Why AI Can Write Code But Can’t Keep It Running (Yet) Transcript, AI Summary & Key Points

00:00 The world that the AI lives in is all digital and there's the actual world and like those things are not necessarily the same. >> When we started doing AI for coding, it was autocomplete. Nobody really knew to what extent you can actually build complex software using AI and then it basically broke through the expectations because we were able to model more and more complexity in the systems.

00:18 >> I just don't know if operations is constrained in the [music] same way. Where's even the data to learn how to like build a system that understands like all of the operations? My assumption is that we have all the data. The question I think is maybe about complexity and like to what extent can you take the things that you've described that are operational and convert them into something that an AI can reason about the dirty secret for the longest time of running like a distributed cluster with just human beings

00:45 managing it. Do you think we're at a spot where you could like use AI to run one of these services? I think that there's enough to be able to All right, Jad. thanks for joining me on the podcast. For those of you who don't know, Jaden is the CEO of Greer, but also we uh went to grad school in the same group. And when I first met you, I we we had a new student.

01:08 [laughter] >> I I walked into the office and there was a motorcycle helmet and a pack of Maros. [laughter] I'm like, "This guy must be unbelievably cool." So my question is is what happened to all your coolness? [laughter] Question number one, [laughter] >> maybe maybe maybe just for a context, maybe give give um a quick um uh overview of your background.

01:36 So just just for those that are listening, you know, dad is one of the best systems thinkers I know. I love talking to him. We're going to go kind of to a deep dive of like engineering from a systems perspective and how AI is impacting us. We're going to kind of especially when it comes to things like operations, but just to ground it maybe a bit about your background.

01:53 >> Sure. [snorts] So, uh, you met me at, uh, one of the peaks in my coolness. >> Yeah, you're very cool. [laughter] >> Motorcycle and Marorrow cigarettes. >> I remember thinking I'm like, "Wow, are cigarettes even legal?" [laughter] It was like so discordant. >> I remember Nick seeing me one day smoking outside of uh outside of Gates and looking at me and going like, [laughter] "G Nick Bun is our adviser.

02:19 Gates was the the building." Yeah, >> actually one of my favorite thing I don't do you remember Balaji Prabakar the professor of course >> my favorite thing about him is that he smoked I was like the [laughter] only professor I just thought it was like great >> just to be clear I don't smoke anymore >> no too old too middle-aged for that now [laughter] um but let's see so you know I went to Stanford with uh with you you were always the guy I had to look into to look to look up to and be like hey >> this is the guy I

02:49 want to be when I grow up >> I was I think it was I it was a I started started four years before you did. So I was ahead >> and so you were four years ahead of me. You graduated a while ahead of me. You went and started Nicer. I stayed and >> stuck around worked more on OpenFlow. Helped uh build a bunch of stuff on the hardware side from a security perspective.

03:08 Actually, >> one of the one of the, you know, most fun projects that I worked on there was uh I built the first set of uh Shaw 3 implementations that people were considering using and um I built them in hardware [snorts] to kind of show whether they work or not and whatever for one of my for uh one of my projects for my thesis. That was fun. >> Yeah.

03:30 Yeah. Cool. [clears throat] And then I decided after a while that I actually don't like networking because it's just about moving bits from one place and putting them in another place. >> How hard can that be? [laughter] >> I think that would have to either be philosophically aligned that it's deep or not. Most people assume it's not. I think it's very deep.

03:49 [laughter] >> I mean there's something interesting about it, but ultimately to me that's like it's not >> Yeah. It doesn't do it for you. >> Yeah. Um and then uh I actually went and did a posttock at MIT with Nikolai Zelda who's also amazingly deep and super smart. >> Um >> he's phenomenal. >> I did some application security work at the time. Started a company um based off some of the research I had worked on with Ramsh Chandra over there who was a PhD student at the time and ultimately technology was too early.

04:23 >> What was it? So what we did was um well you know ultimately a lot of the work was actually really Romeshes and his PhD was on well hey if you get attacked and you get exploited can you recover from that by rolling back the state of the whole world that's related to that application and then roll it over on top of a patched version of the application to remove the exploit.

04:47 >> Yeah. And so we started doing that in the real world and we're like ah that sounds really hard. [laughter] It's actually kind of relevant to what we're talking [snorts] about which is like to what extent can you actually bound the state for like operational systems which I would argue is very hard but >> yeah I we almost got it working [laughter] >> almost I feel like for all longtail disciplines this is why people get in like the fly track for smart people is like you see a bit of progress but you're as toing but

05:20 anyways I digress. Let's keep going. Let's keep going with your uh >> So then I joined App Dynamics. I was there for about three and a half years. They went from about 200 people to 2,000 people. Got acquired by Cisco. I was a Cisco employee for one day. And I said, and then you're like, Jad, >> I know. I [laughter] love you. Come work with me. >> Come do this VC thing with me.

05:40 >> I'm like, what the >> Yeah. >> What What is VC? What? Like I'm an architect. I do like engineering stuff. What do you He's like, great. >> Yeah. >> The less you know, the better. >> Yeah. Literally what you said. >> It's true. And then the first day I started here, you said slow down. [laughter] >> It's actually people that move from operate like from like actually an operating world to VC.

06:01 It requires a lot of reworking the brain. >> Yeah. >> Yeah. Just like going from a PhD to operating by the way. Like you have to like undo the brain damage. [laughter] >> This is my experience. >> This is what real world is like. >> I know. >> And then um I stayed here for a couple of years. Did a bunch of work with you. It was a really fun time. Worked with a lot of great people like Jennifer, Mark, and Ben.

