OpenAI’s New Agent Stack: Computer Use, Decisions API, UltraFast, Dots—Ari Weinstein & Nikunj Handa

1. 0:00 – 1:12

   Hook

   1. AWAri Weinstein0:00

      But now Computer Use is, like, faster at accomplishing tasks than, like, the average human probably in, in most cases. Um, and I think that the next frontier is to have Computer Use be, like, literally superhuman in its performance, where it actually is as fast or faster at using software than, like, expert computer users like us. Um, and I think that'll be really consequential and exciting when that happens because I think we'll be able to, all of a sudden, build products, um, that, um, provide just much more real-time experiences-

   2. SWswyx0:31

      Like, four weeks ago, this was not on the dev roadmap, right?

   3. AWAri Weinstein0:33

      No, not at all, no.

   4. SWswyx0:34

      Okay.

   5. AWAri Weinstein0:34

      This is, like-

   6. SWswyx0:34

      Wow [laughs]

   7. AWAri Weinstein0:35

      ... uh, dev-inspired and, like-

   8. SWswyx0:36

      I think you're officially the first frontier lab to, like, tune in-

   9. AWAri Weinstein0:40

      Yeah, yeah

   10. SWswyx0:40

       ... and adopt this.

   11. AWAri Weinstein0:41

       Yeah, yeah. I, I feel like, uh, OpenAI has such a strong, like, hacker culture, and, like, people are just, like, they get excited about things. And so, uh, guy from Inference, this one awesome guy from, uh, the Infra team are like, "Ah, this is amazing. We're gonna, like, hack on it." They built a prototype, it, like, works, and now we- we're just, like, hill climbing on latency and trying to make this as fast as possible, and we wanna, like, launch it in the coming days. Uh, so as soon as we hit our, like, latency target, we'll, we'll try to get this out. [upbeat music]

2. 1:12 – 4:04

   OpenAI DevDay: Dots, GPT-6.1, Agents API, and Decisions API

   1. SWswyx1:16

      Okay, we're very excited to be here. Today is OpenAI DevDay, special podcast-

   2. AWAri Weinstein1:20

      We're the first podcast after the livestream

   3. SWswyx1:21

      ... first podcast. We have Ari here, who leads the product and engineering team for Computer Use agents. Before we kick in and dive deep on Computer Use, you wanna give a quick recap? What was announced? What's the quick slew of announcements you guys had today?

   4. AWAri Weinstein1:35

      Yeah. Um, yeah, it was a super exciting day. Um, we just got out of the keynote. It was really sick. Um, there were a bunch of Computer Use announcements that I think are worth thinking about. We have, ah, Dots, um, which is the new, um, sort of personal assistant product, and, um, that has some really exciting Computer Use features. There's GPT-6.1 Sol, which is this amazing new model, um, that I think is particularly great for Computer Use 'cause of, um, sort of the cost and- and speed, uh, advantages. I think, I think we shared that it's, um, a fifth of the cost as Astra and a seventh of the cost if you're looking at Computer Use specifically, um, which is really amazing. Uh, sorry, there were so many things. I'm trying to sort through it.

   5. SWswyx2:14

      And the API.

   6. AWAri Weinstein2:15

      Agents API, which now has Computer Use in it, which is really cool, 'cause now developers can build on the same Computer Use, uh, that is, is part of Codex, um, and, and ChatGPT. Um, and then there were some demos of our existing Computer Use features like App Shots, where you can take the context of something you're doing on your computer and bring it into Codex and ChatGPT really fast and, um, and then, like, native Computer Use on your Mac where Roman had it taking screenshots of his app, uh, automatically, and he could do other things on his computer while Computer Use was using his applications. Um, so yeah, really exciting keynote.

   7. SWswyx2:46

      And not to mention, there's Decisions API.

   8. AWAri Weinstein2:48

      Decisions API.

   9. SWswyx2:49

      Uh, uh, off the bat, are they all the same model? Like, this is... Are the same data set distilled to different models?

   10. AWAri Weinstein2:56

       Um-

   11. SWswyx2:56

       Like, basically, like, is Computer Use using Decisions API, or are they, like, kind of separate?

   12. AWAri Weinstein3:00

       So what's really cool about the Decisions API is it, you know, it, it, it has all these new capabilities. It does inference in parallel. Um, it doesn't have reasoning. It's a smaller model, um, than the ones we use for Computer Use. Um, and so those, those capabilities make it really fast.

   13. SWswyx3:18

       Yeah.

   14. AWAri Weinstein3:19

       Um, they also make it a little bit less good at doing, like, long-horizon, um, sort of sophisticated tasks, and so I think it's... I would say it's still an open area of research for how we, like, bring those approaches together. But, uh, yeah, I'm really excited to see what people build with the Decisions API.

   15. SWswyx3:36

       One of the interesting things is Dots now have attached personal computers.

   16. AWAri Weinstein3:40

       Yeah.

   17. SWswyx3:40

       So it seems like they're very much more persistent. You've been using them for a while. How should people push the bounds? Like, what should people aim for? What should they try? Personally, right now, I use it for a lot of customer service like-

   18. AWAri Weinstein3:52

       Ooh

   19. SWswyx3:52

       ... "Oh, this was wrong. I don't wanna sign in, I don't wanna authenticate," find whatever and just get it fixed.

   20. AWAri Weinstein3:57

       Yeah.

   21. SWswyx3:57

       How, how should we push further? What should people try?

   22. AWAri Weinstein4:02

       Dots are a really

3. 4:04 – 6:11

   Dots and Personal Cloud Computers

   1. AWAri Weinstein4:04

      cool product because each Dot has access to its own Linux virtual computer in the cloud, uh, which is different from our other products. Um, you know, traditionally we've have access to a browser in the cloud, or it has access to your own computer, but now you get your own entire Linux computer in the cloud. And so we can run full desktop applications. Um, and it can also use a web browser. And so yeah, you know, I think the powerful thing about Computer Use and the reason why I think it's so, uh, exciting is because it makes it so that the agent can do anything you as a, as a person can do, um, because all the software in the world was designed for humans, and now agents can use that same software, uh, and you can delegate to the agent. So, um, yeah, like anything that you would do on a computer, you can ask a, a Dot to do. Yeah, I think what particularly is useful is gonna really depend on who the end user is and what, what's valuable in their life. Um, uh, but yeah, I would just start by thinking about, like, what are the th- things that you spend time on, and how could you delegate those to an agent?

   2. SWswyx4:58

      Yeah, a lot of flight booking and shopping and honestly even, like, playing a game or whatever, right?

