# NVIDIA Dynamo, Agent Security, and Inference Scaling

**Podcast:** Latent Space: The AI Engineer Podcast
**Published:** 2026-03-10

## Transcript

Agents can do three things.
They can access your files, they can access the internet, and then now they can write custom code and execute it.
You should really only let an agent do two of those three things.
If you can access your files and you can write custom code, you don't want internet access because that's one is the full vulnerability, right?
If you have access to internet and your file system, you should know the full scope of what that agent's capable of doing.
Otherwise, now we're thinking get injected or something that can happen.
And so that's a lot of what we've been thinking about is like, you know, how do we both enable this because it's clearly the future, but then also, you know, what what are these enforcement points that we can start to like protect?
All right, welcome to the Lady Space Podcast in the Chroma studio.
Welcome to all the guests here.
Uh we're back with our guest host, Vivu.
Welcome.
Good to have you back.
And our friends uh Netter and Kyle from Nvidia.
Welcome.
Yeah, thanks for having me.
Yeah, thank you.
Actually, I don't even know your titles.
Uh I know you're like Architect something of Dynamo.
Yeah, I I'm one of the engineering leaders and arch architects of Dynamo.
And you're director of something developer tech.
Yeah.
You're the developer's director, first guy, and NVIDIA.
And we're we're kind of recording this ahead of NVIDIA GTC, which is coming to town uh again, uh taking over town, uh, which uh which we'll all be at.
Um and we'll talk a little bit about your sessions and stuff.
Yeah, yeah, we're super excited for it.
One of my favorite memories for Nader, like you always do like marketing stunts.
And like while you were brev, you like had this surfboard that you like went down to GTC with.
And like no NVIDIA apparently liked it so much that they bought you.
Like, what was that like?
Yeah, yeah, we we um our logo was a shaka.
We we were uh we were always just kind of like trying to keep true to who we were.
I think you know, someone such startups, you're like trying to pretend that you're a bigger, more mature company than you are.
And it was actually Evan Conrad, SF Compute, who was just like, You guys are guessed, yeah.
Oh, really?
Amazing, yeah.
He was just like, guys, you're two dudes in a room.
Why are you bringing that you're not?
Uh and so then we were like, okay, let's make the logo a shaka.
We brought surfboards to our booth to GTC, and the energy was great.
Um some palm trees too.
They they actually poked out over like the the walls, so you could you could see the bread booth and no one else just from very far away.
Oh, so you remember it back then?
Yeah, I I remember it.
Pre-acquisition.
I was like, oh, those guys are cool.
That makes sense because uh we so we signed up really last minute, and so we had the last booth, it was all the way in the corner.
And so I was I was worried that no one was gonna come, so that's why we had like the palm trees.
We really came in with the surfboards.
We even had one of our investors bring her dog, and then she was just like walking the dog around to try to like bring energy towards our booth.
Yeah, Steph.
Yeah, yeah, she's the best.
You know, as a conference organizer, I love that, right?
Like it's like uh everyone who sponsors a conference comes, does their booth, they're like, We are changing the future of AI or something, some generic bullshit.
And like, no, like actually try to stand out make it fun, right?
And people still remember it after three years.
Yeah, yeah.
You know what's so funny?
I'll send I'll give you this clip if you wanna if you want to add it in.
But uh my wife was at the time fiance, she was in medical school, and she came to help us because it was like a big moment for us, and so we we bought this cricket.
It's like a vinyl, like a vinyl uh printer.
Because like, how else are we gonna label the surfboard?
So we got a surfboard, luckily was able to purchase that on the company card.
We got a cricket, and it was just like fine-tuning for enterprises or something like that that we put on the on the surfboard, and it's 1 a.m.
the day before we go to GTC.
She's helping me put these like vinyl stickers on, and she goes, You son of and she's like, If you pull this off, you son of a bitch.
So uh right pretty much after the acquisition, I stitched that within the news of the acquisition, I sent it to our family group chat.
Yeah, yeah.
No, well, she she made a good choice there.
Was that like basically the origin story for launchables?
Is that we maybe we should explain what Brev A is and Yeah, uh I mean Brev is just it's a developer tool that makes it really easy to get a GPU.
So we connect a bunch of different GPU sources.
So the basics of it is like how quickly can we SSH you into a G into a GPU?
And whenever we would talk to users, they wanted a GPU, they wanted an A100.
And if you go to like any cloud provisioning page, usually it's like three pages of forms or in the form somewhere there's a drop down, and in the drop down, there's some weird code that you know to translate to an A100.
And I remember just thinking like every time someone says they want an A100, like the piece of text that they're telling me that they want is like stuffed away in the corner.
And so we're like, what if the biggest piece of text was what the user's asking for?
And so when you go to brev, it's just big GPU chips with the type of thing.
With beautiful animations that you work on pre like pre, you can like now you can just prompt it.
But back in the day, handcrafted artisanal code.
Yeah, I was actually really proud of that because uh it was an I I made it in Figma, yeah, and then I found I was like really struggling to figure out how to turn it from like Figma to React.
So what it actually is is just an SVG, and I I have all the styles, and so when you change the chip, whether it's like active or not, it changes the SVG code, and that somehow like renders like looks like it's animating, but it would we just had the transition slow.
But it's just like the a JavaScript function to change the like underlying SVG.
And that was how I ended up like figuring out how to move it from from Figma.
But yeah, that's our artisan speaking of marketing stunts though, he actually used those SVGs or kind of used those SVGs to make these cards.
Oh, yeah, like a GPU gift card that he handed out everywhere.
That was actually my first impression of that.
Yeah, yeah, yeah.
Yeah, I think I still have one of them.
Yeah, they look great.
Yeah, I have a ton of them still actually in our garage, but just they don't have labels.
We should honestly like bring bring them back.
But um, I found this old printing press here, actually just around the corner on Veness, and it's a third generation San Francisco shop.
And so I come in, an excited startup founder trying to like, and they just have this crazy old machinery, and I'm in awe because the the whole building is so physical.
Like you're seeing these machines, they have like pedals to like move these saws and whatever.
I don't know what this machinery is.
But I saw all three generations, like there's like the grandpa, the father, and the son, and the son was like around my age.
Well, it's like a holy holy trinity.
Yeah, yeah.
Like it's funny because we so I just took the same SVG and we just like printed it, and it's foil printing.
So they make a uh a mold that's like an inverse of like the A100, and then they put the foil on it and then they press it into the paper.
And I remember once we got them, he was like, Hey, don't forget about us.
You know, I guess like early Apple and Cisco's first business cards were all made there.
And so he was like, Yeah, we we get like the startup businesses, but then as they mature, they kind of go somewhere else.
And so I actually I think we were talking with marketing about like using them for the city.
Yeah, yeah.
Yeah, I you know, I I remember, you know, as a very, very small brev investor.
I was like, What why are we spending time like doing these like stunts for GPUs?
Like, yeah, you know, I think like as a you know, typical like cloud hard hardware person, you go into AWS, you pick like T5XXL, whatever, and then it's just like confirmed this and you look at the specs.
Like, why animate this GPU?
And and I I do think like it just shows the level of care that goes through outbrev and yeah, and not and also Dynamo.
And NVIDIA, I think that's what the thing that struck me most when we first came in was like the amount of passion that everyone has.
Like, I think um, you know, you talk to you talk to Kyle, you talk to like every VP that I've met at NVIDIA goes so close to the metal.
Like, I remember it was almost a year ago, and like my VP asked me, he's like, Hey, what's cursor?
And like, are you using it?
And if so, why?
And I'm just like surprised at this.
And he downloaded cursor and he was asking me to help him like use it.
And I thought that was uh, or like just show him what he you know why we were using it.
And so the amount of care that I think everyone has and the appreciation passion and appreciation for the moment, right?
This is a very unique time.
So it's really cool to see everyone really like uh appreciate that.
Yeah, one thing I wanted to do before we move over to sort of like research topics and uh the the stuff that Kyle's working on is just tell the story of the acquisition, right?
Like, not many people have been through an acquisition with Nvidia.
With NVIDIA, what's it like?
Uh what yeah, just anything you like to say.
It's a crazy experience.
I think uh, you know, we were the the thing that was the most exciting for us was our goal was just to make it easier for developers.
We wanted to find access to GPUs, make it easier to do that, and then all oh yeah, actually, your question about launchables.
So launchables was just make one-click experience, like one-click deploys for any software on top of the GPU.
And so what we really liked about NVIDIA was that it felt like we just got a lot more resources to do all of that.
I think uh, you know, Nvidia's goal is to make things as easy for developers as possible.
So there was a really nice like synergy there.
I think the you know, when it comes to like an acquisition, I think the amount that the soul of the products align, I think is gonna be is gonna speak to the success of the acquisition.
Yeah.
So it in many ways feels like we're home.
This is a really great outcome for us.
Like we, you know, I love brev.nvidia.com.
Like you should, you should use it.
It's the front page for GPUs.
Yeah, yeah.
Do you want GPUs?
You go there again.
And it's like internally is growing very quickly.
I don't remember you said some stats there.
Yeah, yeah, yeah.
It's uh I I wish I had the exact numbers, but like internally, externally, it's been growing really quickly.
We've been working with a bunch of partners with a bunch of different customers and ISVs.
If you have a solution that you want someone that runs on the GPU and you want people to use it quickly, we can bundle it up uh in a launchable and make it a one-click run.
If you're doing things and you want just like a sandbox or something to run on, right?
Like open claw, huge moment, super exciting.
Uh uh, and we'll talk into it more, but you know, internally, people want to run this, and you we know we have to be really careful from the security implications.
Do we've let this run on the corporate network?
Security's guidance was hey, run this on brev.
It's in, you know, it's it's it's it's a VM, it's sitting in the cloud, it's off the corporate network, it's isolated.
And so that's been our stance internally and externally about how to even run something like open claw while we figure out how to run these things securely.
But yeah, yeah.
I think there's also like you almost like we're the right team at the right time when NVIDIA is starting to invest a lot more in developer experience or what whatever you call it.
Uh UX or I don't know what you call it, like software.
Like obviously Nvidia is always invested in software, but like there's like this is like a different audience.
It's a wider developer base.
Yeah, right.
Yeah, yeah.
You know, it's funny, it's like it's not uh what is it called internally?
