Why Amazon is hiring 11,000 junior employees

Matt Garman runs AWS, and he says replacing junior workers with AI is one of the dumbest ideas he's ever heard. So why did his own company cut 30,000 jobs over the past year? That's this week on Platformer. Welcome to Platformer, I'm Casey Newton, and today we have the final episode of this first little miniseries we've been doing on AI and jobs. But good news, after a couple weeks away, we're going to be back with a new miniseries, so stay tuned for more on that later. This week on the show, AWS CEO Matt Garman joins us. Matt, I think, has one of the best seats in the world to watch what is happening with AI and work. AWS runs the infrastructure behind much of the AI economy, sells the agents that are starting to do white-collar work, and sits inside a company whose CEO has said he thinks AI will shrink its corporate workforce over time. But Matt is one of the most outspoken optimists on AI and jobs, and we are going to test that optimism. First, though, as always, we begin on checking in on recent news about AI and jobs. And that means it's time to bring in Platformer fellow and Gen Z AI correspondent, Ella Marcianos. Ella, how are you this week? Well, as of like an hour ago, I'm furious because I was researching for this segment I'm about to present to you, and I saw this ad on Corsicle, an app that is a personal enemy of mine, which said, make TikToks for Corsicle. We're paying $500 a month to make videos about Corsicle, an app that actually helps students. Tap here to apply. And first of all, what is Corsicle? Okay, so Corsicle is a like course management software that you use to, like, in particular at Columbia, which used to when I was there, you can use it to plan the courses you're going to sign up for, and then do sign up for those courses. The thing about it is there are hundreds, thousands of courses you might want to take, and the interface for searching them makes absolutely no sense and has been overall like negative helpful for, like, finding the course I was looking for. Like, I literally would search three keywords for the title of a course I knew existed and knew no other courses contained those keywords, and that course would not show up. So, I mean, that sounds infuriating, but what was it about seeing this ad for making TikToks about them that sort of pushed you over the edge? Well, also, it's like, we as, you know, this is near and dear to my heart because I graduated college a year and a half ago. So in some sense, in some sense, I feel I never left. I still, like, have dreams about finishing my problem sets last minute. And I think, like, to say to, like, college students who are, like, stressed about finding an internship, like, here's a great opportunity for you. Chill for this platform that is making your life worse on TikTok for $6,000 a year. That's money for a college student, but, like, is it worth what it does to your soul? Sounds like the answer is no. Well, shame on Corsicle, and shame on TikTok. We'll be interested to see what happens to that company. In the meantime, what have you learned this week about AI and jobs? Yeah, so there's been some recent reporting, partially from the New York Times, about how colleges are increasingly offering not just CS majors, but majors that are specifically AI majors. And so in 2021, there were only, like, five schools that offered a major that was, like, called an AI major. Now, like, a report from Northeastern University's Center for Inclusive Computing says that there are 74 AI-specific degrees in the U.S. And now, like, that's a five-year period. In academia, this is lightning speed. You have no idea how fast they're creating these degree programs. Yeah, do we assume that is because there is now just suddenly a lot of demand in the marketplace for pieces of paper that say AI on them that come from top universities? Yeah, and it's also not just top universities. I would say the percentage of top universities and also the percentage of, like, small private universities and state schools that are starting to offer these degrees is, like, pretty similar. Like, there's demand, like, from students on, like, all sides of the pirate education for an AI degree. Because AI is, like, very hot right now, and people are like, what will keep me employed? The only thing booming right now, apparently AI. So what do I get if I earn an AI degree, and how is it different than, you know, whatever these degrees might have been called before AI became so hot? Yeah, so I've spent a bunch of time staring at the syllabi of these degrees, like, from Ivy League colleges, from smaller schools. And, like, basically every one of them, it's like a slightly modified CS degree, which honestly, that's correct. If it were not a slightly modified CS degree, I would be mad because, like, it wouldn't give you the fundamentals you need to prepare. But it's like, basically, if you're a CS major, you have, like, a certain programming load. This, like, squeezes the programming load slightly while still keeping the conceptual fundamentals, and then adds some math, like some linear algebra, some stats, some optimization that typically is, like, very useful if you're doing any sort of, like, AI application. So the idea is basically you can spend less time learning some of the basic, like, writing code by hand skills that we know you don't need anymore, but we will teach you more sort