The Twenty Minute VCGuy Podjarny: The Future of AI Software Development - What is Real & What is BS | E1232
EVERY SPOKEN WORD
100 min read · 20,096 words- 0:00 – 0:57
Intro
- GPGuy Podjarny
... I think that's bullshit. Like, I don't think that's, uh, true for a variety of reasons. You know, first, SaaS businesses are far more than just the software that they create. In fact, you have a SaaS business and your only differentiation is, "I've written all this code that nobody else can do," then your- your days are numbered. Expectation of when would, uh, GPT-5 scale models arrive. I now think that those are further away. You've sort of seen Sam Altman and others change their tune to talk about reasoning and things like that as the progress. And my sense is there are actually some architectural problems, so it's possible there's a year or two in which these things are stuck.
- HSHarry Stebbings
Ready to go? (instrumental music plays) Guy, I am so excited for this. Listen, I always learn so much from our conversations, so thank you so much for joining me today.
- GPGuy Podjarny
Oh, thanks for having me back on. Clearly, uh, I said a few interesting things last time, if you were to do it. (laughs)
- HSHarry Stebbings
Do you know, I- I- I always learn from ours. Now, I've got a little bit spicy,
- 0:57 – 2:12
On NVIDIA’s Market Position
- HSHarry Stebbings
and I wanna start with, like, a quick fire on what people said and how we think about it. Now, Masa Son said, "NVIDIA is undervalued today." Agree or disagree, and why?
- GPGuy Podjarny
I think there are really three questions in one in that one. Uh, the first one is, is the market NVIDIA in- is in, you know, going to continue to grow? And I think that is absolute yes. You know, the semiconductors for AI, no doubt, it's gonna grow. I think there's gonna be a lot more, there are gonna be a lot more players in that space. But, uh, I think the second question is how much would NVIDIA capture of that market? I think they are going to continue to dominate. I- I don't know percentage wise, but I think they're doing some brilliant things like taking advantage of their momentary, uh, lead, right? Not that momentary (laughs) , they have, you know, quite a substantial, uh, hard-to-, uh, capture lead, but they know that the clouds, for instance, have distribution advantages, and so they're building a cloud, they're using their semiconductor advantage. So, I think they will continue to be, like, the market leader by a margin for a long time. I think the third question is, what do you think about, what is it now, like, 35X multiple on these revenues? Uh, that one's a bit trickier 'cause it- it really isn't a question of would they grow. The question is, uh, should you put that money into other stock and would they grow faster? That one's a bit harder
- 2:12 – 4:40
Will We See a Trough of Disillusionment in AI
- GPGuy Podjarny
for me to answer.
- HSHarry Stebbings
So, uh, just... The only one I have a question on there is actually the market itself, which I think everyone would actually uniformly be surprised at me questioning. Everyone seems to also acknowledge that actually we're gonna go through this peak of, uh, or trough of disillusionment, that companies will realize that actually the RIO on a lot of these AI tools hasn't proven out in this first batch, and actually, there's gonna be this kind of, uh, depression, so to speak-
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
... in the AI world which will lead to a reduction in demand for NVIDIA chips potentially in the next year.
- GPGuy Podjarny
Uh, well, I think first of all, they sold a lot of commitments already. And so I think a good amount of their revenue is pretty guaranteed. Um, but, uh, I also think that the- the core technology here compounds, it keeps evolving and becoming better and better. So, it's quite likely that they are also the ones best able to produce things in the cheapest fashion or that, uh, achieve output with least compute or other types of savings. It's not just about being able to deal with the biggest and biggest models, which is a part of it as well. Uh, so I think those are just between the manufacturing, the IP, the, uh, the- the processes to be able to deal with that, um, uh- uh- uh, CUDA and the whole kind of development environment on top of it, I think all of those are pretty durable advantages.
- HSHarry Stebbings
Do you think we will go through this trough of disillusionment with regards to enterprises thinking that actually it hasn't delivered the value that it once was claimed to?
- GPGuy Podjarny
Uh, I think so, but, uh, it's not because AI is- is not gonna be as promising as- as it is before, it's just because the numbers are a little bonkers, uh, at the moment, in other-
- HSHarry Stebbings
The numbers are bonkers or the timing with which the value should be created is bonkers?
- GPGuy Podjarny
Well, I think a lot of AI budgets at the moment are- are very, um, uh, kind of non-resilient, right? They're- they're coming in, people are spending money, they're trying things out, and they expect a lot from them. And I think in the long run, that would be correct, but it's hard to think that people will adapt the processes, that companies that, you know, the way that would work will adapt quickly enough to really return something in- in the periods of time. And so there will be some amazing winners, but yeah, I think a lot will- will go sideways. I think the biggest one also is all of these tiny startups where there's like 1,000, I think in some cases there might be even 10,000 companies (laughs) doing the exact same thing. So, it almost doesn't matter how good they are and how good the use case is, a lot of that money is just that redundant, you know? A lot of these companies are building the same thing, uh, and so that would surely, you know, a lot of that is waste.
- HSHarry Stebbings
I cannot tell you how many note takers for wealth managers-
- GPGuy Podjarny
(laughs)
- HSHarry Stebbings
... for dentists,
- 4:40 – 6:51
Is AGI Worth the $9 Trillion Investment?
- HSHarry Stebbings
for do- I'm literally like, "Oh, for fuck's sake."
