Comment by regularfry
1 day ago
The bottleneck has moved from producing a thing that works to knowing that the thing was the right thing to build. The more of the latter they can take on, the fewer knowledge workers are needed at all. So rather than 5% of every knowledge worker's salary going into tokens, 100% of the knowledge worker's total employment cost goes into tokens and you get a 20x productivity boost as a theoretical minimum across those tasks.
That's the game. There's a view you could take of this that this is just a growing of the pie: with those cost dynamics a lot more "small businesses" get a vast amount of leverage, so the overall economy grows without replacing the knowledge workers. I'm not sure I trust the MBA class to have that view.
>The bottleneck has moved from producing a thing that works to knowing that the thing was the right thing to build
I would argue that that's been the case for quite some time before AI. As an example, what innovative amazing world-changing products have Google or Meta launched in the past decade with their very high numbers of very talented and highly-compensated engineers? The issue with most big tech companies are leadership, strategy, and product direction. I'm not saying that they don't make any profits, just that they probably aren't "building [the right thing]".
AI for product development and management would be far more impactful than automating rote coding tasks / building React UIs that mirror API structures IMO.
> AI for product development and management would be far more impactful than automating rote coding tasks [...]
Yeah, if this stuff actually worked that well already, OpenAI et al. would just run AI CEOs and engineers. Why get some other company to pay you at all when you can automate every other company out of existence and take all the money they make?
The fact of the matter is that while the tech has some uses, it sure as hell isn't a full scale replacement and you almost always actually have to massage the input into LLMs to get anything decent back out in practice. Some CEOs and managers can learn to do this, of course, and some already are... but that quickly turns into a second full time job. A "programmer" is still needed. The job might change from mostly hand-writing C++/JS/Python to prompt engineering + some manual coding to fix all the stupid fuck-ups that the bots can't solve themselves, but you still need someone to actually prompt the bot.
When that changes, it won't just be engineers losing work; there will be no reason to even have a human CEO any more.
> When that changes, it won't just be engineers losing work; there will be no reason to even have a human CEO any more.
The human race isn’t ready for that world IMHO. The only reason there is a middle class is because people have leverage in the form of their labor. When that becomes worthless … the people who own stuff and make their living from doing so won’t hesitate to get rid of everyone else - whom are now worthless to them.
3 replies →
I don't know, if you've ever tried to build something at companies of that scale you run into incredibly boring problems "what data table do I need for X" and "who is the right person to reach out to for Y" and "they aren't answering me I guess I'll have to escalate"
I don't think there is any shortage of great ideas at these companies, they are just extremely bloated. And I don't think its something like indecision or bad PMs, it's "we have a finite amount of time and resources so we need to be conservative but also not too conservative"
If you have AI systems that can simply build out POCs in days, backtest on real data, show reliable results and numbers, you get a suite of product options you were never able to get before. If you have coding agents that can speed up implementation, you can build more stuff and choose the things that stick.
It changes the cost/benefit calculus of the entire business. I think you are exactly right in that: PMs/leadership are by their nature orchestration machines. Other roles are as well, but I think PM's are at a particular advantage here in that it will be quite awhile I would expect before core product decisions and creativity can be delegated to an AI, but not quite awhile until virtually everything that they're blocked on (legal approvals, POCs, wire frames, etc etc etc) will become less and less of a blocker
>If you have AI systems that can simply build out POCs in days, backtest on real data, show reliable results and numbers, you get a suite of product options you were never able to get before. If you have coding agents that can speed up implementation, you can build more stuff and choose the things that stick.
I'll also add this: within a large organization, you often need to interact with many different codebases owned by many different teams. Agents have made it much easier to wrangle by having the ability to deploy one to scope out your web of dependencies to learn about what would be needed for feature X, and how that integration can happen.
We've been doing far more away team work simply because it makes things move faster. It's easier to convince a team to sign off/review something than it is to get them to commit to the planning and eventual work.
It genuinely is helping things move faster inside large organizations. Or at least, it is for us, particularly since we're getting organizational prioritization to actually build the scaffolding to make those agents more effective at search.
1 reply →
Legal approvals won’t be in that category.
You still want someone whose ass is on the line if they get it wrong.