What what is this that people should be aware of that is going on there?
Uh what like developer experience?
Yeah, yeah, it's just it's called just developer experience, or is there like a broader strategy here?
In NVIDIA, um NVIDIA always wants to make a good developer experience.
The thing is a lot of the technology is just really complicated.
Like it's not, it's uh I you know, I I think um the thing that's been really growing, or the AI is growing, is having a huge moment, not because like let's say data scientists in 2018 were quiet then and are much louder now.
The Pi is right.
There's a whole bunch of new audiences.
My mom's wondering what she's doing, my sister's learned like taught herself how to code.
Like the um, you know, I actually think just generally AI is a big equalizer, and you're seeing a more like technologically literate society, I guess.
Like everyone's everyone's learning how to code.
Uh, there isn't really an excuse for that.
And so building a good UX means that you really understand who your end user is.
And when your end user becomes such a wide uh variety of people, then you have to almost like reinvent the practice, right?
Yeah, and actually build more developer UX, right?
Because the there are tiers of developer base that were added, you know, the the hackers that are building on top of OpenClaw, right?
For example, have never used GPU, they don't know what CUDA is, they they they just want to run something.
Yeah, you need new UX that is not just hey, you know, how do you program something in CUDA and run it?
And then and then we built, you know, like when deep learning was getting big, we built we built Torch and but recently the amount of like layers that are added to that developer stack is just exploded because AI is become ubiquitous, everyone's using it in different ways.
Yeah, it's moving fast in every direction, vertical, horizontal, you're seeing you guys you even take it down to hardware, like the DGX Spark, you know, it's it's basically the same system as just throwing it up on big GPU clusters.
Yeah, yeah, yeah.
Yeah, uh, we saw the preview at the last year's GTC, and that was one of the better performing uh videos so far and video coverage so far.
Awesome.
This will beat it.
Um, even when Grace Blackwell or when um uh DGX Spark was first coming out, getting to be involved in that from the beginning of the developer experience, and it just comes back to you were involved, yeah.
Okay, yeah, yeah.
Say more, same more.
Yeah, I mean, from it was just like I I got an email, we just got thrown into the loop, and suddenly, yeah, I do I it was actually really funny because I'm still pretty fresh from the acquisition, and I'm getting an email from a bunch of the engineering VPs about like the new hardware GPU chip like we're or not chip, but just GPU system that we're putting out.
And I'm like, okay, cool, Natter's now involved with this for the UX.
And like, what am I gonna do here?
So I remember the first meeting, I was just like kind of quiet as I was hearing engineering VPs talk about what this box could be, what it could do, how we should use it.
And I remember uh one of the first ideas that people were ideing was like, Oh, the first thing that it was like I think a quote was like the first thing someone's gonna want to do with this is get two of them and run a Kubernetes cluster on top of them.
And I was like, Oh, I think I know why I'm here.
I was like, the first thing we're doing is easy SSH into the machine, and then and you know, just kind of like scoping it down of like once you can do that, everything you like the person who wants to run a Kubernetes cluster onto Sparks has a higher propensity for pain than then you know someone who buys it and wants to run open claw right now, right?
If you can make sure that that's as effortless as possible, then the rest becomes easy.
So there's a tool called Nvidia Sync that just makes the SSH connection really simple.
So, you know, if you think about it, like if you have a Mac or a PC or whatever, if you have a laptop and you buy this GPU and you want to use it, you should be able to use it like it's uh a GPU in the cloud, right?
Um, but there's all this friction of like how do you actually get into that?
That's part of Brev's value proposition, is just you know, there's a CLI that wraps SSH and makes it simple.
And so our goal is just get you into that machine really easily.
And one thing we just launched this CES, it's in it's still in like early access.
We're ironing out some kinks, but it should be ready by GTC.
Um you can register your Spark on Brev.
And so now if remote managed local in a glass, yeah, yeah.
Because Brev can already manage other clouds anyway, right?
You can use the Spark on Brev as well, right?
Yeah, but yeah, exactly.
So so you you so you you set it up at home, you can run a command on it, and then it gets it's essentially it'll appear in your Brev account, and then you can take your laptop to a Starbucks or to a cafe, and you'll continue to use your sp- you can continue to your spark just like any other cloud node on Brev.
Yeah, yeah.
It's just like a pre-provisioned in your home.
Yeah, exactly.
Yeah, yeah, yeah.
Tiny little data center.
Tiny little data center.
One more thing before we move on to Kyle.
Just have so many Jensen stories, and I just love mining Jensen stories.
Uh my favorite so far is SOL.
Uh what is SOL?
SOL is actually I I think of all the lessons I've learned, that one's definitely my favorite.
It'll always stick with you.
Yeah.
Yeah.
I you know, in your startup, everything's existential, right?
Like we've we've run out of money, we were like on the risk of of losing payroll, we've had to contract our team because we ran out of money.
And so, like, um, because of that, you're really always forcing yourself to like understand the root cause of everything.
If you get a date, if you get a timeline, you know exactly why that date or timeline is there.
You're you're pushing every boundary, and like you're not just saying you're not just accepting like uh a no just because.
And so as you start to introduce more layers, as you start to become a much larger organization, SOL is essentially like what is the physics, right?
The speed of light moves at a certain speed.
So if light's moving something slower, then you know something's in the way.
So before trying to like layer reality back in of like why can't this be delivered at some date?
Let's just understand the physics.
What is the theoretical limit to like uh how fast this can go?
And then start to tell me why.
Because otherwise, people will start telling you why something can't be done.
But actually, I think any great leader's goal is just to create urgency.
There's an interesting thing.
Right, right.
Yeah.
SOL is a term that Nvidia is used to instigate a compelling event.
You say, This is done.
How do we get there?
What is the minimum, as much as necessary, as little as possible thing that it takes for us to get exactly here?
And it helps you just break through a bunch of noise.
Yeah.
One thing I'm unclear about is can only Jensen use the SOL card?
Like, get the bullshit out.
Because it looks obviously it's Jensen, but like, can someone else be like, no, like frontline engineers use it.
Yeah, every I think it it's not so much about like get the bullshit out.
It's like it's like give me the root understanding, right?
Like if you tell me something takes three weeks, it's like, well, what's yeah, the first principles.
It's like, what's the what like why is it three weeks?
What is the actual, yeah, what's the actual limit of why this is gonna take three weeks?
If you're gonna if you if let's say you wanted to buy a new computer and someone told you it's gonna be here in five days, what's the SOL?
Well, like the SOL is like I could walk into a Best Buy and pick it up for you, right?
So then anything that's like build beyond that is and is that practical?
Is that how we're gonna you know, let's say give everyone in the company a laptop?
Like, obviously not.
So then like that's the SOL, and then it's like, okay, well, if we have to get more than 10, suddenly there might be some, right?
And so now we can kind of piece the reality back.
So this is the Paul Graham, do things that don't scale.
Yeah, and this is also the what people would now call behind agency.
Yeah, yeah, yeah.
It's actually really interesting because there's a there's a second hardware angle to SOL that like doesn't come up for all the org.
So SOL is used like culturally at NVIDIA for everything.
I'm also mining for like I think that can be annoying sometimes.
I mean, like someone keeps going SOL, SL, and you're like, guys, like we have to be stable.
We have to where's the fucking plan?
Like it's industry balance.
Yeah, I encountered that with like actually just with Alec, right?
Because we we have a new conference, so we need to launch where we have we have goals of what we want to launch by uh by the conference, and like yeah, at the end of the day, is this GTC?
Um, this is like so we I mean we did it for CES, we for GTC DC before that.
We're doing it for GTC San Jose.
So I mean like every you know, we have a new moment um and we want to launch something, yeah, and we want to do so at SOL.
And that does mean that some there's some level of prioritization that needs to happen.
And so it it is difficult, right?
I think um you have to be careful with what you're pushing, you know, stability is important, and that should be factored into SOL.
SOL isn't just like build everything and let it break, you know, that that's part of the conversation.
So as you're laying layering in all the details, one of them might be, hey, we could build this, but then it's not gonna be stable for XYZ reasons.
And so that was like one of our conversations for CES was like, you know, hey, like we we can get this into early access, registering your Spark with Brev, but there are a lot of things that we need to do in order to feel really comfortable from a security perspective, right?
There's a lot of networking involved before we deliver that to users, so it's like, okay, let's get this to a point where we can at least let people experiment with it.
We had it in a booth, we had it in Jensen's keynote, and then let's go iron out all the networking kinks, and that's not easy.
And so uh that can come later.
And so that was the way that we layered that back end.
Yeah, but it's not really about saying like you don't have to do the the maintenance or operational work, it's more about saying, you know, it's kind of like highlights how progress is incremental, right?
Like, what is the minimum thing that we can get to?
And then there's SOL for like every component after that, but there's the SOL to get you get you to the the starting line, and that that's usually how it's asked.
On the other side, you know, like SOL came out of like hardware at NVIDIA, right?
So SOL is like literally if we ran the accelerator or the GPU with like at basically full speed with like no other constraints, like how fast we'd be able to make our program go.
Yeah, yeah.
Right.
So in in training that, like, you know, then you work back to like some percentage of like MFU, for example.
Yeah, that's that's a great example.
So like there's an there's an SOL, MFU, and then there's like you know, what's practically achievable.
Cool.
Uh shall we move on to sort of uh Kyle's side?
Uh Kyle, you're coming more from the data science world.
And uh I I mean, I always whenever I meet someone who's done work in tabular stuff, graph neural networks, time series, these are basically when I go to New Reps, I go to ICML, I walk the back hauls.
There's always like a small group of graph people, small group of tabular people, and like there's no one there, and like it's very like you know what I mean.
Like yeah, no, like it's it's important, interesting work if you care about solving the problem that they solve.
Yeah, but everyone else is just LLMs all the time.
Yeah, I mean it's like it's like the black hole, right?
Yeah, has the event horizon reached this yet in nerves?
Um, but like you know, those are those are transformers too, yeah.
And and those are also like interesting things.
Anyway, uh, I just wanted to spend a little bit of time on on those that background before we go into dynamo uh property.
Yeah, sure.
I took a different path to NVIDIA than Adder.
I joined six years ago, seven if you count when I was an intern.