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
"Not again." Uh, okay. Uh, staying on Masa, he said, "The cumulative cost to achieve AGI is 9 trillion in CapEx, but the benefit would be a shift in GDP to 9 trillion per year." Agree or disagree?
- GPGuy Podjarny
Yeah. I think th- the investment would be substantial, but I think people underestimate the non-technology parts of adopting AI. I think AI is a transformative or a disruptive technology, not a sustaining one, and it requires people to change, it requires, uh, uh, kind of accountability, uh, to change, insurance and things like that. And so there will be a lot of delays on how does society accept these things. I mean, the- the- the reason or the- the holdback for adopting an AI lawyer wouldn't be the tech, it would be, uh, the law, right? Or insurance, about what happens if that lawyer (laughs) kind of gave you the wrong advice, who do you sue? Uh, same with seeing this with self-driving cars, right? You know, it takes forever. Yeah, the tech is a part of it, but really, a lot of it is society. And so, I think there's a lot of investment, but I think there's also a time horizon. So, I- I think the volume is probably a- about right. I don't know, it's hard to assess. Is it 9 trillion or 15 trillion or-
- HSHarry Stebbings
Sure.
- GPGuy Podjarny
... 3 trillion? It's like very large amounts of money. I think the end result would be very much worth it. Um-... the time horizon is much harder to assess, not because of the tech evolution.
- HSHarry Stebbings
I, I mean, Sam Altman said AGI was coming in 2025, so, uh, it's, it's a busy year.
- GPGuy Podjarny
Yeah. Yeah.
- HSHarry Stebbings
I'm drawing my Google calendar.
- GPGuy Podjarny
Yeah. It's, it's, it's a, it's a very convenient thing to say because AGI is very poorly defined, and so you can basically say that and pretty much, uh, claim it is correct (laughs) at any time.
- HSHarry Stebbings
I think it was Tom Hume, our mutual friend, who said to me, and I may be outing him here, but it was very smart. He said to me, "Actually observe the people who say AGI is coming soon," and it's largely, you know, your, your Sams and your Elons of the world-
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
... who need to continuously raise money, "and observe the people who don't need to raise money," which is largely Zuck and Demis, who say, "It's much further away."
- GPGuy Podjarny
Yeah. Yeah. Yeah.
- HSHarry Stebbings
And that is the core correlation.
- GPGuy Podjarny
And I think, yeah, it's just v- I, I think that's absolutely true. Uh, and those, there's incentives. Incentives are not always nefarious. Sometimes it's just your view of the world is biased by the things that you want
- 6:51 – 9:19
Is $100 Billion Necessary to Compete in Frontier AI?
- GPGuy Podjarny
to believe.
- HSHarry Stebbings
100%. Uh, Larry Ellison said, "To enter the frontier model race, you need $100 billion." Do you agree or do you think we're seeing the democratization of cost to enter that race?
- GPGuy Podjarny
Yeah, I don't think we're seeing democratization of (laughs) of costs-
- HSHarry Stebbings
(laughs)
- GPGuy Podjarny
... when it comes to coming to the models. Um, generally, I agree. I think you need vast, uh, quantities or vast amounts of capital to be able to properly compete in the foundation models. There is, there are kind of two, uh, competing theories though. One is that scaling laws will continue to, to show themselves, and as a result of that, you'd need more and more money to make progress. Uh, and the general models, the generic, the, the global models, they will, uh, they will continue to grow and they will eat everybody's breakfast. The counterargument is that you would actually really benefit from specialized models. Uh, and so if the big guns are really trying to get, like, the, the massive models and they're trying to scale and they fail, which there are all sorts of rumors around how the latest sort of GPT-5 and equivalent in Gemini and Anthropic and all that are failing with some tests.
- HSHarry Stebbings
Mm-hmm.
- GPGuy Podjarny
But if you, if that is correct, and they spend a lot of money and time on that, then the ones that are building, you know, maybe code-specific or robotics-specific, uh, models of it, they may have a shot at being able to train with relatively small amounts. Um, I still think that's a short lift though. If you, if you take a five-year, a 10-year time horizon, those are just, like, maybe you, you get some gain for a few years. But over time, like, capital plays a huge role here.
- HSHarry Stebbings
So you would fundamentally say in the short term we'll have specialized models which are better for efficiency and accuracy, but it will lead to longer-term, more generalized models?
- GPGuy Podjarny
Yeah, I think, I think I am m- more willing to bet that the specialized models will kind of prove out in the sort of, call it three-year timeline, uh, than I did before, just because I think the, the big models, the, the big generic models, uh, are apparently, and this is, again, a bit of rumors, um, they are running into some limitations on it.
- HSHarry Stebbings
Did you contemplate building your own model with Tassel?
- GPGuy Podjarny
M- not very long. You know, I think, I think fundamentally you also, you wanna tap into the innovation of the market. You know, you see now Anthropic and OpenAI, you know, some of them might be better at reasoning, which for software development might be important when you design the piece of software. You know, Anthropic is currently perceived as better at the actual code generation piece of it. Other models might be better at visuals. And, uh, and so why choose? You know, I don't want to, I definitely don't wanna compete with all of those, but I also don't even wanna pick amidst them. I wanna be able to use the best technology available to me, uh, at any time. If you draw the cloud analogy, I want to be the best user of the cloud as opposed to actually trying to compete with the cloud players.
