From Periscope to Macroscope: Kayvon Beykpour’s Vision for AI-Powered Development

26:42Like, cause when we launched, I think we had our first million users within like four days, which at the time was like in today's like chat GPT growth craziness, like that's not unprecedented. But at the time that was like an insane amount of growth for a consumer app. Um, and for infrastructure that was like pretty untested and for live video of all things. It's not like, you know, these aren't static photos that you can put on a CDN. Like there's a lot of like sophisticated infrastructure that we had not tested at that scale. And so, yeah, that was it was a crazy first. It was crazy. I don't think we slept for the first like six months.

27:16Yeah. I remember right out of the gate the day it launched. I'll never forget it. You know, I was working for a company called Aviary. Actually, I think we had sold to Adobe by then, but we were in our office in the East Village. And I remember it launched. And I think within a few hours of it launching, maybe not even there was a building in the Lower East side that exploded. And I pull out Periscope. I had never used it before. And like literally just tap the button. It was like the first time I had ever gone live. I just pointed it out the window towards this explosion and it instantly went viral.

27:47There were tens of thousands of people watching within seconds and the hearts were fluttering. And it was this crazy moment where it really like for me, it like really lived up to this promise that you could tap a button and you could have thousands of people watching instantly. So that was like really fascinating for me. but you know i think this format of live video and live audio which i i later thought much about when i was at spotify i became convinced that this format is flawed and i know you've talked a little bit about like the downfall of periscope um and you also obviously dabbled in things like spaces which i know was actually a big success for twitter talk to us about periscope like why maybe it didn't work long term why live video didn't work long term and really like how you feel about this format as a strategy?

28:33Yeah. So a few thoughts on this one, I would say there's one fatal flaw that we made. Yeah. And then there's a few other reasons as well. I think the fatal flaw that we made is, you know, I mentioned to you that the origin story was all around teleportation, seeing what was happening. And I think that's always spiritually the thing that motivated us. And still to this day, like my romanticized vision for Periscope is still is teleportation. The mistake we made is that what we we didn't realize that the vast majority of people who are using the product were not motivated by that. They were just trying to talk to people.

29:06On the demand side or the supply side? On the supply side. Okay. The motivation of the broadcasters, which is essential, right? They are the supply. Yeah. To go out and collect videos of the world. And their goal is not unlike other social media platforms. Like they either want to be famous yeah they just want to talk to people right they're just bored and they want to have conversations and you can't really become famous easily with this right like my experience was totally random luck i happen to be in a place at a time i think you i mean you can though it's it's not unlike twitter like people become twitter famous people become instagram famous okay just for doing some shtick that resonates with them and they become popular right for that shtick right could become like a talk show or something yeah yeah okay man on the street type thing totally And I think those core needs, like just having conversations because you're bored or making money or, you know, becoming famous, those we sort of we lost our eye on those balls.

30:06And instead, I think for too long, kept focusing on the teleportation thing where that that wasn't the pull from the community. So that that was like the fatal flaw that we made. I think the other dynamic that happened here is one, and I believe this personally, maybe I'm too jaded, but I don't think that short form live video can exist as the sole medium within a social network. I agree. It's too difficult to make the synchronicity work and have critical mass and like durable retention without including some asynchronous networking features. And this is why live in the context of IG, Instagram, and live in the context of TikTok or live in the context of Facebook became so effective.

30:51One, they had much larger distribution because you had these scaled social networks. And B, the network that formed could retain their relationships outside of the bounds. Yeah, outside of the bounds of the live video. And I think Periscope, it took us too long to add those features. And our competitors were more quickly able to add our features into their larger platforms. Combine that with the fact that all the thesis around us integrating with Twitter took way too long for reasons that are annoying, but had to do with like Twitter's priorities and our ability to actually make the integration come to life.

31:26And so we just got outcompeted is the short answer. and you know to our competitors credit as frustrating as it was to see this like you know zuck woke up one day and decided we're going to make it a code red like 300 engineers boom go go build facebook live um and we did not have that kind of priority at twitter but you could argue facebook live suffered you know the same uh fate as periscope did in terms of the the format considerations you just mentioned like synchronous content on the internet short form it's a math problem. Like the beauty of the internet is it's on demand. You know, most of the content on the internet, you can view it anytime you want.

