Austen (00:06.112) All right. I believe we're live. awesome. Thanks everybody for joining and some more folks will trickle on in. but yeah, Vivek, thanks for joining us. Vivek Sodera (Supercharge.vc) (00:17.919) Yeah, thank you for having me. Austen (00:20.002) Awesome. First, what are the helmets that are behind you? are those meaningful in any way? Or is that just what what is those daft punk helmets? Vivek Sodera (Supercharge.vc) (00:23.609) Yeah Dap, Dap on comments, yeah. Austen (00:29.73) Very, very cool. Awesome. well, yeah. So today we're gonna talk a lot about kind of what's changed in the last little bit. I think both you and I view the world very fundamentally differently today versus several years ago. and we're seeing kind of a bifurcation of companies, some that are adopting the new ways, some that are not, and you know, it's just a a crazy world. But before we jump in, I think it makes sense to kind of take through take everybody through the story of how we met. I don't know if we'd ever met beforehand, but but take yeah, yeah. We we knew each other. We kne definitely knew of each other. but I believe the first time that we met face to face may have been when you visited Gauntlet. let's let's talk a little bit about what that was like and yeah do you want to take us through that experience a little bit? Vivek Sodera (Supercharge.vc) (01:06.859) We've exchanged emails. Vivek Sodera (Supercharge.vc) (01:12.501) Yeah. Vivek Sodera (Supercharge.vc) (01:17.899) That's right. Vivek Sodera (Supercharge.vc) (01:24.479) Yeah, sure. So for context, I've been a founder of a few companies. First one was LiveRamp that went public in 2018 and then was just acquired a couple of months ago for two and a half billion. Second company was a failed startup. Superhuman was the most recent one. We exited that company to Grammarly. Grammarly rebranded their entire business as Superhuman with the intent of going very horizontal in the knowledge worker productivity space. And so I've been actively angel investing. During my tenure at Superhuman, started to get my appetite wet around being a venture investor. And so I've since transitioned into venture with my own fund called Supercharge, which is a very early stage fund, at times inception focused pre-seed in the domain of full second-of-prize AI, physical AI, NeoLabs, AI data, software infra. But I had moved to Austin, Texas early 2021, circa 2025. I saw early 25, I saw Austin already here, had moved to Austin, Texas and was like, hey man, I know we've exchanged many emails in the past. It would be good to properly catch up on what you're working on with Gauntlet. And yeah, you know the story. You gave me the tour, my jaw dropped. I was like, holy shit. Sorry, excuse my language. Holy crap. This is the future. This is where the world is going. you know, I think I told you after you gave me a tour, I sat in my Jeep for about a good 15, 20 minutes processing what I had just witnessed and what were the ramifications and what... Vivek Sodera (Supercharge.vc) (03:17.233) What we could expect as society as operators as builders as investors as users and so forth So yeah Austen (03:27.288) So I mean, first what what did you see? I mean, I I was there, but what what struck you? Take us through that experience. Yeah. Vivek Sodera (Supercharge.vc) (03:34.892) Yeah, I think it was the sum total of this Vibe coding kind of ethos, married with the proliferation of the AI tooling at the time. I think this was pre-Windsurf acquisition. And then specifically walking to the different teams. that were sitting in Gauntlet HQ. And in particular, one that you showed me where you had this idea for a procurement app on a Sunday night, you gave the prompt to the AI engineer Monday morning. He built it by end of day Monday. I was there the next day Tuesday. And I had product feedback and feature requests that would have otherwise have taken a 10X engineer. four to six hours to build and you were like, Vivek, don't go anywhere. Arpen, how long is it gonna take to ship? Okay, two minutes, we'll wait. And then it shipped and it blew my mind. And so seeing the Gauntlet program that pulled in all these different parts into this machine, that's what kind of blew me away. Austen (05:01.389) Yeah, I think the I r remember the exact moment we were looking at some graph, some display of information, and you were like, Yeah, man, I I almost wish that you could view this as like an exportable CSV. And I was like, Okay, just to help us all understand what's going on right now, ARPEN built that. You've got like a minute. I need to see a CSV export. And it it wasn't quite one shotable at the time. I'm sure it would be now. but Vivek Sodera (Supercharge.vc) (05:08.117) Mm-hmm. Vivek Sodera (Supercharge.vc) (05:17.013) Yeah. Vivek Sodera (Supercharge.vc) (05:24.117) That's right. Vivek Sodera (Supercharge.vc) (05:28.776) Mm-hmm. Austen (05:30.38) We hadn't seen anything like that, right? Like back then the notion that you could build a CSV export while somebody stood watching over your shoulder was completely unthinkable. I for some companies, for some people, it still is now. so okay, so basically you you see all this craziness that I mean it was early. I'm pretty sure this may have been early cursor and windsurfed. cloud code, I don't believe existed yet. I mean, we're Vivek Sodera (Supercharge.vc) (05:32.96) Right. Vivek Sodera (Supercharge.vc) (05:58.687) No. Austen (06:00.641) This stuff looks like ancient history today relative to where we are now. And so you go out to your Jeep, you're sitting in the parking lot or wherever it may be. What's going through your head? And like how did you begin to process all of that? And what what did you think it meant? Vivek Sodera (Supercharge.vc) (06:05.361) Yeah. Vivek Sodera (Supercharge.vc) (06:22.897) it disrupted for me. almost 20 years of mental models and pattern matching I had built as a multi-time founder operator. Everything from what does a great team look like? What does product market fit look like? What does business models look like? What does OpEx look like? What does GTM look like? Every aspect of company building. And that moment was like, okay, every, is, we've kind of seen the tech Twitter and all this buzz around obviously with this AI paradigm shift and AI cogen and this, particular philosophy around five coding and whatnot. you know, just like what were the second and third order effects for that? And so for me, it was, holy crap, everything that I known as a founder now. as an investor has just got upended. I circa 2022, 23, I just kind of sat in the sidelines and just watched how the world was evolving and to