06:24 >> Satish. >> Satish. Um Matt was uh here when I first uh what almost towards the end when I was leaving. >> Yeah. >> And then um and then joined Apply as their chief product officer. Um helped them build uh their engineering product teams. Kind of scaled from about engineer like 15 people to over 90 across product engineering infosac and design. And after about 3 years, um, I kind of had a head for each one of these pieces that I owned and I kind of was sitting around just coordinating and doing meetings all day and

06:57 wanted to go do something else. So left and then um started talking to a lot of engineering leaders about the problems that they're facing and what they're trying to do. Everybody at the time, this was in 23, everybody was saying, "Hey, we're trying to reduce spend." And ask them, "What are you spending money on?" Everybody would say observability. And you know well first actually it was infrastructure and then observability.

07:20 And so I started looking at that problem trying to understand like what is why do people think it's too expensive? [clears throat] Turns out there's a few things. First people were spending somewhere between 5 and 20% of what uh they spend on infrastructure on observability. But ultimately what they were really trying to spend it on was reliability.

07:38 They wanted to buy reliability and um >> of spend like predictability of spend >> reliability of their infrastructure and applications. >> I see. I see. I see. >> Right. And so they were putting money into observability as a proxy for reliability. And for the longest time, the more money you spend, >> the more reliability you got out. But as architectures changed and turn into microservices that are much smaller, that are much more vertically integrated.

08:04 The volume of data that's being collected at every layer expanded. there was much deeper market penetration for observability and kind of the uh volume of data that you're collecting and the cost of observability was actually starting to grow faster than the units of function that you're outputting. And um and the marginal rate of or like the marginal increase in reliability for every additional dollar you were spending on observability was just flattening out.

08:32 [gasps] Um, and so that's why people thought it was expensive, but people were still willing to pay 5 to 20% on reliability. And so the question was like, can we actually solve that problem and bring back the um that increase in reliability, that marginal increase in reliability, >> not necessarily by spending more money on observability, but some other way.

08:54 And so the thinking was like, okay, great. If we can figure out what is the signal, what is the noise, can we actually turn the observability data that you're collecting that it's mostly fluff and mostly insurance at this point and do that kind of extraction of information upstream, do whatever you need upstream and be able to ultimately massively reduce the cost of observability downstream.

09:15 And if you can extract that signal in real time upstream then you can one reduce the cost but two you can now feed it into other systems and use it for driving reliability operations afterwards [clears throat] >> and the distance at the time there was something that was called AI coming coming into the picture and we're like well great if we are in the data stream and we're extracting the signal then we are in a strategic position to feed that to whatever is going to come in the future because that's going to be

09:46 necessary. >> Yeah. >> And so that's what we built. And um so Greer helps companies, you know, we started with helping them reduce their spend on telemetry massively. And so you put us in the data stream, plug and play, pass the data through, reduce your cost by 90%. And um and now we're building out that second layer where we're feeding those signals into the AI layer that's now going and troubleshooting and figuring out when things are going going wrong and fixing them.

10:14 I'd love to just talk through kind of your experience on how AI is changing dev generally which is like a lot of people get asked this question a lot of people answer but I just feel like kind of like the engineers engineer like you've been an architect you've run product you run engineering teams like you know like actually how big is your company now how many people >> we're t >> yeah so you know and you've got a I mean you're building a lot of software with a relatively efficient team so >> maybe just a highle

10:42 sketch on like you Well, if you're a senior engineer like you, you know, that has architectural experience, like like what sort of gains do you think people are seeing? Maybe talk through like how you use AI, what gains you're actually seeing, what is the stack, what is your process, how you think about it. >> So maybe I can give you like a historical view of like what's been happening in terms of like running a team and so on.

11:07 >> And so what you know when we first started AR was still kind of autocomplete, >> right? I remember like at the time we were using uh co-pilot and uh intelligj and we're like oh my god look it autocomp completed these tests >> uh we're like yeah but it's just autocomplete and very rapidly things started evolving and what what I've noticed is that really good engineers are able to somehow put it into their workflows in a way where they the AI starts training them.

11:41 Oh, interesting. >> In terms of like how to talk to it, how to get the best out of it. >> And what I do, as an aside, do you think that's model specific? Like is it do you learn how to use AI, you know, writ large or do you get efficient at a given model? >> I think there's like some nuances that different models bring. Um, I'm sure looking at your tweets you've [laughter] um there's like nuances, but there is a a way to work with AI and like I think that there's also a way to work with different harnesses and I think

12:17 like you know we tried for example using open code or like different harnesses or like using codecs versus claw like claw code like different ones have different ways of managing context or thinking through problems or when to spin up the right sub agents and what to tell it like the kind of prompts like for example >> uh with cloud code I find that if I don't tell it don't spin up >> you know 20 fable agents to do basic research it'll go and spin up 20 fable agents to do research and burn all your tokens and so like a

12:48 lot of these things kind of you start learning about the harness and the and the models and like how they interact with each other and then how they can interact with the rest of the world but I digress like I'm just going in my mind like this is kind of what I started noticing is that like some people were adapting and learning and kind of doing it that way um and others weren't as adaptive.

13:13 >> Yeah. >> And then as we started hiring more people into the team that disparity between you know okay engineers and really good engineers just multiply 10x. >> Was it can you predict who will who could you predict who would be good? Was it like good engineers got better and bad engineers didn't or was it it really? Okay. Yeah. So, like if if you're a good engineer, you're just a better good engineer.

13:36 It's not like there's some like bad engineers all of a sudden became good engineers. >> No. >> Okay. >> No. No. It it always ends up being like you have to really understand ultimately the the setup because the more complexity that your brain can handle, the more you can break down the tasks and kind of assign them. I think this might actually be uh like non- consensus opinion that like uh I mean I I do hear it sometimes but the idea that like the better engineers like are much better [laughter] those are are less and

14:07 you you found this to be pretty universally true [sighs and gasps] I mean look my team is still small we haven't done it across 100 people >> yeah [clears throat] >> in my experience for the things that we do that is the case because we build distributed systems and distributed algorithms and like your brain has to handle the complexity because like >> AI doesn't do it, >> right?