32:03You can consume it anytime you want, but with live, you can't, right? Like you have to have a critical mass of people who are ready to consume that thing the moment it's happening. And so, yeah, I also believe the format is kind of flawed, which, which is kind of interesting because, and it's kind of interesting you feel that way too, because a few years later I heard, I heard another rumor. I don't know if this is true, that Twitter tried to buy Clubhouse, another live synchronous format platform. So what's the thinking there? Well, first of all, is that true? And then if it is, what's the thinking there in that the Periscope format was flawed, but then you all tried to do it again?

32:39Well, what we felt was flawed was a dedicated live platform. Okay. By the time all this started happening with Clubhouse taking off, it became clear in our minds that live should be a feature within Twitter. OK. To connect to the asynchronous content. You've got a timeline of asynchronous content. And just, you know, as I mentioned, like the thing that we realized, like the vast majority of reason people were using Periscope was to have conversations. And guess what's another conversation platform? Twitter. Yeah. Right. It's just Twitter. You know, we can have this conversation later. But Twitter, I think, had taken its eye off the ball on giving people more ways to have conversations, right?

33:21Like it's this incredibly powerful, iconic, scaled platform that only let people talk in 140 character tweets. And so at the time, you know, I was then sort of leading the product strategy at Twitter. And one of our theses, one of the things I felt very strongly about was we need to give people more ways to have conversations. And live is just one piece of that, right? You need to be able to do long form writing. You need to be able to not just broadcast tweets. You need to be able to have like civil conversations in the replies because conversations aren't just like one way doors. Right. And so there's so many feature ideas we had and product strategy ideas we had around activating on this goal of like let people have more kinds of conversations on Twitter.

33:59One of them was audio. Yeah. You know, so at the time, you know, we had our own effort to build audio conversations in Twitter. Can I just pause you real quick? Yeah. So just like, I just want to say, so everyone's caught up, obviously Periscope got acquired, it launched. Fast forward, you're leading huge team at Twitter at this moment, right? Like you're leading a team of, I don't know, thousands of engineers. After my Periscope chapter, I had two chapters at Twitter. One was head of product, head of consumer product, where I was not leading the engineering team. I was just leading the product.

34:33Right. And then GM of consumer. And then I was the product and engineering and design teams. So at this point, you're... At this point, I was just leading product. Okay, got it. Another sidebar for later, but Twitter's org structure was very strange. And one of my lessons learned from how to not fuck up big company is like org structure stuff. And so this was like a particularly challenging for me personally, part of Twitter, where I had the responsibility of the consumer product strategy and all of our metrics, but was only empowered with the product management team. But anyway, in that capacity, we were super excited about audio and we were building something ourselves.

35:12What would later become Spaces, but we had not launched it yet. And yeah, Clubhouse was taking off. And our thesis, our interest for either building or buying something was always to integrate it. You know, maybe a separate app could exist as a complimentary thing. But like our goal was like, no, we want to make Twitter the product better. um and you know i think you know we did talk to the clubhouse team um at multiple points about trying to convince them to come on board and that didn't work out why not um i mean the first time we were too late and didn't offer enough um and the second time what they wanted was not something we could give and i'm glad the offer yeah was not big enough i mean i thought it was big enough but they didn't.

36:02And they also were able to raise a pretty insane fundraising round and they, you know, they chose whatever they chose their path. Got it. So classic, like, you know, founder goes to either raise or sell and, you know, they explore all options at the same time, whether it's selling or raising and they picked the raise path. Yeah. As one, as one should, obviously. You know, I think every person has their own takeaway from this. Our take, Like my takeaway was, I'm really glad we did what we did, which is we had an internal team that was so passionate and excited about audio that like deeply understood what Twitter was and how audio should exist within the context of Twitter.

36:43And a lot of this team was the Periscope team. Yeah. And we used the Periscope infrastructure, which made it super efficient for us because it's all the same stuff. It's just like take the video out. It's still, you know, it's live streaming. and so we were able to build Twitter spaces and launch it and have really great impact. It's probably the feature I'm most proud of having shipped at Twitter. The speed at which we did it like that project really operated like a startup within Twitter and it jolted Twitter to think that things were possible in a way that no one assumed it was possible before just like the speed at which we launched that.