   3. AWAri Weinstein5:03

      Totally.

   4. SWswyx5:04

      Yeah, yeah.

   5. AWAri Weinstein5:04

      Um, yeah, I don't know. For me, um, something I did recently, um, I've been working on-- I, I've, um, subscribed to a meal prep service 'cause I was trying to, like, eat healthy, you know? And I really like this meal prep service I found because it lets me customize the meals I order to, like, a high degree of granularity. So I can say, like, "I want this many grams of chicken and this many grams of, of, of rice." Um, but it was so complicated, it took me two hours to do an order, and I found that I could ask Computer Use to do it for me, and it did it in 15 minutes. Um, so I, I actually saved two hours. Um, it [laughs] both did it eight times faster than I could, and it saved me two hours on, on GPT-6.1 Sol. Um, so-

   6. SWswyx5:44

      Yeah

   7. AWAri Weinstein5:44

      ... those are the kinds of tasks that I f- feel like, uh, are, are really powerful.

   8. SWswyx5:47

      As a creator, I can tell you automatically, immediately, my number one use case is automating YouTube-

   9. AWAri Weinstein5:52

      Nice

   10. SWswyx5:52

       ... uh, because, uh, YouTube doesn't expose a lot of things via API.

   11. AWAri Weinstein5:55

       Yeah.

   12. SWswyx5:55

       And you have to just put it in a VM and just, like, run it, uh, for like, let's say, let's say their A/B testing feature or making community posts. None of this is available via API 'cause they hate developers.

   13. AWAri Weinstein6:04

       [laughs]

   14. SWswyx6:05

       Um, anyway, so, uh-

   15. AWAri Weinstein6:07

       I've heard that from our developer experience team too.

   16. SWswyx6:08

       [laughs]

   17. AWAri Weinstein6:08

       They, they use it with YouTube a lot, yeah. It's really

4. 6:11 – 7:16

   Why Computer Use Is “180 Degrees Different”

   1. AWAri Weinstein6:11

      awesome.

   2. SWswyx6:11

      Um, so I wanna draw f- you know, uh, let's say I wanna get a little bit spicy. One of our, the leading AI podcasts, our, our, our friends, uh, is famous for saying that Computer Use hasn't advanced in the last two years.

   3. AWAri Weinstein6:24

      Yeah.

   4. SWswyx6:24

      Uh, which is a very interesting statement, and I think you're one of the best people in the world to, uh, to, to talk about this, like how have things have progressed, right?

   5. AWAri Weinstein6:31

      Yeah, you know, they said that a few months ago, I think, and I hope they have a different perspective now because Computer Use is, like, 180 degrees different than it was.

   6. SWswyx6:38

      He's a, he's a tough guy to impress.

   7. AWAri Weinstein6:39

      Yeah, okay. [laughs]

   8. SWswyx6:40

      Um, but you know-

   9. AWAri Weinstein6:41

      We're working on it

   10. SWswyx6:41

       ... you worked on-- y- you've, like, basically spent your whole career working on, like, some kind of computer automation, right?

   11. AWAri Weinstein6:46

       Yeah.

   12. SWswyx6:46

       Like Shortcuts-

   13. AWAri Weinstein6:47

       Yeah

   14. SWswyx6:47

       ... at Apple, and then Sky, and then, and then joining, joining OpenAI. Can you draw, like, what your through line is for, like, what is driving you and what-

   15. AWAri Weinstein6:57

       You, what wasn't possible back then maybe-

   16. SWswyx6:59

       Yeah

   17. AWAri Weinstein6:59

       ... and like what your sort of milestones were.

   18. SWswyx7:01

       I, I guess to add on to that as a follow-up question, what's the major change from using Codex Computer Use from like last week-

   19. AWAri Weinstein7:09

       Yeah

   20. SWswyx7:09

       ... through to today? Is it model? Is it Dots? Is it harness? So all the history plus what, what really just changed in today's announcements?

5. 7:16 – 10:36

   From Sky to Self-Debugging Computer Use Agents

   1. AWAri Weinstein7:16

      Yeah. On the through line, I guess I've always been excited about automation and helping people automate tasks because then you can like save time in your life and focus on things that are more important to you than like operating a computer very intricately. And so, um, yeah, that, that was why we worked on some of those products. I was at Apple before. Um, we, we made a company called Sky. Um, we ended up joining OpenAI, which is really exciting. Um, and I think something that was-

   2. SWswyx7:39

      And almost-

   3. AWAri Weinstein7:39

      Yeah

   4. SWswyx7:39

      ... like you have to hack around Apple until Apple was like, "Fine," like, "we'll just hire you and you can just work-

   5. AWAri Weinstein7:44

      [laughs]

   6. SWswyx7:45

      ... on the inside," right? Like [laughs]

   7. AWAri Weinstein7:46

      It was, it was a cool place to get to work.

   8. SWswyx7:48

      [laughs]

   9. AWAri Weinstein7:48

      Um, what was really interesting looking back at Sky is we were, we were working on computer use there as well, and the models were so much less capable. And now the models just in the last one year have become extraordinarily capable at computer use. I think the biggest delta that I see is before they could like reliably start tasks, but then they would run into problems, and now they're really good at debugging. They're really good at trying again, introspecting what is and isn't working. Um, and I, I think we've also brought the computer use... The, the computer use field itself has moved forward. I think we're using more techniques. Um, now computer use, um, often writes code. So if you actually look at it in Codex and you expand the tool calls manually, you can see that it's not just doing one action at a time, it's actually writing JavaScript code that it executes, th- that, that the computer executes to perform sometimes many actions at once, which is a great, you know, speed up and great capability. We use more accessibility sort of multimodal interfaces. So, um, the model may use screenshots, it may use accessibility, it may use Playwright, um, it can use a lot of different mechanisms, uh, based on the task at hand. Um, and then yeah, the, the, the model acceleration has been, has been just amazing. So, um, yeah, what's different today? I think we're making computers better all the time, so I think just like one day's difference, um, is, is probably a little bit less consequential than like even the past month or the past two months. Um, but, uh, yeah, I think the, the computer use in Dot is really exciting as well as, um, the, the new model that we came out with.

   10. SWswyx9:14

       On the keynote, Tejal was mentioning 7x improvements in computer use speed, um, a lot better on a few benchmarks. How do you guys think about measuring it? Computer use is one of those things where, as you say, you know, it's improvements over time.

   11. AWAri Weinstein9:31

       Yeah.

   12. SWswyx9:32

       Is it harness? Is it model? Is it post-training?

   13. AWAri Weinstein9:33

       Right.