1 reply →
Pieces of concept and other prototypes have always been cheap (see hackatons). The main issue is that as soon as you’re touching customer data or modifying process they’ve paid you for, you have to be really careful. No one wants to be responsible for an outage that cost the company its biggest customer.
3 replies →
Yes, that exists at the wider business level. No question. I think what needs to get asked is are we talking about a bottleneck within the business as a whole, or a bottleneck within the scope of the knowledge work in question. Within software delivery there's a very clear shift when it's suddenly trivial to drop a 100kLoC plausible-looking PR into code review within an afternoon. Producing working code with a whole bunch of tests which make a very clear assertion that it does, in fact, work has had (if you're going that way) all the human-scale thinking time taken out of it, down to a rounding error. It still needs to be checked by a human, which was previously assumed to be a comparatively quick task in comparison to producing the thing. At least, it does where I am, and I don't think that's a silly position today at all.
If they can crack that latter review/spec-check/assurance step, checking that what was built was what was demanded of the problem such that we don't have humans in the loop at that step either, then the bottleneck moves again. Then I think it moves to requirements capture and to product development, but that might depend on the industry.
Trusting CodeRabbit for sign-off is "just" a small matter of configuration.
1 reply →
> As an example, what innovative amazing world-changing products have Google or Meta launched in the past decade
Kubernetes is at 11 years ago, and is huge enough to be included there. The Google Pixel was just under 10 years ago. So... not nothing haha
Numbers I see put Pixel at less than 5% of iPhone sales. Not nothing, but world-changing? I doubt the world would look significantly different had Google not done Pixel.
If you really think this you simply have no theory of mind for this stuff. There are tons of immensely successful products in the ad space that both of those companies have launched. They don't need to innovate in the product or technology space (doing so certainly makes a big difference in having more placement for ad real estate), but to suggest there have been no real innovations (specifically engineering specific innovations) related to ad tech would be completely ridiculous to suggest. You don't need to change the world to get rich, just look at wall street where major innovations have been made in the pricing models of fixed income securities.
Second to this are countless other areas that have a major impact on the companies bottom line that are entirely engineering driven, especially at google given they are a cloud provider and have meaningfully grown the workspace business and launched waymo in this time.
>I would argue that that's been the case for quite some time before AI.
I would agree but it's really minimized the building. More and more time is being spent on pre-coding work.
Google's internally developed and sometimes even launched plenty of innovative new products in the past decade. Stadia, Fuchsia, federated learning, and the whole transformer architecture that underlies this AI boom are good examples.
The problem is they get killed by some other executive who is afraid of their department looking bad by comparison.
I think this is fairly illustrative of the challenges in AI becoming as impactful as the Internet. The bottleneck is not making things. There are plenty of people who are really good at making things and can easily be 10x or 100x as productive as the average corporate worker. YCombinator was founded on that premise - small teams of founders and early employees could be orders of magnitudes more productive than the 1000s of corporate employees at their competitors.
The bottleneck is on bringing your product to market. If your innovative new product is built within a corporate environment, it'll get killed unless the executive you work under can get a promotion out of it, and you'll be denied all sorts of help with approvals, launch process, PR, marketing, branding, etc. If it's a startup, they'll try to shut you out with exclusive distribution deals, legal threats, lobbying efforts to change the legal environment, PR campaigns, FUD, etc.
The Internet was revolutionary because it let millions of people bring products to market without asking permission. Instead of having to bid for retail shelf space among dozens of entrenched competitors that all had sweetheart deals with the retailer, you could just put up a website and sell it to anyone across the globe. Instead of following hundreds of regulations that governed existing commerce, you could just launch something and sort it out later. AI doesn't really have that property - if anything, it makes things more centralized, with more gatekeepers, and so seems more likely to destroy economic value than add to it.
Google does not follow through in the long run on many of their pet me-too follow projects, however they do not stray away from their core remit making their real customers happy the ones who buy the ads…
Obviously that includes whatever needs to be done to hoover in data from their marks and Meta also does the same thing without fail and both are really good at it. But outside their remit not so much.
What I think is happening is that the scale of thing you can hope to build at a below-corporate scale should radically grow. Corporate environments should suffer for this, being that inefficient.
> YCombinator was founded on that premise - small teams of founders and early employees could be orders of magnitudes more productive than the 1000s of corporate employees at their competitors.