of CS-adjacent topics that could be useful to you if you ever need to train a large language model? Yeah, or train a small neural network. Like, I still, like, the vast majority of people employed as machine learning engineers are not training frontier large language models. Well, so tell us, as somebody who took your share of computer science classes in school, how are you feeling about CS these days in general? Yeah. So unfortunately, if you're, like, looking to AI-proof yourself or something, there's not, like, a totally airtight case to be made for CS. Like, some of the best econ literature so far, especially this study from Stanford called Canaries in a Coal Mine, that, like, indicates there's some sort of economic displacement effect from AI, shows this effect where, like, young person hiring in CS just, like, is worse compared to, like, a lot of other fields. Like, even, like, for that particular study, they compared, like, the decline in CS and the decline in, like, something like communications, which is also pretty, like, AI-exposed, and, like, CS is doing, like, the worst. But I will say, I feel like the evidence, it's just, like, it keeps shifting back and forth. It's like every two months, I find a new econ paper that, like, kind of changes my mind. And so one reason for hope, perhaps cope, for me and my brethren, is there's this paper from some people at the London School of Economics that I liked a lot that came out last month, which basically, it does something kind of complicated, but basically, it tries to isolate the effect of underemployment in, like, people in AI-exposed professions and just people who are working from home in general. And it does this partially by, like, isolating the times at which both of these things, like, became relevant. And basically, what they find is, like, the correlation between exposure in work from home and, like, specifically AI-exposed stuff is 0.77, where the highest correlation you can have, obviously, is 1. Like, that's, for any social science research, that's, like, really high. And then also, when they try to isolate those variables, work from home is a bigger explainer. I see. Well, so land that point for me a little harder. What does it mean that these two things are so correlated? I get what you're saying. It's sort of harder to understand what is AI exposure when, but sort of elaborate a bit. Yeah, so I would say, like, there are obvious theoretical arguments why, like, CS majors can get replaced because, like, AI just can write code. But at the same time, like, on a macro perspective, like, the current data doesn't sufficiently distinguish between several effects of work from home. One, companies hiring a lot of people during the pandemic, tech companies, when they had, like, comparative booms. Two, it's just harder to train young people remotely. And so it becomes more and more expensive to hire them in general in a non-CS-specific way. And that hypothesis, you should probably think is pretty strong and explains a lot of the stuff you're seeing, even when you're a CS major and you're like, nobody wants me to write code anymore. Right, right. Okay, so that is interesting and is maybe a reason why if you're entering college this fall, you may want to consider getting one of these AI degrees after all? Yeah, it's like at least an update against no hope for AI degrees. And okay, another piece of cope I would like to share, or potentially hope, is I think on a more conceptual level, like, as AI gets better at a lot of tasks, including programming, including, like, potentially, like, including machine learning engineering even, which is, like, something that is, like, a very specific AI skill that is pretty hard to train people to do, AI is getting better at that. At the same time, we don't trust AIs. We've never trusted AIs Before AWS even launched, when maybe some saw it as a strange side project for a bookstore, 20 years later, it's a $130 billion business. So it strikes me that you've already lived through a cycle of people initially underestimating something that turned out to be really big. But also maybe people at various times during that transition overestimated how quickly it would happen. So I wonder, as you're looking at the AI landscape, does any of that history rhyme for you, or does this just feel like a totally different moment? Yeah. No, it's actually a really good analogy, and it's exactly that. I think, if you roll back 20 years ago and we were first starting AWS, we had to explain to people what the cloud was and why Amazon was a part of that. And it was a completely different shift into how work was going to get done. And what we had to explain to people was like, Okay, it used to take six months for me to get a server, and now I could get one in five minutes. And if something fails, I can just shut it down and launch another one. And it was just that didn't... It's not how the world worked. So it explained kind of this new way of thinking about building applications where you just... You did it differently. You didn't think about how do you have one big SAN and a couple of servers that you kind of took really good care of and then slowly scaled, but rather you kind of had this disposable infrastructure that was serverless that you didn't really worry about the underlying pieces of it and focused on software and scaling. I view a lot of what's happening right now similarly. I think