   14. SWswyx9:34

       How do you guys look at it internally about measuring how good it is, and w- what were the changes with the new model?

   15. AWAri Weinstein9:40

       We actually have a bunch of different ways of measuring it, some of which are on different permutations and configurations of the harness. It's a bit of a complicated story because, you know, our production products have, you know, some more safety checks, um, and, uh, you know, those are configured differently based on the, the, the needs of the, of the task at hand. So there's a lot of ways to, to measure it. But I think regardless of how we measure it, we find pretty consistent gains. Um, and those gains are, um, sometimes in the harness and sometimes in the model. Um, and yeah, I, I, I was really excited by this result that GPT 6.1 is even more cost-effective for computer use than its baseline cost improvement as compared to Astra. It's like really cool to see.

   16. SWswyx10:21

       Yeah. I mean, one of the visuals I really liked from the livestream was that, uh, you're sort of improving the Pareto frontier of, uh, your, um, curve, and there was a lot of talking about how you're improving it together with the harness.

   17. AWAri Weinstein10:35

       Yeah.

   18. SWswyx10:35

       Can you give

6. 10:36 – 13:21

   How Computer Use Sees and Operates Software

   1. SWswyx10:36

      some examples of aha moments that you had, uh, whether it's on like model driving the harness, harness driving the model, whatever?

   2. AWAri Weinstein10:43

      I don't mean to repeat myself, but I think like introducing more modalities has been really powerful.

   3. SWswyx10:47

      Okay.

   4. AWAri Weinstein10:47

      One, one more specific example of that is, um, in the past, I think we saw a lot of computer use, um, products had to spend a lot of time like scrolling, you know? So it would like take a screenshot, it would try to do something. It would be like, "Oh, I gotta like scroll down to the next page of results," and then it would take a screenshot, and then it would try to do something. It would scroll down again. And so I think, um, with, with accessibility and other, and, and, uh, direct access to the DOM and other things like that, now the language model can actually see like an entire page or an entire application. It can write code that can do multiple steps at once, and so I think those have been probably the biggest single aha moments. There's like a lot of tiny ones that are less exciting in comparison. But actually, we do find also that a lot of speed improvements are driven by like a lot of little paper cuts that we gotta go in and introspect.

   5. SWswyx11:32

      Yeah. A lot of really hard engineering.

   6. AWAri Weinstein11:34

      Yeah.

   7. SWswyx11:34

      I mean, app shots in general, like, like, uh, I think people don't quite get the difference b- because there's like a nice visual in Codex when it-

   8. AWAri Weinstein11:41

      Yeah

   9. SWswyx11:42

      ... you take an app shot, but they don't maybe get the difference that, um, you are able to actually drive each button, and you have the, you have each text, um, in, in, in a very optimal representation.

   10. AWAri Weinstein11:52

       Yeah. Yeah, exactly. Yeah. It's kind of fun actually. [laughs] If you want to be like really nerdy about it, you can go into Codex, take an app shot by hitting the two command keys. So you grab the content from whatever app you're working with, bring it into, into the, uh, Codex or ChatGPT chat, and then the... If you click on the attachment and you click on this like little tiny button in the top right, you can see the raw text, and you see the raw accessibility representation. And yeah, we've put a lot of work into, um, putting-

   11. SWswyx12:15

       Just dumping everything out. Yeah

   12. AWAri Weinstein12:16

       ... dumping it out, but also making it token efficient, um, uh, d- doing it efficiently. There's like a bit of an art to it. And, uh, you know, it turns out that the same technology that was invented for humans, um, you know, who maybe have accessibility needs, who want to use a screen reader technology, that technology is really helpful for them to be able to use computers. It's also really helpful for LMS to be able to use computers, so that's been like really fun to get to work on.

   13. SWswyx12:38

       For context, I feel like a lot of people don't understand app shots. They don't even know it's a feature.

   14. AWAri Weinstein12:42

       Yeah.

   15. SWswyx12:42

       It's when you double hit command, it pulls in what looks like a screenshot.

   16. AWAri Weinstein12:46

       Right.

   17. SWswyx12:46

       And you're like, "Oh, why have I opened up just a screenshot and thrown it in?" No, it's actually pulling all the metadata, all the code, everything.

   18. AWAri Weinstein12:51

       Yeah, exactly. Yeah. So it's like, you know, if you take a screenshot of a webpage that has a link- The screenshot doesn't include where the link goes. It doesn't include, you know, maybe you take a screenshot of your calendar, the ca- event ti- titles are truncated, you know? But when you take an App Shot, it gives, like, the language model, like, full context about everything, and, um, that lets it, uh, just sort of, like, do, do much more.

   19. SWswyx13:13

       Yeah.

   20. AWAri Weinstein13:14

       I don't-

   21. SWswyx13:14

       For those who want to see more, uh, Jason Liu, I invited him to do a full workshop on this, uh, at AI Engineer.

   22. AWAri Weinstein13:19

       Amazing.

   23. SWswyx13:19

       He did a great job.

   24. VIVibhu13:20

       I have

7. 13:21 – 17:15

   From Faster Than Humans to Superhuman Computer Use

   1. VIVibhu13:21

      a broader vision question on-

   2. AWAri Weinstein13:23

      Yeah

   3. VIVibhu13:23

      ... Computer Use agents. So your example of take a screenshot, scroll page, take a screenshot is where we were.

   4. AWAri Weinstein13:29

      Right.

   5. VIVibhu13:29

      Today, they can automate a lot. Uh, what are the bottlenecks? Is it models? Is it harnesses? What... Where do you see it going in, like, two years? Do you see it just running for hours? How do we get there? Any, any predictions on where Computer Use goes?

   6. AWAri Weinstein13:43

      Yeah, I mean, I, I think what's really crazy that I think, you know, the team's accomplished over the past couple months is that now Computer Use is, like, faster at accomplishing tasks than, like, the average human probably in, in most cases. Um, and I think that the next frontier is to have Computer Use be, like, literally superhuman in its performance, where it actually is as fast or faster at using software than, like, expert computer users like us. Um, and I think that'll be really consequential and exciting when that happens because I think we'll be able to, all of a sudden, build products, um, that, um, provide just much more real-time experiences, and I think it'll also... Lowering the barrier to entry of, um, or the activation energy, I suppose, of using Computer Use, I think will make us start to default to doing certain things in agents that we've become accustomed to doing manually, and I think that's exciting also 'cause it'll save us a ton of time. Um, and I think there's a, the, you know, there are a lot of different little paper cuts and bottlenecks that are sort of standing in the way of that. Um, um, I think that there's, um, yeah, there's things on the model side, there's things on the inference side, there's things on the harness side, there's things in the, in the representation. You know, we find that as Computer Use gets faster, we're increasingly bottlenecked by just, like, the speed of doing an operation. Like, for example, um, you know, a non-trivial amount of time in our benchmarks of Computer Use tasks is actually, like, like, let's say you're automating a task on doorDash.com. Like, a lot of the time is actually waiting for doorDash.com itself to load, you know? [laughs]

   7. SWswyx15:16

      Yeah, then you just write a wait, and then you execute the wait.