actually have these predictions start to play out in front of me in real time, but then also start to think through, okay, what does that mean moving forward for... for us, for society, for those that are building these companies, for those that are operating these companies. So yeah, was very jarring to say the least. Like I was kind of shocked when I came back home, I told my wife what I had witnessed and told some of my investor friends as well. Yeah. Austen (08:13.002) Yeah. I mean, y you're also someone who spent decades perfecting the craft of software generation. You you know, you probably you're one of the experts that I look to when it came to how do you generate product market fit with a software company. How do you build software the right way? How do you really care about product? and obviously some elements of that remain, right? And we'll we'll talk about what remains true. and some elements of that change. you shared with me, and we won't force you to share all the details, that shortly thereafter you kind of sent a letter to all of the LPs, so all to all the investors in your fund saying, like, hey, things are gonna be different from here on out. And here's a little bit about how I think about what that is and why. yeah, could you can you share a little about that? Vivek Sodera (Supercharge.vc) (09:07.637) Yeah. Vivek Sodera (Supercharge.vc) (09:11.305) Yeah, yeah. In preparation, I pulled together some of those predictions and what I thought second, third order effects are going to be. So this is a... an annual letter update I sent to my LPs and my fun supercharge and I specifically wrote an essay about the Gauntlet AI experience and this idea of software at the speed of thought. And so I extracted, not gonna share obviously the LP update, but extracted the specific predictions. And so. At the time, this was about 18 months ago, was like, okay, programming will be bifurcated between AI engineers and non-AI engineers. And I think that's come true. It's become this default hiring frame. You still see these labels of L4, L5, L6 engineers, but that seems to be a bit more decorative. This idea of the, yeah. Austen (10:07.916) Do you see I mean I live in a weird world where I don't honestly encounter very often non-AI engineers unless it's like they're the stick in the mud at the company that won't like get on board. what are you seeing there? I mean with people that aren't adopting AI. Vivek Sodera (Supercharge.vc) (10:30.794) I mean, I don't look at those companies as an investor. It's just, it's... Austen (10:33.568) Sure. Vivek Sodera (Supercharge.vc) (10:39.244) It's likely going to be some of the more antiquated technologies. You know, have legacy engineers that work at Pentagon that are coding in COBOL or something to that effect. But for me, because I'm investing in the enterprise AI space, it's expected that everyone is an AI engineer. Now, are they in reality an AI engineer? That is subject for scrutiny and debate or analysis, but everyone should be adopting this paradigm. I've seen this with a couple companies that I actually angel invested in before I started my fund. When I talked to the founders and I asked them, this was last year sometimes, like, okay, what percent of your code base is AI generated? Austen (11:08.598) Vivek Sodera (Supercharge.vc) (11:37.709) versus handwritten. And there's a consistent pattern here where the ones who answer anything less than, if they're saying 40, 50 % of their code base is handwritten, they're not gonna make it in this market. The one in particular, Austen (11:59.799) That's wild up here. Yeah. Vivek Sodera (Supercharge.vc) (12:04.396) I recall he said 70 % was handwritten, 30 % was AI generated. And I told him, you are shipping at a rate of every couple months, the new release. It should be on the order of days. And your cohort of the market of what's going on, they're shipping at such an astronomical rate that you are going to become irrelevant. And that's actually what's happened. That company is on the verge of dying. It's like a few months away from shutting down. And it's like, yeah, okay, this is what I saw last year. Austen (12:39.166) and is that yeah, is that just all the competitors? The competitive landscape is all moving, they've all sped up and yeah. Vivek Sodera (Supercharge.vc) (12:45.792) Yeah, yeah, yeah, you just have to adapt to the market. And if you're not adapting and try to stay ahead of the market, you're going to die. You're going to become irrelevant. And predicted that was gonna be the case, saw that play out. was like, yep, I knew that was gonna happen. But what's wild is this notion of the unicorn engineer, the 10x engineer, is replaced by the 100x engineer, soon to be the 1000x engineer. We've heard stories of folks getting compensated with eight figure, nine figure comp packages. I think it was OpenAI compensated the Thinking Machines co-founder, I forget his name, paid him a billion dollars, like comp. Was it Meta? No, I forget if it was Meta or OpenAI. Whoever it was paid that particular individual closer to 10 figures, which is insane. But one of the predictions was, okay, agents won't replace the AI engineers in the near term, but instead will become pair programming partners. This is like yes and no. it's... Austen (14:01.161) Yeah, it's different than how the way it's written kind of implies like an i you're going back and forth on a smaller cadence of yeah, it it's interesting because it's it's directionally correct, but it doesn't feel like pair programming partner is the right paradigm necessarily. Vivek Sodera (Supercharge.vc) (14:19.604) Right, yeah, yeah. I assumed that it would be baby steps towards full autonomy, and it's blown past those baby steps. And so now we're seeing agents have truly become autonomous. So it was like that prediction was partly right, but partly incorrect in terms of this framing. Austen (14:30.293) Mm-hmm. Vivek Sodera (Supercharge.vc) (14:45.612) cybersecurity risk vulnerabilities. We're seeing this is under resource risk. Company OpEx, yeah. Company OpEx is like, okay, this is gonna decrease due to headcount reduction, but will be offset by the rise of software spend. Yeah, compute infrared data. It's like, yeah, we're now seeing this and token maxing is a very real thing. Austen (14:49.579) It's all over the place. Yeah. Vivek Sodera (Supercharge.vc) (15:06.334) Enterprise companies in particular are trying to think through what budgets should we set aside for our EPD talent and not just EPD but the entire organization in terms of token utilization and spend there because it's it's starting to offset in many cases the OpEx reduction from the