there's some differences too, which I'll call out too, but in the similar ways, I do think AI is going to also completely change how people build applications. And it's not just, do the exact same things you were doing before, but a little bit faster, a little bit cheaper. It's completely, like fundamentally changing business processes. It's changing workflows. It's changing business outcomes and customer outcomes and customer experiences and what was possible. And it's moving so much faster than that last change where it took people... It was really through the first five, 10 years, like you said, we thought was really, really rapidly, but we looked up five years, 10 years into AWS and still today, the vast majority of workloads still run on-prem. Today, even 20 years later, there's a massive amount of workloads that still run in on-premises environments. And I don't think AI is going to take that long to transform many businesses. And so I talk to customers and really everyone out there. I think that's probably the biggest difference is the speed at which this transition is happening. It's obviously a totally different technology, but it has really wide-ranging impacts across a lot of different industries. And even slow-moving industries are thinking, this is not something that I have 20 years to get ready for. This is something I have 20 months to get ready for, maybe, maybe that long, maybe not, but it's something that's going to happen much faster. Why do you think it's happening faster? I imagine that some of it is just businesses feeling like, hey, there's just a lot of competitive pressure for me to go fast. But is there something else in there that I should think about? I do actually think, in many ways, these two technology shifts that you're talking about that we've been really intimately involved with are quite complementary and compounding of each other. I think without the cloud, AI doesn't take off in the same way. If everybody is still running in their own data centers and you had to install the new update to the next model and then you had to scale your own GPU capacity in order to run it on that, and you had to deal with network bottlenecks, that just was never going to happen. But now because of the cloud, and because enough of workloads and enough of your data is in a cloud world and these models are available in a cloud world, that cycle is that much faster. And so I think those things probably compound upon each other and have led to it happening at a much faster rate than a previous shift. Well, so I was so excited to talk to you because of this unique vantage point you have as the person running AWS. Basically every company experimenting with AI is running through you. You have millions of customers. Every major model is on Bedrock. You can see what the enterprises are actually doing. Sometimes we hear enterprises have run a bunch of early experiments. They aren't seeing a huge return on that investment. Other companies seem like they're starting to figure it out. Give us a sense of like what you're seeing in that range. Yeah, I will say almost everyone ran a huge number of proof of concepts, right? So that was the common thing that everyone did was two or three years ago, they just told everybody, go see what this thing does. It's some magical new technology. We don't really know what it does. Go experiment. And this is what people did. And so people built a lot of really cool things. They didn't always have a good idea of what they wanted to get out of those experiments. They just wanted to kind of see what the technology could do. And so not surprisingly, there's not, across the board, most of those experiments didn't show great returns because they didn't actually have a plan of what they were going to have coming out the other side. And so, you know, the proof of concept of an interesting chatbot they built or an interesting content generation thing or interesting workflow, like it was cool and it worked, but it wasn't actually driving real business value because, you know, appropriately people had to learn about the technology first. Now what we're seeing is, okay, we understand, we get directionally where this technology could go. Now you kind of see enterprises and companies, all companies, like kind of saying, where can I get that return and how do I roll that into production? And it turns out that they're not the same thing. A lot of these proof of concepts, even if the proof of concept actually was directionally where they want to go, now you have to think about how do I think about data security? How do I think about governance and who has access to this data? How do I make sure about security of these agents where they can go and where they can't go? How do I think about compliance? There's a bunch of like actual, you know, regulations and rules and operating pieces that you have to think about before you get in. And how do I think about real business value? What is the cost of this thing? What's the either cost savings or a revenue increase because I rolled it out? And so as we, and interestingly, that's when, and we have this, long had this hypothesis that when customers started going into production, this is when they're going to really want this to be in AWS because this is the parts where AWS is great. Number one, it's where their data lives. It's where their production applications live. And it's where their security