   8. AWAri Weinstein15:19

      Yeah, totally.

   9. SWswyx15:19

      [laughs]

   10. AWAri Weinstein15:19

       And, and you wanna get... Yeah, actually, it's, it's actually really important that you get that de- like, you want as little delay as possible between when it finally finishes loading and when you go and-

   11. SWswyx15:28

       Yeah

   12. AWAri Weinstein15:28

       ... trigger the LM, do the next action, which is actually-

   13. SWswyx15:30

       It's, it's-

   14. AWAri Weinstein15:30

       ... itself a statistical science.

   15. SWswyx15:31

       Like, an event-driven way maybe to do that.

   16. AWAri Weinstein15:34

       When possible, you want it to be event-driven-

   17. SWswyx15:36

       JavaScript has some load event

   18. AWAri Weinstein15:36

       ... and, and JavaScript has load events for, or the web browser has load events for web navigation, but there's other types of events that actually really can't be event-driven, so there's a lot of complexity, uh, to doing this correctly.

   19. VIVibhu15:46

       The one that comes to mind is, like, chatting with customer service.

   20. AWAri Weinstein15:48

       Yeah.

   21. VIVibhu15:49

       Reply could take 30 seconds, could take three minutes.

   22. AWAri Weinstein15:51

       Oh, right. [laughs]

   23. SWswyx15:52

       [laughs] I have dealt with so many bots with Codex. Uh, it's great, but I also wonder if the other side knows that they're talking to a bot, 'cause I'm, like, answering in complete sentences, like, I'm capitalizing correctly.

   24. AWAri Weinstein16:02

       That's hilarious.

   25. SWswyx16:03

       Like, I'm giving full num- full reference numbers and everything. Like, it's too, it's clearly too good. I, I don't care, like-

   26. AWAri Weinstein16:08

       [laughs]

   27. SWswyx16:08

       ... it's, like, I'm just, like, trying to get my support case.

   28. VIVibhu16:09

       I've prompted it to, like, you know, "Don't pretend you're a bot. Be, be very annoyed human."

   29. SWswyx16:14

       Yeah. [laughs]

   30. VIVibhu16:14

       Short one-liners, like-
8. 17:15 – 18:43

   Agents API: Trust, Permissions, and Safety

   1. AWAri Weinstein17:15

      First of all, I'm just really excited that we brought Computer Use into the Agents API. I think this is, um, really great because obviously a lot of developers are building applications that want to be able to work with third-party websites and, and, and, and services. And so Computer Use has this universality to it. It can work with anything. So now all of a sudden, developers can build using the same Computer Use implementation that we're building on. I think there's great work to be done if you want to build your own Computer Use harness, but it's hard. And also, we train our models on our Computer Use harness, so there is, um, an advantage to using the one that's in distribution for the model. There m- actually might be a speed and cost and, and, and accuracy advantage. So I think it's really great for people to get to build on top of that. And, um, um, yeah, you know, I think kind of to, to the point that you were making, like, I think we're all sort of still in the process and maybe, like, some of us are ahead of many people in the world of, like, getting comfortable with this technology and trusting it. And so I think it's incumbent on us to, um, sort of build that trust over time by making sure we're building things that are reliable, by building, um, the right kinds of, of safety checks, by asking for the user's consent before doing something consequential like making a payment, by, uh, asking, you know, maybe depending on the application, making sure you're only letting it access the websites or applications that it actually needs for the task. So that's, I think, something important to, to think about. Um, but yeah, I'd really encourage people to try the, the new Agents API, build all kinds of cool stuff on it. We'd love to hear your feed- feedback if, if, uh, you know, depending on how it goes.

9. 18:43 – 20:26

   Computer Use for Coding, Testing, and QA

   1. VIVibhu18:43

      Have you seen any changes in the way it affects dev workflows? So one of the things with Dots is, you know, you're seeing it in Slack. You're seeing-

   2. AWAri Weinstein18:50

      Yeah

   3. VIVibhu18:50

      ... people use Voice and Build. Um, the example Roman showed of change this app and send me screenshots along the way and all this.

   4. AWAri Weinstein18:57

      Yeah.

   5. VIVibhu18:57

      Um, is anything that you're seeing there in adoption about how people are using Computer Use for coding workflows? Any tips people should take from that?

   6. AWAri Weinstein19:06

      One of my favorite use cases for Computer Use actually a- and, and one that we see a, a lot in the wild, is Computer Use letting the agent

   7. SWswyx19:15

      Actually test the software that the agent has built, which is far more consequential than it sounds because traditionally, you know, you'd build something in Codex, and then the, the... And then Codex builds it for you, and then you have to test it, and you are now like QA for the agent, right? So with Computer Use, you can complete the develop- the software development life cycle, um, where, uh, the agent can build software, it can test it. So I have a lot of fun, uh, you know, building stuff, having the agent test it. By the time it comes to me, it's already working. Um, I have, uh, extra fun because sometimes I'm, like, developing Computer Use itself, and so now I have a Computer Use agent that's using my Computer Use agent that's using something else. Uh, so yeah, I, I, I really, I really think this is a super powerful class of use case.

   8. NHNikunj Handa19:55

      I have a visual play test skill that I've developed that, um, really catches a lot of design issues-

   9. SWswyx20:01

      Nice

   10. NHNikunj Handa20:01

       ... um, that, you know, uh, normally when you just look at code, you wouldn't really pick it up. Uh, it's also really good for cloning apps, though. If you're using a shitty SaaS and you want to kill the SaaS, you just clone it screen by screen by screen. Uh-

   11. SWswyx20:11

       [laughs]

   12. NHNikunj Handa20:12

       ... and obviously Computer Use can completely drive everything, take screenshots, note it down, and then clone everything with Codex.

   13. SWswyx20:17

       It's really cool.

   14. NHNikunj Handa20:18

       Uh, but yeah, thanks for all your progress. I think, uh-

   15. SWswyx20:20

       Absolutely

   16. NHNikunj Handa20:20

       ... that is our time. Um-

   17. SWswyx20:21

       Yeah

   18. NHNikunj Handa20:22

       ... this is not the last that we're, we're gonna talk [laughs] .