headcount reduction So this is going to continue to be a problem for companies in general as they try to figure out how to best manage their spend and manage their burn. But, yeah. Austen (15:40.928) Yeah, I remember, I mean, I I talked to a a a founder the other day. who, you know, he they had one of those like there are people that token max for the sake of tokens and maxing token maxing, and they're like, you know, just I check my email with Fable and do everything with Fable. And you can run up a crazy bill that way. but I've had a couple experiences in the last little bit where Basically, the mental model of the founder or the CEO is like, hey, I have this really special person. And a lot of what I just need to do is just find that person, give them unlimited token budget, and let them take off. And you know, I'm not gonna hire 10 more people the way I used to need to, but that person has a different token budget because of the way they use AI than like, you know, if you gave the other person the same token budget, it would just be wasted. This person really knows how to to use it. So it's really interesting to see that play out. And I mean, to give you an idea, he was talking about token budgets for an individual that were almost a million dollars a year. And the person beforehand had been making, you know, less than a hundred dollars an hour. So it's it wasn't like they had they had a raise since then, but it wasn't a clear trade off of token for, you know, Vivek Sodera (Supercharge.vc) (16:49.205) Mm-hmm. Vivek Sodera (Supercharge.vc) (16:55.614) Wow. Austen (17:03.403) people necessarily or like you know take the highest paid person and give them token budget. And so I commented like, man, are you like, I would start to get nervous if somebody on my team were spending a million dollars a year. And I'm pretty AI pilled, you know? and he's like, yeah, but I've also lived through like I've lived through the world where I know what it's like to have to hire the 25 people that would have done the work that that person is doing with the million dollar token budget. Vivek Sodera (Supercharge.vc) (17:05.292) Mm-hmm. Vivek Sodera (Supercharge.vc) (17:16.746) Yeah. Yeah, yeah. Vivek Sodera (Supercharge.vc) (17:30.273) Yeah. Austen (17:31.639) And you know, I'm probably saving three X what I'm spending. So the more the the direct quote that's stuck with me is the higher my token budget is, the lower my HR budget will eventually be. And it it isn't close to offsetting. But it is, you know, if you think about it from an economic standpoint, you can get to a point where the marginal token approximates what the marginal cost would be of stuff. So that's super interesting to watch from my end where We give everybody you know, we have students on max plans because if they were on enterprise our economics would change fundamentally at Gauntlet. yeah. Vivek Sodera (Supercharge.vc) (18:12.428) Yeah, yeah, I'm starting to see this metric emerge of companies tracking revenue per million tokens or tracking revenue per, you know, X amount spent. the idea isn't to token max, which is a vanity metric, but rather generate and utilize AI to actually produce like a true output. But that's a metric that I'm seeing emerge more and more. Austen (18:20.677) I love that. Austen (18:34.955) Fascinating. Austen (18:40.617) Fascinating. Okay. and now we're to the those were all pretty darn good predictions. they were accurate. and I mean, to be fair, some of these look pretty obvious now. At the time, none of these were like if you would have asked 10 investors if these are true, they would have agreed with maybe half, you know, like it was not this was not the the norm or the status quo. Vivek Sodera (Supercharge.vc) (18:43.456) Yeah. Vivek Sodera (Supercharge.vc) (18:57.068) Mm-hmm. Austen (19:09.751) and this next one is super, super interesting to me because if there's anything that was the most widely predicted from people that saw AI, it was, hey, obviously the teams are gonna become one tenth the size of what they are today. And to some extent that's been true, but take me through what you're seeing in in that world. Vivek Sodera (Supercharge.vc) (19:25.748) Yeah. Yeah. Vivek Sodera (Supercharge.vc) (19:32.621) At the large enterprise level, we're not seeing that yet. At the startup level, we're seeing this adoption of this philosophy. Maybe the plan wasn't originally grow headcount to 30 and instead you're gonna keep it to three people, but it's very much a significant, by X multiples in terms of... of what the head count is. But as you know, as you've seen, we've seen, junior talent was one of the first casualties. And middle management is also starting to come on the chopping block. And the senior architects are holding up. There was this thinking I had at the time, which was, OK, a team of 30 to 50 engineers. over the coming years, gonna get cut down to three to five AI engineers who are domain experts or they're working in some coordination with their pod mates and they're reporting to an architect or they're reporting to the CTO. That was the vision I had, which was, okay, this is what it's gonna look like. We're not there yet, but I absolutely believe we're trending in that direction. Austen (20:58.89) I mean is interesting. I mean, I look, you know, I don't invest nearly as much as you do. but when I look at five years ago, if I were investing in a company, the expectation was that you're gonna raise two or three million dollars, you're gonna hire four or five people, that's gonna get you to like VP and like prove the market. And now I'm like if you need two or three million dollars to get an VP out, you're completely incompetent, right? Like that's not even like Vivek Sodera (Supercharge.vc) (21:01.484) you Vivek Sodera (Supercharge.vc) (21:18.154) Yeah. Yeah. Vivek Sodera (Supercharge.vc) (21:24.766) Yeah, right, right. Austen (21:28.948) That's not a question that you would ask anymore. But it does start to break down at the enterprise level to some extent. Like you're not seeing you're not seeing teams stay small in the same way that I would anticipate. If just be you know, looking at startups, like that said, I also don't see, I mean, there's a time when if you're DoorDash, you need to go hire 10,000 engineers, right? That's Vivek Sodera (Supercharge.vc) (21:32.264) It's outdated. Austen (21:55.529) I don't think a startup that raises today and gets to the scale of DoorDash goes out and hires twenty thousand engineers anymore. I they maybe they will, but I don't I don't see that. I don't know. Vivek Sodera (Supercharge.vc) (22:09.308) You don't know and and if you are Right. Yeah, I think