policies are, their credentials about access, and they really trust AWS with their data, right? And so they trust us to not leak their data out. They trust us to make, to when we say you're going to define the guardrails of what a model can do, that we actually hold it inside of that. And so security, operational excellence, and the existence of all the rest of their production applications is when customers said, great, now that I've done proof of concepts and I've done them everywhere, sometimes on-prem, sometimes on my laptop, some other places, like this is where I want to deploy them. And that's the cycle that I see us in now, where those customers are now deploying those workloads into AWS and you're seeing real business value. In fact, I was talking to a room full of CIOs just a couple of months ago and asked, okay, raise a show of hands. It was probably a room about a hundred people. How many of you either today are seeing materially positive ROI or have a path in the next couple of months to really high ROI to your AI investments? 90% of hands went up. And so there, which is totally different than a year before where they're like, no, it's just a cost model for me, right? And so the people are starting to see that, right? It's not, and by the way, it's the first couple of percentage points of that transition of where they're going to be, but people are starting to see that return to the business. They're starting to see, okay, this is where I see real value. There's plenty of other proof of concepts, by the way, where they're shutting them down because they don't actually see that. And so you have just as many and probably way more where people are shutting them down. They're not seeing those returns, but they're really doubling down on the ones where they see that benefit. The obvious place where people are seeing returns is in coding, software engineering. It seems pretty clear that there's just a ton of value there and the labs are selling tokens as fast as they can make them. But because you're looking across the entire industry, I'm curious if there are other sectors where you're starting to think, oh, like these folks have really figured it out. Like they cracked something. Yeah. I think like coding is a clear and obvious one. And I think I actually still think we're in the early stages of where that's going to evolve to. But the whole software development life cycle is, we're seeing massive improvements and efficiency and gains there. The other ones though, is if you take that a half a step further, is those gains in software development are enabling agents to be much, that reasoning, ability to write code and to get work done is then the next stage is it's enabling agents to autonomously go do business processes things. And so we're starting to see people roll this out where it's, you know, telcos thinking about network optimization or financial services companies Yes, that is definitely true. I don't know which ones, by the way, or I would guess be in the VC world. But that's going to be true. I think we're in a position where, you know, I feel very good about our investment, but it's a different position. And I think everybody in the industry is slightly different spot. Yeah. This capital expenditure that we're seeing, I think it's fair to say, has not been universally well-received. Amazon's hometown of Seattle just passed a moratorium on new data center construction. We're seeing a similar backlash around the country. Amazon wants many communities to say yes over the next decade to data centers. What do you think the objection really is here? And do you think that these folks are right about anything? Well, look, I think there's a broad swath of things that people are appropriately worried about. I think they're worried about, you know, I have a big, ugly building next to me that I didn't used to want to have. Or is it going to be really loud? Or is it going to be bad for the environment? Or is it going to... And actually, one of the most common things you hear now is, is it going to make my electricity rates go up? Because it's a natural conclusion. It's not right, by the way, but it is a natural conclusion to say, if there are, you know, 100 units of energy and they used to cost a dollar, and now somebody comes in and says, I want 50 units of energy, supply and demand says the price goes up. Practically speaking, it's not quite how it works. And actually, we've, because there's a number of things, price actually depends on your peak to average load. So actually, if you bring in more average load, which data centers bring in a really stable average load, it actually shrinks your peak to average and actually reduces the cost of an incremental unit of electricity to a consumer if you're operating on that same grid. And so, in fact, we have one of our very largest data center implementations or construction projects is in Indiana, which we've talked about Project Rainiers. We've talked about publicly. It's a very large project. And we've announced with the governor there that we actually reduced the average cost of energy for the citizens of Indiana because of this project. So their cost of energy actually went down because of that. And I think that's, you know, it's harder for people to understand that because you have to understand actually how energy is charged and things and other things like that. But that is more what it is. And we've, we've