   19. SWswyx20:23

       Yeah, cool. This has been really fun. Thank you guys for having me.

10. 20:26 – 24:33

    GPT-6 APIs, Async Tool Calling, and UltraFast Inference

    1. NHNikunj Handa20:26

       All right.

    2. SWswyx20:26

       All right.

    3. VIVibhu20:27

       Okay, we're strict cutoff. We're just gonna dive right in.

    4. NHNikunj Handa20:29

       [laughs]

    5. SWswyx20:29

       Let's do it, yeah.

    6. VIVibhu20:30

       Okay, so, uh, Nikunj, we're very excited to have you. You've shipped a lot on the API side, like we just-

    7. SWswyx20:37

       Yeah

    8. VIVibhu20:37

       ... talked about with Ari. You can now build with Computer Use agents. Anything you wanna highlight the API side of changes and introduce yourself a little on what you do?

    9. NHNikunj Handa20:45

       Yeah, for sure. My name is Nikunj. I lead product for the API team. Been here for roughly three years. Um, been working on launch and models. I feel like that's just been, like, a thing, constant thing throughout, throughout my time, uh, here at OpenAI. And, uh, with every new model, we try to, like, basically work super closely with the post-training team, the research team, to figure out what's new in it. Um, and then we, like, expose those capabilities in the API. Uh, so that's, like, the, the basic way of putting it. Um, and if you just look at, uh, everything that's new with GPT-6, the, the cool new capabilities that we launched were firstly async function calling. Um, so what you see with like, um, uh, like a lot of the things that you're seeing in like Codex and Dots and everything is that tool calls take so long that you don't have to, like, pause the model's execution while, uh, the tool is running. So you could just like kick off a tool call, keep running, keep reasoning, and then check back in. Uh, so we launched async tool calling. We launched like mid-turn steering, so now you can like inject messages while the model is reasoning, uh, in the middle. Uh, so as your tool call finishes, you can put the- put in that instructions into the context.

    10. SWswyx21:55

        And that's also partially a model alignment capability, right?

    11. NHNikunj Handa21:58

        Yeah.

    12. SWswyx21:58

        Like, they have to train in the ability to train.

    13. NHNikunj Handa22:00

        Exactly, yeah.

    14. VIVibhu22:00

        I feel like we've had it in the app. You could always, as it's reasoning, you could steer.

    15. NHNikunj Handa22:06

        Yes.

    16. VIVibhu22:06

        It wasn't the best. It's gotten much better.

    17. NHNikunj Handa22:09

        Yeah.

    18. VIVibhu22:09

        I'm excited to see how it does this iteration.

    19. NHNikunj Handa22:10

        Yeah, and then like our main-

    20. VIVibhu22:11

        And now-

    21. NHNikunj Handa22:11

        ... goal in, uh, our main goal in the API is to like put things in the API once it's trained into the harness. And so we kind of wait for that moment until it's good enough. And a lot of that is like actually being powered by WebSockets, which we launched, uh, a, a few, uh, I wanna say months ago. And so WebSockets just opens this like whole bi-directional like communication thing with the model. This is not, uh, the GPT Live thing. I'm just talking about GPT-6. Uh, and you can do all these like async tool calling, async reasoning, injecting messages. So it's a really fun API to work on. I think, uh, like really enjoying.

    22. SWswyx22:44

        Yeah. This is why we are the engineering podcast because we get to talk about WebSockets.

    23. NHNikunj Handa22:47

        Yeah [laughs] .

    24. SWswyx22:48

        Uh, this also pairs very well with UltraFast, right?

    25. NHNikunj Handa22:50

        Oh, yeah.

    26. SWswyx22:50

        Like that, that is now like, I think for the first time ever av- available in the API.

    27. NHNikunj Handa22:54

        Yes.

    28. SWswyx22:55

        Uh, this is just basically the, the theoretical fastest speed you can ever get at frontier intelligence.

    29. NHNikunj Handa23:00

        Yeah, yeah. It's been so exciting to work on that project. I think, uh, before I g- go into the API, the most fun part of, uh, UltraFast has been just watching the, the inference team cook with Astra. Like, they're just like constantly having these like Codex agents running, trying to like squeeze out more performance. And, uh, I would say like, uh, at least for a couple of months, a lot of it was focused on efficiency and driving the cost down, which is how we, like, were able to cut the Luna price by like 80%. It was like a lot of that was driven by like, uh, all the, the inference improvements they landed. And then now they've, like, shifted gears towards like, how can we make this run as fast as possible? And so UltraFast has just been like amazing to see on a mo- on a model like Astra, like to go that fast has been really cool. And yeah, WebSockets is like, uh, actually it was like the, the first time we launched WebSockets, it was for GPT, uh, 5.3 Codex Spark, which was-- can't believe we named a model that. But, uh, you know, that, that's what we launched it for. And obviously it helps so much because, like, you gotta have the tool calls. Um, you gotta like really reduce the overhead, uh, of, of going back and forth with tools. And, and so, uh, WebSockets is awesome for that.

    30. SWswyx24:09

        Yeah. Uh, it's always cute to see like I have my reset usage limit, and then I have my Spark usage limit that I never use.
11. 24:33 – 26:44

    The Rapid Story Behind Decisions API

    1. SWswyx24:33

       and for, uh, 5.3 Spark is explicitly attributed to Cerebras. You guys are not confirming or denying that, uh, UltraFast is related to Cere- Cerebras, but people are... I'll, I'll just say that people do, do care and, uh, are wondering about it. And you have your own silicon as well. Um, elephant in the room, decision models.

    2. NHNikunj Handa24:49

       Oh, yeah.

    3. VIVibhu24:49

       Decision API. We were the first podcast to do a big JEV, uh, deep dive with, uh, Diogo, and I also, you know, featured him at AI Engineer. Uh-

    4. SWswyx24:58

       How quickly did you see Jev and go like-

    5. NHNikunj Handa25:00

       Oh my god.

    6. SWswyx25:00

       [laughs]

    7. NHNikunj Handa25:01

       Yeah.

    8. SWswyx25:01

       Yeah.