that approach is very outdated and so where I'm seeing the best teams approach this is they are Austen (22:13.374) Like that scale doesn't make any sense, yeah. Vivek Sodera (Supercharge.vc) (22:29.868) delegating and automating out through AI for a particular function and a hiring of that function is only done if it's essential, if it's necessary. It's not the default position. It is the, all right, we've exhausted our options here. We now have to hire for this role. Austen (22:53.832) Yeah. This next one is super interesting. quality assurance will become irrelevant and replaced by reprompting. I I understand why that was the prediction, right? Like, hey, if you can just tell AI to do it again, or you can tell AI to test. and to some extent, I think people are worse at QA now than they were before. but Vivek Sodera (Supercharge.vc) (22:58.57) Yeah, I assume. Vivek Sodera (Supercharge.vc) (23:08.812) Yeah. Vivek Sodera (Supercharge.vc) (23:18.284) 100%. Austen (23:21.03) If anything, it feels more important, I would say, now than it was before in the past. Vivek Sodera (Supercharge.vc) (23:26.476) Yeah, it's a function of what I'm seeing, which is there's a lot of slop in user experiences that we engage with, whether it's content, code, product, whatever it may be, and you just have a lot of human-driven laziness. And so this is where it's important to continue to have a QA culture. especially on the EPD side. and it's, but it's been rejiggered. So it's not traditional QA. It's, it's, there's very much the eval engineer or the, the, the folks who are responsible for evals and like now agent harnesses is all the rage and like test infrastructure, et cetera. And so, when you're now immersed in the market of a lot of noise, What separates amongst the noise is the product, is the company that has taste that is actually high quality. And that comes from having this QA-esque culture in place and not just like one-shotting it or just being okay with slop. And this, you can't get away with doing that nowadays. You could a year ago, maybe, but now everyone's doing the same thing. And so now it's, there's just so much slop in the market. Austen (25:02.45) Yeah. What I mean, then you know, your next thing on the list next couple of things on the list are about marketing and distribution, which is still interesting to me because marketing teams haven't changed nearly as much in the way that they operate as engineering teams, I would say. Like maybe it's just because I'm, you know, swimming in AI engineering land. Marketing feels like Most teams are still doing pretty much the same stuff, maybe a little bit faster, maybe with a little bit of AI. But I haven't seen it like given the impact that it has on the amount of stuff that people are able to produce, I am surprised that it hasn't changed that as much. And I don't know I don't fully understand why, but it yeah. Vivek Sodera (Supercharge.vc) (25:47.083) Yeah. Vivek Sodera (Supercharge.vc) (25:54.093) I think GTM and marketing, they tend to be more laggards. What I'm seeing are the GTM marketing teams. There's a couple different shapes. One shape is, okay, AI is a tool and we will use this tool for productivity purposes and we'll supercharge my productivity, our team's productivity, et cetera. And maybe we are... 30 % more productive, 50%, 100 % more productive. think that's consistently I'm seeing that with more growth stage, more enterprise type companies and businesses. Where I'm seeing this other shape emerge on the marketing team side and the GTM side is you have this individual who is early in their career. They are AI native. They have a approach around. speed, adoption of technology, experimentation, customer centricity, et cetera. And there was a story I heard from a friend of someone on the ads team. I forget which company it was. Well, I wouldn't name the company anyways out of privacy purposes, but there was a team of I believe 13 and 14 individuals and this one employee. replicated the work of the 13 and 14 people in very short order. The manager of that team fired many of the ICs there and promoted this particular individual like you have done this at a fraction of time against our entire marketing team or advertising team. And you've done this at a fraction of what our OPEX is. Here you go, right? Vivek Sodera (Supercharge.vc) (27:47.357) So I don't think that's the norm. I think that's more the exception. But like I said, think those type of teams tend to be a bit more laggards and not as early adopter like as engineering product and design is. Yeah. Austen (28:05.533) Yeah. I I have I mean I I have a mental model that I can't get out of my head that somebody shared with me. and it's basically look, if you if you look at what all the best companies in the world are doing. So the the best companies in the world are operating in a completely different way than the average company or the median company. And then there's the, you know, the laggards that are just hoping that, you know, they don't get crushed. But if you look at the median company, it's Kind of that, hey, speed stuff up a little bit. These AI tools are nice. They can help me write a little bit faster of a blog post and post the exact same blog posts I would have posted before. and the best companies in the world, I think, are the the mental model that this CEO shared with me was like, I I find the absolute smartest people and I give them huge token budgets and I'll pay them way more than they would have previously been worth. And, you know, you have that force multiplier of AI where people, you know, it's it's one thing to use AI as like a a tool that will go back and forth with you and do some of your work. But if you can figure out how to orchestrate these systems and how to, you know, have AI systems actually in place and doing the work, you know, then I feel like I'm talking to someone who showed up at a tractor and everybody else is manually hoeing the ground. And it's like there's just no like there's no way you can compete. So wh how do I go find all of the tractor people and give them all the tokens? I think it's really, really difficult for someone who has spent 20 years hoeing the ground and perfecting the craft of hoeing the ground, you know, slightly better than the next person to make that shift because it's a fundamentally different thing. I do think that's part of why we're seeing Product builders and engineers adopt AI a little more quickly because it's easier for that type of mindset to abstract. And I mean, it's building software for yourself. It's building systems the same way you've always built systems. but I am so surprised at even like we