14:27 Yeah. Yeah. Yeah. >> And so for the problems that we solve, [clears throat and cough] none of the models can actually think through the things that we do. >> Like there's a lot of for example, we have a query translation layer. We were just talking about query optimization. >> Yeah. >> And so you have to really understand the semantics of the query and how to get to the semantics of the query and then explain it to the model >> and then have the model go and build the translation layer, let's say.

14:52 Yeah. So for for for the people that are listening that that may not be so deep in computer science, what do you mean by semantics? [laughter] >> Uh so semantics are really the meanings of things. Like the semantic of a dog of the word dog is the animal that actually walks on four legs and barks. >> And the syntax is the fact that it's just three a threeletter word >> and it has three different letters.

15:19 the shape of the word is different and that shape can be different between different words. [clears throat] Um, but different words can actually also mean the same thing. And so they might have the same semantics, but they might have a different syntax. So a dog and a puppy might be slightly different semantics. [clears throat] They refer to the same, >> but they're still called a dog.

15:43 So the way that and just for for like computer science, the way that I describe it very simply is syntax, you know, is like the the characters you type and semantics is what the machine actually does, right? And so it's like like the actual runtime properties of the the system. >> So you did computer science, I did electrical engineering, [laughter] >> you became a computer scientist and I became a VC.

16:08 So [laughter] there we go. >> That's the problem. I see. Yeah. >> Anyways, I just wanted for people that are that are listening, I just want to understand like there's the actual syntax which is writing the code, you know. Um, and then when you're talking about syntax, you're just talking about the code itself. Then when you talk about semantics, you're literally talking about you have a machine that interprets that code whether it's compiled or not.

16:31 And that machine does something. The semantics is what the machine actually does. And these things are, you know, clearly very different, right? Like so for example, some semantics you can't even derive by looking at the syntax, right? you don't know what the machine is going to do because whatever you kind of have to test it you have to test it you know like and it can go all sorts of different ways when you test it and like a very classic version of this is is um uh is SQL for example like you can write you know some

16:56 SQL query but like how it actually performs how long it performs depends entirely on like say what the schema looks like or what the data looks like in the database etc but that's I would not call that semantics though I would call that performance characteristics. [clears throat] >> Oh, sure. Yeah. I think that would be like >> like I like I would say like a the semantics part of that would be like um depending on like what the shape of data is or like what the configuration you might end up with different semantic

17:26 because like you said you want null to be treated as false. Um and you won't know that unless like you also know the configuration of the system and and and so on. Okay. Okay. Okay. So, for the purpose of this conversation, I think we should just say that there's like the actual code you write and there's how the system behaves. And we'll call the second one semantics just so we can have a conversation.

17:45 [laughter] >> There's a >> Well, what do you want to call them? >> We need to dive into the semantics of this and make sure they're clear. >> That's [laughter] right. >> Because there's a lot of >> I know. Okay. Okay. We'll keep it high level. Okay. So, so sorry. So, let's go back. So, um, you know, actually before we started the podcast, you and I were just shooting the [ __ ] and you were talking about like how you can do products, projects now that take a week where in the past would take a team and months.

18:14 Can you maybe walk us through a little bit about like how you personally >> kind of well do projects and then kind of the gains that you see? [snorts] >> So like a little bit about like your stack and your process. So, I use um cloud code. Um I'm going to say something that's actually pretty funny, but I use uh Windows. >> What [laughter] was that created by this company Microsoft [laughter] and I use WSL?

18:45 So, I use Linux for everything like development and so on. There's no like X or UI for it. I just use the command prompt. Um, >> how do you use Windows? >> Because I just never really wanted to learn the keyboard shortcuts. >> Are you a gamer? [laughter] It's like you tab tab over to Counter Strike or whatever. [laughter] Like why else? I'm not hating on Windows.

19:08 I mean, you know, like I used it in the '9s. >> That's really like the only reason is I never wanted to learn the the shortcuts. The I I hate the onebutton mouse. I hate the one button stuff. I hate the number of modifiers you have to type for shortcuts and stuff like I I just I just never I'm like why why why >> why waste time. [laughter] >> Okay. >> So like yeah use Windows.

19:32 Yeah. >> Um and so WSL I use cloud code. >> What is WSL? >> That's Windows Subsystem for Linux. >> So it's uh they actually run the Linux uh kernel within Windows in a hypervisor and provide you access to it. It's great. They've done a great job. >> Great. Windows WSL >> and Cloud Code. >> Cloud Code >> and uh like my latest the latest thing that I do is I use Opus 5.5.

19:59 I tell it >> which is a really good model. >> It's an amazing model. >> Really good. Yeah. >> And it's I've shipped in the last two weeks these massive 20 PR things that I just dump on my team to go and and review and they're like [laughter] I'm like don't worry about it. >> Horrible. It's like is like the new reporting structure like the people below review the PRs for the CEO.

20:21 Is that is that where we become [laughter] like >> No, everybody's like writing the same PRs too. >> Yeah. I I review my people PRs and like everybody's like just just PRs flying around in PR. >> Yes. And but like the thing that has changed a lot in the review cycle is that like we started much more focused on to focus much more on like the the tests and like doing things like golden expectations and like are they shifting in different ways and so on.

20:53 But um Opus 5 and like the the trick that I've recently learned about is that you instead of using sub agents, you basically tell Claude um to spin up command line sub aents because those maintain their persistent session over time. >> And so you can have it spin up these agents separately. You can actually run them from other sessions and and so on.

21:15 And uh that allows you to be much more efficient from a context perspective. >> How many agents do you run? How many >> how many >> how many what agents >> agents um typically I work on like three or four session windows at the same time because I found that more than that I just start losing context um between the different things because like they they start finishing faster and then I I then they wait.

21:40 >> And do they interact via the do they interact via the PR? Is that how it works? >> The sub agents like together? No, they interact directly through the command line. >> Oh, wow. Okay. >> Yeah. So, like I asked like the the main like the main thing that does the implementation to have a tech lead reviewer and like they basically ask it questions. Does this make sense?