    9. NHNikunj Handa25:02

       Firstly, like huge props to Diogo and like the Jev team for like really inspiring the-

    10. SWswyx25:07

        Yes

    11. NHNikunj Handa25:07

        ... like a whole segment in the market. Like, obviously Jev comes out, everyone's like losing their minds over it. Our users are like hitting us up, but also like our internal teams are like, "We need like a, a much faster classification system. We can..." Uh, I, I don't wanna like, um, get ahead of some of the, the Dots features that are, are gonna come-

    12. SWswyx25:27

        Ooh

    13. NHNikunj Handa25:27

        ... but you're gonna see like some cool like really snappy, fast things built on top of the Decisions API. Um, but you know, like, yeah, props to Jev for like inspiring this whole thing. Uh, obviously a bunch of people at OpenAI get nerd sniped by that, and they're like, "How can we like make this work? We're not gonna like-

    14. SWswyx25:45

        Okay

    15. NHNikunj Handa25:45

        ... train a new model." But-

    16. SWswyx25:46

        Like four weeks ago, this was not on the definite agenda, right?

    17. NHNikunj Handa25:48

        No, not at all, no.

    18. SWswyx25:49

        Okay.

    19. NHNikunj Handa25:49

        This is like-

    20. VIVibhu25:49

        Oh.

    21. SWswyx25:50

        Wow [laughs]

    22. NHNikunj Handa25:50

        ... Jev inspired and like full props to you.

    23. SWswyx25:52

        I think you're officially the first Frontier Lab to like clone and-

    24. NHNikunj Handa25:55

        Yeah, yeah

    25. SWswyx25:55

        ... adopt this.

    26. NHNikunj Handa25:56

        Yeah, yeah. I, I feel like OpenAI has such a strong like hacker culture and like people are just like, they get excited about things and so, uh, a guy from Inference, this one awesome guy from, uh, the Infra team are like, "This is amazing. We're gonna like hack on it." They build a prototype, it like works, and now we- we're just like hill climbing on latency and trying to make this as fast as possible, and we wanna like launch it in the coming days. Uh, so as soon as we hit our like latency target, we'll, we'll try to get this out.

    27. VIVibhu26:25

        It's interesting. At the same time of hacker culture, you also, as Sam said, like 99%, one of the most reliable APIs with-

    28. NHNikunj Handa26:32

        Mm-hmm

    29. VIVibhu26:32

        ... I think probably the most usage, which is your team directly. Um, how should people see Decisions API? I feel like a lot of people saw Jev, heard the buzz, haven't built with it. You're making it very mainstream.

12. 26:44 – 31:36

    What Decisions API Is and How It Works

    1. NHNikunj Handa26:44

       Mm-hmm.

    2. VIVibhu26:45

       What should people see it as? How should they use it?

    3. NHNikunj Handa26:47

       Yeah. I think the, the main use cases we've seen is like really fast classification. Um, all the computer use demos have been amazing and really cool. I think there will be limitations, of course, um, in terms of, you know, having Astra like, uh, write like a JavaScript like script to control your computer versus having Luna pick like one action at a time. I think, uh, it's not gonna be at the same intelligence level, but like maybe there's some computer use tasks that this is good enough for, so excited to see that come through. Um, the other cool prototype I've seen internally is people hooking it up with GPT Live. So GPT Live is like, you know, our, our bidirectional like real time, um-

    4. VIVibhu27:25

       Voice thing

    5. NHNikunj Handa27:26

       ... A- API. And, uh, it's built on this like model of front-end models and back-end models. So GPT Live is this like-

    6. VIVibhu27:32

       Think or Talker

    7. NHNikunj Handa27:32

       ... super fast, yeah, think or, um, talker thing. So GPT Live is the talker, uh, super fast, really good at delegation, uh, and you have something like Astra sitting at the ba- at the back. But tool calling has always felt like really slow in, in GPT Live. Um, and so people have been like putting together these like tool calling demos of GPT Live controlling a computer. It just feels like so much more snappy and, and natural. So I'm like kinda excited to see like what people do with Live and with, uh, with, uh, with Luna on Decisions API. Um, so that'll be pretty exciting. Yeah.

    8. SWswyx28:06

       So I wanna iron this out for people, especially from the product side, because a lot of people have been putting out Jev clones. There's been about 100 in the last two weeks.

    9. NHNikunj Handa28:13

       Oh, really? That's amazing. [laughs]

    10. VIVibhu28:14

        Marketed in the first two days.

    11. SWswyx28:15

        But like, but like, uh, they, like they can clone the Jev API, which is honestly structured outputs.

    12. NHNikunj Handa28:20

        Yeah.

    13. SWswyx28:20

        Which OpenAI was first to.

    14. NHNikunj Handa28:22

        Yeah.

    15. SWswyx28:22

        Right? So like I think let's, let's iron out for people what is a decision model a- as far as you-

    16. NHNikunj Handa28:28

        Yeah

    17. SWswyx28:28

        ... as far as like what is important. It is not just latency, it's not just structured output, right?

    18. NHNikunj Handa28:32

        Mm-hmm.

    19. SWswyx28:32

        Because I could just have Luna as it's the decision model is priced the same as Luna, right?

    20. NHNikunj Handa28:38

        Mm-hmm.

    21. SWswyx28:38

        Um, have turned off reasoning and then have structured output. Do I have a Jev? Uh, d- you know, no, right?

    22. NHNikunj Handa28:44

        Yeah.

    23. SWswyx28:44

        And that's the, that's the real-

    24. VIVibhu28:46

        There's a confidence there.

    25. SWswyx28:47

        Yeah.

    26. NHNikunj Handa28:47

        Yeah, totally. I think, uh, the, the way that... So we haven't trained like a new model for this.

    27. SWswyx28:51

        Yeah.

    28. NHNikunj Handa28:51

        We're like building this purely on top of the same Luna weights that we have.

    29. SWswyx28:55

        Oh, okay.

    30. NHNikunj Handa28:56

        Uh, so yeah, this is like really just Luna and, uh, on top of that what you're doing is you're constraining, so like structured output's a big part of it. Uh, you're really optimizing the, the inference stack to like, uh, get, get very fast on TTFT. And because you can have multiple questions, what you do is like you basically run those in parallel, uh, so like how these-
13. 31:36 – 33:35

    What OpenAI Is Building With the New APIs

    1. VIVibhu31:36

       being on the platform team, a lot of your job is to empower builders.

    2. NHNikunj Handa31:39

       Mm-hmm.

    3. VIVibhu31:39

       What do you think people should build with Decisions API and also computer use agents? Any, any stuff that you've been building with internally that you think really opens up after the new change?