- HSHarry Stebbings
I wanna move into
- 9:19 – 11:23
Is Benioff Right to Criticize Copilot?
- HSHarry Stebbings
that kind of sub stack lived beneath it. You know, there is a huge amount of AI dev tools as well. I mean, we spoke about kind of note takers.
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
Fuck, there's a lot of dev tools.
- GPGuy Podjarny
Yeah. (laughs)
- HSHarry Stebbings
Benioff came out fighting against Copilot. He said, "It just doesn't work and it doesn't deliver any level of accuracy." Gartner says, "It's spilling data everywhere and customers are cleaning up the mess." We've got Benioff coming on in a couple of weeks, so that'll be fun. Um, is Benioff justified in his criticism of Copilot, Guy?
- GPGuy Podjarny
Well, I think GitHub Copilot, uh, and rather, coding assistants as a whole, Cursor and GitHub Copilot, it's hard to say that they are not providing value when they are in a place in which a lot of developers say, "This is now the way I develop software. Uh, I don't want to go back to not having them." I do think that it develops software of, uh, questionable quality in many places because you're, you're just kind of reviewing the codes that get generated as opposed to, uh, actually writing it. And so just naturally, the amount of thought you put into it, uh, is, is, is far lesser. And as a result of that, uh, you don't notice anything. And I think, like, L&Ms as a whole, they, they, they kind of average everything, right? Like, you, they, they generate things that unless you give them very specif- specific instructions, the code that gets generated is quite average. So it does help you generate that more. And most code is totally fine being average. Uh, but, uh, it is a bit of a concern that it's duplicating a lot of code that is average.
- HSHarry Stebbings
Why does everyone love Cursor? Everyone loves Cursor.
- GPGuy Podjarny
Yeah. When you think about AI solutions as a whole, if I think kind of broader mental framework, the, uh, the solutions that are easiest to adopt right now are the ones that really don't require any change in how you work today, and they just provide you with some magic, and they don't really need you to trust the result. They just n- just need to work often enough. And so coding systems and Cursor specifically are really, really nice. And Cursor has taken that further in, hey, I'm gonna make changes in various places in the (laughs) in the code and things like that, in the, in multiple files. Um, and they just make it very easy, very easy to eyeball to say, "This is correct. This is not correct." So you're not worried about them doing something incorrect and it's, it's in your flow. You're just
- 11:23 – 14:05
Are AI Dev Tools Actually Any Good?
- GPGuy Podjarny
sort of coding away.
- HSHarry Stebbings
We mentioned kind of the proliferation of AI dev tools and just how many there are. Are they any good? I, I don't mean that ri- ridiculously-
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
... but it's difficult to know.
- GPGuy Podjarny
I think they are providing value in very specific places. Uh, and so where they've been helpful is reducing toil, is when you are trying to do... Toil is this notion of, of just repeated work that you have to do to, uh, uh, decorate and sort of describe your code over time. So creating documentation, creating tests. Um, and so I think they, they help because they... It's kind of like a template for your email, right? Like, you, instead of needing to start from scratch, they give you something to start with. And those have been quite helpful. Um, I think the code completion piece, because of that sort of low cost of verification, they've also been helpful. Beyond that, they haven't been dramatic yet.... mostly because they are still unreliable. Um, they're, uh, you know, we, uh, our, uh, founding AI engineer calls it the jagged edge of AI. You know, it is amazing at some point, and then it bombs. It, like a, a complete disaster the second. Uh, and so it's very hard to-
- HSHarry Stebbings
Are those bombs predictable?
- GPGuy Podjarny
... rely on them. They're not. They're, they're, th- (laughs) you know, there's a, there's sort of two types of mistakes that, that might happen. There's the ones that humans might make and that AI sometimes makes or maybe never will, uh, but humans are, are kind of quite forgiving on those. And then there's mistakes that AI will make that humans will never make, and that makes people mad. That makes them, like, viscerally annoyed if they, if there's a leaf in the middle of the road and the car thinks that that's, you know, whatever, some creature and it's not willing to move, people get upset. They get annoyed by that. And so they say it's stupid, right? Like, this is like, what is this?
- HSHarry Stebbings
A lot of people today say to me, "Oh, we can actually replicate the majority of SaaS companies and actually serve as titan. It's not that complex a product. We can just build it in seconds with, you know, w- co-creation tools and spin it up."
- GPGuy Podjarny
Yeah. I, I, I think that's bullshit. Like I don't think that's, uh, true for, for a variety of reasons. You know, first, SaaS businesses are far more than just the software that they create. In fact, if you're, if you have a SaaS business and your only differentiation is I've written all this code and nobody else can do, then you're probably, your, your days are numbered. Um, so you might have data, you might have, uh, uh, you know, distribution, you might have, um, uh, you know, specific, uh, switching costs as you've sort of built in. That's when you're already successful of it.
- HSHarry Stebbings
Partner relationships.
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
Customer relationships.
- GPGuy Podjarny
I think there's a, there's just a lot more to the machine that is a SaaS company, and specifically a SaaS service over time, um, which is a lot more than just the code. To think that you can just replicate the SaaS business because you can tell an agent what it would do, I think is,
- 14:05 – 18:08
What Is Agentic Development vs. AI Dev Tools?
- GPGuy Podjarny
uh, I think is silly.
- HSHarry Stebbings
What is agentic development? How does that differ from AI dev tools?
