.
Today’s world is an interesting dynamic but I’m pretty excited about it.
Java itself was a much slower standard workflow. Dynamic languages in general, like Ruby, Python, Perl or PHP were much faster to work with than the burdensome recompile workflows. Think instant feedback vs waiting for your CI to finish after every single save. The speed of iteration was critical.
Beyond that, when the speed of development is slower, more people get hired to do the work. When more people get hired to do the work, there's more communication relay as the team size grows. Teams coordinating between each other. Decisions going to committees. Middle management layers.
You'd routinely see companies with 3-5 Rails devs outperforming shops with 50 Java devs. Communication and iteration speed was huge. The other piece of it was the convention over configuration aspect because it removed so many things you'd have to discuss and make decisions about as a group which further accelerated communications.
10-20x multiplier isn't an exaggeration at all. If anything I could be underselling it.
On one extreme, a solo dev could prompt a frontier SOTA agent with not much more than "Build Uber", give it the necessary hardware/software/permissions to build itself a useful feedback loop and act autonomously, put it in goal mode, and let it cook, and as long as the tokens kept flowing they'd eventually end up with something reasonable in actual orders of magnitude less time than it would have taken that dev to build the same thing — especially if you're counting calendar days rather than hours worked, and extra especially if they wouldn't have had the luxury of investing full-time hours into that particular thing. At the opposite extreme, BigCo full-time SWE #5000 tasked with an extremely specific ticket for a critical revenue-generating brownfield system might experience even "20%" as liberal, and correctly observe that writing code was never the bottleneck for their job. And then there's the massive spectrum of circumstances in between those two ends.
None of that means that the gains aren't real, or that anyone in particular is wildly exaggerating in either direction. It just means the gains aren't evenly distributed.
If you're allowed to use the newer libraries and tools, it has actually been a perfectly fine language to write software in, even relatively quickly, since ~Java 8 or 11.
I have had elaborate fantasies of what I would do to the designers of JavaEE if I ever found them, and I had assumed that because writing JavaEE is so unbelievably horrible that the entirety of the platform was like that. Years later I had a different job that was much more amenable to me using "modern" Java, so I used Vert.x and Disruptor and properly learned how to use NIO and and I actually had fun.
I think Java has been the "regression to the mean" language primarily because it's taught to frequently in schools and it allows you to write in a very verbose and irritating style, meaning that it was a great fit for mediocre-to-incompetent engineers who want to look like they're doing a lot.
sigh
Not some greenfield dinky todo list startup project with 0 users
Why do HNers always talk like these are the same as real products
I tend to be more on the "Jevons" side of things, where I do think that demand will more or less expand. There are lots of projects I've been putting off for years because there would be an aspect about it that I hate doing and don't have the money to hire people to do for me (e.g. frontend web development). Those are projects that otherwise wouldn't exist that do now exist and potentially might make money or something (though with my business acumen that's debatable).
Then inflation eats up the salary, but doesn't lower the productivity requirement.
Also, there are fewer secretaries.
From https://www.bls.gov/ooh/office-and-administrative-support/se...:
"Overall employment of secretaries and administrative assistants is projected to decline 2 percent from 2025 to 2035.
Despite declining employment, about 314,400 openings for secretaries and administrative assistants are projected each year, on average, over the decade. All of those openings are expected to result from the need to replace workers who transfer to other occupations or exit the labor force, such as to retire."
Inflation is about the real price for a dollar, ie how many goods a dollar can buy. It's not about the relative price between, say, your labour and the goods and services you want to afford. Inflation is a red herring in this discussion.
It does not.
Productivity increases and salary increases have been uncoupled for half a century
We’re talking about corporate compensation to labor vs shareholders and admin.
"Increasing productivity drives salaries up"
To which I said: No, that's not the case; productivity and salary have been uncoupled for a long time. That's a different claim from "salaries did not increase", which is what you are countering.
The wage-productivity gap is well known and I won't attempt to school anyone on it. Just look it up. Its really disingenuous to claim that the gap is going "to the robots". Robots don't give a shit. The gap goes to the _robot owners_.
No it doesn't. Salaries have been decreasing in real terms over the last decade despite productivity improving. There was a period of time in which productivity increased alongside salaries but that does not prove causation.
Do you have a source for that?
https://www.marketplace.org/story/2026/09/04/wage-growth-can...