committed, you know, with the administration that we'll pay for all of the power that, you know, we'll make sure that we bring power. We'll make sure that we bring cost upgrade grids when necessary. There's a bunch of those things that we've committed to. We just announced this week, actually, from an environmental perspective, that we use 7x less water in our data centers than most data centers do. And we're well on our path to being water positive by 2030. So from an environmental perspective, people worry about it appropriately so. Like it's a, it's a big worry. And we're very committed to carbon zero and water positive impacts in the environments. And you know, I think, look, a lot of the places where we operate, we bring a lot of really high paying jobs. And so in a world where people are excited about high paying jobs, where they want some of these skilled labor, where they want high paying, you know, they, in communities that that are looking for, for more employment, they actually quite welcome us and like to have us there. So, you know, are we going to put a data center inside of the city limits of Seattle? Like, probably not, but that's not really where you put them anyway. And so, you know, I think that's okay. And I think you have to be thoughtful about where you put them and how you impact the communities. And not everyone is as thoughtful as we are, by the way. And so there's another problem where everybody kind of gets swept up if there's a couple of bad actors who, that maybe don't follow some of the environmental regulations or don't kind of work with the local communities. But we spend a lot of time making sure that we do that. If you're able to do what you say, give people good jobs, reduce the environmental impact, maybe make sure the building isn't ugly. How optimistic are you that that actually changes the view here? Is there a fear that maybe people are just sort of just uncomfortable about AI in general and they're just going to kind of reflexively resist anything that feels AI adjacent? You know, look, I think there will be advocacy groups that will say things all of the time. That's true for now and will be true forever. Our view is, look, I do think there's a risk of not everyone kind of, not everyone necessarily acting in that same way and thinking about the community. And I think that could, could cause problems for everyone. And so I think it's important that everybody actually takes those paths and goes towards carbon zero and goes towards water positive and thinks about their impact to the community and makes sure they pay great wages. So I think that's a super important thing. And, and we would advocate for everybody doing that, even if it adds a little bit of cost to your operating. I think it's important. You know, look, I think there's, I haven't yet talked to the consumer who wants to turn off their Netflix and doesn't want to use Airbnb and doesn't want to call an Uber and doesn't refuses to use Gload or ChatGPT or doesn't want to search on Google. Like at some point, like, you know, you can't have your cake and eat it too. And those things are all powered by data centers. Most of them use AI to accomplish what they're doing. And, and so, you know, I think there's, there is kind of a reality there where there's no, there's no like, I don't want any data centers, but I still want all of the things that come out of data centers. That, that's not really a reasonable trade-off. And at some point people have to make that call and realize that. Let's talk more about jobs. You called replacing junior employees with AI, quote, one of the dumbest things I've ever heard and told Wired in December that never hiring junior people is a quote, a non-starter for anyone trying to build a long-term company. There are other views in the industry. Dario Amadei, CEO of Anthropic, has said that AI could wipe out half of entry-level white-collar jobs in five years. What's he missing? Well, here's the thing that I think that there's subtle differences in there that I think are important. I do think that half of white-collar jobs may change, but it doesn't mean wipe out and change are different, right? And so, you know, Excel wiped out all of the jobs of people who were hand calculating things, but it, they, those people then learned how to use a computer and there you go. They, they, now they had a job again. And so I think that the key thing is that not to look at a, a still picture of the world and say like, that job is not going to exist, so I guess those people won't have jobs. New jobs will be created. And I firmly believe this fact because I will say that, you know, like, I think if you believe that half of jobs get wiped out, like the whole economy collapses on itself and everything goes away and then you're not going to have AI and then you have to go back to those other jobs at some point. Like that, it just, like the math doesn't work out. And so, but, but, but more than that, like these, these jobs, these, these, um, the, the AI, it's a new technology that is inventing new things and we're already seeing new jobs get created. They're different jobs. And so what I tell people at Amazon is, look, there are going to be lots of jobs. And then you talk about entry-level jobs. Number one, they're your cheapest employees. They haven't learned bad habits. You can teach them the culture. They're willing to learn the new tools. They're some