- GPGuy Podjarny
So, uh, it depends on at what level do you, uh, sort of delegate a task to the machine. And so do you... Let's say you want a, whatever, an, uh, a shoe e-commerce shop, right? Do you just basically tell that to the LLM and you trust the LLM to go do the product research, uh, explore it, figure out, you know, what are the types of shoes to, uh, to handle in there, uh, build the application, verify that the application behaves correctly, you know, like do all of that process, and, and you're just this slightly less sophisticated, you know, customer (laughs) interacting with this system on it? That's the extreme of an agentic system. There are a lot of decisions there that you don't even know what are the sets of steps that need to happen for that application to be created. You're just giving the instruction on it. Maybe you get a bit more specific. Um, and, uh, and I think companies like, uh, uh, Cognition with Devin are trying to do that. I think Magic, uh, Dev in, uh... they're trying to build it into the model and things like that. I don't know. They're quite secretive, so maybe I'm misrepresenting them. They're trying to build this whole process. Um, and, and I think that's different than most sort of AI dev tools today that, uh, try to give you a bit more control. So they're still assuming that there is a software developer involved, uh, and, and they, uh, give you kinda more tools to be able to create with them. There is a bit of a chasm in between. You know, you kinda have the... you can go to Anthropics artifacts or, you know, to the new OpenAI interfaces and just sort of say, "Hey, create an app." And for small apps it's kind of like agentic behavior, but it's very tiny. And then the question is, how do you break through to something bigger? Do you, uh, do you build a new software development kind of methodology to be able to say, "Okay, let, let's sort of piece together what it is, like let's work together with the machine, uh, to define what is being created"? Um, that's kind of more the, the, the workflow approach, you know, frankly more the Tessell path. And then there's the kind of outsourced, uh, lens, which I think agentic tends to be just embrace the chaos a bit or embrace the customer perspective, uh, and work from outside-
- HSHarry Stebbings
I've seen there be quite a lot of skep- skepticism, I think it was towards Cognition, uh, and their abilities. Is that fair?
- GPGuy Podjarny
Well, first of all, they slightly overplayed their hand with the, uh, with the initial video of it. So it, it just... It's very easy to create amazing videos in (laughs) AI, amazing demos, uh, and it's, it's... There's a big, big gap between that and creating a reliably working product, hence the jagged edge of AI. Um, so I think most of the flack that they got was that they just overrepresented what they're doing. From everybody that I talk to that has tried Devin, they all think it's really cool and that it doesn't really work.
- HSHarry Stebbings
When I look at Magic, I look at Cognition, honestly I even look at Tessell, um, you guys have raised so much money so early. Do you need to or is this not just peak enthusiasm from a venture crowd that has too much money?
- GPGuy Podjarny
Well, I, I think it's, it's twofold. You know, one part of it, and I think that's true for Magic for sure, and I think that's true for Cognition, is that they use, uh, a big chunk of that cash for GPUs and sort of, you know, training models on it. I can tell you that at Tessell, that's not our path. We talked about not wanting to train a foundation model. AI is still costly. You, you know, just, uh, before coming here, the team has run a quick evaluation, cost them 1,000 bucks. You know, they, they... because you need to run statistical evaluations, and it just, it does take more money. Um, that said, you know, I think a lot of these are go big or go home type propositions, so when you're trying to say, "Hey, this will either flop or it would change the world," then you, you wanna have good reserves so you'd be able to, to run the distance, uh, but also, you know, you're, you're trying to be ahead, and right now these systems require a lot of iterations to, to get it right. So, I think there's no single right or wrong. It definitely reduces optionality when you raise a lot. For me, that wasn't an issue, and I'm very much sort of not looking to, you know, make some sort of quick one, two billion dollar exit.
- 18:08 – 21:29
Open vs. Closed: The Future of Software Development
- GPGuy Podjarny
- HSHarry Stebbings
Now, we spoke about some of these products from Magic to Cognition to many of the others before. Uh, a lot of these platforms are pretty kind of, uh, closed off gardens, magic boxes, whatever you wanna call it. How do you think about open versus closed in the future of software development?
- GPGuy Podjarny
Yeah, so I'm actually quite worried about it. So closed environments, closed- development platforms, closed, um...... uh, uh, just, sort of ecosystems like, you know, the big platforms. Um, all of those, they, uh, when they add AI, it's a little bit easier to think about how to make it useful, how to live within that. And eventually, they create these magic boxes and you have no ability to interact with it. And I think to me that's a concern, right? If the web becomes two, three, four companies in the world that have these very, very powerful large compute models, uh, to which you can give an instruction and get a result, and you, you really have no ability to build tools that would plug into them or modify some things, uh, with the-
- HSHarry Stebbings
So why would that ever be the case? Like, if you look at, like, an OpenAI, f- for example, they are clearly doing a platform play where they are not expected to build all the different tools around it. They are fully trying to develop an ecosystem and it will be multi-participant who service very different fashions of the customer base.
- GPGuy Podjarny
I think the, uh, the platform themselves actually want to be... The LLM foundation models actually want to be platforms 'cause they want everybody to build on top of them.
- HSHarry Stebbings
Yeah.