So I joined NVIDIA like right out of college, and the first thing I jumped into was not what I'd done in during internship, which was like, you know, like some stuff for autonomous vehicles like heavyweight object detection.
I jumped into like, you know, something I'm like recommenders.
This is popular, and yeah, and you did Rexis, yeah, Rexis, yeah.
I mean that that was the tabula data at the time, right?
You have tables of like audience qualities and item qualities, and you're trying to figure out like which member of the audience matches which item or more practically which item matches which member of the audience.
And at the time, really it was like we were trying to enable uh recommenders, which had historically been like a little bit of a CP-based workflow, into something that like ran really well in GPUs.
And it's since been done.
Like there are a bunch of libraries for XS that run on GPUs.
Uh the common models like deep learning recommendation model, which came out of Meta, and the wide and deep model, which was used or was released by Google, were very accelerated by GPUs using you know the fast HPM on the chips, especially to do you know vector lookups.
But it was very interesting at the time and super super relevant because like we were starting to get like this explosion of feeds and things that required recommenders to just actively be on all the time.
And sort of transitioned that a little bit towards graph neural networks when I discovered them because I was like, okay, you can actually use graphical neural networks to represent like relationships between people, items, concepts.
And that that interested me.
So I jumped into that at NVIDIA and and got really involved for like two-ish years.
Yeah.
Uh and something I learned from Brian Connezzaro, yeah, is that you can just kind of choose your own path in NVIDIA.
Oh my god, yeah.
Which is not a normal big corp thing.
Yeah, like you you have a lane, you stay in your lane.
I think probably the reason why I enjoy being in a big company.
The mission coming from a startup guy, yeah.
The mission is the boss.
Yeah.
Uh it feels like a big game of pick-up basketball.
Like, you know, if you play one, if you want to play basketball, you just go up to the court and you're like, hey, look, we're gonna play this game and we need three.
Yeah, and you just like find your three.
That's honestly for every new initiative.
That's what it feels like.
Yeah, yeah.
And it also like shows, right?
Like NVIDIA is just releasing state of the art stuff in every domain.
Yeah.
Like, okay, you expect foundation models with Nemotron.
Voice just randomly like top tier parakeet just comes out.
Another one, uh, the voice team has always been producing.
There's always just every other domain of paper that comes out, data set that comes out, and it's like, I mean, it also stems back to what NVIDIA has to do, right?
You have to make chips years before they're actually produced, right?
So you need to know, you need to really forget that.
The design process starts like exactly three to five years before the chip gets to the market.
Yeah, I'm curious more about what that's like, right?
So like you have specialist teams.
Is it just like, you know, people find an interest, you go in, you go deep on whatever, and that kind of feeds back into, you know, okay, we we expect predictions, like the internals at NVIDIA must be crazy, right?
You know, you know, you you must not even without selling the people, you have your own predictions of where things are going, and they're very based, very grounded, right?
Yeah, it it's really interesting.
So there's like two things that I think that NVIDIA does, which are quite interesting.
Uh one is like we really index into passion.
There is a big sort of organizational top sound push to like ensure that people are working on the things that they're passionate about.
So if someone proposes something that's interesting, many times they can just email someone like way up the chain that they would find this relevant and say, like, hey, can I go work at the end of the day?
That's actually like I worked at a big company for a couple years before uh starting on my startup journey, and like it felt very weird if you were to like email out of chain, if that makes sense.
Yeah.
The emails at NVIDIA are like mosh pits.
Shoot.
And it's just like 60 people, just whatever.
And like they're doesn't think I'm messy, like reply all the time.
Oh, it gets in it's insane.
It's insane.
It does help, you know, match the context.
But that's actually like I've actually so this is a weird thing where I used to be like, why would we send emails?
We have Slack.
I am the entire I'm the exact opposite.
I feel so bad for anyone who's like messaging me on Slack because I'm so unresponsive.
So your email max is a good thing.
I'm I'm email maxing out.
Email is a different thing.
Email is perfect because we can't work together on Slack.
Email is great because important threads get bumped back up, right?
Yeah, yeah.
Um, and so Slack doesn't do that.
So I just have like this casino going off on the right or on the left, and like I don't know which thread was from where or what, but like the threads get and then also just like the subject.
So you can have like working threads.
I think what's difficult is like when you're small, if it's not 40,000 people, I think Slack will work fine.
But there's I don't know what the inflection point is.
There is gonna be a point where that becomes really messy, and you'll actually prefer having email because you can have working threads, you can CC more than nine people in a thread.
You can fork stuff, you can fork stuff, which is super nice, and just like yeah, and so but that is part of where you can propose a plan.
You can also just like start honestly, momentum is the only authority, right?
So, like if you can just start start to make a little bit of progress and show someone something and then they can try it.
That's I think what's been you know, I think the most effective way to push anything for forward.
And that's both at NVIDIA and I think just generally, yeah.
There's the other concept that like is explored a lot at NVIDIA, you want to go and start a zero billion dollar business?
Jensen says we're completely happy investing in zero billion dollar markets.
We don't care if this creation is a big thing at NVIDIA.
It's important for us to know about this market.
We think it will be important in the future.
It can be zero billion dollars for a while.
I'm probably mangling his words here.
But like, you know, like I'll give an example.
NVIDIA's been working on autonomous driving for a long time.
Like an Nvidia car?
No, no, no.
They've used the Mercedes, right?
They're on the HQ.
And I think it finally just got licensed out.
Now they're starting to be used quite a bit.
Yeah, but for 10 years, you've been seeing Mercedes with NVIDIA logos.
Okay.
Yeah, if you're in like the South Mayor Santa Santa Clara, it's a it's just this track is right.
Yeah, yeah.
So um zero billion dollar markets are are a thing.
Like, you know, Jensen.
I mean, okay, look, cars are not a zero billion dollar market, but yeah, that's a bad example.
I think I think he's he's messaging uh zero today, but or even like internally, right?
Like, like it's like uh an org doesn't have to ruthlessly find revenue very quickly to justify their existence, right?
Like a lot of the important research, a lot of the important technology being developed.
That that's kind of where you're talking about.
Research is very ideal ideologically free at NVIDIA.
Like they can pursue things that they were you research officially.
I was never in research officially, I was always in engineering.
I'm in an org called deep learning algorithms, which is basically just how do we make things that are relevant to deep learning go fast.
That sounds freaking cool.
And I think a lot of that is underappreciated, right?
Like Time Series.
This week, Google put out time FX, a new time series paper.
Rexis, uh semantic IDs started applying Transformers, LLMs to Rexis.
And when you think the scale of companies deploying these, right?
Amazon recommendations, Google web search, like it's huge scale, and you want fast.
Yeah, yeah.
Yeah, actually, it's it I there's a fun moment that brought me like full circle.
Like uh Amazon ads recently gave a talk where they talked about using dynamo for generative recommendation, which was like super like weirdly cathartic for me.
I'm like, oh my god, I've I've supplanted what I was working on.
Like I you're using LLMs now to do what I was doing five years ago.
Yeah, yeah.
Um amazing.
And let's go right into dynamo.
Uh maybe introduce sort of top-down and yeah.
I think at this point a lot of people are familiar with the term of inference.
Like, funnily enough, like I I went from you know, inference being like a really niche topic to being something that's like discussed on like normal people's Twitter feeds.
It's on billboards here.
Yeah, very, very strange.
Drive driving seeing just an inference ad on 101.
Inference at scale is becoming a lot more important.
Uh, we have these moments like you know, open claw where you have these agents that take lots and lots of tokens but produce incredible results.
There are many different aspects of test time scaling, so that you know you can use more inference to generate a better result than if you were to use like a short amount of inference.
There's reasoning, there's re-quarrying, there's adding agency to the model, allowing it to call tools and use skills.
Dino sort of came about at NVIDIA because myself and a couple others were sort of talking about these concepts that like you know, you have inference engines like VLM, SQLang, Tensor TLM, and they have like one single copy.
They sort of think about like things as like one single copy, like one replica, right?
Like one version of the model.
But when you're actually serving things at scale, you can't just scale up that replica because you end up with like performance problems.
There's a scaling limit to scaling up replicas.
So you actually have to scale out to use a maybe some Kubernetes type terminology.
We kind of realized that there was like a lot of potential optimization that we could do in scaling out and building systems for data center scale inference.
So Dynamo is this data center scale inference engine that sits on top of the frameworks like VLM, SHL LANG and TensorHLM and just makes things go faster because you can leverage the economy of scale of the fact that you have KV cache, which we can define a little bit later in all these machines that is like unique and you want to figure out like the ways to maximize your cache ships, or you want to employ new techniques in inference like disaggregation, which Dynamo had introduced to the world in in March, not introduced.
It was an academic talk but beforehand, but we're you know one of the first frameworks to start supporting it.
And we want to like sort of combine all these techniques into sort of a modular framework that allows you to accelerate your inference at scale.
By the way, Kyle and I became friends on my first date NVIDIA, and I always loved because like he always teaches me new things.
By the way, this is why I wanted to put two of you together.
I was like, yeah, this is good.
This is gonna be great.
It's very different, you know.
Like we've we've we've talked to each other a bunch.
Actually, you asked like why why can't we scale up?
Yeah, model you said model replicas.
Yeah, so you so scale up means assigning more heavier, yeah, heavier, like making things heavier, adding more GPUs, adding more CPUs.
Scale out is just like having a barrier saying I'm gonna duplicate my representation of the model or representation of this some microservice or something, and I'm gonna like replicate it many times to handle the load.
And the reason that you can't scale scale up uh past some points is like you know, the there are sort of hardware bounds and algorithmic bounds on on that type of scaling.
So I'll give you a good example that's like very trivial.
Let's say you're on an H100.
The maximum MV link domain for H100 for most DJX H100s is H GPUs.
Right.
So if you scaled up past that, you're gonna have to figure out ways to handle the fact that now for the GPUs to communicate, you have to do it over InfiniBand, which is still very fast, but is not as fast as NB link.
Is it like one order of magnitude, like hundreds?
It's about an order of magnitude.
Um not terrible.
Yeah, I I need to I need to remember the the data sheet here.
Like I think it's like about 500 gigabytes uh a second unidirectional for NVLink and about 50 gigabytes a second unidirectional for Infiniband.
Uh it it depends on the the generation.
I just want to set this up for people who are not familiar with these kinds of like layers and the tractor speeds.