37:22And so I'm really proud of not just like the product, but like how that shifted the culture of the company. It was really healthy for us to go through that. So on this strategy, you know, we just talked about how you didn't believe synchronous content could work if there wasn't an opportunity to then immerse the users in asynchronous content, right? This is, this is like merging live into Twitter where you have all this other content. Then why go out and spend all of this money on this platform, Clubhouse, which at the time just was a synchronous platform why not just do the the build path i mean there's there's there's trade-offs right there's in theory when you buy something it's the same reason why twitter decided to buy periscope rather than build with themselves like they're bringing in talent bringing in people who care deeply about a problem who can sort of bring that founder je ne sais quoi of like i'm gonna will this thing so you want the talent yeah and the user base bring them over and people who have like thought about the problem deeply um and and yeah the user base as well yeah um and you know so those are the reasons to do something um and then there's lots of reasons not to do something it's hard to generalize these things right it was a there was a time in that debate where it was like a kind of up in the air like it could have been a good idea to acquire and then eventually became like very clear that it's just not like it would be foolish to do anything other than leverage the start that we have to go build it ourselves.

38:46So that, and that's why we did that. Okay. So now let's talk a little bit about Twitter and your role there, especially managing such a large team. You know, I think Twitter, I hope this doesn't offend you. I think famously got a lot of shit on the internet for many years about not shipping fast enough, not, you know, not getting enough stuff out the door. And as you've spoken about Macroscope, you highlight how you and your co-founders experience the challenges and the pain of communication and collaboration in a really, really large organization. Is that connected at all to the feedback like many had about Twitter over the years?

39:28I think in part, for sure. It's not the only reason. To understand the Twitter dysfunction requires a PhD dissertation. But we can go through the high levels. One is it was just a really large company, like any large company, there's inefficiencies. That part of it is very connected to Macroscope and we can go into detail there. There were some other things that were uniquely sort of Twittery, which was like, there was just a stated strategy for years, which was, and I quote, refine the core. Like the strategy was like, let's not take big swings. That's fascinating. This is right around the time that Periscope joined.

40:04We were a separate startup. We weren't really involved in Twitter land. And we were kind of like a little frustrated from the sidelines because we were trying to get the attention of Twitter folks to help integrate live. And we sort of couldn't get the time of day because at the time the company's strategy was like, no, no, no, we're not taking big swings. We are just like going to turn the knobs on the home timeline to make recommendations better. Like that is what we're going to do. And we're not going to go ship other random features that are like riskier and more speculative, which is ironic, by the way, because they like bought a whole company.

40:35So like, you know, why would you do that and not do the integration part? That's fascinating. Refine the core. Is that known that that was the strategy? Is that publicly known? This was very explicit. Yeah. I think I've talked about this in previous interviews as well. But yeah, I mean, this was a huge mismatch in selfishly what we wanted, but also, you know, and by the way, this is nuanced. There's pros and cons to this, right? The cons were Twitter had a reputation, rightfully so, for not having like meaningfully changed the product or innovated. that is a deserved reputation for over many years.

41:07The upside was that focus actually brought the company back to growth because at the time, like the consumer product stopped growing. DAU stopped growing. It was that refinement of the timeline, literally that one product strategy focus that got Twitter back to like double digit year over year DAU growth. What do you mean? Like, what did you do? Going from a reverse chronological timeline to a rank timeline. Got it. That like literally no single feature at Twitter has been responsible for more DAU growth than that. I guess it makes a lot of sense. And then B, turning the knobs on that timeline, which by the way is still like what the XAI team is doing is just like constantly refining the timeline.

41:45I see tweets from Nikita all the time. Yeah, we got to get him on. Yeah. I mean, that's real. Like at that scale with hundreds of millions of DAUs, like making better recommendations drives DAU growth. Yeah. And so Twitter was on to something with that focus. I think that they didn't personally, I didn't agree with the portfolio balance of like having that be a hundred percent of your focus and having zero percent on like building new features that can get people to use the product in new ways that are true to its essence. I would have shifted that portfolio a little bit, but, um, as much as I was throwing stones from the sideline, from the sidelines, like it ended up being very impactful for the business.