- GPGuy Podjarny
So I don't know how deep they will go into the application layer. I think that's a debate for them, right? You definitely see OpenAI getting into the application layer and the question is, how far do they go? Um, I think the, the next layer of the big companies are the ones that are more concerned about, you know, whether you... Think, think about the cognition, they're not doing that yet, right? (laughs) Like they're, they're still at the beginning of it, right? But if that model works and you just give it the instruction and that succeeds, and again, they're building great stuff, what happens next? You're sort of building, uh, building those models... Uh, sorry. You're building these platforms and you're just giving them the instruction and they go off and they, uh, they build it out. How does the rest of the tooling ecosystem plug into that? If the core of software creation becomes dependent on this one magical understanding of your app, of your code, of the applications, of all the domains, and you just become a customer to that system, I worry that we go into a place in which there are a small number of players that have that broad capacity, uh, and that the rest of the dev tooling ecosystem becomes kind of minor and delegated versus today's world where we have many thriving developer tooling companies.
- HSHarry Stebbings
How likely is that, the concentration of these platforms being the sole providers or kind of few providers to really dominate this space? Is that like a 10%? Like a 50%?
- GPGuy Podjarny
Um, I, I think there is a, a pretty high chance that we're going towards this path. Ease of use of software development is already, you know, amazing and you're seeing platforms like, uh, Vercel for instance, you know, make it very easy to generate applications and go end-to-end with them. And that's great because again, they represent a subset of the ecosystem, a specific type of application. Uh, when you think about GitHub now for instance, going to actually deploy applications, they're, they're going from the ADE to the deployment and they're also the engine that writes much of that code. I think that's a pretty high probability. Uh, and the more you lean into the agentic magical creation of it, the less control that the developer
- 21:29 – 27:15
When Will Enterprises Move Beyond Experimental Budgets?
- GPGuy Podjarny
has.
- HSHarry Stebbings
How long do you think it will be before we move out of the experimental budget phase for large enterprises?
- GPGuy Podjarny
Uh, I think it's incremental. So I think assistants are already providing value today, and some of them are moving out of the, uh, the experimental budget into real budgets. I think the resolution ones, the ones that are like true outcome-oriented, autonomous activities, I think they're in very specific fields. Support that we talked about right now, I think is, is already real, uh, because, uh, it is a, it is a, a lookup environment. Um, I think to an extent the SDR ones, although there's like a lot of blast radius, uh, (laughs) over there. If you mess that up, uh, you could really hurt your brand. Uh, but I think some of those sales tools are getting out. I'd say they're probably still somewhat experimental. Mostly I would say, uh, small... M- mostly it's the assistants that are getting the dollars. I think the rest are probably a couple years away. And as I said, I don't think the LLMs are the limiting factor. I think it's more the processes and the systems around them to make the LLMs more reliable.
- HSHarry Stebbings
You said the assistants are the ones getting the dollars. I, I recently interviewed Sam Altman at OpenAI's dev day. Um, how... We said earlier, um, how far do the models go in terms of into the application layer? It's pure speculation, it's unfair, but how far do you think they go in terms of entrance into the application layer?
- GPGuy Podjarny
I think as long as we are dealing with the same data and just a different way to interpret it, I think they will go all the way. And so search is a good example of that, right? There's no reason why they wouldn't go all the way to actually replacing web search because it's the same type of data, uh, that they need. They don't need a different type of expertise. Um, to an extent, user experience of code generation or image generation might fall under that bucket as well because it's the same data, it's the same ingested data, uh, that they would go into. And so I think those domains are there. I think when it becomes different data or very dedicated elaborate workflows, like it's not about the magical make a call, get back a response, but a, a whole system around how do you process those, I think that's outside of their domain and wouldn't make sense for them because they, there are... It doesn't really threaten their core business and they actually would be more better off if many, many different applications were built on top of them.
- HSHarry Stebbings
Okay. Y- you have three options, right? You got Chat... or OpenAI at 150 or 160, you have Anthropic at 40, or you have x.AI at, I mean, the latest pricing says it's 50.
- GPGuy Podjarny
Yeah, I think-
- HSHarry Stebbings
You can only buy one. Which one do you buy?
- GPGuy Podjarny
I think I'd go Anthropic. Uh, I feel generally the, uh, the qua- the... It, it still holds to the quality of the team, uh, or maybe, uh, you know, the stability of the team and their opportunity to grow. Uh, and yeah, the value arbitrage is, uh, uh, it's not hard.
- HSHarry Stebbings
Do you not think it's nuts, the safety and alignment team moving so much out of OpenAI?
- GPGuy Podjarny
I don't know what's happening inside of there, but it's hard for me to imagine that the organization is not suffering substantially from the churn in leadership. Even, even regardless of whether the reasons are legitimate or not, having such a big turnover must really rock the boat.
- HSHarry Stebbings
Final one on this, and I... You said about search kind of being an obvious one for them to do, and obviously they've got, you know, GPT-Search I think it is. Um, would you rather invest in Perplexity at nine or...
- GPGuy Podjarny
Or, or...
- HSHarry Stebbings
OpenAI?
- GPGuy Podjarny
... um-
- HSHarry Stebbings
Like, who will win the search game there?
- GPGuy Podjarny
Yeah. I still think Perplexity is not going to, uh, to go as far as people think. Um-
- HSHarry Stebbings
Why?
- GPGuy Podjarny
Mostly, not because of OpenAI but because of Google. I think people, again, we, I think we underestimate-
- HSHarry Stebbings
Because of Google?
- GPGuy Podjarny
I think we underestimate how hard it is-
- HSHarry Stebbings
The- the self-driving car company?