Also, maybe even just going like a few steps back before that.
Like most people are very familiar with you see uh, you know, you can use on your laptop whatever these SD Lang, VLLM, you can just run inference.
There's all I'm there's you can run on laptop, then you get to okay, uh models got pretty big, right?
GLM five, they doubled the size.
So uh what do you do when you have to go from okay, I can get 128 gigs of memory, I can run it on a spark.
Then you have to go multi-GPU.
Okay, multi-GPU, there's some support there.
Now, if I'm a company and I don't have like I'm not hiring the best researchers for this, right?
But I need to go multi-node, right?
I have a lot of servers.
Okay, now there's efficiency problems, right?
You can have multiple eight H100 nodes, but you know, is that as a like how do you do that efficiently?
Yeah, how do you like represent them?
How do you choose how to represent the model?
Right?
That's like that's like a hard question everyone asks.
Like, how do you size?
Oh, I want to run GLM5, which just came out, new model.
There have been like four of them in the past week, by the way.
Like a bunch of new models.
You know why, right?
Deep sequence.
No problem.
Yeah, but GLM5, right?
We we have this new model, it's it's of like a large size.
And you have to figure out how to both scale up and scale out, right?
Because you have to find the right representation that you care about.
Everyone does this differently.
Let's be very clear.
Everyone figures this out in their own path.
I feel like a lot of AI or ML even is like is like this.
I think people think, you know, I I was there was some tweet a few months ago that was like, why hasn't fine tuning as a service taken off?
And you know, and like that might be me.
It might have been you.
Yeah, but people want it to be such an easy recipe to follow.
But even like if you look at an MOE.
It's specific to you.
Yeah, yeah.
And the model.
And then there's so much tinkering.
Like, like when you see a model that has however many experts in an MOE model, it's like, why that many experts?
You know, they tried a bunch of things and that one seemed to do better.
And I think when it comes to how you're serving inference, you know, you have a bunch of decisions to make.
And there you can always argue that you can take something and make it more optimal, but I think there's this internal calibration and appetite for continued calibration.
Yeah.
And that doesn't mean like, you know, people aren't taking a shot at this.
Like Tinker from thinking machines, you know.
RL as a service.
Yeah, totally.
It's it also gets even harder when you try to do big model training, right?
We're not the best at training MOEs.
Uh, when they're pre-trained, like we saw this with Lama 3, right?
They're trained in such a sparse way that Meta knows there's gonna be a bunch of inference done on these, right?
They'll open source it, but it's very trained for what meta infrastructure wants, right?
They want to they want to inference it a lot.
Now, the question to basically think about is okay, say you want to serve a chat application, a coding copilot, right?
You're doing a layer of RL, you're serving a model for X amount of people.
Is it a chat model, a coding model, dynamo, you know, back to that?
It's like sorry.
So you we we sort of like jumped off of you know, jump jumped up on that topic.
Everyone has like their own journey.
And I I like to think of it as defined by like what is the model you need, what is the accuracy you need?
Actually, I talked to Nattery about this earlier.
There's three axes you care about.
What is the quality that you're able to produce?
So, like, are you accurate enough or can you complete the task with enough performance?
High enough performance.
Yeah.
Uh there's cost.
Can you serve the model or serve your workflow?
Because it's not just the model anymore, it's the workflow, it's the multi-turn with an agent cheaply enough.
And then can you serve it fast enough?
And we're seeing all three of these like play out.
Like we saw we saw new models from OpenAI that you know are faster.
You have like these new vast versions of models.
You can change the amount of thinking to change the amount of quality, right?
Produce more tokens, but at a higher cost and a higher latency.
And really, like when you start this journey of like trying to figure out how you want to host a model, you you you think about three things.
What is the model I need to serve?
How many times do I need to call it?
What is the input sequence link was that what is the workflow look like on top of it?
What is the SLA?
What is the latency SLA that I need to achieve?
Because there's usually some this is usually like a constant.
You you know the SLA that you need to hit.
And then like you try and find the lowest cost version that hits all these constraints.
Usually, you know, you you start with those things and you say you you kind of do like a bit of experimentation across some common configurations.
You change the tensor parallel size, which is a form of parallelism.
I'd say it goes even deeper.
First kind of thing, what model is that?
It's like it's like a multi-step design process because as you said, you can you can choose a smaller model and then do more test time scaling, and it'll equ equate quality of a larger model because you're doing the test time scaling or you're adding a harness or something.
So yes, it goes way deeper than that.
But from the performance perspective, like once you get to the model you need you need to host, you look at that and you say, Hey, I have this model, I need to serve it at the speed.
What is the right configuration for that?
Do you guys see the recent uh there's a paper I just saw like a few days ago that uh if you run the same prompt twice, you're getting like double.
Yeah, exactly.
And you get a lot, yeah.
But the the key thing there is you give the context of the failed try, right?
So it takes a shot.
And this has been like you know, basic guidance for quite a while.
Just try again.
Because you know, try this is just try again, dude.
You try again.
All advice in life.
Just try again.
It's a paper from Google, if I'm not mistaken, right?
I think it's it's like a seven page little short paper.
Yeah, yeah.
The title is very cute, and it's just like, yeah, just try again.
Give it as context.
You just like say, like, hey, like, you know, like take take a little bit more, take a little bit more information.
Try and fail, fail.
And that basic concept has gone pretty deep.
There's like um self-distillation RL where you you do self-distillation, you do RL and you have past failure, and you know that gives some signal.
So people take try it again, not strong enough.
Uh for for listeners uh who listen to here, uh, V Bo actually and I, and we run a second YouTube channel for our paper club where we would just cover this self-dissillation and all that.
That's that's why he's so up to speed on it out.
Yeah, it's it's just a good practice.
Like everyone needs like a paper club where like you just read papers together and the social pressure just kind of forces you together.
We have there's like a big inference reading group.
I feel so bad every time I I he put it on like on our he shared it.
Yeah.
One of your guys uh is is big in that.
I forget.
Yeah, Sean.
Yes,han's on my team.
Actually, funny.
There's a there's a there's an employee transfer between us.
Eshan worked for NATO at Brev, and now he's he was our head of AI, and then yeah, once we got in the case.
And uh Ishan was like, I'm always looking for like nudge Isan into like is there something here?
I mean, I don't think there's this new inference techniques every day.
So it's like it's you would you would actually be surprised.
Um the amount of blog posts you see.
And if you there's a period where it was like Medusa, Hydra, what Eagle, like you know, now we have new forms of decode spec uh we have new forms of speculative decoding or new.
What do you expect?
And it it's exciting when you guys put out something like NemoTron, because I remember the paper on this Nemo Tron 3.
Uh the amount of like post-training, the amount of tokens that the GPU rich can just train on.
And it it was a hybrid state space model, right?
Yeah, it's co-designed for the hardware.
Yeah, co-designed for the hardware.
And one of the things was always, you know, the state-space models don't scale as well.
When you do a conversion or whatever, the performance and you guys are like, no, just keep training.
And Nemo Tron shows a lot of that.
Yeah.
Also, something cool about Nibotron, it was released in layers, if you will.
Very similar to Dynamo.
It's it's it's essentially it was released as aggregated.
You can the pre-training, post-training data sets are released, the recipes on how to do it are released, the model itself is released full of the same.
So you can just benefit from us turning on the GPUs.
But there are companies like uh Service Now took the data set and they trained their own model.
And we were super excited and like you know, celebrated that work.
Zoom, the frontier model.
Zoom is Zoom is AGI.
I think uh, you know, also just to add, like a lot of models don't put out base models.
And if there's that, why is fine-tuning not taken off?
You know, you can do your own first training, but you guys put out base model.
I think you put out everything.
I believe basically can put out base can be canceled cancelable.
Yeah.
Safety training.
Do we get a full picture of dynamo?
I don't know if we can.
What I'd love is you mentioned the three axes.
Like break it down of like, you know, what's pre-fill decode and like what are the optimizations that we can get with dynamo.
Yeah, that's that's that's that's a great point.
So to summarize on that three-axis problem, right?
There are three things that determine whether or not something can be done with inference cost, quality, latency, right?
Dynamo is supposed to be there to provide you like the runtime that allows you to pull levers to you know mix it up and move around the Pareto frontier or the Pareto surface that determines is this actually possible with inference and AI today?
It gives you the knobs.
Yeah, exactly.
Gives you the knobs.
Uh and one thing that like we we use a lot in contemporary inference and is you know starting to like pick up from you know, in in general knowledge is this call concept of disaggregation.
So historically, models would be hosted with a single inference engine.
And that inference engine would ping pong between two phases.
There's pre-fill, where you're reading the sequence, generating KV cache, which is basically just a set of vectors that represent the sequence, and then using that KV cache to generate new tokens, which is called decode.
And some brilliant researchers across multiple different papers essentially made the realization that if you separate these two phases, you actually gain some benefits.
Those benefits are basically A, you don't have to worry about step synchronous scheduling.
So the way that an inference engine works is you do one step and then you finish it and then you schedule, you start scheduling the next step.
It's not like fully asynchronous.
And the problem with that is you would have uh essentially pre-fill and decode are are actually very different in terms of both their resource requirements and their sometimes the runtime.
So you would have like pre-fill that would like block decode steps because you'd be you'd still be pre-filling and you couldn't schedule because you know the step has to end.
So you remove that scheduling issue.
And then you also allow you or you yourself to like split the work into two different types of pools.
So pre-fill typically, and and this changes as as model architecture changes.
Prefill is right now compute bound most of the time.
If a sequence is sufficiently long, it's compute bound on the decode side, because you're doing a full pass over all the weights and the entire sequence every time you do a decode step, and you're you don't have the quadratic computation of KV cache.
It's usually memory bound because you're retrieving a linear amount of memory and you're doing a linear amount of compute as opposed to prefill where you retrieve a linear amount of memory and then use a quadratic map.
That's funny.
Someone Exolabs did a really cool demo where for the DGX Spark, which has a lot more compute, you can do the pre the compute hungry pre-fill on a DGX Spark and then do the D code on a on a Mac.
And so that's faster, yeah.
Yeah.
So you can you can do that, you can do machine stratification.
Yeah.
And like with our future generators generations of hardware, we actually announced, like with Ruben, this new accelerator that is pre-fill specific.