Obviously, the above is talking about labor generally, not just SWEs. But speaking as a professional computer scientist at a place with decent salaries and excellent benefits, my raises over the last 5 years have barely outpaced inflation. It’s pretty annoying to realize that I probably would have earned a lot more money had I NOT gone back to school for a PhD. Not that I would make a different decision given the choice again, but money is much more abstract in your 20s than it is in your 40s.
If it wasn't absolutely essential that a person was there, it would have been automated away long before AI.
My point is that MS Office has really automated everything about the job that can be automated. I don't think LLMs will really dent secretarial/PA work much further. Those people have their jobs not because the work hasn't been automated, but because their bosses are mediocre and relatively disorganised, like most people.
Silly Jarvis fantasies aside, an LLM cannot make your boss less shit at returning calls or managing their work life. The people who need support humans will always need support humans, until they themselves are replaced.
My bank account disagrees with that statement. We are just expected to do more for the same pay.
I can name 5 other pieces of software I have open right now, more likely than not there are 100x supercharged engineers in this thread who work on them, and they're also the exact same piece of shit they were 2 years ago.
As a software user, when can I expect to see results of this unprecedented productivity boost? Are you people completely detached from reality, do you not see the output of the industry hasn't moved in the slightest?
You mistake “I don’t care about <feature>“ with “<feature> hasn’t been built”.
Are you posting from 1990?
If they’re needed still, they are really needed!
You can look at what happened to hardware engineers starting in the 1990s (if not earlier). Improved software allowed 1 HW engineer to do the work that was previously done by 3-4. Did this drive HW engineer salaries up? No, no it did not. It did the opposite.
As for secretary salaries, what's your basis of comparison? Women earn a lot less than men doing the same job. This used to be a lot more. So, if secretary pay went up at all (and I'm not sure it has in real terms) then how much of that is just eliminating the gender pay gap?
The last 40+ years of data on wage growth vs productivity growth says otherwise.
Thinking human skill will remain necessary for this is a huge bet against the AI companies.
If productivity increases lead to wage increases, being a weaver would be one of the best paid jobs today.
The business of software still exists and it needs to people to run it. It will require fewer people than before, and as a result cost of software will come down. There will be increased competition, and big SaaS companies will face fresh competition from upstarts. As a result there will be pressure to innovate.
AI can't innovate. This is one of the key limitations of AI. Not all humans can either. But those who can will be in increased demand.
As software becomes cheaper, the total consumption of software increases (See Jevons paradox.) Cheaper development makes smaller markets viable. There will be software available to meet needs that are unmet today because the market is too small. This will increase the number of software jobs. It won't all be coding jobs, it will be jobs that run the software business (see first paragraph).
And, what do you mean by "AI can't innovate"? I suspect that for any definition you pick, either most humans also can't innovate, or examples of AI innovation can be found. And this is only going to get better.
See Chris Lattner's observations about compiler written by Claude: https://www.modular.com/blog/the-claude-c-compiler-what-it-r...
2 options:
A) A real person with a job title that's not SWE. Will it be managers / product managers / business owners / something else? Whatever the job title, those people need SWE skills and experience to manage the agents. It is a very high ask for non-SWE job title to absorb both the skill and time commitment to drive AI-first software work.
B) AI Agent. But how does the AI Agent get setup with what it needs to run in a loop, who investigates and steers the agent course when it deliver something that doesn't work? It will needs someone SWE skills/experience.
In either case sounds like you need someone with SWE skillset in your company... so not sure why the job would go away.
AI can write syntactically perfect code but barely has a sense of architecture.
It needs tons of 'guide wires' to keep it from going off track.
Unless you're building basic SaaS, it's still a 'research and coding assistant'.
AI has never produced non-messy architecture that I did not have to go in, re-think and guide a re-write.
SWE's have to guide the AI, look at all those AI PR's, many of them make very bad assumptions.
Second point should be obvious to anyone actually building something even slightly new/complex.
So many executives have turned all the safeties off and given over control of their business to a probabilistic system and ignored every engineer who said its dangerous and risky. The engineers will be the ones held accountable though, because since AI is "so effective" it must be operator error
Ofc unless you call some data modelling “architect” but I’m sure AI is fine for that. Guess how I know.
2) I think u might be right about brute force - but probably some methods/tools will come along first - and I don't think they will solve the problem though.
I think SWE is actually very safe.