- GPGuy Podjarny
... e-eh, it, we underestimate how hard it is (laughs) to, uh, change people's habit. Uh, and I think Google is so deeply ingrained and have the distribution around, uh, around search. And Google is generally pretty terrible at product, uh, on many fronts, but the one product in which they actually do a pretty good job (laughs) is the search product, and they have all the traffic.
- HSHarry Stebbings
I don't understand what you mean. I- I- I l- I think the world of you, you're smarter than me, you're an engineer, I'm not. The- their product sucks compared to Perplexity.
- GPGuy Podjarny
It does, it does. I think Perplexity's, Perplexity's a lot better.
- HSHarry Stebbings
And like, my mother now uses Perplexity over Google. I think we underestimate how fast people can move-
- GPGuy Podjarny
Yeah.
- HSHarry Stebbings
... if the ease of moving is- is bluntly very obvious.
- GPGuy Podjarny
Yeah, you might be overestimating how many moms have, you know, a son who is a tech podcaster (laughs) -
- HSHarry Stebbings
(laughs)
- 27:15 – 28:28
Why Would Companies Like Snyk Choose to Go Public?
- GPGuy Podjarny
acquire a lot more easily.
- HSHarry Stebbings
Why would Snyk, or companies like Snyk, ever go public with the extension of private capital and private markets meaning you don't have to?
- GPGuy Podjarny
Well, I mean, in part because if you want to build a long-term sustainable company, then the reality is that in vast majority of cases, you- you should go public. Uh, that gives you some brand recognition, it gives employees constant liquidity, uh, it gives some enterprise customers some, uh, assurance around being able to see your financials and work with you. And so eventually, you have to do it. When? That's a different question. That's really, uh, you know, that- that's math, where you compare all of that to the, uh, heavy toll, uh, that- uh, that is public companies.
- HSHarry Stebbings
Would you like Snyk to be public now?
- GPGuy Podjarny
I would like Snyk to be public at some point. Uh, I- I, I, whether now or not is a constant conversation.
- HSHarry Stebbings
Do you miss being CEO?
- GPGuy Podjarny
Uh, well, that's why I went back to Tesla to be CEO. I, uh, I do. I think it's tricky handing off the sort of, the CEO reins. I think there was a lot of great implications of that that came, uh, to Snyk because of that. I think Peter is incredible. Um, but, uh, it is- it is hard to transition, and it's hard later on to sort of find your place in the company a bunch of years later.
- 28:28 – 30:29
How Will the Role of Software Developers Evolve?
- GPGuy Podjarny
- HSHarry Stebbings
Listen, I do wanna go back to some form of structure, 'cause we- we've mentioned all- kind of the change in how products are created, the changing structure of teams. What is the future role of a software developer, though, when we look at, like, assistants, but then also code completion platforms? Like, how does the role of a software developer change?
- GPGuy Podjarny
Well, I- I think the best software developers are not the best because they're the best coders. Uh, it's because they think about development as a whole. They are systems thinkers, they, uh, understand the important bits about, um, about, you know, the- the requirements of it and they emphasize around it, they can anticipate trade-offs and what will happen later. And in fact, the progression path that oftentimes developers most want is- most like is the architect path. And architects don't write that much code. Uh, and so I think the coding piece of software developers' work will diminish substantially. I think in 10 years time, coding would be, uh, uh, very much alive and well, but it would be the edge case. It would be the thing that you do because, you know, the- like, it- it's something where you have to be close to the bare metal, uh, or some older technology. Um, I think most developers will either progress up the architect path, and they will really invest in those trade-offs and system thinking.
- HSHarry Stebbings
So- so what does that actually mean, go up the architect path, for someone who doesn't
- NANarrator
(clears throat)
- GPGuy Podjarny
Yeah-
- HSHarry Stebbings
... know very much?
- GPGuy Podjarny
... well, it means that they will think about s- uh, software and sort of systems in slightly bigger paths, and they will make some, uh, decisions. Every time you build a system, you, uh, have to trade off, for instance, um, uh, how extensible it would be versus how simple it would be. The more extensible it is, the more complex it is. Is it important? Is it not important? Uh, you can, uh, super optimize it to a specific environment, say, like, AWS, or you can say, "No, I want it portable between the different clouds." So all of those are architectural decisions that have implications downstream, and I think they will continue to be important because the- the L&M doesn't know, the humans don't know either, but they can make an assessment on...... uh, what is more likely to change with the software in the future.
- 30:29 – 34:54
How Many Devs Should Move to Architectural Thinking?
- GPGuy Podjarny
- HSHarry Stebbings
The challenge is, with that transition up to kind of more architectural thought processes around core strategic decisions, like, you know, which architecture are we doing? Which trade-offs are we making? That seems to be, like, a progression towards leadership style thinking, and there are, by definition, only so many big decisions or leadership positions to be had. My question is, like, does everyone progress up the architect path? 'Cause you don't want 20 devs all going, "Well, I've got this architectural opinion."
- GPGuy Podjarny
Well, I think you do because you are creating software, uh, like substantially faster-
- HSHarry Stebbings
Mm-hmm.
- GPGuy Podjarny
... and so you will, you will have more software for which you need to make those decisions. I, I gave you a few examples here, many decisions are more immediate. I do think there's another progression path for developers that is more towards the product path, more understanding of user empathy, uh, uh, whatever the Henry Ford, uh, you know, "Don't, uh... If, if, if I did what customers told me, I'd build a, uh, a faster horse."