I think this is still true, but the theory is:
1. You don't need YC-type funding to do YC-type business any more; 2. You don't need to scale the business past those small teams any more, you just buy more tokens.
For clarity YC still obviously has a place as an incubator, mentoring, and networking function. I just think that what was previously the inevitable conclusion that you have to hire all the people the second you hit PMF to keep up with scaling the business as you scale sales is no longer inevitable. If you didn't want to go that way before AI, you were a "lifestyle business" and not worth investing in. As more and more knowledge functions get capably implemented by AI, it's the preferred position: humans are vastly more expensive than tokens, so you want them doing the stuff the AI still can't do.
I don't think this necessarily translates to mass unemployment. I think it translates to masses of smaller businesses that are radically more efficient because the handoffs between business functions are tool calls, not emails to someone who doesn't want to help.
> The Internet was revolutionary because it let millions of people bring products to market without asking permission.
Think about it this way: if I am a small business owner but I think it makes sense to do something that previously only a team in a corporate environment could do but is now within the reach of AI, not only can I do it now, but I also don't have to ask anyone for permission! Who wins between the corporation and the small business in that scenario?
> AI doesn't really have that property - if anything, it makes things more centralized, with more gatekeepers, and so seems more likely to destroy economic value than add to it.
I think this will turn out to be backwards. I can see a version of this where the number of things you can do without needing to turn to a gatekeeper for help increases to the extent that the balance completely inverts.
The vast majority of businesses are small, and AI can give them tools which previously required corporate scale to make sense, without the inefficient hand-offs between busy, political humans. Which is also something that the internet did! Getting an advert in front of a national market pre-internet was Hard but sometimes you had to do it because your target market was "all Canadians who buy toothpaste" or whatever and that meant saturation-bombing the physical environment with physical billboard ads, posters, flyers, and so on. So you only did it if you were P&G-scale. Now you, personally, can do it, trivially, for better or worse.
3 replies →
Google & Meta are illustrative of late-stage capitalism -- it's all about distribution, not innovation. Their job is (mostly) to just acquire the products that have passed the gauntlet, then scale up their monetization through their distribution-focused machine. The same dynamic plays out in virtually every industry (not just tech).
You'll find that most internal "innovation" teams are just lip service. In most cases, the "mothership" will be incapable of reproducing true innovation -- from a statistical perspective, culture perspective (mega corps are anti-scrappy; internal politics), and motivation perspective (startups aren't 9-to-5). It's much easier to have big M&A budgets, a VC arm, and some handwavvy internal innovation group.
Every now and again, you'll get real innovations (Waymo, transistors, GUIs), but even those have a spotty track record of commercialization when created internally.
The one I'd point out for that list is Kodak and the digital camera.
This is the same argument that has been historically made for outsourcing developers. Get 20 more devs for the cost of 1 dev in the US.
I suspect that AI will fail to pan out to the same extent for the same reason why outsourcing hasn't fully panned out (even though every company tries it after getting big enough).
The problems that will come up will be and always have been ongoing maintenance. AI is great at writing new code without a brain behind it, but once you get to the point where you need to refactor code, you start really needing someone with coding experience to guide the AI or veto it's mistakes.
I don't think that's really fixable even with a lot better AI. It's not something that ultimately comes out of the likes of github data.
I'm not saying that AI isn't going to make things better, btw, I just don't think we'll see a 20x improvement. Probably more like 1.5 or 2x.
Outsourcing of knowledge workers didn't work out because at large enough scales, the geographic arbitrage disappeared. Companies mostly always got what they paid for.
The determinant of success was only whether the task needed American-tier labor or could make do with sub-American quality labor.
I am not sure this feels right. I agree that the US currently has leading minds in terms of tech, but I am not sure it is too big of difference with the EU knowledge workers. EU is still a lot cheaper then US in terms of wages you would need to pay.
6 replies →
That's certainly part of it. But the other part that I've heard time and time again is that in order for outsourcing to be successful you basically needed an american engineer in the mix hand holding everything, clarifying requirements, and vetoing bad code.
That part of dev work, the requirements gathering, attention to details, clarifying requirements, is something AI also struggles with. A lot of companies basically waste time and money on outsourced devs because without a clear path forward they effectively will sit and do nothing, waiting for a prompt.