    4. NHNikunj Handa31:50

       Yeah. Um, okay, let's think. Um, Decisions API, uh, use cases internally have been pretty obvious. Like, the user ops team was, like, jumping on it, being like, "We gotta classify all of our support tickets." Um, w- what else came up? Uh, uh, obviously there were, like, the really cool GPT Live demos. I'm sure, like, the Codex app team might, like, pick this up and try to do something cool with it. So, you know, like, this whole thing started, like, a week ago, so it's, it's, like, very, very early and-

    5. VIVibhu32:18

       Oh, one week

    6. NHNikunj Handa32:19

       ... we're excited. Yeah.

    7. VIVibhu32:19

       There was a big push-

    8. NHNikunj Handa32:20

       Yeah, pretty much. [laughs]

    9. VIVibhu32:21

       ... in, um, evals, LLM as a judge, having really low latency there.

    10. NHNikunj Handa32:25

        Right. Yeah, yeah. That'll be interesting to see. Um, and then, um, uh, uh, with, with the Agents API, uh, we, we have... We're basically, like, having a bunch of first party products, like, uh, at, at OpenAI built fully on top of it. Um, we've had the Codex security stuff that just went out that's fully built on top of, uh, the Agents API. We have, uh, sort of the... We, we are having, like, a meetings type of thing launching today.

    11. VIVibhu32:52

        Meetings?

    12. NHNikunj Handa32:52

        I think there was, like, a demo. Um, do, do you remember, like, the plugin extensions when Sam was showing it? There was like a, a demo for, like, you're in a calendar, you can sort of like, uh, have your meeting notes, like-

    13. VIVibhu33:02

        Like drop into his, uh-

    14. NHNikunj Handa33:03

        ... flow into, like, your space

    15. VIVibhu33:04

        ... a Zoom call, like a Granola type thing.

    16. NHNikunj Handa33:05

        Yeah. And so the... all of that stuff is, like, fully built on top of, uh, the Agents API. Um, and yeah, I'm, like, just excited to see. Like, we're just getting this out, and let's see what people build on top of it and, yeah.

    17. VIVibhu33:15

        I, I think you showed it off very well, the whole edit spaces, pages, collaborate, add in your dot.

    18. NHNikunj Handa33:21

        Yeah.

    19. VIVibhu33:22

        Like, that's, that's a lot. So-

    20. NHNikunj Handa33:24

        Yeah

    21. VIVibhu33:24

        ... there's a lot of inspiration people can build from. [laughs]

    22. NHNikunj Handa33:26

        Yeah, yeah. All possible with Astra, you know? Like, thing, things just move so fast now. Like-

    23. VIVibhu33:31

        Yeah

    24. NHNikunj Handa33:31

        ... people go from idea to execution so quickly, it's amazing.

    25. SWswyx33:34

        Is there

14. 33:35 – 36:32

    Prompt Caching, Pre-Warming, and API Performance

    1. SWswyx33:35

       something that you want, uh, people to focus on to give you feedback? Like, what... Like, you know, maybe you're um, you're, you're just putting this out there and you want... Y- and there's, like, a fork in the road and you want developers to help you decide.

    2. NHNikunj Handa33:47

       So I think Agents API and Decisions API, they're, like, these are our newest products. Would love, like, any and all feedback on that to figure out where to take them. I think, uh, over here we're, like, very open. On Responses API, uh, which is sort of like our, our workhorse over here, um, like, really focused on performance right now, and the performance comes in, like, two main ways. Uh, first is just, like, latency. We've been, like, rewriting the whole Responses API stack to, like, make it as fast as possible from a DTFT perspective, DVT perspective. So there's, like... That's, like, continues to be, like, a main area of focus for us. The second thing we've been trying to do is, like, really go deep on caching, particularly with these, like, personal agents that are, you know, like, um, uh, basically like a single thread that just goes on and on forever. We've been, uh, trying to, like, really up our game on caching. We provide now guarantees of, like, cache hits within, like, 30 minutes. We're actually, like, we... For one of our users, we just launched, like, a much longer cache window, so we have, like, a 12-hour caching guarantee, uh, that we offer so that you, you, you have, like, guaranteed cache hits for a, a-

    3. SWswyx34:52

       Is that a public API?

    4. NHNikunj Handa34:53

       Not yet.

    5. SWswyx34:54

       Ah.

    6. NHNikunj Handa34:54

       That's in preview. We're gonna, like, uh, try to get that out to everyone as soon as possible. But, like, just pay a little bit more for the cache write and we, like, guarantee, um, like, cache reads for, like, a much longer period. So even if, like, your instinct thread, for example, like, you just, like, do something on it and then come back to it like three to four hours later, you're, you're still getting the caching performance out of it. We've launched-

    7. VIVibhu35:16

       And you cut the cost there quite a bit too, right, with the new model?

    8. NHNikunj Handa35:18

       Oh, yeah. That's right.

    9. VIVibhu35:19

       95% cheaper, so-

    10. NHNikunj Handa35:19

        Yeah, we're, like, driving on cache reads, yeah

    11. VIVibhu35:21

        ... for builders, they should implement because-

    12. NHNikunj Handa35:25

        Yeah

    13. VIVibhu35:25

        ... it's significantly cheaper.

    14. NHNikunj Handa35:26

        Yeah, yeah. Just, like, building your apps with, like, uh, to be very cache aware and, and sort of like use our prompt diagnostics, our cache diagnostics tool to figure out, like, where things are, are, are dropping off. And, and, um, uh, so, so the, the caching part is, like, really important. Um, yeah, I also want to talk about pre-warming. We have that in the API now. So, like, if you know that, "Hey, I'm gonna get a cache," like... Uh, sorry, "I'm gonna get this prompt. I just want to, like, pre-warm the cache."

    15. VIVibhu35:54

        Mm-hmm.

    16. NHNikunj Handa35:54

        Pay, like, the cache write fee right now, and then, like, uh, have it sort of ready to go for the next 30 minutes for whenever things come through.

    17. SWswyx36:00

        And it can spawn many instances of that thread.

    18. NHNikunj Handa36:03

        Exactly, yeah.

    19. SWswyx36:03

        Yeah, yeah.

    20. NHNikunj Handa36:04

        You can just keep going and have, uh, tons and tons of that.

    21. SWswyx36:05

        Yeah, just keep assisting from there. Yeah.

    22. NHNikunj Handa36:07

        Um, and so, uh, yeah, like, I'm very excited about getting feedback on, like, the low level performance things so that we can keep making Responses API the most performant and reliable way to, like, build on top of an LLM. And then you, you basically have our, like, new products where I'm just looking for, like, any and all feedback.

    23. SWswyx36:26

        Yeah, just use it, right?