It's called Ruben CPX.
So I have a question.
When you do the scale out, is scaling out easier with Dynamo because when you need a new node, you can dedicate it to either the prefill or uh decode.
Yeah.
So Dynamo actually has like a Kubernetes component in it called Grove that allows you to do this like crazy scaling specialization.
It has like this hot it's a representation that I don't want to go too deep into Kubernetes here, but there was a previous way that you would like launch multinode work.
It's called leader worker set.
It's in the Kubernetes standard.
And leader worker set is great.
It served a lot of people super well for a long period of time.
But one of the things that it struggles with is representing a set of cases where you have a multi-node replica that has a pair, right?
You know, pre-fill and decode, or it's not paired, but it has like a second stage that has a ratio that changes over time.
And pre-fill and decode are like two different things.
As your workload changes, right, the amount of pre-fill you'll need to do may change.
The amount of decode that you you'll need to do might change, right?
Like let's say you start getting like insanely long queries, right?
That probably means that your pre-fill scales like harder because you're hitting these this quadratic scaling growth.
Yeah.
And for listeners, like pre-fill will be long input, decode will be long output, for example.
Right.
Yeah.
So like decode decode scale.
I mean, decode is funny because the amount of tokens that you produce scales with the output length, but the amount of work that you do per step scales with the amount of tokens in the context.
Yes.
So it both scales with the input and the output.
That's true.
But on the pre-fill decode side, like if suddenly like the amount of work you're doing on the decode side stays about the same or like scales a little bit, and then the pre-fill side like jumps up a lot.
You actually don't want that ratio to be the same.
You want it to change over time.
So Dynamo has a set of components that A tell you how to scale.
It tells you how many pre-fill workers and decoded workers you it thinks you should have.
And also provides a scheduling API for Kubernetes that allows you to actually represent and affect this scheduling on your actual hardware on your compute infrastructure.
Not gonna lie, I feel a little embarrassed for being proud of my SVG function earlier.
No, it's really cute.
I like it.
It's all it's all engineering.
It's all engineering.
Uh sort of technical.
One thing I'm I'm kind of just curious about with all with you see at a systems level everything going on here.
And we're, you know, we're scaling it up in in multi in distributed systems.
Um I think one thing that's like kind of of the moment right now is people are asking, is there any SOL sort of upper bound in terms of like let's call just call it context length for one for a better word, but you can break it down however you like.
Yeah.
I just think like, well, yeah, I mean, like, clearly you can engage in hybrid architectures and throw in some state space models in there all you want, but it looks still looks very attention heavy.
Yes.
Uh yeah, long context is attention heavy.
I mean, we have these hybrid models.
Um most models like cap out at a million context, and that's it.
Like for the last two years has been it.
Yeah.
The model hardware context codesign thing that we're seeing these days is actually super interesting.
It's like my my passion, like my secret side passion.
We see models like Kimmy or GPT OSS.
I'm gonna use these because I I know specific things about these models.
So Kimmy2 comes out, right?
And it's an interesting model, it's like like a deep seek style architecture.
It's basically deep seek scaled like a little bit differently, um, and obviously trained differently as well.
But they they talked about why they made the design choices for context.
Kimmy has more experts, but fewer attention heads.
And I believe a slightly smaller attention uh like dimension, but I need to remember I need to check that.
Uh doesn't matter.
But they discussed this actually at length in a blog post on Jihu, which is like our Jeepu, which is like um in Chinese Reddit, yeah, it's a good idea.
Yeah, so it's a it's actually an incredible blog post.
Uh like all the MLSIS people in in in that I've seen that on Jeep are like very brilliant.
But they they they talk about like the creators of Kimi K2 actually like talked about it on on on there and in the blog post.
And they say, We have we actually did an experiment, right?
Attention scales with a number of heads.
Obviously, like if you have 64 heads versus 32 heads, you do half the work of attention.
You still scale quadratically, but you do half the work.
And they made a very specific like sort of barter in their system in their architecture.
They basically said, Hey, what if we gave it more experts?
So we're gonna use more memory capacity, but we keep the amount of activated experts the same.
We increase the expert sparsity, so we have fewer experts act the ratio to of experts activated to number of experts is smaller.
And we decrease the number of attention heads.
And kind of for context, what the what we had been seeing was you make models sparser instead.
So no one was really touching heads, you're just having a lot of people.
Well, they did they implicitly made it sparser.
Yeah, yeah for for Kimmy they did.
Yes.
They also made it sparser.
But basically what we were seeing was people were at the level of, okay, there's a sparsity ratio.
You want more total parameters, less active, and that's sparsity.
But what you see from papers like the labs like Moonshot Deep Seek, they go to the level of okay, outside of just number of experts, you can also change how many attention heads and less attention layers, more attention layers.
So and that's all basically coming back to just tie it together, is like hardware model co-design, which is harder model context co-design.
Yeah, right.
Like if you were training a model that was like really, really short context, uh or like really light is good at super short context tasks, you may like design it in a way such that like you don't care about attention scaling because it hasn't hit that like the turning point where like the quadratic curve takes over.
How do you consider attention or context as a separate part of the co-design?
Like I would imagine hardware, or just how I would have thought of it is like hardware model co-design would be hardware model context co-design.
Because the harness and the context that is produced by the harness is a part of the model once it's trained in.
Like, even though towards the end you'll do long context, you're not changing architecture through training.
I mean, you can try.
You're saying everyone's training the harness into the model?
I would say to some degree, or there's co-design.
I know there's a small amount, but I feel like not everyone has like gone full send on this stuff.
I think I think it's important to internalize the harness that you think the model will be running into the model.
Yeah, interesting.
Okay.
And like bash is like the universal harness.
Yeah, right.
Like I'll I'll give an example here, right?
I mean, or just like a like a it's easy proof, right?
If you can train against a harness and you're using that harness for everything, wouldn't you just train with the harness to ensure that you get the best possible quality out of the Well, the uh I can provide a counter-argument, which is what you want to provide a generally useful model for other people to plug into their harnesses.
Yeah, but so if you're harnesses can be open open source, right?
Yes, I mean that's that's effectively what's happening with codex.
Yeah.
But like you may want like a different search tool, and then you may have to name it differently.
I don't know how much people have pushed on this, but can you train a model?
Would it be have you have people compared training a model for the for the harness versus like post-training for it?
I think it's the same thing.
It's just okay.
Extra post-training.
I see.
And so I mean cognition does this, of course, it does this, where you you just have to like if your tool is slightly different, um, either force your tool to be like the tool that they train for or undo their training for their tool and then re-retrain it.
Yeah, it's it's really annoying.
And like I would hope that eventually we hit like a certain level of generality with respect to harness training.
So you can use new tools.
It's not AGI.
Like it's just like it's really stupid, like learn my tool, bitch.
Like I don't know if I don't know if I can say that, but like, you know.
Um I think what my point kind of is is that there's like I look at slopes of the scaling laws, and like this slope is not working, man.
We we're at a million token context, okay, maybe next year two million.
We're not going to a hundred trillion.
You know, like this just so many interesting ways to do that.
It doesn't work.
It doesn't work.
What's kind of funny is whenever there I feel like we always want to see a trend that we can predict, but every time something's come, it's been like a leapfrog.
So I I imagine I I don't know how we go from one to two, but I imagine what what's likely to happen is we break through that from some new Yeah.
There's actually there's an interesting formalization of this.
There's an essay, it's pretty interesting essay by Leopold Aschenbrenner called Situational Awareness.
Okay, yes.
He introduces a concept into awareness called an unhobbler, right?
So he you know, Leopold in this essay details, hey, I want to get, you know, like I want to get to this point in intelligence.
And I think that it is four orders of magnitude worth of like compute and data and training away.
And you know, he says, Oh, yeah, I think data centers can scale up by about this much.
I think that you can do scale up the data and some other things by this much.
But one of the things that like makes the rest of that order of magnitude growth pop possible is this unhobblers, like these scientific discoveries that are discovered during you know, model architecture search or training that really, really, really impact how how you are able to scale.
Like a a good example of this might be that like we see like a model, a lot of models that are, and this is probably a very tiny unhobbler, but is important for the performance perspective.
We see a lot of models that are like trained with multi-token prediction natively in during pre-training.
And per deep seek in their paper, they say, hey, this actually helped us ensure state more stable convergence.
But they're like unhobblers that are like that, and then there are like rather large unhobblers, right?
Like architecturally, a lot of our models, like we had different types of attention.
And one of the problems with attention is like you have a lot of KV, but people have found like different forms of attention, like group query attention, and uh like MLA in Deepseek, multi-head latent attention that like decrease the burden that KV has on the model, which allows you to grow like longer in context.
Yeah, and that that was very drastic for Deep Seek.
Yeah, yeah, it for context, like the the total, I think the total context length of Deep Seek is 128,000 tokens or might be 256,000 with rope extension.
That entire context, I think it's 128,000, fits into eight gigabytes.
And previously, context, like I think the the Llama 405B context of a similar size was like 40 or 80 gigabytes in the same precision.
Yeah.
Um, so like those unhobblers like really decrease the stuff of that size.
And I wouldn't be surprised if we do see the ability to like break through to like 10 million, 20 million, 100 million context through the an unhobbler showing up.
I see.
And it's just science.
More deep learning algorithms is what we're doing.
Yeah, more deep learning algorithms.
Um, pick up and he has room for two.
I could actually give you an example like of like a a theory, not a theory theory, but something theoretical that you're excited about or an unhobby that I mean I haven't seen.
So it could be a tar pit and it could not just not work.
But uh I I would be really excited to see a model that does prefill and decode differently.
So a model that does uh pre-fill like locally like document wise pre-fill like it doesn't in chunks and then you do decode globally across like the entire sequence.
Because it logically to me it doesn't seem like you would necessarily need to have KV be associative between documents that have like no mutual association.
But that like places a lot of burden on prefill to like or sorry on on decode and pure attention within the decode phase to like make those connections since the KV is like static at that point.
And you see other techniques that are interesting like this too.
But if if you're able to do that like if pre-fill becomes local and decode is is still global you solve that pre-fill quadratic scaling problem because you have a bunch of like small chunks that you pre-fill independently.
Okay.
All right.
Well, let's uh wait and see but I I think it'll be pretty exciting.
Fingers crossed.