3 replies →
Outsourcing of knowledge workers is still ramping up. The issue in the past was the skills were few and far between internationally. Facilities were also not built. That has changed now in a lot of fields. New campuses have been built in places like Bangalore and Hyderabad, even Singapore. The skills are there now, the training is decent, and you can see that the hiring is going on for very skilled titles in these cities. It is a different animal than just 10 years ago in this.
The “American tier” labor of course is distributed across the world and the top performers in every nation find ways to get paid at something approaching American salary levels. Look at all the international FAANG offices paying high salaries, in purchase pricing parity terms.
> I suspect that AI will fail to pan out to the same extent for the same reason why outsourcing hasn't fully panned out
My mental model for that is that outsourcing fails where the work is being done organisationally far from the knowledge needed to do it. We know that's true of teams inside organisations, there's been a lot of research on how distance in the organisational tree negatively impacts productivity. Outsourcing is a pathological worst-case of that.
The promise (promise! We're not there yet!) of AI is that I can have a cross-functional team on my laptop. Organisational distance is zero. Where previously the outsourced team has to wait for the time zones to roll round so I can answer their blocking question when I get to my email STRICTLY AFTER I have had my coffee, now it's a prompt in a chat window with a button I can click to make a choice in 5 seconds. Delay is gone, cost of delay is gone.
> The problems that will come up will be and always have been ongoing maintenance. AI is great at writing new code without a brain behind it, but once you get to the point where you need to refactor code, you start really needing someone with coding experience to guide the AI or veto it's mistakes.
Oh, absolutely. That's a minefield. Today. It will be, right up until it isn't. There are ways to set up agents and projects right now that make a dramatic difference to how this part of the picture plays out, but those will sink into the harnesses as time goes on.
But also the big problem with maintenance and outsourced teams tends to be the commercial structure around the contract. You get a Build team, who Build the Thing and then: no more features for you, anything you want to add past the original spec costs extra. They hand over to the Run And Maintain team, who get to fix all the bugs that the Build team left but without the knowledge gained from building the thing, but are scaled and located to be absolutely as cheap as the supplier can get away with so probably don't have the skill, inclination, motivation, or permission to take on any restructuring to make the bug fixing easier and they're on the wrong end of the globe so there's a 24-hour latency on any queries. It's a terrible way to set teams up, but it looks good on paper.
Again, that's peculiar to outsourcing and completely goes away if I have the same team that built the thing own the thing long-term. That's true if it's humans or AI!
> I don't think that's really fixable even with a lot better AI. It's not something that ultimately comes out of the likes of github data.
No, it's a harness problem. You need to start from a maintainable point and keep standards in place. It'll take work to get the harnesses there and it's not ubiquitous. You might also need better models, but I've already personally seen big differences in outcomes between projects that took certain steps and others that didn't; it's nothing revolutionary, mostly stuff that works for humans also works for AIs but you need to know to ask for it.
> I'm not saying that AI isn't going to make things better, btw, I just don't think we'll see a 20x improvement. Probably more like 1.5 or 2x.
I think people radically underestimate the cost of delay. I don't know if 20x is realistic for the AI itself, but I think it's not impossible once the inefficiencies of having to go to other humans is factored in.
> mostly stuff that works for humans also works for AIs but you need to know to ask for it
I'm most curious about this sentence. What have you noticed about the similarities? I'm getting really good at asking for confidence levels, tests and pushing back, but I'm curious what you found
Outsourcing also fails because it’s a pathological case of adverse selection. The businesses that outsource projects are ones who are organisationally incapable of managing those projects well internally. But, that inability extends to their oversight of outsourcing shops as well.
End result is that many outsourcing firms are borderline fraudulent in the way they treat their customers.
Who pays for that value, and from what, if all knowledge workers lose their jobs?
It sounds like the economy would largely reduce to the small minority class of independently wealthy people.
The more time I spend using agent tools the less I worry about knowledge worker job loss.
It takes a skilled knowledge worker to use these things.
Yes, but I do worry about junior knowledge worker job loss. These models are very good (and getting better) at the vast dark matter of "donkey work" that happens in knowledge-based industries -- work typically done by junior devs / analysts / lawyers / consultants, paralegals, admin assistants, customer success / support, etc. -- and those roles comprise the bulk of the workforce.