- HSHarry Stebbings
Build a faster horse, mm-hmm.
- GPGuy Podjarny
And so, you know, there's examples like that that are more towards the product manager. And so I think those two paths, they evolve up the, kind of one of those routes, and I, I do think that we will produce more software than ever, um... I think, uh, uh, consider the web of whatever, like, 2000. It was all boxy and, you know, messy, and to build a very fancy website, you know, of, like, what we consider today even a legitimate website, it would have taken a ton, a ton, a ton of work. Um, and-
- HSHarry Stebbings
And money.
- GPGuy Podjarny
... and money, and expertise, and it was very, very hard. And today, it's much, much, much easier to, uh, build software that, you know, creates a pretty website. And therefore, expectations of consumers have changed, and now we expect websites to have that level of functionality, performance of website. We used to be much more tolerant of latency. Now we don't have those. And so some of it is just the systems will allow us to create software better, right? Will give us more leverage. Developers will have more leverage in their ability to, uh, produce high-quality software. Uh, and consumer expectations, business expectations, will increase accordingly.
- HSHarry Stebbings
How does the role of PMs change? We have a lot of PMs that listen.
- GPGuy Podjarny
Well, I, I think part of it is, they will be more autonomous and maybe there will be some blurriness of the lines between the product manager and the, uh, the software developer, because oftentimes, the product manager will be able to say things. But product manager is like a very ill-defined job (laughs) and some PMs are, uh, quite technical and they're close to the software. I think those roles will definitely merge. Uh, and some PMs are very much the conduit between the technology and the user, and the understanding of those users. And I think the role of those individuals will actually stay fairly similar. Like, they spend most of their time with users, with customers, observing what happens, trying to kind of rationalize those. That would probably be AI-assisted as well, they'd be able to kind of get some support. But tho- that's a, a decisions and an accountability, uh, uh, position that, uh, will continue to need to remain humans, because we want humans to make those decisions.
- HSHarry Stebbings
So if we project forward then, say, five years out, and we think about that evolution of the role of the PM, and the evolution of the role of the software developer, what does the structure of a technology company look like?
- GPGuy Podjarny
Uh, uh, it's a good question. I don't, I don't know. Uh, I think there are actually-
- HSHarry Stebbings
You're not, you're not a VC, are you?
- GPGuy Podjarny
Uh, yeah. (laughs)
- HSHarry Stebbings
Y- you, we make bold statements without knowing.
- GPGuy Podjarny
Yeah, yeah.
- HSHarry Stebbings
That's the, the core tenet of beginner's startup.
- GPGuy Podjarny
Yeah. I, I think sometimes, I like, I like to think about first principles, you know, for these things. And so what would we still need? Well, we would still need, uh, uh, to make choices between multiple options. So you'd still need a, a decision power arbitrage, choices that try to say, "This is my strategy, this is my business, and therefore this is what I need." And so you'd need that at multiple levels, at the business level, and someone that can translate that into the sort of the product side of it. Uh, you would still need, uh, people responsible for the system running, for keeping the system running. And they would be probably, their span of control would be much bigger. And so I think those would be the, the development teams that can produce more software. Software would be more, uh, uh, adaptable. It would be, um, more personalized even. But I think you'd still have the strategy, you'd still have the, uh, the developers, uh, and, you know, you have, you have teams, right? And they're sort of building together, and so naturally you'll need management. So I don't think a lot of the structure will change dramatically. I think the roles of the different positions, the sort of the scope of what they can do, uh, are the things that will change.
- 34:54 – 36:21
Why Is Security Becoming Harder with AI Dev Tools?
- GPGuy Podjarny
- HSHarry Stebbings
Security is kind of hotter than ever in, in the amassing landscape. Um, and the threats seem to be more prominent than ever. Moving forward in this world of code completion, AI dev tools, why is security an ever-more-increasing problem?
- GPGuy Podjarny
The real challenge with LLMs right now is that we lose all control. Uh, we, we make a request and we get something back, and we hope that it is correct, and we're reliant on a, a very kind of unreliable review of what was produced to say that it's correct. Uh, and that's specifically happening in software a ton. There's a lot more unreviewed or poorly reviewed code that gets deployed. And, and I think that's a real security risk. The other thing that's happening is people produce software, like a lot of these coding systems, they create code, but they don't maintain code. And so that code will stick around, you know, technology never dies (laughs) . It'll, you know, be another entry point for attackers, uh, and code rots over time. These systems will become, uh, uh, more prone, you know, to, uh, to, to attacks. And so I think the, the problem is really this messiness. We have to figure out layers of control to be able to say we, alongside the creation of software, we have to think about maintenance, we have to think about ownership, and we have to be able to have guardrails at various points to, to say, "This thing that we created, is it, is it any
- 36:21 – 39:21
Spicy Questions
- GPGuy Podjarny
good?"
- HSHarry Stebbings
So we're gonna move into a very interesting new round. Uh, I essentially have some super spicy questions here.
- GPGuy Podjarny
(laughs)
- HSHarry Stebbings
Um, and you have a choice. So you can either donate to a charity of your choice, uh, or you can answer the question. What would you like the, the donation to be per...... card. (laughs)
- GPGuy Podjarny
(laughs) Per card, uh, I don't know, uh, a thousand bucks?
- HSHarry Stebbings
Okay, let's do a thousand bucks per card. This is, this is good. (laughs) So these were submitted by your friends as the spiciest questions.