42:24Now, fast forward a couple of years to when I became the head of product, the first thing that I did and felt very strongly about was shifting that portfolio balance. We still had an enormous amount of resources focused on improving the home timeline. We were not ignorant to the fact that that is extremely compounding, valuable growth to the product and improved user experience. But we started doing a bunch of things that we felt were good long-term bets, things that turned into community notes, super follows, Twitter blue, you know, paying out creators, spaces, letting people, you know, talk to fine-grained communities of their interests.

43:05Like all this stuff like started as bets. Some of them didn't work, by the way. Fleets, we tried, you know, fleets was like the Instagram stories. Oh, yeah. Chat stories. I forgot about that. Yeah. But we just like. You shipped a lot. I remember when you got elevated into that role, the pace just from the outside was instant. We were like literally like this. Like I one of the things that I tried kind of going the exact opposite side of the spectrum on are like we had a product strategy, which was like, what are all the sacred cows that like everyone thought we couldn't do? Let's go do those things.

43:35Yeah. Like just like make a list of the sacred cows and let's come up with compelling product ideas that like solve customer problems. So that I think that process of like cultural transformation was really important. I felt very strongly about and was frankly like the hardest chapter in my career because it was not not building stuff. It's like influencing change and like politicking when you kind of don't have complete authority. Right. Like I was just leading the product team. I couldn't. Yeah. I had to wait to sort of influence the engineering team to come along, influence the design team to come along, which is like not it was just challenging to do.

44:09but that chapter of my time at twitter really exposed me to a lot of the like inefficiencies of working at big companies and a lot of what influenced like how we think of macroscope i'll give you one simple example when you're the head of product of a you know company that has 3 000 engineers and a product that has hundreds of millions of dau like literally one of the most annoying but important parts of my job is just understanding what the fuck people were working on. Like 3 ,000 engineers. What are they doing? You ask the engineering leaders, they don't even know. They have to go ask their engineering managers who then go ask their engineering managers who then go ask another layer of engineering managers who then go ask ICs, who then just tell them what they want to hear.

44:49So you get this like game of telephone where by the time some clarity gets shared to an executive who's asking the question, it's been so whittled down. The sort of problem that was being solved has probably been lost in translation and you're getting a lot of sugar-coated like well here's what we're working on and like you know 40 of our team is working on like keeping the lights on what the fuck does that mean you know and like how much of that is just like projects you want to work on versus actually keeping the servers running so but this is important when like if you're trying to build new features you need engineers to work on those features and if the team is saying those engineers are busy working on other important things it brings about these questions of like no no but what are you actually working on?

45:32Like, let me make, let me assess whether that's important or not. And so this game, this process happens at every company. Totally, yeah. It's just at smaller companies that are startups, the CEO, the CTO just knows. They can sit around the table with their team and they know what's happening. But at bigger companies, whether you're 30 engineers or 100 engineers or especially thousands of engineers, you have processes in place to solve these problems. And that's where like, even the most sophisticated AI companies that like we all would know their names, they are using spreadsheets to track like they might be using linear for their backlog or some planning stuff but like they are using spreadsheets to understand what their portfolio allocation is which is insane and it's all grounded in this like self-reported information and so that part of our experience and lots of other things like you know to sort of get to the bottom of like project status we would have meetings and those meetings at twitter were enormous it was like 60 person meetings which is like shameful to even say out loud but like we had we had to sit down with the tech leads and the pms to be like what what did we work on what progress did we make so did you have the idea for macroscope while you're at twitter or this this is only sort of in hindsight where you're like there's got to be a way to solve this with ai well we asked the like there's got to be a better way question many times but i i cannot claim to have thought of this idea because prior to llms like you just couldn't there was no better way that i we could think of to solve this problem yeah um and and the reason the thing that made this different is that the source code has always been the source of truth for all this stuff right like for a software development company if you build software the source of truth for what has happened is the code period prior to llms it just wasn't really feasible to use that source of truth like what are you gonna get to open the commit hash you're not gonna do for every you know commit to a repo as large as twitter's mono repo with 3 000 like no way and like not even the the most experienced engineer at the company couldn't do that it's just like infeasible llms changed that yep and so this problem was percolating even you know i had been retired from twitter for a year comfortably playing tennis still annoyed by like the shit that we had to go through to get things done and so it did start sort of formulating in our brains like wait a second like this annoying thing that we lived with there's got to be a better way now than lms and there's been so much energy and effort and and capital thrown at ai coding but no one to your question earlier no one had really solved the sort of management or or put solutions in place to the pain felt by the management layer, or the pain felt by engineers solving questions for the management layer.