Yeah, fingers crossed.
Yeah, yeah.
I'm excited for pre-fill decode on separate hardware.
So like Grok acquisition, right?
Can we decode on the Grok?
Can we get super fast?
I don't think I'm allowed to comment on this.
Mark is going to shoot arrows at us.
Uh he's got a little dark yeah, he's in the side of the room just like go to sleep.
Yeah, yeah.
I'm I'm super excited to see the team come in and like, you know, I've gotten the the pleasure of working with some of the the grok people coming in.
So you know I I know Sunny, we've had him uh at the same conference that you were at.
Yeah.
Um and uh I I think you're you guys are gonna be doing some sessions at GTC.
I don't know if you want to, this is a good place to plug them.
Yeah, yeah, yeah.
So I can't speak to any LPU related sessions at GTC.
I have no idea about that.
Oh no, that was not the UQ on the on the GROK side, yeah.
I use the associative NVIDIA you.
Um on the on the NVIDIA dynamo side, we're we're giving it there are a large number of sessions.
For those that aren't aware, you can actually search all of these sessions for GTC online and just go to the GTC website.
I don't know what the URL is, but go there.
Google it.
Yeah.
Uh and you can just look up Dynamo and you'll get all the sessions.
There are about 20.
There are a couple that are hosted by the Dynamo team.
There are a couple that are hosted by people that use Dynamo that want to show off the results they've been able to get.
But there are two that I'm really excited about.
Uh one is just the general dynamo tutorial.
And this is the I'm going out with Harry, who's our lead product manager for Dynamo, and we're sort of talking about like how to use Dynamo to get better performance and also like where we see dynamo going in the future.
And then there's another session that I'm doing with one of our agents teams at NVIDIA to talk about sort of the future of agents in production inference.
So we're talking about there's like this new horizon with respect to agents because we have these harnesses that actually impart structure among upon calls.
Like if you if you compare like the past and the and the present with respect to like how LM calls work, like in the early days when there were chatbots, like every call was like very different.
There was basically no structure.
You could assume that like people you if it's conversational, there might be like some implicit structure because you have you know a multi-turn conversation.
But agents, you have this this harness that like abides by rules, right?
So it imparts direct structure onto the context.
And you see this, there's an interesting Twitter post about how Claude Code like structures its context so that you get as many caches as possible.
And I think it was by one of the PMs for Cloud Code, and he he wrote about it.
And that type of structure that the harness can impart actually like goes hand in hand with the inference co-design.
So I'm doing a talk, I I don't know the session name or the session number, but I'm I'm doing a talk.
Uh, you can look at me up by name on on the GTC website on how we accelerate agents and where we see specific optimizations for agents going in dynamo and in inference in general.
Yeah, I think there's only one PM for cloud code, and it's kind of woo.
There is uh there's there's DevRel, there's Boris.
Maybe it was maybe DevOps.
Yeah, exactly.
I mean, let's go into agents.
I think this is like the last part of the the discussion we planned.
How have we not talked about agents?
Also with you guys.
We scheduled it.
I was like, okay, you know, like let's have like cohesive sections where I mean there's the big news, right?
The Nvidia is a huge like deployment of codecs.
Yeah, NVIDIA uses everything.
I mean, you use this cursor and we use this code.
But that's that's a pretty big deployment, right?
Like that's tens of thousands of people.
Totally.
Yeah, we're just curious.
Yeah, I mean it goes back to the mosh pit of emails we kind of mentioned earlier, or just the like um how fluid the org feels.
So when there's new technology, people will just email it out and everyone will try it.
And if it if it's making people's lives easier, it'll spread like wildfire.
A lot of times Jensen will get it and be like, let's make this work across the company.
Let's make this work right now.
Honestly, uh, if I was a startup, I feel like a cool hack.
If you have something that's gonna save an NVIDIA's time, they'll spread it to a couple in the same thing, right?
It'll just spread like wildfire.
Be careful before your email blows up from startups.
Well, you gotta have to know the person, right?
But no, I um I yeah, so I mean we I love using codecs, it's been a ton of fun.
Uh I've been using it personally, been using it at work.
It's been um, yeah, I don't know, it's been great to see the rollout.
Something really funny.
Uh on a day that we got uh codecs and clawed code access.
I found this person, uh his name's Carlos at the company, he wrote an Outlook CLI.
Oh, yeah.
And uh just a CLI for email.
And this was I've been using that.
Yeah, like four or five weeks ago, and uh the site.
So once I got like codex access, I installed the CLI, it had a skill, and I just asked it to go through all of my emails, which it's very messy.
So I don't respond to your email, I'm really sorry.
But I asked it to give me a summary, highlight any escalations that I should look at, put any thread that it thinks I should respond to in a folder and then archive everything.
And it did.
So if I missed your email, that's because it didn't give.
So I should put a prompt injection in my emails to what you should do is just FaceTime, which is yeah, yeah.
Um yeah, my SLA's highest on FaceTime.
But that was it was magic.
And so sent it in a big email thread to like 500 people, a bunch of folks tried it out.
I started like FaceTiming whoever I could at the company to get them set up with this.
Yeah, um that specific example, you guys deal with like some pretty sensitive emails.
Yeah.
Is there a security review with this?
Because I one guy made it for himself, but like it's not meant for all the security.
Like, shout out to them.
They're they're they're trying to.
We have an amazing security team because they're progressive and they know that this is really important technology, and we have to bring it in.
If you think about it, like if you work at a big company, your laptop's usually very locked down.
If that you can only access certain things, NVIDIA engineers have those restrictions aren't there.
So you're expected to understand the risks when you try things out.
And so very quickly, you know, made sure to chime in security on what we were doing.
There's actually a lot that we've been thinking about, especially with open cloud, right?
Like there's you know, agents can do three things.
Yeah.
Agents can do three things.
They can access your files, they can access the internet, and then now they can write custom code uh and execute it.
And you should really only let an agent do two of those three things.
If you can access your files and you can write custom code, you don't want internet access because that's one is the vulnerability, right?
If you have access to internet and your file system, you should know the full scope of what that agent's capable of doing.
Otherwise, now we're can get injected or something that can happen.
And so that's a lot of what we've been thinking about is like, you know, how do we both enable this because it's clearly the future, but then also, you know, what what are these enforcement points that we can start to like protect?
And is there any directive of like, hey, we have a company account or a company agreement with OpenEI, we use open EI models here, or like choose whatever.
Um, no, no.
So so I would never put any company data in a model that's not either that we don't either it has the most security.
Yes.
Yeah.
Like how to how that goes.
Uh, you know, obviously, you could run your own models, you know, NemoTron and we did open.
Yeah, we have an we have an internal cluster, so we you know, uh of course running the abandoned.
Uh yeah, yeah, I think we're Dynamo's first customer.
Actually, uh, there's a funny story about like how I got the experience that informed what we needed for Dynamo.
At one point, there's a website called build.mvidia.com, and also for us inference.com that is allows people to try models.
Like it gives an API service, you can call the model with like a REST API and you know, you get a response.
I ran the model side for that, and it was at one point the largest inference deployment, and still may actually be the largest inference deployment NVIDIA.
I've I've since like handed it off to some people and they're doing a wonderful.
This is an extremely underknown or less known resource.
Build.
And and the SLA on getting models, day zero models up is like a day.
Yeah.
Like they're they're incredibly good at like figuring out the right way to host the model to get it up there as soon as it comes out.
You ran this?
Yeah, I ran.
I ran it a long time ago.
It was originally called NVIDIA AI Playground, and then it was called Oh, yeah, I found the original answer.
Yeah, and then it was called build.mvidia Call.
And I I ran the model side of it.
So there were there was a large multi-organizational team.
I ran how which models should we host?
How should we host them?
And like what's the proportion of them?
And then of course there was like an SRE team that like made sure that things ran well and scaled the models as well.
But I ran like, you know, model how do we get the model to silicon, and then which also worked with our product team to determine like which models were important a very long time ago.
Yeah, yeah.
There's also like a middle ground in between though.
This is like for the hacker, try anything.
There's the Brev console, then there's Dynamo.
There was also NIMS, right?
Yeah, I remember it had its little moment like a year or two ago.
Is it still?
Yeah, no, no.
Nim is uh, you know, inference uh oil.
I think it looks like for something it's it's no longer in acronym.
Yeah, it's just an NIM.
Um yeah, NIM is uh how enterprises can take our uh any of the any of this technology and run it with support and all of that, and so that includes Danimo, that includes I don't know, all of our other optimizations that are package of for enterprise.
Yep, yeah, yeah.
Uh anyway, so so you got you got a bunch of experience like running the sort of internal inference gateway playgrounds.
Yeah, yeah, I'm Bill also built how Nvidia's first internal like VS Code thing.
We called it MB code.
It's what I the uh extension, right?
Yeah, it was a VSL like the fork VS code and we jokes absolutely not.
It just a while back be like we should have a fourth VS Code hackathon where you that's for the best fork V VS Code.
We are doing a how do you make a billion dollars?
Someone from VS Code was there, and he was like somewhat done to get involved, and I was like, Oh, you should do that.
That's all I said.
Then did a cool thing became for Chrome Hackathon.
Chrome.
And no, no, no, IDs are not cooling.
I also thought what's it called?
I was talking to Joseph uh from Roboflow.
And uh your partner in crime.
We were talking about how with the new Alpamayo model, so NVIDIA just released an open source uh the the Mercedes cars that you saw.
Drugs on Frazy, yeah.
Release will you open source a autonomous driving model?
Uh I yeah, so we were thinking like, could we hackathon a driverless car?
Like, I have my old car, let's just try it.
We'll take it take it to like click trail.
Yeah, what it was treasure island in the middle of the day.
Just like just see let it run.
Yeah, like how many how many cameras do we need, right?
Like one, two, three, four.
I don't know, maybe five space.
I don't know.
I yeah, but um, I think we're gonna try you just do it with us.
We can see we could even have a race, it's like the first person to automate their the driving.
I mean, over a weekend, we do have an autonomy track at Wilsfair.
Yeah, uh, we mow was there, like yeah, Nvidia did send people those for GUT, and because he didn't have the driving thing yet, yeah.
Yeah, it's that's cool.
Yeah, I think also has a version of this.
They've done a fun hackathon on it.