21:59 Does this not make sense? Review my PR, etc. And so they iterate on these PRs together. Um, [snorts] and then they have a PR doc that basically they work on with a design doc and everything in it. Like the first thing that I do is I ask that agent to do it. Ask the tech leads to review the PR doc. I go through the the PRD. I go through the design doc.

22:20 Make sure it makes sense. Go through each piece in like detail. Make sure that all the algorithms work and like all the pieces work, the division of labor works, the division between PRs and their relationships work. And then I tell it just go start start doing it. Figure out if there's like paralization that you can do and then spin up other agents to actually go do them.

22:39 M >> um but ultimately you have to kind of put it all back in one stack and be able to go through it as a human one by one because otherwise it just gets messy and bushy and you can't really tackle it, you know? >> So how much time so can you give a sense of like a project you do now versus I don't know six years ago like time differences. So the latest project that I shipped was something >> by the way this is like to be clear this is production code.

23:10 This is demo mockups whatever. So like the statement you're about to make is the gains that you get for like shipping production. >> Yeah. So I'll give you the example of the last there there's the last PR that I worked on or like the last set of PRs was what we call auto auto promotion in a data lake. >> Yeah. So in a data lake you store all the data and um typically the data is um column oriented parquet files and iceberg and so on.

23:37 Um and when you get data from um logs they're semistructured they're arbitrary columns and so on. In order for it to to have the best layout of the data and to be able to query it efficiently and fast, you kind of the ideal situation is that every field in the log event becomes a column that you can filter on independently of the rest of the rows. >> And so we call that autopromotion or shredding.

24:03 >> Um, and doing something like that is incredibly complicated, right? like you have to go and figure out how does the the metadata system work, how do the commits of the metadata system work, how do you coordinate across different uh writers and so on and um that took maybe about 10 days. In previous days that would have taken maybe about 6 months.

24:30 >> Wow. >> For a single person. >> Wow. >> And was this just you or >> that was just me insane. Yeah. Sort of like factor of 10, factor of 20 improvement. >> Yeah. >> For production code for new products. >> Yes. >> You know, one, you know, one one um question I always have is like I can very much understand, you know, the these models are trained on all the code ever written, let's call it, right?

24:59 So, I can understand why they're able to generate code, which is the syntax piece. Um but then like for you know you're doing core infrastructure let's say you're doing distributed algorithms like like like the properties like how it handles itself when it's running is very very different like often you just can't even look at the the syntax to know what's going to happen right like a lot let's say let's say you're I don't know you're you're let's say you're writing some distributed system and then like the way that it

25:28 like it will perform the the availability um the correctness even is is going to be dependent on what the workload looks like you know like like what are the access patterns what is the traffic >> but are you talking about like the performance aspects of it though or are you talking about like this the >> well I'm think like availability like you may have to lock things so it's not available for a period of time I'm talking about correctness like maybe you've decided to do eventual consistency and so like if you do a

25:55 read it's not going to be consistent because there isn't a lock right >> um and then performance like you know maybe you cache maybe you don't c these are all kind of trade-offs in the design space, right? >> And like unless you know like the access patterns, you won't know how it's going to perform in these things, right? And so I just think to myself, okay, I've got an AI, I can understand how it will write, you know, like like code for a generic distributed data store, for example.

26:21 What I don't understand is how it will know to like make the trade-offs needed because it doesn't even know what the access patterns are. And it's not like these things can like like where's even the data to learn how to like build a system that understands like all of the operations right like today we rely on human beings to like manage clusters and to like you know optimize databases etc.

26:46 And so can you maybe talk about how you think of like I get the writing the code but like you have these deep operational systems like how do you apply AI to that problem and like a have we made any progress at all or is this still an outstanding problem? The question I think is maybe about complexity and like to what extent can you take the things that you've described that are operational and like that happen in production and convert them into something that an AI can reason about >> and uh exactly like how can an

27:12 AI reason about this system >> well a first off like to to what extent do you see that it writing the code that it can actually predict what's going to happen like how good of a job does it do at that and the second one is like Okay. Now, how do we make it reason about that? So, let's ask the first one like like if I if I use your method, the the JAD method of writing code and I built some big distributed system like how often does it actually get it from performance correctness?

27:40 >> How how often is it like actually operationally like pretty sensible? And by the way, I I work on this this distributed Gaussian splat render and I would say it's [ __ ] Like it's horrible. like because like you know you've got millions of Gaussian splats that you've got to page in and out and stream and like it just totally depends on like the running system and so like none of the models do a good job.

27:57 >> So in our world what I found is that we've defined basically really well what performance benchmarks look like and how the system should run >> and uh what the constraints of actually building within like our stream processor look like to get the right performance. But every once in a while it'll do like my la latest project has been all about query optimization and query performance.

28:23 And so I was like hey >> the classic example of something that's >> I was like we are basically generating code that corresponds to SQL and we want to run it in a pipeline. >> Yeah. >> And we're like okay well you're generating [ __ ] code fix it. >> Yeah. And uh it started basically going through and what it didn't realize was that it wasn't actually doing initially any optimization on the code.

28:46 It's like oh it's fine. It looks like what's the problem? And I said well look there's these uh these conjuncts over here you're doing and and like this particular term is repeated and it should have been taken away and then it's like ah okay let me figure this out and so on. And then like it basically went on a thread to try to figure out what are all like it built a bunch of hypotheses around like what are all the things that it can go and optimize to make this better.

29:12 >> Yeah. >> Um but ultimately it was able because it was able to ground that in performance benchmarks and I was like look here's here's a data set of the queries and like what the users are asking and so on. Go work on it. >> Um and so it can basically do a relatively good job that way. But when it comes to saying something around like how much memory should I give this system?