We're going to need fewer 'JS coders' and technically we could 'save' a few SWE jobs - but - given the $ being made, businesses will want to expand, and so 'Jevon's Paradox' and I think SWEs will have jobs and probably even the JS jocks as well.
I've been building for the past few weeks a fancy text editor for myself to work with AI in a more sustainable and sane way. I don't have much time for this, so I've entirely vibecoded it: I don't look at the code and I've provided only some initial design and engineering decisions, debug hints, etc. on top of natural language feature requests.
For about a week using the strongest Claude models (Fable 5.2 and Opus 5.5) this went great, and I had a minimally functional tool that worked very nicely. It seems that I hit some sort of complexity ceiling for those models though. The second week, every feature I try adding creates a new type of jank that can (at least by me) be only vaguely characterized. Most recently inserting into a line (even in the middle of the viewport) causes the viewport to jitter about.
Our work will change, we will spend less time on tedium, we will become more ambitious. Barring an architectural paradigm shift we will still need people with strong engineering and design instincts, and even more importantly the ability to write reference code and formal specifications codifying desired behaviors.
When systems get complex, someone needs to know them and know where to burn the tokens. You need to know what events need to be produced because you have specific consumers. You need to know your business invariants and why.
If you’re working on something that isn’t complex, yeah I can see why you’d think you can be replaced. If it’s just a front end? Yeah I can see that. A basic CRUD service? Sure. A business critical domain service with events, data migrations, complex business logic? I don’t see it.
However it will be similar to cybersecurity in that there will be no 'entry level' product engineer. It will probably have requirements having worked in customer service or finance or some other adjacent field to get a larger picture of what you are delivering. You will also not need 3000 of these individuals anymore (did we ever really??)
I can see a world where you go from "AI Shepard Customer Success rep" -> some middle role with increasing 'product' responsibilities -> Product Engineer.
The new roles will also probably be more generic - instead of backend and frontend you have 'Backend' (Infra + API) and 'Frontend' (Frontend + Design) and both of these roles will require genuine product thinking
No, and it's the same with leads, they can be given a system they understand but need to reference it.
I'm very certain these are people brushing up against their lack of processes, workflows, or engineering practices.
The context are large enough. You do not need to understand the world to build a chair.
I see the very opposite in many different domains happening now, and some codebases are monsterous. They are figuring out ways that do indeed work well.
The engineers that are figuring out these problems when working with a new tool, will have jobs.
Those who continue to doubt, and distance themselves are the ones that will be behind.
currently working on an initial product for a startup and while I'm substantially further along that I would be without LLMs, I'm at the point where the LLMs are now randomly breaking things.
There's typically a combination of issues that are occurring when the codebase goes off the rails.
One is that cruft builds up. Large vibecoded codebases need regular "review cleaning". Where you review the whole thing top down for hygiene, contradictions, bad comments, style, redundancy, design. It's a standard thing for me now. A full review and cleanup pass once every couple months.
The second thing that tends to go wrong is that after a certain complexity threshold, the models cannot keep track of the relevant details for a task very well. They lose sense of project directionality. Every task gets disconnected from other tasks and they start pulling the system apart at the seams, exacerbating the cruft issue above and accelerating collapse.
I've found that separating my documentation into three distinct lobes helps: goaling docs, guardrail docs, and implementation docs. Goaling docs describe in prose the goals of the project which includes scope, certain design targets, and other things that you are aiming towards. Guardrails are for design and implementation choices you have explicitly ruled out, the "what not to do" part. Impl docs describe the state of the system as it is.
The top-level agent documents refer to these and make their purpose explicit.
You'd think a ticketing system might capture this, but it doesn't really provide the richness of expression a collection of documents in a directory does.
These approaches aren't perfect. The process still has a tendency to wander off the rails, but it happens less often now and I catch it more quickly. Every session misses little bits of the goaling alignment, but they also find and address small issues missed by previous sessions. Overall it does a better job of keeping the codebase hovering around some threshold of coherency.
I'm having a lot of fun to be honest. Yes it's bizarre and nothing like what development used to be, but it finally feels like there's some new things to learn now.
These days they know how to handle their own context in review. So I will use that to find improvements too in retrospectivea and a small usage script hook it runs each push that alerts if high token usage. Hint the subagents are getting excessive context, which is also good hint of drift. So then I can measure when it starts happening typically via that script that Claude checks and readjusts when one goes over budget.