Is it he and I had because I really what I really want is a Tesla with Tesla level self-driving, yeah, but as a smart car, like a two-seater that's the basically a wheelchair with a roof.
I don't even think they make them into a bunch of.
The demand has been there.
They're like this for like five years.
Yeah.
Really?
Yeah.
They were a different manufacturer raising the link.
I thought it was one of those things we'll s where we'll see someone buy the brand and it'll be revived.
I I would buy it.
Like I probably would be for go.
Someone hears this, go buy your car.
Yeah, yeah.
That's crazy.
No, because Mercedes because that they're like, I think most can bring it says Mercedes.
Uh I wouldn't know.
They'm in their saying you're used to make them, yeah.
I don't know.
I feel like they own the brand.
And you out.
That's actually Your Dream might come true.
Okay, we're time to be a small.
And it was another like the every time I see I try to park in San Francisco, I have to buy a smart car.
Because like 20% of the parking lots in San Francisco only fit smart cars.
Yeah.
Really?
That's where I mean all.
You remember that it was late here trying to this comfort.
That's what the the Vespa was a life hack.
Yeah, exactly.
Yeah.
You know what happened to the Vespa?
Um I used to have this yellow Vespa.
Uh I left it outside the hacker house when we moved out.
This trend um it's just it was always there and then like a month ago, it's not there anymore.
I've been meeting to be I don't know.
You could let us actually be like a DV here.
You forgot about it.
Yeah.
And left.
Okay.
Yeah, yeah, yeah.
No, it's it's probably hazard.
And speaking of hacking class, I also wanted to be give a big shout out to the World Shortest hackathon.
Let's go.
Uh you did twice.
Yeah, there's gonna be one at GTC.
Oh, we're doing that one.
Pretty much we have a bunch of challenges that no we haven't released and you get to bring your agent to come and attempt to uh go through those channels.
Because it's like a zero the zero minute hackathon idea is you just you just bring your I have remote either a long a long time ago.
You just bring your agent and then you press the go button.
You're not allowed to code.
It's just the agent doing hackathon.
It's a good hidden email, right?
Yeah, you make a JRO and you make like this I mean I would love to see from Cognition or someone else be like come bring your agent, like drop it in.
Because you don't know the ISO profile will it be uh you know operate a browser order a pizza will it just see you like that snake game you know and you don't know what the task is yeah I don't know what the task is like it we're just like you don't even know what the judging categories are and then you give it the judging categories like trying as much as possible it's great though it turns into like yeah so let's build something on dynopod it's a great business anyway funny story actually we have a couple of people at NVIDIA we've been working with security to like bring agents really close to compute so we now have like stuff where we can like tell dynamo like go run some experience with dynamo like on X cluster and just like try it right now like queue up once you get queued like send this request load and we've actually been able to like just like you know like one shot problems like we used to have this problem where you know with with dynamo you have to like find the right configurations and we sort of do it automatically for some parts of it but you have to like a good initial configuration that you want to use and we've just had like an agent just completely one-shot that it goes it gets the compute it like runs a couple experiments it's like this is the best.
This is this is these are part of the Prada frontier.
Go run this.
And then we just like give that to people, and it's like faster than anything that they have.
Agent UX and agent marketing are super important.
The stuff that we've been thinking a lot about.
Um Alec is like redoing the entire Drev CLI um so that you can fetch all the different compute types that are available.
I don't know, it's gonna be really soon.
But then you can you can just browse what GPUs are available and then provision one SH to it right there, and you can pipe all the commands.
But I think it goes back to like the Alex DLI.
Like if you coding agents are it's kind of funny.
I feel like coding agents have been so much more effective than a general purpose agents.
And I think a large part of that is it just has access to the terminal, like you said, and that means it has access to everything that you've installed into your terminal.
It can run it so you know it would write code and then it can compile the code, and if there are errors, it can fix it, it can run your suite of tests because that's all just in your terminal.
And so that you know, then that's for the idea or what I've got me really excited about the outlook CLI.
We're now just churning through building CLIs for the entire like for the entire business suite.
Slack building.
Also work based CLI and AP go.
I I've also done it for myself for some.
Really?
Yeah, yeah.
Um, we're gonna we're gonna open source all of this.
And like, yeah, all the the I mean, they're just they're the yeah, CLIs for the business applications.
We would love for someone to run with this and like build like I don't know, like open CLI foundation in or something.
Yeah, we NVIDIA would love to support uh anyone that's doing this.
Like every dev tool should really have good CLI support at this point.
Like at one point it was you want your docs to be like accessible by it LLM, right?
You want LM good docs.
No, every everything needs some CLI tool.
Yeah, it's kind of funny, right?
Like we like computing began with a terminal with a shell, but we said that it's not empathetic to uh humans, so we don't these nice user interfaces, and then now we have LLMs navigating our user interfaces, and ironically, we're not empathetic to the machine anymore.
Yeah, just give the the LLM access to the shell.
One thing that's slightly makes me unfortunately is like why do we have to build CLIs?
Why can't we just expose APIs?
Like I I have I have an interesting answer to this.
So there are a couple of reasons.
Like there's there's like you know, portability is like one issue.
Like, like you know, like sometimes APIs are not like discoverable or like reachable, right by some you know types of things.
There's some element of locality, right?
Like uh like the CLI is like literally you interfacing with your like local system, which is a little bit different.
You could still do it by API, but like there's this highlighting of like what is the difference between like a CLI and an MCP, right?
Like they kind of occupy the same purposes and you call them, it does something on the system and that's done.
I think that in pre-training, there's just an enormous amount of command line data.
Yeah, yeah.
Like even let's ignore our let's like let's ignore RL.
Like you're doing no harness, you're doing no harness post training.
Just the amount of like CLI versus API documentation for just like navigating this world of the CLI in your file system through that is just enormous.
Yeah, yeah, right.
I think there's there's a couple of things too.
Like, if let's say we want to so one, I think your intuition is right.
The CLI is just wrapping the API, right?
So functionally functionally, right?
Yeah, and I think it's nice because one, you're you're being very uh specific and pedantic, even um, of what and that's really good because you're describing the problem space.
So you know what the I don't know, I don't want to call it like what the space for vulnerability, you know what network calls you're making, it's not arbitrary and that's not decided on the fly.
That's like pre-decided, which is important from the security perspective.
But then if you were to write a bunch of API requests, you would probably do that.
I don't know, would the model like use Python to do so?
I kind of like that everything like a CLI is just dash because it's ubiquitous, like it's just there, and you don't have to make sure that there's certain environment variables that are set up.
Like if your Python version is different than my Python version, we're using the same model to go do the same thing.
Is it gonna write like different code?
It probably would.
And so it's kind of nice to go work, right?
With human as well.
So I think just like making those decisions happen ahead of time versus yeah.
One last thing on this sort of agent, I guess, maybe co-location or whatever you call it.
Uh, one pattern of tracking for this year.
I always try to think about what's the theme of this year gonna be.
Last year, definitely coding agents.
This year's definitely coding agents breaking out of containment into broader rate real worlds.
I go definitely has to be here rent a human.
Yeah, what's you do?
Yeah, I'm on there.
Are you really?
When I say I'm like five thousand dollars, I'll do anything.
I think so.
I need I need uh my uh my bowels from Costco.
Uh but I think the best part is only the aging can book me, you know.
Yeah, it's very usually like it's just like another labor marketplace.
Um Mechanical Turk was this.
So this way I have a weird story with why I did it.
So back to your example of just giving agent access to compute, right?
Yeah, you guys are GPU rich at NVIDIA.
I hooked up, he's not shy about it.
I have I have a 24-7 agent running.
I hooked it up to run pod.
It doesn't shut down instances, and I'm like, I've tried prompting it, I've given it instructions, shut down when you're done.
It's like I need to keep it warm, I'll need it soon.
And it's horrible on time estimates too, because like they realize it's like, yeah, I'll need it in 45 minutes, 45 minutes, I'll shut it down.
Forty five minutes of human time is actually three minutes of agent time, so it's like I'm booting it up, I'm waiting.
I'll just leave it on all night.
And mode modal is good at shutting down after some inactivity.
I had it on my local server, like a little dual GPU thing.
It just stays on.
I have a little space heater at home now, but careful.
So basically you know they don't care about the concept of money.
Just burn it.
I need it.
It's useful.
And another way with DGX Bark will be really nice.
Like I I think I'm looking at it as it's super useful for agents because yeah you you buy it once you plug it in and they it can rip.
I'm gonna make a I'm gonna make an Nvidia ad here.
Okay.
The Blackwell like RTX six thousand cards pro are only like I think it's eight thousand dollars slightly cheaper.
Yeah well it's much it's much cheaper than the data center cards.
Yeah.
And it's got ninety six gigabytes of VRAM.
So if you and your your crew want to go like run a local agent for you you know you you in the home I feel like it's got a significant amount of VRAM.
I've thought about purchasing this and running in my basement except my neighbors and hate me.
It's just a single like two, three slot GPU it's yeah it's a PCIe.
Yeah it's C A GPU.
You can go by that I mean the big difference against like the RTX like gaming GPUs is it it I mean obviously it's like black bulb like it's a pro GPU and it has a lot of ERAM, which means you can run pretty large models on it.
You can stack four of them for the max Q in a system.
But that's a base.
It's beefy.
You can run uh what is it 96 zigger and anything?
96.
Uh you don't know you seek.
Uh but also they they are slow.
They're not I mean performance of speed will be somewhat slower button to API, like oh yeah that that's true.
So again, the big learning economy of scale allows you to do things that allow you to get both speed and throughput.
Like you can run, I'll give you an example.
There's an optimization called wide ep.
I'm not gonna go into it fully, but like it featured heavily in in inference max for deep seek.
And there's a there's a great set of stories from Nvidia and from semi-analysis about like why YEP is important.
But for like MOE models, it's like basically essential, and you run it like the a level of parallelism, the level of scale up parallelism used for it is like 32.
So it goes beyond that eight barrier, and it like really, really, really is important to have that M MVL72 GB200 MV Link to serve at scale.
And like it's like I don't remember like the you know cost improvement.
I think against Hopper, right?
Like against Hopper with this MVL72 system, you're getting like 35 times cheaper per token for like a lot of the curve, yeah, which is crazy.
Yeah, and normalized per GPU, obviously, because part of the GPU's cost or the code the cheapest part of the cost.