29:39 >> Yeah. >> It has no idea and that has to come from actual experiments that run in production. >> Yeah. >> And so what I think actually these systems seem to be doing well at this point in terms of operational stuff is that it knows that hey I probably won't be able to answer this question. Let me run an experiment somehow. and come up with that information.

30:02 But of course when it comes to operations like production systems don't behave the way that you expect and ultimately that's why incidents happen. >> So have you found or are you doing any of like the AIS stuff where you actually have a running system that has logs and you feed it to an AI and that kind of helps you like tweak it? Yes. And how how well does that work?

30:22 >> So we're just >> because anecdotally uh my experience with other people doing this has not been that great. Like the reports I've been getting anecdotally have not been great. So, a couple of things. The first one is we're doing that for both operational pieces and uh and for security, but the way that we're tackling it is not the way that other people are doing it from an AISR perspective.

30:41 A lot of those systems basically start from a lagging indicator like, hey, this metric spiked or like this alert happened. Go back through a very complex system and try to determine why that happened, >> right? >> Like, >> yeah, sure. It's, you know, that's why those don't work because like backtracking into a root cause is actually very problematic.

31:03 And so instead, what we're doing is doing it the other way around, which is can we actually start by figuring out if there's something different in your environment that has changed? Like can we do a large scale clustering, a large scale kind of figure out a needle in a haststack, something that is different? And then so you're assuming that it was working fine before and now it's not.

31:25 So something happened, but it could be just like a change in like normal traffic patterns, right? >> That's fine too. That's fine too. And so the the the thing that we're looking at is like when things go wrong, usually the system will tell you something went wrong somehow. Yeah. Right. Um if it's working fine and scaling as normal, nothing will happen.

31:48 But sometimes you'll see some like message which says error connections ran out. >> Yeah. >> Or memory exceeded or resource buffer is emptied or like some kind of indication like some kind of really early warning like something happened and that's the one message that we want to capture and then go from there and see >> is that an actual problem that we need to investigate and deal with.

32:10 So how does it do? So far it's doing pretty well. I mean we're running it internally in our systems. We run it uh we ran it in uh in a couple of our customers environments and in the cases where there was a couple of cases where we ignored it. >> Yeah. >> And those turned into production incidents. >> Really? Oh wow. >> Yeah. It's kind of just just from an intuitive perspective like I can understand how these models understand syntax because again like I said they're trained on all of the co code ever written and it

32:39 just turns out that covers a lot of useful cases. It's hard for me to actually even think of a bound on the actual operations of software, right? I mean, it feel like, you know, this is a ridic it's a ridiculous statement, but it's actually kind of true. It's like every time you add a bit to a program, the state space of like what it can do operationally increases by a factor of two.

33:00 So, you very quickly get to like, [laughter] you know, like more more, you know, states than atoms in the universe. And arguably, this is why This is why like product approaches to operations have never really worked, right? Like the chaos engineering stuff never really worked and we've always kind of relied, you know, actually the dirty secret for the longest time of running like a distributed cluster was just human beings managing it.

33:26 And so I'm just wondering and then you know and then in order to like even train an AI to do it, you have to have like enough data and enough configurations. And so like I'm just kind of curious like do you think we're at a at a a spot where you could like use AI to run a distributed cluster or run one of these services or do you think that's still on the come?

33:47 I think that [snorts] there's enough to be able to have it start that loop, right? Like I mean ultimately like I think that I agree with you like there's infinite space there's a lot of different things but ultimately humans also operate on runbooks and like sure there's a lot of things that happen in the um in production that you ultimately need some human to go and like hey this is too complex there's like some state that's hidden that I know of because I talked to this other guy like at the water fountain the other

34:20 day and he told me about this thing he deployed like there's a lot of stuff that potentially isn't visible in the state of the system. Yeah, >> but I'll argue that like the goal of what we need to be able to do is to like get to have enough telemetry and enough understanding to like make everything really visible from a hey here's what the system looks like, here's how it's working, here's like whatever.

34:44 And then I think ultimately most of the issues that end up really showing up in operations assuming you have the system previously running okay is due to some kind of change. And if you're able to just focus on the goal of figuring out what the change was and like what and kind of sorry can I just ask a stupid question on this which is I I I never think of these generative models that good at predicting stuff and it sounds like you're talking a lot about predictive modeling like you predict a change or you baseline and

35:16 you detect an anomaly. Like am I thinking about this wrong? >> I no you're not. Okay. Because the goal is not actually having the uh agent try to predict but have it uh get a signal that something changed. >> Oh deterministically that's something I see. I see. >> Yeah. >> Wait. So so there there sorry. So so there is a separate model that does the predictive stuff and then that will alert the AI.

35:43 >> Yes. >> Okay. Yes. >> So yeah. So like what we do like what we do is we basically you give us your pabyte of logs per day. >> Yeah. and we scan all of them and we we like every time we see a new kind of pattern emerge in your logs, a new thing that we haven't seen before, we lift it up to the AI. >> I see. So that's that's a predictive model that actually does like does what predictor models are good at which is determining out of norm things or whatever whatever and it's like the that goes to the AI and the AI kind

36:08 of uses that to figure out what to do. >> Exactly. So the AI basically can then reason and drive that >> I see >> that down. So it can say, "Oh, this message, let me actually look at the code and figure out where it comes from and understand what the code does. Let me see if it's actually happened before. Let me see." So it basically runs all these hypotheses and heristics to try to understand does it make sense?

36:28 Does it need meaningful in some kind of way and then it can determine what to do with it. So how far away do you think we are for let's just go back to query optimization, not like logs or something like so. How far away do you think we are from having an AI basically do the job that a human used to do when it comes to query optimization? I think very close.

36:47 >> Really? >> Yes. Wow. >> I think uh >> running a distributed cluster. >> I just think of like okay you've got like James Cowing PhD from MIT who runs a distributed cluster and like if the cluster's down like the company doesn't work and then if it's up it's running. I don't know like are we at the point where you would say that an AI can do that? >> No, not yet.