of the very best employees you can possibly have. But I tell all of our employees, if you, if, if you look at what your job was two years ago and you look at what your job is going to be in two years, it's going to be vastly different. You're going to have a job. You're going to have probably a more exciting job and an interesting job, but you're going to have to be willing to learn. And I actually think it's one of the things we start to look for in employees is not what are the skill set you have, but do you have the ability to learn? Do you have the willingness to, to dive in and learn new things and the agility to reason about problems, to think about how do we solve customer problems and how do I apply that to the capabilities that we have? And I think when you have employees like that, that have that mentality that I want to learn, I want to lean in and new, do new things. There's a crazy amount of opportunity in front of us to run really fast, build new businesses, deliver value to our customers, reduce costs. All of those things are huge opportunities. And I think I find, and there's a reason we're, we're hiring 11,000 interns and new college grads this year at Amazon. And so, because they come in with an energy and excitement, a new view on things. And if you just have the exact same people that you've had for the last 15 years, you don't get that energy and excitement and new ideas. And, and so I, I think there's a number of reasons why that personally, I find that logic A couple of months that intersect with jobs. In April, you launched one called Amazon Connect Talent, which is an AI recruiter that autonomously schedules calls and conducts voice interviews with job candidates. It can do it around the clock with no human involved. That sounds like it's pretty close to automating a full job to me. Do you see it that way? Yeah. Yeah. And I would say, this is, look, I think this is something that Amazon has been doing for a really long time, is we automate work so that our employees can start to do higher-value work and we reduce costs for customers. And this is not a new thing. This is something that we've literally been doing for probably 25 years since we started the company. And it's something that we push our employees to think about. Like, how do you automate parts of your work so that you can have broader scale, so you can get better leverage on what you're doing and do more? And so, you know, we don't pack boxes in the same exact way that we did it 20 years ago. There's a lot of efficiencies. And we think about, how do you organize differently? How do you think about supply chain differently? How do you think about robotics differently? We think about the software development lifecycle. We think about, how do you, you know, automate parts of this job so that you don't have to do it? This has always been part of what we do. Now, AI is a great new tool that's allowing us to do more here. But this has always been true. And so, you know, if you think about those things that you just said from a recruiting perspective, our recruiters would prefer to, like, really handhold candidates once they're through that pipeline, answer some of those specific questions that they have, go out and really source great talent or kind of think about... Once they're, like, you know, inputting, like, inputting, like, all your details and things like that, it's not what makes up a recruiter's day. It's not what they get excited about. It's not why they got into that job. And if you can automate that away, you know, now we can get more talent into the pipeline and get better leverage and we'll find other things for those folks to do. So it absolutely is automating away jobs and replacing them with new things to do. Right. So you still think that in this world, like, you will still have recruiters. They'll just be working on different kinds of things. Our recruiters are some of my most important partners. Like, you know, we're recruiting every single day. And I would much prefer our recruiters to focus on going and finding great talent than doing the blocking and tackling of, you know, inputting details and making sure that they get the email of, like, what time their interview is going to be. Like, those are things they don't need to spend time on. Yeah, I mean, I think the whole question is just, you know, across how many different categories of jobs is there this kind of adjacent work to move into? Like they say that when spreadsheets came out, we lost something like 400,000 bookkeeping clerk jobs, but we created 600,000 accountants because making calculations cheaper created more demand for analysis. On the other hand, like when word processors came out, there were whole pools of typists who did have to find something else to do. So I'm just curious, like, kind of what that mix winds up being. How many current professionals have something kind of right next door that they can still do, you know, once AI takes away some of the other parts? Yeah. I don't know the answer to that other than, like, every time I've actually dug into this, I see us hiring more, right? Like the current one, the biggest people are worried about is, like, maybe there won't be software development jobs. There are more software development developers being hired this year than ever before in history. So, like, just the evidence doesn't point to it. And, you know, you can go through every single job