It's not that I don't think those skills are important, but they betray a fundamental misunderstanding of what SWE is, which is about managing systems, juggling complexity, and having some wisdom about how those things should fit together.
Hopefully we start to better formalize some more of that now that the grunt work can be taken out of the equation a bit. And teach it in schools, find ways to measure it in interviews. etc
What we the people need to do is gain control and hold all the power of AI in our hands - not the filthy rich's hands! Devise systems that flip the power into our hands and create economies and jobs around it (i have ideas and surely 1000s to more do to around this).
LLMs won’t always guide you to do this correctly so you need to understand how important it is from the start and always enforce it.
> Barring an architectural paradigm shift we will still need people with strong engineering and design instincts, and even more importantly the ability to write reference code and formal specifications codifying desired behaviors.
The people building those specifications and the tight validation loops you describe are the technical software workers of tomorrow.
If the software was capable of creating high complexity software I, or anyone, should be able to replace and build anything I want with just some prompts and guidance.
Anyone who has worked with LLM heavily knows this is far from the truth.
Here be lions, since these things tend to lie to you or themselves when skipping the enforcement part.
Same goes for AI slop comments. I can just prompt AI into more or less writing like me.
It is hard to say how things will develop the coming years, but this sums it up for me today. Almost constantly, LLMs (e.g. Opus 5.5/Sonnet 5.5) make choices that a junior programmer would think is ok, but I know that will create technical debt, subpar designs, etc. Note that I'm not saying that LLMs program like a junior programmer, in many ways they program like senior engineers who have happened to memoize all API docs as well. But they are often missing the larger context or don't understand the time dimension (what will be hard to maintain, what will introduce an API that we can not easily backtrack from, etc). So, I am not surprised that vibecoding breaks down as a project becomes larger.
At the same time, current LLMs make me a lot more productive (which was not true for me ~one year ago). I can do in 30-60 minutes what would take me a day 1-2 years ago. In many ways, programming also became more interesting, because I can focus on design, the really challenging parts, while the LLM can quickly implement it. LLMs really feel like a different, in many ways complementary intelligence for development.
My other firsthand observation is seeing the difference between what quality of software that I, as someone with many years of experience, can produce versus my coworker, a talented designer without SW experience who is very AI pilled and excited to be making things, is able to produce.
Through a combo of me and it over about a month we got it working well enough to roll out though. The entire time I had to hand hold the model through the changes though, and it loved to try and throw it away for quick fixes that would break something else.
For example I don't see any mention that you ran debt reduction loops, code simplifications loops, and in general asking the model to clean up the code after it's initial phase of "organic growth".
For sure, the guys who lost out when automobiles displaced horses were the ones who were certain they only knew how to make buggy whips.
Adapt and thrive.
After about a dozen failed iterations I reverted to the last known working commit (albeit with the new module only partially implemented), isolated the issue as best I could, then asked Gemini again and this time it correctly identified a latent timing bug that was the cause of all the problems.
In short, I had to manually reduce the complexity of the problem for this magic infinity-dimensional-processor to deduce what the problem was. Maybe it would have gotten there eventually, but without AI I would have just reverted the commit a lot sooner and likely figured it out on my own with some basic debug logging.
It's still a net gain, as it's successfully one-shotted coding errors and bugs that might have taken me hours to find before, but it can miss basic problems as well.
Then after i understand it i can make a targeted investigations for that issue and after that it will be obvious for the AI, and all of those we still didn't change even one line in the code.
After that, we fix that issue (which is the root of the cause), release and congratulation the bug was fixed this way works even in lowest powerful models.
But, as @sama said, these models you see today are the worst they'll ever be.
It's only a matter of time before someone figures out how to maintain focus and drive a product to a polished end state over many iterations.
LLMs have massively improved the quality of our company's main app. However, someone less skilled at our company tried to vibe code an internal tool and it was pure crap.
Not sure what’s going to happen but it’s not pretty. I have been pushing my wife to sell our house and downsize but she doesn’t agree. I was telling her that once that happens no one is going to buy our house even at a 50% discount.
What Anthropic has excelled at isn’t engineering. It’s convincing your boss that you can be replaced with a Claude Code subscription. Just look at the leaked Claude Code source. Or read their “engineering,” blog. It’s all marketing glitz.
The reason for all this marketing is because they rent their compute and they need every software developer to need them. They need us to depend on them to do our jobs. Because this isn’t a profitable venture. It’s resting on a house of cards.