37:08 >> Okay. I think it's I think I mean I think look look very very simply when we started doing like AI for coding it was autocomplete. >> Yeah. >> And nobody really knew to what extent you can actually build complex software using AI. >> Yeah. >> And like when we were looking at it we're like ah like AI is never going to be able to build >> X or Y or Z or whatever.

37:30 And then it basically broke through the expectations because we were able to model like more and more complexity in the systems. I just I just feel like like I'm a very simple thinker on this stuff and it to me these models are distribution in distribution out and it turns out like you can actually train on all the code ever created and it's actually like that manifold is pretty manageable and I just don't know if operations h is constrained in the same way if that feels to me more like kind of like the universe

37:58 [laughter] like heavy tailed and exceptionbased and power laws and all the tough stuff for these models to learn. But I don't I don't know. I mean you're the one that's in the field so maybe that's not the case. Maybe they're manageable, but >> but I I I will tell you the history of systems as you know is is you always want to build a smart algorithm that does the efficient thing.

38:17 The smart virtual memory management, the smart congestion control, the smart routing. And what do we do instead? Every single time [laughter] >> we build the >> We overprovision. No, we overprovision every time. No, literally we every single time. Like like remember like swap memory where you like swap to disk? Dude, if you did that, you were [ __ ] Like that's it.

38:36 Your computer was dead. So like we literally overprovision RAM and like routing you know how many like adaptive routing algorithms there are like we will respond to traffic but you have oscillations like [ __ ] it we overprovision we just use you know so I just feel like the history of systems is you don't predict stuff you overprovision because you can't and I feel like I mean you know so I feel like in the operation side so either we build this you know we we'll build the systems up front that are overprovisioned so

39:02 you don't have to like do kind of complex operational stuff or maybe AI can do the way that a human being, but I think it's very different than code because you're dealing with like exceptions, there's not a lot of data, etc. I just feel like a different domain, but I don't know. I'm a I'm a VC. I'm just like, >> you know, I I I I kind of like the reason that I'm more uh optimistic here is because I live in the world where like what we do is actually deal with the observable world.

39:28 Like >> I was going to say you're the literal expert on this. So, it's silly for me to even like ask you these questions. >> No, not at all. Because like in in my world, we We h like my assumption is that we have all the data is that like we know exactly when like a new thing happens or like an exception emerges or like a workload changes or like whatever.

39:50 Like in my world, all of that has already been collected and streamed through and like >> but but the model that is not trained on a grounded source for that stuff like that. >> No, you're not. >> That would be really hard, >> right? But the but the story here isn't like, "Hey, Mr. Claude, here's a bunch of logs. Is something going on?" Right? And it has to be much more uh able to run experiments and kind of dive in and see like what are the different shapes of possible causal chains that actually fit in the

40:24 environment that exists. >> I see. So yeah, this is what you said before now. Now I see what you're saying. You're saying that like >> the universe is heavy tailed. The the number of bad things that can happen is is probably ex an exception thing, but like the set of steps used to debug it is a manageable thing. That's what you can you can teach one of these models and that's what you have and therefore if you give it the right signals it can go through a number of steps to actually constrain this problem.

40:48 >> Exactly. >> Interesting. >> Exactly. So it really is taking the role of the human because remember remember the whole remember the whole chaos engineering wave where like >> and and the reason it didn't go anywhere is like you're very so quickly like trying to like you know reduce like a state space that was atoms in the universe like so like super quickly like and actually make that manageable and it was like the human being that would go through a steps that would do that on the actual debugging and [snorts] this

41:13 is what you're saying the AI would do. >> Yes. Right. So the goal isn't to like have the model just be like this is what happens. It's hey run through this workflow to determine what happened. And like we're trying that also on the security side for example where like hey in security you have a something different changed in your like statistics like some behavioral difference in terms of what entities exist the cardalities of various things that exist and so that becomes a perturbation in the >> uh whatever the

41:45 universe and so we take that signal and then we feed it into an AI and we're like hey AI >> come up with a bunch of possible causal chains that exist and then go through and verify if any of those fit the pattern of the data that we actually see. And so that becomes a much more like hey here's a here's like a uh I don't know like fit this square into find the peg or like find the hole that fits this peg.

42:15 And that seems even harder to me cuz the operations one the argument is >> the system can get in states in into broken [clears throat] states that are kind of like very rare and there's a lot of those rare states >> and so it's just not something that you can you know train a model on just because there's so many rare states it can get into that are that are bad and so maybe you can reduce that by doing the steps.

42:39 The security one seems even harder because this isn't just like the natural order of rare states. You actually have an adversary that is >> trying to like intentionally find rare states and intentionally find stuff that you you you can't do. And so it may be the case that like code becomes solved. It takes a little while, you know, longer to do operations and it take a lot longer.

43:01 >> Yeah. But like also like think about I mean I don't know to me like the whole world is sort of quantized to a certain extent. the plank length. [laughter] Is that what the physicists say? Yeah, I don't know. I'm not a physicist. Constant. >> I [laughter] think there's like actually I I literally don't know what I'm talking about right now, but I literally actually think that like like like spacetime is discretized into a blank length.

43:21 But but yes I I mean I agree like to a certain extent what you just said but for operations and systems in general like there is a discretization of like the possible states and the path through them that are possible. >> Yeah. >> Right. >> I just think when you're dealing with >> there might be there might be a ton of them. >> There might be a ton of them.

43:44 I'm going to tell you remember Sundar who um was in our in our grad school group. >> Yeah. in their ear. >> Yeah. Yeah. Yeah. Yeah. So, I remember he came to me once and he and he he was he was he was looking very sad. He's like, "Martin, I just had this really sad realization >> that, you know, if you take any MP3, remember MP3 is like the [laughter] music for us."