that exists in the world, and that'll take us a long time, probably longer than we have today. But, and I'm not going to know those answers other than history has shown us that this is the case. And if we're really creating value, then that is creating more opportunity to go do more. Let me ask you another return on investment question that I think is on the minds of lots of folks in business today, and that's about token maxing. The FT reported that some Amazon employees had used an internal agent platform called MeshClaw to run unnecessary tasks that inflated their scores on internal AI usage leaderboards that the company had created to encourage AI adoption. Amazon was not the only company that had these leaderboards. I'm told you've since gotten rid of them, but I'm curious what you learned from that episode and how you're measuring AI adoption now. Yeah, I think the most important thing is to make sure that you actually measure the thing that you want to measure. And so we never intended to have those leaderboards that way. It really was a way of pushing people to say, like, I want you to do this AI native piece. You measure it in the wrong way, and people figure out a way around the measurement as opposed to the goal. So it was never the goal for us. I think that's actually, it's a good rule for everybody that you get what you measure. And so you just got to make sure that you measure it. And for us, it's not that we, like, you know, I've seen people who are like pulling back and saying, like, oh, I'm not sure, like, AI native coding is worth it. Like, to me, that is bonkers. It is a great, it is like the biggest efficiency gain that I've ever seen. I'm more than happy for our competitors to make those choices, but if they want to. But for us, that's not it. But the thing you actually want to measure is not, like, can you use the most tokens? Like, that's like a use, like that, the only person that wins there is the person selling tokens. Like, you really want to measure, like, who's getting great output, right? Are you having more code check-ins? Are you creating, are you pushing more features to customers? Are you, you know, closing bugs faster? Are you, like, all of those kind of things. I think those are the right measures. And so we, it wasn't like we were like, ooh, like, we don't want to do this, like, measuring of tokens anymore. We just realized we had the wrong measure in place and just quickly changed to kind of measuring some other outputs. And I think others had this kind of at a much more extreme than we did because ours was never quite, I think others actually did have that as the output. Ours was never really that. It was really more just kind of measuring to make sure that we were, kind of people were using the AI tools in a good way. But I think that's right. And it's a good rule on generally, you know, measure what you want to get. How are you thinking about, like, token spending in general? Where do you want employees to have freedom to experiment and where do you need to rein them in? I think it's like, at some point, it's just a resource and it's a cost. And I think people want to, you're like, you, with anything that costs money and is a resource, you want people to be thoughtful owners. And actually, one of the leadership principles we have at Amazon is ask like an owner. And so we want all of our employees to think like an owner. And if they think the investment is good, then we give people some guardrails and things like that. But generally, you know, if they're thinking about this is a good investment, we don't want them wasting money. And in the same way that we think about running EC2 instances, right? It's not really that materially different. You know, people didn't launch 10,000 EC2 instances just to see if they could, like, because it cost them money and they would see the bill. But we give employees the ability to go launch EC2 instances if they need them. And so in the same way, we want our employees to be using AI to make themselves more efficient and effective and deliver more value to our customers. And so we encourage that and we're building tools to help them with that. Like, I think one of the things that's driven up a bunch of costs is that people have kind of naively just said, great, I'm going to use the best model for every single thing. And it turns out that, like, you don't need Opus 4.8 to generate code. Like the Sonnet model or even the Haiku model does a really good job at that. Now, when you're going to plan and do some of those, like, higher reasoning things, like using Opus 4.8 is a really good model for that. And so actually in Kiro, which is our Amazon coding tool, we do a lot of this for customers where we pick the right model and then we appropriately kind of use the amount of your token budget and things like that. And so we kind of use cheaper models when we can that are faster, that are sometimes they're faster, they're less expensive, but they accomplish the results. And so that's a good way of balancing some of those things where, if we can give tools to customers where they don't have to build a lot of this complicated logic, but they can just get the benefits of those. I think that's one of the things that helps. But for us, we also kind of want our employees to think like owners and to know, and the more you can expose those costs, I think it enables people to do that. Well, let me close by