And worse, these are the tasks that help the junior people eventually grow into the skilled knowledge workers required to operate models, so there's a pipeline problem too.
3 replies →
We'll get around to training job specific models or the equivalent. Thats just lower on the value chain for now.
Sure. I was challenging the parent on how the “game” they are positing would play out.
See https://news.ycombinator.com/item?id=48300427 for an alternative take. I don't think either direction is inevitable, yet.
To follow on from that comment, if the growth in breadth of capacity of AI leads to a decrease in the risk of running a smaller business, which I don't think is an unreasonable prediction, then it's not inevitable people do lose their jobs. Employers get smaller, higher-leverage, and more plentiful.
There were no knowledge workers in the middle ages.
Back then people were mostly farmers, but we already automated that job away.
Not completely, but compared to the middle ages we 50x'd their output. Which is a great illustration what it means to make a job 50 times more productive. We went from 80-90% of the population being required to barely make enough food for everyone to survive, to 4% of the population producing such an abundance that consuming too much food has become a systemic health issue
5 replies →
There definitely were what could be considered knowledge workers in the (high) middle ages, it just wasn't the majority of work like today. The knowledge workers then were just a tiny, elite faction, mostly employed by the church or directly by nobility. Kindgoms were still big bureaucracies and needed scribes, theologians, academics, lawyers.
Relatively few anyway. Monks (who wrote and edited books and managed libraries, and also taught), artists and musicians, bookkeepers/treasury/exchequer, scribes/chancery (who were the administrators of the kingdoms), and bankers all existed in European "middle ages". But a significantly smaller part of economy/society compared to "western world" now, yes.
There wasn’t 20x value to pay for in the middle ages either.
Are you sure? Any functional organization requires keepers to oil the machine. First the government. The best examples were the chinese empire, the catholic church, and the various kingdoms. Or do you think that everyone was either fighting or farming? Stewardship is knowledge work. Bookkeeping is another.
> Who pays for that value, and from what, if all knowledge workers lose their jobs?
They do not care unless these companies can get a bailout.
UBI only exists for companies that are too big to fail. Case in point, 2008 and SVB when there was too much money on the line.
One of the AI companies attempted to guarantee themselves a way for the government to bail them out if they were close to defaulting on the debt from the data center build out.
SVB didn't get bailed out, their investors and creditors were wiped out. You could argue depositors were bailed out -- as they took the undue risk of keeping more than $250k in a single bank (though as part of a requirement for getting a loan from SVB, you had to have your operating accounts with them. So lots of companies had no choice, as SVB was one of the few banks that would lend to them).
Arguably, the main impact of securing SVB depositors above the $250k limit is that it prevented thousands of people from being laid off that week, as their employers wouldn't have had the money to make payroll the following Wednesday.
4 replies →
> UBI only exists for companies
What's the U stand for in UBI?
Producing a thing has always been cheap since personal computers existed. From mail-order software companies' times to SaaS times, producing a sellable MVP was an initial cost that is relatively small compared to the later cost of expansion and maintenance. Marketing and selling was and still is the hardest part.
Why do you think of knowledge workers as a fungible commodity?
What makes you think the people who used to build (or would have built) software will switch into the industry of "knowing that the thing was the right thing to build", as opposed to something cooler like surgery, city planning or experimental physics? The roles within a tech company are not the only jobs in the world.
> Why do you think of knowledge workers as a fungible commodity?
I don't.
> What makes you think the people who used to build (or would have built) software will switch into the industry of "knowing that the thing was the right thing to build", as opposed to something cooler like surgery, city planning or experimental physics?
Because it's probably already part of the job. It's a change of emphasis, not a change of career. Your boss can already ask you to do it. If you're producing code, you're probably also reviewing code, checking it matches the acceptance criteria, testing it, sanity checking that it was the right code to have been written, today.
[dead]
> The bottleneck has moved from producing a thing that works to knowing that the thing was the right thing to build
“There’s more capital than good ideas to fund” has been a complaint from the likes of A16z & other VCs for a long time now. It’s why we ended up with stuff like NFTs getting funded.
That’s very unimpressive return on investment compared to what was promised.
If knowledge workers get laid off in mass, you can expect political curbs on AI adoption.
[flagged]