work with some of the companies that are constantly Vivek Sodera (Supercharge.vc) (30:12.053) Mm-hmm. Austen (30:21.414) pounding their chest about how AI first they are and how on the cutting edge they are. And then you go anywhere outside of the engineering org and it's not even it it looks exactly the same as a year ago, maybe a little bit sped up. And it's it's just fascinating to me. Vivek Sodera (Supercharge.vc) (30:37.386) Yeah, I mean, it seems aspirational. And so it's kind of like a, is this a manifestation of saying you're AI native with the hope that you will truly become AI native within a certain time period. And it's just pure marketing. It's also marketing for hiring. Yeah. Austen (30:52.434) Right. And I get it. Ships are ships are big and it takes a while to shift and it's people based. And like, you know, I know you can't snap your fingers at an enterprise company and have everything shift overnight. but it is yeah, I have been surprised several times to walk into walk into a team and then say, you know, yeah, we're like very AI forward. And then it's like that guy uses Claude sometimes. We're AI forward. And you're like, Vivek Sodera (Supercharge.vc) (31:21.878) Right, right, right, yeah. Well, I mean, yeah, it's, I think if there were some type of user experience that, you know, maybe it's a VR headset or maybe it's shadowing for a day or something, but just experiencing what. Austen (31:22.118) That's okay. that's what we're dealing with. Vivek Sodera (Supercharge.vc) (31:45.043) Excellent looks like is going to absolutely recalibrate your expectations, right? It's just like yeah, we're in AI native We you know, I use Claude to help generate, you know a particular blog post. It's like You're using as a tool. You're not using as an employee, right? That's a there's a particular investment and resource indulgent and indulging of actually putting in the infrastructure in place and having AI employees at every part of the org chart. Austen (32:18.706) Yeah. Yeah. okay, so I think these predictions are fascinating. I think you did probably much better than most people would. And this was it's hard to put this into context of when this was, right? Like now things are so terribly different than I mean back then that was in our first cohort of Gauntlet where we were saying, hey, there are gonna be people that are going to write code without ever manually writing a line of code. Vivek Sodera (Supercharge.vc) (32:19.724) Yeah. Vivek Sodera (Supercharge.vc) (32:34.154) Yeah, yeah, I know. Austen (32:48.688) And we had people like wanted to throw me out of the building for saying something. We had people that like were in the first cohort of Gauntlet and then they heard me say stuff like that, and they're like, You're an idiot. I'm leaving. Literally. it was not But then I'm very interested to know, okay, so you're a you're a VC, you're a product centric guy. you spent a lot of your career in SaaS or stuff that's SaaS adjacent. Vivek Sodera (Supercharge.vc) (32:48.811) Yeah. Austen (33:18.354) For me, when I started seeing all this, I it took me a while before I could like think about SAS as a category again. Like I was just I don't know, you know, the motes are all completely different. Your your moat in SAS used to be, hey, we built the thing. and when that goes away, everything changes and you almost have to but like how did how did you think about that as someone who spent their entire life in Vivek Sodera (Supercharge.vc) (33:34.049) Yeah. Yeah. Austen (33:46.48) software and wildly successfully as a product builder and operator. now now what? How did you r respond to that and what yeah. Vivek Sodera (Supercharge.vc) (34:00.364) The last couple of years have been pretty anxiety inducing. I think everyone's experiencing anxiety. As someone who started their first company in 2006, and then we really started to see SaaS take off in the 2010s, this series of mental models and... Austen (34:05.703) Ha ha. Vivek Sodera (Supercharge.vc) (34:27.052) just patterns around building a SaaS business. And then this AI paradigm shift happened in 21, 2022. For me, moving forward, if I'm being intellectually honest here and I'm looking at what was the defensibility and mode around SaaS, it's the scarcity of what was being built. That is... That was true then of like, you could get away with building a durable business that has defensibility because the thing you're offering is whatever the asset or the resource is that underlies the SaaS UX. The thing that you're building, you're offering is scarce. Now it's not scarce. Saw this to a degree. Austen (35:19.08) When say it, you mean software in general? Vivek Sodera (Supercharge.vc) (35:22.22) the software, user experience, the product, right? So like, you know, we've seen just in the past couple of months, you know, obviously the Airtable acquisition, what happened with Canva, this is like a massive reckoning. I looked up some stats. The average public market SaaS multiple in 2024 was 6X, end of year was 6X. End of last year was about 5x, currently mid 2026. It's about three to three and half x. We're talking market cap to revenue. And so it is. Austen (35:58.152) Crazy. Austen (36:05.765) I mean, you had two decades of the entire ecosystem and infrastructure being built on a ten X assumption minimally, if not more. Vivek Sodera (Supercharge.vc) (36:14.76) That's right. Yeah, that's right. Yeah, that's right. And so, you we saw early signs of this in my last company Superhuman, which led us to join forces with Grammarly and Coda and basically build this horizontal productivity suite. Because as a standalone email experience, yeah, sure, there was a particular moat around the brand itself and the particular, you know, we were anti-slop, we were the antithesis of slop, were perfectionists, craftsmen, etc. But start to see an emerging number of email and other productivity companies just come out of the woodwork pretty quickly. And this is war. When you're building a company and you're penetrating the market, this isn't flowers and daisies, this is war. And you're going to war. and in times of war you shack up with your allies and that's what we did and it was really the next kind of evolution for what we're building as superhuman and it just it made so much sense to do that and it was absolutely the right move because we had we had raised our series C at 825 post, 825 million post. And then we ended up exiting the company slightly above that number, closer to a billion. And we had to grow into that valuation. Like when we had raised that, was... Austen (37:55.208) Yeah, was gonna say that's not that's not the norm. That's like it the