Anthropic has to deal with the very real possibility that their providers go under. And that other harnesses on alternative models get better, cheaper, etc. People seem to be pretty price sensitive after the token-maxing rush didn’t work out.
I think there are still plenty of reasons to believe that our profession isn’t going anywhere. Changed, sure. It always changes. But it’s not over and it’s not, “solved.”
The tricky part is you have to price your offering well below what your potential customers think it will cost to vibe code an equivalent.
In a way it reflects how repetitive our industry is. We train a model based on the github projects and much smaller expert data, yet the model is sufficient to solve many of our current prolbems.
On top of that, we're not hiring less people, we're just building (slightly, the speedup isn't consistent across all parts of the workflow) more products
There's obviously still a requirement that people have a lot of skill and knowledge to produce good software, and I'm not sure that's going to go away.
It's the boot camp/degree/whatever to get the paycheck but never enjoyed the problem solving or learning aspects...
I assume the premise of a software factory is that it can churn out any software idea you have. You won't need to build a new distinct factory for each new distinct piece of software you dream up. Otherwise it wouldn't be called building a software factory, it would just be called building software.
So if somebody gets this right (this remains to be seen, the various Gas Towns of the world have not proven their value), won't one good software factory be enough for everyone to download and use? Wouldn't we all then be done with the labor of making both software and software factories?
For example, I could clearly see having a "Static web page with HTML and CSS only software factory" that just churned out that kind of software. Or a "Postgresql optimized SQL" software factory. And so on.
every company is making a totally different product so its hard to see how every company is its own 'software factory' outside of such components (which they were already doing even before llms)
large groups of people dont have good taste. Its always going to be reversion to the mean. which is just mediocre software
I don't really see a future where human labour at all can still be competitive beyond like 2050 in any field. Anyone who's ridden in a Tesla in FSD knows transport jobs are defacto over, the tech works and it will be copied and deployed en masse sooner or later. Humans don't scale and we haven't meaningfully mentally improved in 100k years, we're limited to 20W of compute, take 20+ years to train and go EoL in 40 after that. It's inevitable that we'll be outclassed and it only needs to happen once. Once a model exists that is slightly better at just one specific task, it only gets cheaper over time and it can be used for that forever.
Unless we organize to prevent it, technology will continue be commoditized like all other technology into the hands of a few players.
And especially with AI it’s going to look more and more like nuclear or biological technology and that’s why you see so much push for regulatory capture.
Literally all I do is spend my time trying to figure out how to prevent this, and still make progress towards AGI for everyone, but it doesn’t seem like people care
Everyone just wants their bag everything else be damned, and that’s what we’re seeing
I would understand this comment if I was telling you how to live frugally because it's not my lived experience. But in this case, it is my lived experience. I literally work at a job where I have to build a software factory to be successful.
Perhaps. But I work right now, and I actually care about people other than myself. I would like to see the industry that made me successful keep making others successful.
We'll have to wait a month or two I suppose to find out if you're right, but I think all companies will be moving to the model of software factories. And It's not something I thought of recently -- I've been building a software factory with AI for a year now, and working with others who are building entire businesses on it.
And FWIW, pretty much every tech company is now moving towards the same model.
This is my first reading about your work but I like your attitude. Databricks is also a good service.
From what I am seeing, those who embrace AI are embracing the 'computer' from the 80's. The landscape and reality is still being defined.
To the AI haters and doomers I say buy the largest house with cash you can to enable your retirement. The stress that people experience is the thought of their crushing debt or obligations that can be met with a 90% reduction in spend. Of course, that requires a lifestyle change that most are unwilling to make.
Success in a prior time does not guarantee success in the present or future as too many variables are different
Demographics have shifted wildly and most users are not a soft spoken grama needing help setting up pop3 in Outlook Express
Look around you; everyone uses their phone at whatever level of expertise they need now
Goodhart's Law has come for how we used to measure need and todos in tech. We hit the top of the S-curve relevant to this form of computing devices and there is nowhere else to go with it
It's like that meme from about a decade back about the millionaire saying that the only reason millennials can't afford houses is that they eat too many avocados. That advice makes sense in a long-gone era where houses are cheap and avocados are expensive, but in the 21st century the price of personal avocado consumption would never add up to a house in a reasonable amount of time no matter how excessive your avocado lifestyle may be. jedbergh is going to be an "avocado guy" from now on (again not that I actually know anything about him).