    24. NHNikunj Handa36:27

        Um, so yeah, just tell us what-

    25. SWswyx36:28

        Use the thing, yeah

    26. NHNikunj Handa36:29

        ... yeah. Define our roadmap for us, please. [laughs] So

15. 36:32 – 38:25

    Context Compaction for Long-Running Agents

    1. NHNikunj Handa36:32

       yeah.

    2. SWswyx36:32

       Uh, I, I think for me, the caching thing, great, right? Like, obviously very needed, but at the end of the day, you're still bumping up against a million token context.

    3. NHNikunj Handa36:40

       Mm-hmm.

    4. SWswyx36:40

       And that's probably not gonna change for the foreseeable future.

    5. NHNikunj Handa36:42

       Mm-hmm.

    6. SWswyx36:43

       Like, you still need good compression.

    7. NHNikunj Handa36:44

       Yeah.

    8. SWswyx36:45

       What is the best practice there?

    9. NHNikunj Handa36:46

       Yeah, yeah, totally. Uh, so firstly, um, OpenAI has its own, like, proprietary compression, uh, com-

    10. VIVibhu36:52

        Which, which is in-

    11. NHNikunj Handa36:53

        ... compaction

    12. VIVibhu36:54

        ... this is in Agents

    13. NHNikunj Handa36:54

        ... it's in the API.

    14. VIVibhu36:55

        Yes, Agents API.

    15. NHNikunj Handa36:56

        Yeah.

    16. SWswyx36:56

        You, you decide for us, right?

    17. NHNikunj Handa36:57

        Yeah, exactly. So in the Agents API, uh, it comes built into the harness. Uh, and if you're in Responses API, there's, like, two ways of doing it. One is what we call server-side compaction, which is you basically tell Responses API that if you ever hit this t- This threshold of tokens, uh, just auto compacted and, and, and, and, like, go back, uh, or sorry, like, um, re-reduce the context, uh, being used. And the second way is, like, /compact, which is, like, if you want full control. So you can, like, /compact at any time.

    18. SWswyx37:27

        I hate it.

    19. NHNikunj Handa37:27

        Have your own logic on when to, like-

    20. SWswyx37:29

        It's not AGI.

    21. NHNikunj Handa37:29

        Yeah.

    22. SWswyx37:30

        It's AJ.

    23. NHNikunj Handa37:30

        [laughs] Yeah.

    24. SWswyx37:31

        Yeah. But it-- I mean, it is the manual override.

    25. NHNikunj Handa37:33

        Yeah, it is the manual way, and, like, uh, I, I don't know, but a lot of the, the big coding agents like to do it manually. I, I mean, like, if you look at the Codex implementation of it in the Cod- open source Codex harness, you can see that they, they use /compact and do it. Um, and, uh, there's also, like, new, um, by the way, new compaction techniques that we are working on. Some of them you will be able to see in the Codex harness. Like, it's already implemented in the Codex harness. And so, um, uh, there are, like, some file-based, uh, systems that we are, like, experimenting with. So yeah, lots of cool stuff going on around in compaction as well.

    26. SWswyx38:08

        Cool. Uh, we are running out of time.

    27. NHNikunj Handa38:11

        Great.

    28. SWswyx38:11

        Um, I think you've talked about-- a lot about performance and talked a lot about, uh, the new APIs that you're launching.

    29. NHNikunj Handa38:17

        Mm-hmm.

    30. SWswyx38:17

        Can you give us any other hints as to things that you're interested in, uh, as, as far as the future of the platform is concerned?
16. 38:25 – 40:23

    Memory, Higher-Level APIs, and the AI Cloud

    1. NHNikunj Handa38:25

       like, very low level. Like, I, I used to work at Stripe before this, and, uh, at Stripe, a lot of the game was, like, building these higher-level primitives and products on top of, like, the core payments primitives. Um, and, uh, I'm always, like, curious about what the best way of doing that is in, in AI, and I think we've had a couple of attempts at that. We, like, had launched Assistants API, like, way back in the day, and, like, wasn't really the right fit. We were sort of, like, going off with this, like, Agents API, and, uh, it gives you the Codex harness, but, like, where's the, like, the, the... What's the right amount of flexibility to give in that? That's, like, an open question. Like, how-- Should we, like, have memory vaults and, like, all of these, like, higher level, like, API objects to, to take away, uh, or so, like, to, to abstract away more, um, um, like, storage concepts? Like, this is, like, a whole, like, uh... There's a whole space that I'm, like, very curious about figuring out how we design. Um, I think a lot of things in AI are just have a low-level API primitive and see an example harness and go and im- have your coding agent implement that. But how much of that do we build into the, the API is, like, a constant question that I'm thinking about. Um-

    2. SWswyx39:35

       Yeah

    3. NHNikunj Handa39:35

       ... so I don't know. If folks have thoughts on that, if, uh, if anyone has ideas, uh, uh, it'd be super interesting to hear.

    4. SWswyx39:42

       Yeah. The, the analogy I always bring back to, and, uh, w-we'll end there, is, uh, you're building an AI cloud, right?

    5. NHNikunj Handa39:47

       Mm-hmm.

    6. SWswyx39:47

       Like, which is, uh, something that, uh, Sam said a year ago.

    7. NHNikunj Handa39:50

       Mm-hmm.

    8. SWswyx39:50

       Where-- And you're-- It's almost like you're sort of doing the AWS invention, and you have to do, "Okay, this is EC2."

    9. NHNikunj Handa39:57

       Yeah.

    10. SWswyx39:57

        "And this is S3, and this is like..." But you're doing the AI-native versions of each of these.

    11. NHNikunj Handa40:00

        There are a lot of analogies.

    12. SWswyx40:01

        Yeah. That's true.

    13. NHNikunj Handa40:01

        So you're pre-warming caches for stuff that you know will [laughs]

    14. SWswyx40:04

        [laughs]

    15. VIVibhu40:04

        Yeah, yeah, yeah. [laughs]

    16. NHNikunj Handa40:05

        Uh, and it's nice that it's all exposed to builders-

    17. SWswyx40:08

        Mm-hmm

    18. NHNikunj Handa40:08

        ... 'cause it just opens up ways that you can build new things.

    19. SWswyx40:10

        Yeah, absolutely.

    20. NHNikunj Handa40:12

        Cool.

    21. SWswyx40:12

        Okay. Awesome.

    22. NHNikunj Handa40:12

        Awesome.

    23. VIVibhu40:12

        Well, that's everything.

    24. NHNikunj Handa40:13

        Thank you, guys.

    25. VIVibhu40:14

        Thank you.

    26. NHNikunj Handa40:14

        Yeah. [outro music]

Episode duration: 40:23