best possible outcome. I mean, if you that's top ten percent of all like for sure. Vivek Sodera (Supercharge.vc) (38:00.363) Yeah. Vivek Sodera (Supercharge.vc) (38:04.011) It took a lot of work on the team's part and credit to the team and my co-founders. But yeah, and then you see, you know, was great timing on our part last year. And then you see in today's market, just like this reckoning that's happening. The SaaS reckoning is absolutely, it's this, I wouldn't say, yeah. It's, Austen (38:23.461) Yeah, today so you'd have to be at five hundred million ARR. So Yeah, that's tough. Vivek Sodera (Supercharge.vc) (38:33.449) Because the future value of a SaaS business is unclear. And pre-AI, you can ascribe a future value by just looking at its growth rate and where it is currently and say, OK, look, I can underwrite an investment at this particular valuation because of its expected value in the future. And that's not the case anymore, which is why it's now being viewed as 2 to 3X at best in many cases. And so this is where it's led me to take a step back and be like, application layer investing to me no longer makes sense because you don't have that durability, defensibility as you did previously because there was such a huge barrier to entry to build a product. and then not just a product, but high quality product, and then get in front of customers. Like that act, that right there did entail raising certificate capital, hiring a team, having product shops, having great engineering chops. Now that's been, the bare to entry has just pretty much gone down to zero. And so that's the next central crisis for anyone building in SaaS. I wouldn't say it's dead yet, it's on the steep decline. for sure, but I don't touch that. no longer, I get founders ping me all the time. I see about 500 companies a quarter. I usually invest in one to two companies a quarter. And even this week, I've had four different founders ping me and I just have to reply back and like, look, I'm not investing in SaaS, I'm sorry. And they're like, we understand with how things are in the market. like, you should adapt or you're gonna die. You're not even gonna get off the ground. So it's definitely changed how I invest. Austen (40:02.107) Yeah. Austen (40:21.54) Yeah. And I I'm of the opinion that I do think there will be pockets where SAS survives and thrives, right? Like I don't think SAS goes away. I think there will but I don't think SAS as a category, just like by virtue of being software that people subscribe to, that's way too that's so broad, right? Like that means so many different I mean we we work with companies where it's like, hey, we're the only people that Vivek Sodera (Supercharge.vc) (40:45.025) Yeah. Austen (40:50.224) deeply understand the workflows that truck drivers do when they pull in and like, okay, you're like, that's not as easy to duplicate as like an email client, which, you know, the modality of email is changing fundamentally at the same time, right? That's hard. like I don't I'm a superhuman user. Maybe thirty percent of the time I'm using email, it's via the interface of superhuman now. And it's not because I don't like superhuman. It's just because a Vivek Sodera (Supercharge.vc) (41:19.53) Right. Your needs have changed. Yeah. Yeah. So I think, I think where they're... Austen (41:19.682) I'm within another tool or I mean AI. Yeah. anyway. so so you basically said, okay, this doesn't make sense as a category for me anymore. And it doesn't help that like the prices and expectations of founders at this pre-seed stage aren't dropping in accordance with the market. They're still it hasn't shifted really. So what do you get excited about now? Vivek Sodera (Supercharge.vc) (41:46.133) Yeah. Yeah. Where I had this bet when I launched my fund end of 2024 was around full stack at a prize AI. So you have different layers of the stack. You have the app layer, you have the data ops layer, you have the MLOps, you have the MLOps layer, the data ops layer, and then you have the foundation model layer. And so I had a thesis back then around investing in full stack. There's a SaaS component, but there are other products that underlie that component. An example from my portfolio, I angel invested in Ryder's C plus round. They're now at a 2 billion series C valuation. They're doing phenomenally well. Austen (42:19.76) Mm. Vivek Sodera (Supercharge.vc) (42:37.708) And they have built multiple, it's an enterprise gen AI experience for the no-long-EPD side of organization. So marketing, customer. Et cetera. And so they built the user experience products. They have the data products. They have the MLOps products. They have their own foundation models called Paul Myra. That's for specific enterprise workflows use cases outperform using anything open AI, entropic, open source, open weight models. Austen (43:08.368) So are they building their own models? Are they okay. Wow. Vivek Sodera (Supercharge.vc) (43:11.37) Yeah, since 2021. Yeah, 2020, 2021. They were pretty early. Ever since the Transformer paper came out, we've seen the CTO pretty much got on it. And now he's very much on the bleeding edge of self-evolving models. So they go in and they sell into the Fortune 500. And it took them a couple of years to really figure out and nail the enterprise GTM. But that's not one of their superpowers. But one of the things they do... is they're going into the CFO, they're going into the C-suite and they're saying, you have these seven vendors, replace them with us. And it's a fraction of the cost with less vendor tooling in your stack. And that is winning the hearts and the minds of the different players in the C-suite. And so... Austen (43:59.269) It's interesting, you almost took I mean, that's almost a net short SAS play, right? Like you took the other side of the bet that you were like, this doesn't make sense. However, this is the thing that could crush that. So I'll I'll invest in this, you know, in the replacement. Yeah. Vivek Sodera (Supercharge.vc) (44:13.868) Yeah, Yeah, adapt or die. this is what I proselytize to founders like, look, like you should be building full stack. Back then, I had my aperture open for founders who were building sequentially. I was like, okay, you're not building full stack today. Austen (44:36.004) Hmm. I'll build SAS first and then I'll work my way down the stack and I'll have the ML next and then maybe I'll build models at some point. Yeah. Vivek Sodera (Supercharge.vc) (44:42.06) Yeah, yeah, that's no longer, that no longer makes sense. You have no excuse in today's climate to not build simultaneously. And so the best founders I'm seeing are the ones who are building multiple