44:04 >> Yes. He said he said MP3s have a finite, let's say, you know, four minutes. It's like four megabytes, you know, and that means there's a finite number of songs in the universe >> that can no because you know because it's it's all made of it's all made of bits, right? And it's four megabytes and that means you can you could just enumerate every one of those concepts.

44:24 So so I'm so sad in this notion of like the the finitness of music, right? Literally, this is like we were walking on the dish when he had this conversation and I'm like, "Motherfucker, dude, you know, [laughter] you know how quickly you've like overcome like the atoms in the universe. You haven't even gotten to a K, right? [laughter] It's like it's like two to the two to the two to the two."

44:43 Like, what are you talking about here, man? Be happy. >> Two to the two to the two four megabytes. >> FOUR MEGA YOU LIKE THIS IS so much state. Like, be happy. That's that's effectively infinite, right? Like what are we talking about here? [laughter] So I I do think at some level like we do underestimate like the number of states that like you know systems can get into and like maybe I will tell you from the perspective of a VC that looks at these companies the ones that are trying to tackle operations are having a

45:12 tough go at it. It is a it seems much much tougher than those that are tackling the code because I think it's a much tougher problem. I 100% I agree. It is a much tougher problem and and it's not just because like the systems themselves are you know there's like external loads that are random that kind of h happened to you or that uh you know somebody just shipped a new thing and rolled it out into production didn't tell anybody or like >> uh there's some state that actually happens when like this thing happens and

45:42 this thing happens and then you end up in a deadlock. Um it's also because it involves humans a lot more and not just like one human who's the coder. It actually also involves >> the pools of knowledge are distributed too. >> Exactly. and the interactions between like what it takes to like >> connect the pieces and put them together like >> but ultimately I mean I think about like where AI is and like what are the possible futures and like the the depth and the forwardlooking capacity of all the things like I think

46:11 code is just to me it's a microcosm of like may maybe I'm just too fatalistic maybe I'm just like too pessimistic about how the world works and and I think of everything is like following some kind of probabilistic path. Um, and so to me like if we can make it work in code, we can continue expanding that level of complexity bit by bit and kind of adding layers and layers and layers until we can handle those problems of >> a lot of people.

46:41 It's not fatalistic. It's the opposite. That's very optimistic is what people would say. For for me, I don't think the universe gives up its secrets that easily, but like we'll see. I mean like this is kind of the big question and and I would say to as yet it's not an answered question. >> No. Yeah. No. But >> I don't know a lot of things that um as humans we think of as being random or >> impossible to tackle.

47:08 Yes. And that's what we saw with with coding like you know initially like people thought like the the possible world of all these things that you can implement in code are kind of limited and you can't really get a system to potentially understand the full complexity of a human brain that's required. But I think over time it'll all It's so funny that I'm actually telling you this and you're like the VC who's looking into the future and being like no well I mean I I'll tell you my view.

47:38 My view is like the that this the study of these models are you train on data and um either that I that you know the let's try to say so you you train on data and that represents a distribution that distribution is useful or not and so like when you [clears throat] train on language the fact that it's useful that to me says that humans are so remarkable because they've looked at the universe which is this very highdimensional system and they've reduced it to a manifold that's actually useful for a model, right?

48:10 >> Yes. >> Right. But that's different than the universe. Like that's there's a reduction from this high dimensionality to the language is much less dimensional. >> And I think that code is a reduction of all the operational states of a of a machine into like the instructions that make it run. And so to me, there's still a disconnect between like the instructions or the the description and the actual like the actual running of the universe and actually running the system.

48:31 I think we can make gains toward it. But I what I will say is like you're kind of converging on simulation. Like there are things that are computationally irreducible. Like you kind of have to run it to get the answer. You can't just like >> you know predict it. And so like maybe we just are at an event horizon where like >> like models will solve anything that you don't have to actually simulate to get the answer from.

48:53 I but why why why do you have to like just think about it in terms of hey it'll have to answer the question like now as opposed to being able to simulate and and sorry being able to plan its own uh execution to gain the information that it needs in order to be able to >> No maybe that's where we're going. Yeah. Like >> that's kind of what I'm fair enough.

49:18 >> Yeah. And and I think like there's a lower dimensional there's lower dimensional space which is like these steps that you would take that we can reduce a whole bunch with AI and then whatever steps that you need that that can't do it can create the code to do or can simulate. >> And the point that you're making though is that like hey look the model can't actually go and figure out the additional dimensions of the world that we need to go and find to like model the universe into its world.

49:42 Like that's probably I mean that's probably true. But can it actually simulate a scientist sufficiently to be able to get to that point? Like maybe, >> maybe. Cool. Can I ask you a question? How much of a of uh the VC work do you think is something that an AI can drive? So my model for VC is it's very relationship driven. It's I think so. Okay. I just I just have a belief that markets are actually pretty efficient, you know, and so it's not about like finding the company that nobody else finds.

50:23 I think people all find like the right company. I think they're efficient. And then I also have the model of VC that [gasps] um, you know, tech tech is positive and I'm kind of optimistic about all tech and I don't worry about TAM. And so I I I think the key is access. It's like raising money. It's access. It's ability to win deals and there's ability to add value after that happens.

50:48 And a lot of that is like stuff that that startups just can't naturally do like you know recruit you know big execs or help with that. And so the way that I think about venture you kind of have to be a human being to do it's like you know um because because that's the actual job. So for that I don't think like an AI if you gave an AI simulated universe where it could like you know like whatever reward hack like building trust relationships fine um but I think that that would have to be to be [laughter] that said I will

51:25 say I have seen anal analysis done by AI that are as good or better than any you know gene analyst I've ever worked with like so for the the the analysis piece um they're great but I think this is more an indictment on VC's ability to like analyze that than it is like how powerful you know the models are. I just think it's all a kabuki show and [ __ ] I just think the right VCs are like you know you know tech is is is net positive.