48:20And so that's where we really started thinking, wait, maybe there's something here. And even still, we had to do a lot of building to kind of figure out what this wanted to be when I grew up. But that turned into Macroscope. You mentioned you had been retired from Twitter for a year. you were famously or infamously fired by Parag, which at least from the outside, and I didn't know you that well then, was a shock to me. I know it was a shock to many others. Do you think you were fired as a condition of the acquisition? Did Elon's team ask Parag to reduce the staff or something? No. In fact, the very first time I met Elon, which was seven months after, basically the day he closed the deal was the first time I talked to Elon.

49:06And his first question to me was like, why did Pro fire you? Really? Yeah. He didn't. And you were like, I don't know. Yeah. I said, I don't know. You'd have to ask him. Yeah. That meeting with Elon was documented in the Elon biography by Walter Isaacson. I remember reading that. So this was a phone call with Elon. Oh, okay. But then a couple of days later, I met him in person. Okay. For the Walter. Yeah. It was really interesting. I was reading. Yeah. I was reading that book. I was like, oh my God, Kayvon and Scott Belsky are now in this story. Yeah. Tell us about that meeting. Yeah. that was that was a wild 48 hours um you know in the my understanding of what happened here is as elon as it became clear that elon was gonna do this deal even prior to him closing he ended up asking a lot of his contacts and his network to like get him up to speed on all the things at twitter who are the great people what are the projects i should go look into what are the things i should cut um and so he started putting feelers out into his into his network one of those people, um, introduced him to Scott Belsky.

50:08Um, and then Scott introduced him to me because, you know, I had just been there and spent a lot of time there and had lots of strong opinions about all those things. Like who are the great people? What are the projects? And so it was in that context that Elon and I talked on the phone. And I think we talked maybe the day he closed the deal on the phone. Um, just had like a FaceTime audio, you know? And that's, that was the conversation where his he was like why did prog hey nice to meet you my first question is like what happened there yeah um and i was like i thought you would know i should ask prog you know we had a fun you know fun but quick you know 30 minute conversation where i i told him like if i was in your shoes here are all the things i would go learn about here are all the people i would talk to like again bias but like don't let these people quit these people are your stars yeah these projects are super interesting and then you know he he suggested we we chat in person i was like can you come in tomorrow and i was like sure why not and you went with scott scott and i went to the office which was super weird because the mood was grim this is like when people were sleeping on the floors and stuff no this is this was before that elon and his people were like very publicly to the company assembling lists to fire people and everyone knew that and it was like It was unclear that the execs hadn't been fired yet.

51:33I think that happened either that day or the next day. And so everyone was just like, what's happening here? What's Elon doing? What are these lists for? Am I going to have a job? It was super weird, understandably. And so it was in that craziness that I strolled into the Twitter elevator. And people obviously, a lot of people recognize me because I've been there. What's Kayvon doing here? Yeah, so there's that whole thing. Do you think he was thinking about making you CEO? Oh my God, no, not at that. At that time, he probably didn't remember the conversation he had with me 24 hours a year. I think he was very much in the mode of like trying to save the company, right?

52:10He had acquired the thing. The market had gone to shit. His financing structure was creative, let's just say. And so for whatever reasons, he was very focused on saving costs. And that was his primary motivation. His secondary motivation was more on the product side to understand like, what are the opportunities here? What are the people here? And so I think it was as simple as that. Like, oh, here's here's here are two people who might have insight on like what I should focus on. And he was, to his credit, very like open to learning and absorbing, you know, like a sponge information. So that was it.