layers of the stack simultaneously. And as a result, they're raising more capital. And as result, the valuations end up being a bit higher than what they've been, not a bit higher, like quite a bit higher than what they've been historically. But that's one area. Another area I'm starting to, that's emerging for me is the physical AI space and robotics. So I think there's a lot of defensibility there. The common denominator of where I... kind of expanded and evolved my investing thesis. And I call this like my metamorphosis. I sent my LP update last month to my LPs explaining this metamorphosis. And it was like, look, my original thesis was in a nutshell, early stage enterprise AI, productivity and DevTools. Now it's full stack enterprise AI. DevTools has now evolved into software infrastructure. because you can have DevTools that are a bit higher up the stack and that ends up basically just looking like SaaS. So really looking at software infra, which is there's a lot of defensibility there. It's really hard to build. You don't see an abundance of YC founders and they've given batch building. Like I just invested in this email infra company called Primitive and they came out of YC. That was like the exception, but you don't see too many of those because it was, it's a very hard problem space to build in and you can't just buy code. You can get away with buggy app layer user experiences. You can't get away with buggy infrastructure. It needs to be. Vivek Sodera (Supercharge.vc) (46:42.931) performant, robust, reliable, right? Especially if your customer are the big enterprises. So that's an area that I've just kind of iterated towards and then. Austen (46:56.527) So in some ways did you just move down the stack, so to speak, because Vivek Sodera (Supercharge.vc) (47:00.903) No, not necessarily. was just, it was just, well, that's, that's okay. That's a good question and a good analysis. Yes, to a degree it is moving a bit further down from the app layer. And, but I'm not, I'm not a hardcore infra. I don't have a hardware in for background. So for me, it's kind of playing to where I can win and play to where I can operate. Then you go further down the stack and you go to hardware in front. And I see a lot of deals on the data center and compute side. It's kind of out of scope for someone like myself. But what is interesting is this emergence of NeoLabs, which I'm seeing as the fourth mode of companies. Austen (47:26.137) Yeah, that's fair. Austen (47:42.265) Yeah. Vivek Sodera (Supercharge.vc) (47:52.556) for AI talent, the first being startups, the second being big tech, and then you have academia, and then the Neo labs are applied AI research labs where their intent isn't to generate revenue or customers, it's to basically produce products and be on the bleeding edge that would potentially at a later point be, would produce a profitable organization, but their goal in the near term isn't revenue or customers, it's just really around applied research. research. And so that's been interesting to see this emerge. I am starting to. Yeah, I've invested in one. I've had opportunity to invest in multiple. I mean, they're very expensive. But those are an increasing number and opportunity. And then data, AI data. there's as long as the data is differentiated, unique, high quality and scalable. Austen (48:26.725) Do you invest in those? Yeah. Vivek Sodera (Supercharge.vc) (48:51.403) and there's like a proprietary to surround it. There's almost instant product market fit with going to some of the labs and saying, look, I have this really interesting financial markets data, or I have this really interesting robotics high precision tracking data that can be used. because the demand for data has produced so many unicorns and dekakorns in the form of Scale AI, Mercor, and Handshake, and so many others, that that's an interesting shape of company where I'm like, what is the differentiator, the high-qualityness, and the scale here? So that's also a type of company that I think will continue to have a foothold in the market and continue to blow up. Austen (49:42.755) Yeah. There's not gonna be a world next year where all the labs say, we're good on all this data stuff. We're yeah. Vivek Sodera (Supercharge.vc) (49:47.485) Now, now, I the question is that starts to look into like synthetic data. I know there's a lot of talk around that, but conversation for another time. Austen (49:55.13) Right. Fascinating. okay, so what excites you? I mean, if you were to make this list again, this list of predictions, it feels maybe it's because it's backwards looking versus forwards looking. It feels a lot harder to make predictions to me about where the next year looks like than it did a year ago. Part of it is a year ago it was kind of incremental, right? It's like, will there be more engineers or less engineers? Will it be AI programming, pair programming? And now it's like. Vivek Sodera (Supercharge.vc) (50:11.667) It's... Yeah. Austen (50:24.867) Will we have data labs in space that just have a neural net and serve software on demand? You know, but what is there anything that you feel really compelled about in the next little bit? Vivek Sodera (Supercharge.vc) (50:27.827) Yeah, it's wild. Vivek Sodera (Supercharge.vc) (50:43.091) I mean, okay, so. What I'm tracking right now is the org chart within companies and historically the org chart has been from company inception to scale has been the found, you have like particular domains. So you have let's say front end, back end and for DevOps, security, product design, marketing, yada yada yada. And at beginning, it's the founders and the founding team who are, then within a particular domain, you have management and then you have IC. And so the difference between founders, founding team, early stage engineers and late stage engineers is that founders have to jump all over each of those domains at a high level, at a low level, cetera. The founding team, they're going a domains, categories adjacent, and maybe they're jumping up and down diagonally. Early stage employees typically will stay within a particular domain, let's say front-end engineering, and they're going up and down within that particular domain. And then late stage employees typically are staying either at the management layer or as an IC within that particular function. And so we're starting to see these pods emerge. We're starting to see different archetypes emerge that's completely orthogonal to that framing. And so... Yeah, I start, I'm starting to see this where I didn't really have labels assigned to it and then Boris Churney from the head of Clod Code. Vivek Sodera (Supercharge.vc) (52:24.075) He