One thing I'm exploring is the sort of this year is also the year at the subagent, um, where you have the main agent, but then that also kicks off tools which are in themselves agents that have limited agents and sort of context, local legals, whatever, right?
Different prompts.
So, for example, what one thing that Cognition does is before you kick off a search, they do is a like a fast context model where you kick off APR ages to search uh across the code base and cause all that that is better than indexing uh a lot of the times, not not all the times, and uh you should still index for some things, but like the idea that agents should be able to command sub agents and probably run them like maybe close to inference as well.
I I don't know if that's like architecturally possible or even yeah, we're we're thinking about that for Dymo.
That's like our big theme for the year.
Because like you're like if you can design that into your stuff, then a lot of people a lot more people will use it right now.
It's like just kind of theoretical because you do pay a lot of like back and forth uh coordination costs.
I think you'll net speed up though, right?
Like even at a basic level, speculative decoding, you're running a small model, you're running two instances, but it's not fun.
That is one example, yes.
Yeah, but you this is like a little bit like different with like agents, agents.
Yeah, this is not spectacular.
I I think I think there's like a summarization of that trend that I like to do where I like to say to my team, it's like this is the year.
So there are two things.
This is the year system as model, right?
Where like instead of having like a single model be a thing, you have a system of models and components that are working together to like emulate the black box model.
So when you when you make an API call to something that's like like a multi-agent in the background, it still looks like an API call to a model.
You're still getting back to under the hood.
Yeah, under the hood, it's like a billion different models, and that's a lot of complexity with dynamo and with other libraries and media where we're looking to help like manage that complaints.
Yeah, it's funny.
We actually for CES, we just released the model router for DGX Spark, where you can have a local model that's running on the Spark, and then also a foundational model, and then the model router decides when to send queries to which one.
So it's no longer this like either or it's use the best of everything that's available to you.
You have a good post-trained model that's running on it.
It's relates to the also the breadth functionality of being able to manage the spark.
Oh, that'd be cool.
Oh, yeah.
I did be able to j request yeah.
I actually have a question.
Like I'd like to like extend and flip over how much longer do you guys think like agents are gonna be running?
Because that's one thing I've been throwing around.
Like, what happens when I mean always more it even affects the like back to the prefill d uh decode, right?
Like codex is I'd say compared to Cloud Code, it's much longer at tasks.
Like that thing will like to run six, seven, eight hours.
I'll run it overnight.
Yeah, and I'll I'll go back and I have like a little crappy logging software I use, and there's just times where it wants to like I'm gonna go deep on v search and it'll eat up 80,000 tokens, go on another, go on another, just eat through tokens, and you know that's part of it.
Like at the end, it does it does hit a long task, and I think you only see that that expensive.
Yeah, I yeah, there's insatiable demand for tokens, and every improvement that comes kind of just makes our demand even higher.
It's kind of funny, right?
Like, if you have like a teammate and you ask them to do a task and they're like, should I save some effort and not think too hard about this task?
We're like, fuck no.
I mean, my favorite literary you can have four shots, right?
Like the original codex before the app, you why do one call?
Like give it four attempts, just just use all the tokens like that, right?
Try more directly.
Try again, try more.
It's like it's like the the meta index, right?
Is the thing that tracks like how long models are able to run.
I expect that we'll just see like log linear, if not log super linear growth.
We will see before the end of a year an agent that is capable of running for longer than 24 hours with like self-consistency the entire time.
I I would also poke at different domains having different desires, right?
Like at a transumer level, I'm getting slightly frustrated at 20 minutes per basic query.
Sure, you can optimize, you know, six, eight hours.
I don't see myself shooting off many one week agents, right?
Someone doing like okay, GPU kernel research or medical or biological, like you know, in in those domains, sure, shoot off a lot that take them out of up.
So, like, I think it will be somewhat domain specifically.
Because you also really need to train that in, right?
That's funny.
When it comes doing your taxes, right?
Like, that's taxing a month.
Yeah, okay.
Yeah, expect get it right.
I wonder if like a smaller school case.
That's sort of like uh speculative decoding is like your agent figuring out what you might be prompting it the next day at night and like pre-fetching.
Yeah, you can already do that.
Yeah, really branch branch prediction.
Oh, well, no, that well, that's that's too that's too low level, but I guess sorry, yeah, yeah, yeah.
Uh one question I gotta get it.
So like uh we actually did record a part with uh the meter folks uh was that right here.
Their chart is the human equivalent work, uh hours of work rather than how long the agents themselves are uh being autonomous and that and uh there's a huge difference, right?
Like human work, five hours, agent work, 30 minutes.
Like it's actually 30 minutes, not uh yeah, five hours, right?
Like, so they that that the chart that you see is them estimating what the human equivalent replacement is.
Um, I think the I think actually Infopic released a more recent chart that showed cloud code autonomy from their production traffic numbers, and that was 20 to 45 minutes.
That's roughly where we are.
So yeah, yeah, that's that's the sort of realistic thing.
I mean, I I do think like there's experimental setups where we can just like Ralph William might just prompt it to keep going with his soft, and obviously you can that can go arbitrarily long.
Fairly from my experience around, yeah, I guess 20 to 40 minutes seems right for when I'm using like codecs or cloud code.
But then like what I always try to just like if I want to spin up like a new there's a net new project, I'll I'll often start with replit and like it'll be unferred a baby anchor.
Yeah, like spin up like the their new, like from the v3 agent, like it'll spin up a web browser and like click around and discover new bugs and just keep churning.
Um I think like my longest was like over an hour that I have been churning.
I think before we see super long running, I think there's gonna be a bit of an efficiency hit.
So sure, you can take an hour and go down paths, but you also want you wanna be more efficient, you want to be smarter in your reasoning, right?
So I think that'll actually go down before we go back up.
Like you don't want to scale non-optimized systems just for the heck of it.
As much as I love saying use all the tokens, um, you know, they are expensive, like going from dense to reasoning models, that's an added cost, right?
You're paying for a lot of tokens, and it doesn't make sense to just scale stuff that's not optimized.
So there's there's always that little balance.
Yeah, but you know, yeah, I think you'll see both sides of it.
Yeah, so 2023 was super exciting.
I think if you were in SF, you were like, okay, uh, I know this is gonna be a huge world-changing moment, but it seemed like you know, no one had known yet, and maybe even before, was it 2022, maybe?
Yeah, yeah, it was a yeah, like Rune had this tweet where like everyone was in SF from like 2021 to 2023, yeah, like understood what it was like to be like RD.
Totally on.
I'm yeah, 2021.
That's when I made my first open AI account.
Yeah, it went um, it was crazy.
And I remember it was so funny because at the time SF had not been doing well.
So pretty much what it felt like was the concentration of founders in the city had wro had risen because um where my neighbors were used to doing a bunch of stuff, those people had all left.
So the only people that were still in the city were people that really wanted to build it was cheap tech.
It was yeah, it was also way cheaper.
I feel really bad anyone uh who is trying to get rent now.
But there was uh cello was they had a huge office.
So blockchain.
It like took over the the old Casper building.
Yeah, they had the showroom and they had the like the what would I think was like the back warehouse.
It was it it was a huge office, and it's right across an opening eye is in the or link.
Yeah, it was in the original arena.
I named the arena because of it.
Yeah, yeah.
And so it was really exciting because like Roboflow, I think uh I forgot there really five.
But yeah, Mintlify, uh Brev was there, you guys were there.
I remember that was actually it was there that you bought the AI.engineer domain.
Yeah.
I didn't know what I was gonna do in AI.
I thought I wanted to do something.
But it was kind of this it was a really fun moment where we were kind of all in this cello space and it um I don't know, it was it was a really cool community, especially being so early.
Yeah, and so it's then you got me early cruise access.
Oh yeah.
So there was a going period of time that both crews and Wimels were just free.
Yeah, always singing.
If you had an A, I mean they're they're sell back.
Sello is opened again.
Yeah, so nature Zoop Zooks is doing Sucks a robot taxi, yeah.
So totally uh eight.
Um but yeah, and so it's actually really cool that you guys have this studio so close to uh cello.
Yeah, this rock climbing gin right around the corner.
It's like um yeah.
So uh yeah, it's a it's an awesome vlog.
Well, yeah, just and you're a bit of a serious country, but I do think what uh one thing I try to do with the podcast is like bring like what is like to be in San Francisco to the rest of the world.
Yeah, and also just like maybe give uh El Tepa Takari uh Yeah.
My favorite talk was in the city.
Uh and uh yeah, stick and shrimp.
I know it's very good.
Yeah, and I guess what it's like to be in San Francisco, I think is just everyone seems to be super supportive.
Uh sometimes I feel like the city believes in you more than you do.
And even uh I don't know if you remember, but I remember posting my first blog post and I had met you on Twitter and you gave me like an hour of your time super randomly and you kind of coached me through uh writing content for developers and I was trying really hard not to come off salesy or plug myself, and so I kind of stripped all personality out of the blog post.
Yeah, and you you brought that out.
You're like people don't it's it's okay to talk about what you're doing, like you don't have to be weird about it.
And I remember just that I think that really helped me kind of figure out what our voice is and not shy away from it.
And so always really grateful for you.
Hey, you inject your voice into like everything now.
It's actually actually a huge advantage to be like very genuine about what you care about.
Yeah.
Yeah, like imagine like some re someone interested in DMC and like it's like, can you give me feedback on this blog post?
And it's pretty boring, and you're like, fine, like you know, he looks interesting, I'll just do a Zoom call.
And then you meet this guy, yeah, right?
He's so energetic.
Just be right there.
That's uh but and but I think people are trained to write a certain way in school, and yeah, they never totally see there's like a bro broader well.
Variety writing is thinking, and like everyone thinks differently.
So like you might as well just like write your way.
Cool.
Well, thank you for uh in indulging with us.
Uh really broad breaking discussion.
But I love like you guys are like sort of like the sort of young faces on NVIDIA with so much energy and but like also a lot of typical death, and I think uh people learn about for this session.
So thank you.
This is awesome.
Thank you guys, and thank you for everything that you've done and you could talk.
Yeah, NG, the podcast, all the abum.
And uh C O G T C forward to it.
Yeah.
Cool.
Thanks.
Awesome.
Thank you guys, thank you.