What is falling away is that we just won't need as many SWEs as we used to. The way to make sure you don't "fall off the ladder" is to make sure you're the best at the new way of doing things.
It doesn't matter how good you are at anything when you can't even get a reply from a person. And absolutely forget about learning anything on the job. The ladder is gone. Combined with the war on the middle class, am looking at a retirement in poverty. Even social security is scheduled to be empty in a few years when available.
Sorry, I just checked and I don't see anything from you. Where did you ask? I try to respond to everything I get, but sometimes gmail doesn't surface random messages.
edit: Body contains Fry's Canoga Park etc.
See web dev jobs tanking. Backend is next.
Backend dev opinions/arbitrary strings are being compressed out of the stack.
Devs that don't understand hardware development are naive about the endgame; black boxes that boot to a Google Opal like GUI.
Boot to Blender-like/WYSIWG environment loaded from a model itself. All the context we put in files added as vectors on the presentation layer.
Users will conjure a word processor or email UI and then a video game.
Network gear will boot a model to handle the specific job the model was trained on. That's what all the Stargate stuff is about but obfuscated behind marketing to hide implementation goals details from knowledgeable labor to avoid public scrutiny. Vague scrutiny without concrete proof is fine but if we could point to the details too they'd be naked.
50 million devs globally among 8 billion that do not want a subscription service to an expert when they have no such leverage over the expert themselves.
This isn't empty speculation. Hardware industry executives are responding to social pressure at scale.
America was able to leverage its clout to empower IT to pretend it's more than just math relative to memory and display matrices but America no longer has the global clout to assure that. To Chinese engineers it's just math.
American software devs were engaged in sophistry and social moat building. Ya got disrupted by the world. Believed they would ignore their eyes for the rhetoric and hype forever; rather Orwellian really.
https://en.wikipedia.org/wiki/Rebound_effect_(conservation)
YMMV I guess.
My worst fear is that I will have to do an absurd job where my brain rots under people with less skills and ethics but better at playing the new game.
Software just needs to be good enough to make a profit before it gets replaced by a slightly better or cheaper alternative, and we're probably going to see a plethora of those alternatives.
However I'm just some rando, I don't really know.
I don't really know either, just my mental health is torn to shreds due to that impending doom on the horizon.
If we step outside of our IT world for a second though, I see people who have it much, much harder already. At least in my country, compared to last couple of years, jobs are much harder to come by and right now it's hard for a lot of my friends in: marketing, social media, accounting, even management and business majors who recently graduated have difficulty to find anything entry level...
What's gonna happen in we go through a job erosion that suddenly dissapears half office jobs? Are we ready for a surge of unemployment and the ripple effects it will have to the rest of economy?
If it makes you feel any better, there are a lot of us in this position. Far more than talk about it.
I mean, I talk about it all the time but the main response I get is people who are somewhat relieved to learn it isn't just them. (Along with a different bunch of people who need to point out I'm being a doomer to suggest that it is remotely possible that things could get pretty ugly for our profession very soon).
It seems like many people are afraid to talk about how shaky the ground is underneath us for fear that mentioning it will manifest the worst outcome. And a lot of other people seem to be in denial that we could even really be at such an inflection point, clinging to the "AI is normal technology" idea.
I fully accept that my view may be overly pessimistic and things won't be as bad as I think, but I really don't understand how the people who are 100% sure everything is going to be just fine are able to maintain that certainty unless to some degree they are self-soothing as much as anything else.
I'm not scared for the future of software engineers and computer scientists.
I can think of several bigger moments in the last few months alone
That could be as simple as raising taxes on corporate profits and using them for a universal basic income, or something more clever, but I hope governments manage to build something before it leads to a dystopia and social unrest.
Not everyone gets to work on the most interesting engineering problems, I think most of us felt more like glorified digital plumbers. The money was too good to pass up that CRUD work but I can't say it was very fulfilling, I'm glad a machine can do that drudgery for me now.
Massively sad moment, a few handful of already well to do individuals got even richer, allowing business to cut wages. Sure it’ll create new jobs / industries but the short term impact is not something I look forward to.
Coming from the service industry, yeah there's some long stressful days doing software development but those barely held a candle to average weekends / holidays working at restaurants and cafes.
Then the whole boot camp thing happened, where it's like sure, I guess since you spent 4 weeks learning just React you're worth as much as someone whose been doing it for a decade? I figured that would have been the end of inflated software developer salaries.