had this tweet about how he's seeing this and he labeled it as Prototyper Builder Sweeper Grower and Maintainer, which I thought was really interesting. It's like, yes, I'm seeing some incarnations of that. The TLDR's Prototyper comes up with ideas. They buy code, they quickly test, maybe get some early PMF signals. The Builder turns a prototype into production. The Sweeper, they refactor, they're doing optimization. The Grower is doing a of GTM engineering, maintaining PMF and then... the maintainers like maintaining and scaling. So I think that's gonna be interesting to see and to track over the next six, 12 months, how companies are evolving their org charts around this potential new paradigm. And a given employee can span multiple roles and the ideal team has a given employee can span two or three of those roles. I think that's really interesting. In terms of where... I'm playing and what I'm seeing... It is not common for a venture fund to evolve its strategy. Like pre-21, 2022, it's like, I have this strategy. I'm investing in crypto or I'm investing in biotech or I'm investing in SaaS. Right. You can't have that thesis for a decade now. And so it's... Austen (53:46.35) Here's my thesis for a decade. You're investing in my thesis. Yeah. Vivek Sodera (Supercharge.vc) (53:57.788) It behooves anyone who is an investor and anyone who is an operator who should be thinking and critiquing and analyzing companies like an investor because their time, time is the most scarce resource. We can't buy time back. So their time is incredibly valuable and scarce. So. whether they're IC engineers, operators, whoever they are within our chart, they should be thinking like investors and evaluating companies in the market accordingly. But it's important to stay on top of and bake in the derivative of the change that is happening. And I've sent this letter to my LPs where it's the best investors likely are updating their strategy annually. And I think in the next few years, the next two or three years, it's probably going to be quarterly. That is my prediction. Meaning this strategy... Austen (54:54.616) Do it does make sense to me. If the half life of everything else is getting cut in half all the time, why would a venture portfolio not have to change, right? Like yeah. Vivek Sodera (Supercharge.vc) (55:05.055) Right, yeah, and so it was very, you know, it's like, okay, this no longer makes sense to invest in this particular area or have this particular strategy. This needs to evolve. And so I've been having catch-up calls or just calls with LPs just to kind of walk them through this. And they're like. Okay, this is interesting. It's like, yeah, know, other VCs that you've invested in, I hope that they're evolving their strategy and staying ahead because complacency is gonna be the death of what they're doing and it's important to continue to evolve. So I think folks, whoever you are in the market should be reevaluating right now what their expectations are, what's important for them, what they look for. all parts of the ecosystem at a minimum once a year but start to develop the muscle to do this more frequently and I would venture to guess the next two or three years like every quarter because of how fast things are changing. Austen (56:06.029) Yeah. Austen (56:11.703) Yeah, you've definitely if there's one been one through line of this entire conversation, it's that you've internalized you you actually have to adapt or die in a way that you may not have had to. I mean, I think about this all the time with you know, we're dealing with people in careers. My dad did the exact same thing when he retired as he did when he graduated from college 30, you know, 40 years, I don't know, what's the time frame, whatever, however many years after. And Vivek Sodera (Supercharge.vc) (56:38.155) Mm-hmm. Austen (56:41.283) I think my job has changed five times, you know, like in completely different ways. And I don't know any but any of my friends or like colleagues, very few of them have just like, this is my track and I stay there for you know for career. Yeah, the the the half life of knowledge that we have is just getting shorter and shorter and shorter. So a lot more is definitionally going to na Vivek Sodera (Supercharge.vc) (56:56.427) Right. Austen (57:10.605) be spent on learning and relearning and adapting and shifting and it it'll be fun. Yeah. Vivek Sodera (Supercharge.vc) (57:18.441) Yeah, I do see some fundamental attributes that have stood the test of time with respect to the dynamics and the changes of what's going on. At a company level, it's really around... what has durability, defensibility, and creates monotonic customer value. then culturally, within an organization, within a company, it's specifically around speed, experimentation, self-driven, like the folks in the organization are self-driven and hiring for that. shipping rate that kind of goes back to speeds and then in today's shape of company, it's this emphasis around e-valves and also everyone being customer centric, not just the customer, not just product or not just the customer success team or the GTM team. It's everyone has to be customer centric. So speed experimentation, customer centricity, like these are, these are things I'm seeing stand the test time. and, yeah, it's, it's something that I look at as an investor, but if I were back into operator mode, I'm trying to decide, is this the company I want to work at? And not just a company I want to work at, but also evaluate vendors that way too. I want to obviously want the best deal for my company. if when I'm trying to choose a vendor, but I also want to have some confidence that a vendor is going to be around. a year, two, three years from now. And the ones that are operating in a particular way with a particular culture, those are ones that increase my confidence, essentially. Austen (59:19.715) Awesome. I think we should should leave it on that. That was fascinating. thank you for spending the time with us, Vec. where can people find you or find I we can see supercharged dot Vc, but what's the best way to to learn more about you or get in contact if if necessary. Vivek Sodera (Supercharge.vc) (59:27.327) Yeah, thank you, Avery. Vivek Sodera (Supercharge.vc) (59:33.428) Yeah. Vivek Sodera (Supercharge.vc) (59:38.623) Yeah, yeah, absolutely. I'm at Twitter, at Vsodera, V-S-O-D-E-R-A, my first initial last name. LinkedIn also, Vivek Sodera. My email is vivek at supercharge.vc. So yeah, those are the best places to find me. Austen (59:54.468) Awesome. Well thanks again. and yeah, good seeing ya. Vivek Sodera (Supercharge.vc) (59:56.438) Alright. Alright. Cheers.