52:44That was like. But I heard he asked you to come work there the next day or. So at the end, so we had like an hour and a half plus brainstorm. It was like all over the place. It was like we talked about products that we thought were cool. We talked about teams. He shared all of his wild ideas for what he wanted to turn the product into and was sort of getting our thoughts on that. We were in like the largest conference room at the Twitter offices. And like in one corner was me, Scott, Elon. I think that just us. And then at the opposite end of the coin, like seemingly 100 yards away was a guy who looked like Walter Isaacson.

53:19I did not know was Walter Isaacson until after this meeting. Walter came up to me and was like, here's my business card. I might have follow-up questions for you. And I was like, great. Didn't realize this was on the record. Not trying to be in a book. Yeah. And then his EA was also, Elon's EA was there as well. At the end of that conversation, Elon was like, you seem like you love the product still. And you have decent ideas. Like, do you want to like come work here? And my question to him was like, what would my job be? Yeah. And I'll never forget his response. He was like, don't know, but like you can come hang out and can swipe right if you like it and swipe left if you don't, which I still think is like an amazing line.

54:00And my response to him was like, listen, I have so much respect. I literally have purchased every single product you've ever made except for a SpaceX rocket. I just don't know if like I just left this place. I was here for eight years, kind of enjoying like being a dad. I want to start a company again. I have to think about it. Let me get back to you. Yeah. And that was the last time we talked. Fascinating. Macroscope wasn't the first thing you guys tried to build. There was another product before Macroscope, which I never tried. I don't know a ton about it, but it was a consumer thing, which in my mind made a lot of sense for you and your co-founders because you guys come from the consumer world.

54:40What was that and how did it lead to Macroscope? When we started the company, we had a few ideas that we were sort of marinating on. Yeah. And the first that we built was very different to Macroscope. It was another idea that was sort of like in our heads for a long time that we felt like was maybe finally possible with LLMs. And it was a essentially a personal assistant app. It was sort of imagine an executive assistant meets a to do list. Right. What if we could give every consumer access to the superpowers of an EA, but in a form factor that looked like a to do list and without the cost of like hiring an EA?

55:19Sure. And so it's sort of like it was a to do list that wrote itself and executed itself. We would infer tasks from your email, from your calendar. We would also let you write tasks. And rather than just sitting in your to do list, we would try and do the tasks for you. We would like we actually hire like a team of EAs. Oh, there were people doing these tasks. It was a mixture of sort of an LLM in the loop. Yeah. And yeah, we had humans. Because if your task was like cancel the doctor's appointment. Yeah. AI can't do that. AI can't do that. And our insight, which is still the thing that I think we got the most right about that idea, was like no one gives a shit how much of it is AI versus not.

55:56What you care about is having your task solved. And so we obfuscated all that from the customer. To you, it looked like a to-do list with an agent. And like it literally, just like ChatGPT tells you it's thinking steps, like we would tell you our thinking steps. It's just one of those steps was, cool, we've queued this up for an agent to go to do for you. And then we would follow up. It was sort of ambiguous to you whether it was a LLM responding or a human responding. Some of those steps were human steps. That's what we were trying to build. And, you know, long story short, in terms of why it didn't work and why we pivoted, The customer we were solving for just did not have a big enough pain point to use this frequently enough for this to be like a repeatable habit.

56:38Yeah. And no less one that you would pay, you know, substantial money for. And more importantly, like we just weren't feeling it. Like we've been building stuff for long enough to know that if we weren't feeling it. Yeah. It's toast. Like we have to be working on something that we're super passionate about. And was Macroscope something you guys were building in the background? Like, were you, had you created a tool to help understand this code base or was it a completely distinct separate? It was just the other idea. It was the other idea that we had that when we started building the first consumer product, the models still weren't, I think, good enough to do what we are doing with them now.

57:17And so, you know, we had thought about this and in our own playing with these LLMs, we're like, it's just not going to, we're not liking the quality that we see just sort of tinkering with this stuff. for the use case that we had in mind. Like if you're an executive trying to grok what's happening with like your company, if you get LLM gibberish back, you're just never going to trust the thing again. Yeah, you're done. The bar is so low, which is very different by the way, for if you're, you know, if you're in cursor and you're in some IDE and you're doing tab-tab to autocomplete, like in that it's so human in the loop, your tolerance for errors, I think is a little bit higher because you're still driving, you're in control, you're not fully delegating.