51:52 You know tech waves always add value like let's find the best company and invest in it. The markets are very very efficient and so I I I just feel like the answer to this question is almost kind of a technical like my answer is almost a technicality. So how like you know one of the things you know I came here I wanted to ask you this real this question which is like >> you know we started talking specifically about ops.

52:16 >> Yeah. >> But to me like the there's this world of AI writing code which is kind of like this abstract simulation world that it lives in. >> Yeah. >> And then there's the rest of the world. >> Yeah. Exactly. I the entire discussion is is this. It's like there's like this >> there's like the AI the world that the AI lives in which is all digital and like whatever it is the training data the indistribution world and there's the actual world and like those things are not necessarily the same.

52:39 They're quite often not. And so the question is like to what extent do you need to train an AI to be able to survive in a world and like what are the things that it actually I I don't want to answer that question like what do you need to to train but like I'm curious about the open spaces. >> So we talked a little bit about operations. You said I don't know if it's going to be solved or not.

53:01 I'm like yeah Martina it's going to be solved. And uh in my mind like there's so many other things like for example self-driving cars like for a long time we thought that that's not going to happen but it's clear that it's actually going to be done. Um and it's going to happen. It's just taking time because of a lot of different things. And so what are like what are the other next stages the next areas that you think are kind of like lined up right after code?

53:29 I mean my my sense is is you know AI is going to continue to take off in a wildfire for a constrained domain where the where the solution space is in a constrained domain right where the distribution of the solution space is the same as the distribution of the training set the distribution of the syntax for code is the same as a distribution of the training set the distribution of the generated images is the same as the distribution of the training set in those you know areas I think it's going to continue to take off

53:58 as a wildfire the domains where like the distribution of the solution is the natural world. I think it's going to, you know, have, you know, it's going to be much harder, take longer, and I think the economic equivalency is going to be much more difficult. Self-driving is a great example. I remember when Sebastian Thrun won the DARPA Grand Challenge when I was at Stanford.

54:16 Remember that? And we're like, hooray. You [laughter] know, self-driving cars is around the corner. Seriously. And it took 20 years and $170 billion of investment to get to worse unit economics than Uber, right? So that's, you know, so I think that like solving like like solving real problems where you have to deal with the the distribution of the real universe is just much much harder.

54:42 And so I view as AI as this general tool. It's a very effective tool. It's one where like you have you know a distribution of data and then hopefully you can apply that to like you know some problem and we're going to apply it to a bunch of stuff and I think economic equilibrium is going to determine what works and what doesn't work because I think it's a very very general tool it's like a universal function approximator >> and wherever that nets out I mean again >> you know one of the things that I said earlier was

55:07 like how AI was training humans to like use it. >> Yeah. And I very often think about like you know some of the problems that seem intractable was is because the world is kind of that way and then we're trying to fit AI or the tools that we're building to manage the world as it is. But in some situations we actually tend to change the world to fit to what we can actually do from >> 100%.

55:30 Of course I think this is the story of robotics. Yeah. >> The the automated data warehouse. And so the the question like that always comes to my mind like when we were looking at these data warehouses and whatever and like the only one that worked was the one where it was like you know what just automate the whole [ __ ] thing. Yeah. >> Um >> and um and I wonder to like what extent as we kind of improve or change how we uh or kind of expand the domain of what we want to solve with AI tools and so on that we actually

56:01 start modeling the world or changing the the world and the things that can actually happen to fit >> what AI needs it to look like. >> Yeah. I mean I listen I I think that right now a lot of AI just operates in the digital world. Yeah. >> And for that is like literally the solution set is the same distribution as a training set in a sense. And so that's fine.

56:22 And I I think that again like in the case of equilibrium like I think like when you're dealing with the real world like you can solve any problem you want. It's just a matter of like compute and money and that the equilibrium point is going to be an economic one. And I think one way you can impact that function that you're doing equilibrium over is to constrain >> the environment is constrain the system.

56:46 And so I I think that we're going to like we're in now in this meta economic equilibrium flywheel where we're kind of optimizing this super general but very inefficient process with a solution set that people value. And I think you listen we've got 30 years of this and a lot of value is going to be created. >> You know one you want to hear a funny story?

57:07 Yeah. >> So, you know, we built this AI thing and we're like talking to people. We're like reaching out to customers. We're like, "Hey, we built this AI thing. You know, we're helping customers become more reliable. Here's like what we're doing." They're like, "Great. We want to talk to you." And then we go on the call and we're like, "Hey, here's what we do.

57:23 You know, we built this system. Helps you optimize your data. It helps you reduce your costs. And then it helps you become more reliable because now we can find these problems and solve them before they hit your incident." They're like, "Wait, wait, wait, wait. Roll back. You said you can save 90% of our data dog bills. >> Yeah. >> Like that's what I want to buy.

57:40 But but but the but the AI thing. And they're like, "No, no, no, no. That's what we want to buy right now." >> Yeah. >> And it's uh it's interesting like a lot of these uh like for code, you know, I always think about it. It's like, "Oh, for code, people just get it. They're like, oh [ __ ] it just works." >> Yeah. Yeah. Yeah. >> And on the op side or like anything that's beyond that, >> Yeah.

58:01 it's very to understand the actual value and put a dollar value to it and be a pay out the door. 100%. So, and I also think from from a coding standpoint, literally it's engineering productivity which is one of the highest line items on these things. So, it's just very straightforward. >> Exactly. But you can also think about it, it's S productivity or it's like your business going down.

58:17 Maybe you can save an incident and what the cost is. >> I don't know. It's it's weird. I think like how people think about it outside of the >> Yeah. I I also I also just think like like this code was so magical because it was so good and nobody thought that could happen. And I think the operations stuff now it's still like >> pretty early. >> It's still pretty early. So, >> yeah. [snorts] >> All right. Well, listen, it's great to chat. Thanks, J. Thanks,