Now we're here.
1: https://en.wikipedia.org/wiki/IPhone
More prosperity than you've ever had.
I grew up in a third world country, so I can say first-hand that the internet has helped the average person around here gain opportunity and open doors to the possibility of prosperity.
A dystopian vision that keeps popping into my head is from the movie WALL-E, where the robots are running everything and the humans are overweight blobs gliding around on recliners with entertainment pumped into their eyeballs and endless soda refills. Is that what a world of "abundance" but no meaningful work, looks like?
Everybody is an architect now.
Don’t worry.
The programmers were/are simply capable and in relative short supply. Even with that they earned the companies massive profits. (I have completely different issues with my high-IQ nerd peers... more what they were working on.)
Now the parasite companies can simply harvest all of that value for nothing. Big Tech is moving from a gravy train to a free ride.
Defintely a very Capital Guy perspective top comment.
I have seen coding agents get better as both the models and harnesses improved. I used to have to manage context by hand, I used to review every line. I no longer do.
However, I will tell you this from my experience with AI. Software is not any easier. The same complexities are still there. What's different isn't the kind but the scale. A 2 person team can do what a 20 person team did 5 years ago.
What is still required is a shit ton of feedback to build what you want. It's still hard to build complex software even with AI. In fact in some ways harder because AI agents will go down rabbit holes and create entropy faster than you can say Bingo.
Productivity is waste, it has to be, it's thermodynamics. The person's job is to fight entropy in the system. That has been true before AI and true today.
We have to remember, software is there to serve us and fix our problems. AI doesn't change that.
Speaking from experience I see areas where the need for software is indeed saturated after a year of AI-accelerated development (where the devs are really competent and proactive), but in other areas I see the demand for it skyrocket, even with no limit to Claude Code.
I also wonder what part of the "gravy train end" is the general economic rough patch we're in due to geopolitical conflicts & crises in the world e.g. Iran/Ukraine
Unless demand for software somehow increases 10x, this leads to 90% of developers being out of work.
This turn of phrase really amuses me (whether I'm in complete agreement or not).
?
They certainly did not. LLMs are way worse than a competent human programmer.
Just today Opus 5.5 tried to send a Slack message to one of my teammate when it had to edit a code file. When asked why, it said it just "picked the wrong tool".
LLMs are borderline useless, AI has led to 0 increase in actual value produced. It's all powder in front of our eyes.
I genuinely cannot take anyone seriously when they say they're afraid of being replaced by AI. The only explication possible (unless they're astroturfing), is that they literally never look at what LLMs produce and always run in "Trust all tools" mode without ever looking.
Don't oversell it.
It’s a dark coincidence that “Turkey Day” is next week in the USA. Many turkeys that will be on our plates never imagined THAT fate.
The gullibility of people is astounding, and they never learn.
Eventually we just accept the new status quo, and the QOL hit that comes with it.
Anyone who hasn't FIRE'd yet should aggressively be saving as much as possible before shit hits the fan.
Good luck everyone! It's going to get ugly.
We have to balance "a lot of people can get the product" with "product is high quality" somehow.
But... I've also spent an incredible amount on poorer quality shoes and regretted it often. Many £50-150 shoes that just didn't last. The obvious planned obsolescence of eg trainer (sneaker) soles is just so obvious.
Damn it, why can't I learn the lesson!
£300 is equiv to £485 today too
Similar Loakes go for ~£320 and Cheaneys ~£500 today! https://herringshoes.co.uk/collections/oxford-shoes?filter.p...
The work of software engineers has enabled a (temporary?) profession of mostly busy-hands sort of work to thrive, mostly achieved by automation. For any software to presently work entails a substantial amount of things working in an automatic or "factorized" fashion. Thus, the engineering aspects will now move onto enabling a different sort of work, before it factorizes it, also.
This is multi-generational change happening in a decade. This is disruption across nearly all industries simultaneously. We can't just waive that away. There are going to be consequences.
(And I'm getting lots of job offers. But I'm in Singapore, and that's presumably not where you are.)
"if its used correctly and broadly"
It's not enough to just have a claude code subscription for every engineer.
The powerband of AI comes through automation, integration, and tooling. When literally everything is touched by AI, and when most tasks get passed through AI before seeing a human, only then are you actually at an AI-first shop.
Only then will you actually realize how little you do in this AI-first world.
"You are holding it wrong."