58:00This is obviously different now with all the agentic tools, but the models are way better now. But for our use case, it's like trust is so imperative that the gibberish bar is very, very low. And so it just didn't pass the smell test at that time. But in that first seven months that we were tinkering with the first consumer product, a lot changed with the models. Totally. It was like the difference between cursor didn't exist and then it did exist. Yeah. And all these, you know, GPT, I think, 4 came out. So anyway, a lot changed. And the combination of now we had sort of reminded ourselves what it was like to like be building software with a team.

58:38It wasn't a substantial team, but we had a large enough team that a lot of these problems sort of represented themselves to us of like, what's everyone working on? And like, what did we get done this week? And so all those stars kind of aligned in a way that made that idea come back to the forefront. And we started prototyping it and it started feeling good. like, oh, wow, there's something here. Like we can write technical summaries about code changes that we felt the AI could write better than we ever could. And that's still the thing that we hear from our customers today that makes us know we're on the right path.

59:08It's like - This isn't something a human could do. Really competent engineers who are like senior experts in their code base will say, wow. Like getting an engineer to say, wow, about like a summary is, you know, I think is a good signal. Yeah. So product is launching now. I, of course, think it's going to be a huge success. Assuming it is, how does this evolve over the next 12 months? We said at the beginning of the conversation that we expect more and more agents within organizations. So I imagine that'll have an impact on the product. But how else do you imagine this product and this space evolving over the next 12, 24 months?

59:43I think there's a couple dimensions that we're really excited about. One is fundamentally, Macroscope has built this perception layer that tries to understand what's happening at your company. And primarily today we use the code base as the sort of core tenant of that perception layer. And we've got a bunch of techniques that we've built that leverage the AST of the programming language to really build a graph of the code base and do all this magic to sort of set up the LLM to be successful. But it's using the code base as that source of truth. I think one of the sort of dimensions you can expect us to innovate on is using additional sources of data to understand what's happening.

1:00:22Oh, interesting. The code base is one. Your issue management system is one. Figma is obviously extremely important to the extent that you're building a sort of customer-facing product that has pixels to it rather than just like an API, let's say. Figma is a really important source. Being able to actually run your code and visualize it not just through the mock-ups, but like how the code actually looks when it's rendered is another aspect. There's many other data sources that matter to you. For example, you might want to understand what the impact is of certain features you've launched. The impact of your work is not in the code base.

1:00:55It's not even in a Google Doc. It's probably like in your experimentation system, like LaunchDarkly or StatSig. So there are many systems. There's a patchwork of tools that companies use in their product development stack. And you can expect Macroscope over time to integrate with more and more of these, which I think will improve the perception layer and will improve the sorts of insights that you can get. about what's happening. So that's one layer. And then the other layer that I think is further out, but it's part of our aspiration is we're not just trying to storytell what's happening. We're trying to help you build your product.

1:01:28Today, we focus more on helping you understand what's happening, but we have some ideas that we're really excited about in terms of like actually being on the development path for you to not just understand, but orchestrate the future evolution of your product. And our thesis is that, you know, the system that can best do that is also the system that has the best understanding of like what your product is, where it's come from and how the context, all the context. We're starting with humble beginnings. And there's a lot that as we've learned from our customers so far, like what we have today is valuable in and of itself.

1:02:00And we don't want to, we don't want to take our eye off of making that great. Um, but there's a lot more we can do moving forward. Awesome. Well, I'm personally really excited to see how people use it starting today? And I guess, where can people get it? You can go to macroscope.com and learn a lot about the product and you can actually sign up to use the product as well. Cool. Awesome. Well, thank you so much, Keva. Thanks, Mike. Thank you for listening to Generative Now. If you like this episode, please rate and review the show and of course, subscribe. It really does help. And if you want to learn more, follow Lightspeed at LightspeedVP on X, YouTube, or LinkedIn.

1:02:39Generative Now is produced by Lightspeed in partnership with Pod People. I am Michael McNano, and we will be back next week. See you then.