- GPGuy Podjarny
Well, I'll know who to come back to after this. (laughs)
- HSHarry Stebbings
How much liquid did you take out from Snyk?
- GPGuy Podjarny
Uh, I sold about a third of my ownership, and it's in nine figures.
- HSHarry Stebbings
Got you. Fucking boom. Um, what were the acquisition offers for Snyk, and how much were they for?
- GPGuy Podjarny
Oh, um, I don't know if I can disclose all the numbers, but we had more acquisition offers early in the sort of, when we were still sub-one billion, uh, where people thought they could get us cheap, and I wasn't interested in selling. Um, and I think after that there was a lot of dancing, but generally the investors were a lot more bullish than the acquirers. You know, there being a little bit of sort of the... (laughs) You know, the hype of it. Uh, and so I think investors were always ahead of the numbers. I think the last acquisition that really I sort of felt like we were talking in, in concrete numbers was, was early in the journey, was in sort of 200 million range. And I think later on we talked about multiple billions, but we would... That they didn't get to, like, concrete, "These are the numbers," because we were always well ahead of them.
- HSHarry Stebbings
We're smashing this guy. Uh, I'm enjoying this. Ooh, I hadn't seen these also, so I literally got beaten with these before.
- GPGuy Podjarny
(laughs)
- HSHarry Stebbings
Uh... Oh, I like this one. What was the worst investor meeting you ever had, and why?
- GPGuy Podjarny
Ooh, the worst investor meeting. Um, like, b- trying to fundraise or, uh, or with existing investors?
- HSHarry Stebbings
No, no, trying to fundraise.
- GPGuy Podjarny
Yeah. I have a clear one, I'm debating whether I should name the, uh... So it was a, a firm that has since disbanded here in the UK.
- HSHarry Stebbings
Oh, you're fine.
- GPGuy Podjarny
It was a series A, uh, series A investment firm. And I, I sat down there, uh, I really liked one of the partners, he ended up sort of, uh, investing in my, uh, in my series A, uh, like as a, as a small investor. But then the other two were just on their phone, uh, you know, just felt like they really couldn't care less that I'm in the room. I had high demand at the time, (laughs) like I really didn't need them on it, and that was just, uh, insulting. So, I think to me, the worst thing is to have, to be in a room with someone you bothered going to and have them not pay attention to you.
- HSHarry Stebbings
(laughs) I love Mike Schaufane. The joys of my job is I can say things. Uh, okay, so I wanna do a quick-fire. I
- 39:21 – 48:29
Quick-Fire Round
- HSHarry Stebbings
love that, that's such a good addition.
- GPGuy Podjarny
It was fun to, wasn't it?
- HSHarry Stebbings
(laughs)
- GPGuy Podjarny
As, as... I don't know, like, I'm a pretty transparent guy, so I'm not that, uh... (laughs)
- HSHarry Stebbings
You're very, very... I, I was not expecting that. Um, that was great. You know, we're t- totally doing that moving forward.
- GPGuy Podjarny
I'm gonna donate some anyway. (laughs)
- HSHarry Stebbings
Uh... (laughs) Okay, so quick-fire. What have you changed your mind on in the last 12 months?
- GPGuy Podjarny
I think maybe the, this, like, expectation of when would the GPT-5 scale models, uh, arrive. I now think that those are further away. Uh, you, you've sort of seen Sam Altman and others change their tune to talk about reasoning and things like that as the, as the progress. Uh, and m- my sense is, there are actually some architectural problems, so it's possible there's a year or two in which these things are stuck.
- HSHarry Stebbings
What in AI does no one focus on that everyone should focus on more?
- GPGuy Podjarny
I think everybody's very focused on the short-term gain, on these kind of low friction, you know, just optimize my existing workflow, and people are really not imaginative enough. They're not trying to think about, "If I stop and think about this domain from first principles, which things just shouldn't be here in the first place?" Uh, I wish I would see more of those.
- HSHarry Stebbings
What's the most contrarian or unorthodox advice for founders listening?
- GPGuy Podjarny
I think founders should look for ideas that they need to work to convince people, uh, are useful. Like, you need to find something that people don't immediately believe. Uh, because when you, when you pitch them things and everybody's like, "Oh, yeah, that would be awesome," there are basically a thousand other companies figuring out the same thing. And you have to have something that's a little bit harder to understand that really requires some depth of thinking or a different angle, uh, and that is very valuable if you're successful, if you're, if you're right about it. Uh, otherwise, you're just gonna get lost. It, it doesn't matter how good you are. If there's a thousand other companies doing the same thing you are doing, it's going to be very, very hard for you to succeed.
- HSHarry Stebbings
Do you like competitive markets? I just had this debate internally with the investing team, and I was just, "I don't wanna be in the 50th note-taker for wealth managers." And they're like, "Yeah, but that shows that it's a space worth going after."
- GPGuy Podjarny
Yeah. I, I don't love them. I, I think it depends on the levels. I think a market of one is not good, because, you know, often- oftentimes there's a reason, the reason that... So, th- there's nobody else in that domain, and you have to be careful. Um, I think a market of five or 10 companies is actually pretty good. That's healthy, everybody's pushing along, people are, you know, raising awareness for you. I think when it's a market of 50 or 100, again, you could be amazing, but it's just everybody gets a few of the dollars in the market, and you have to really be kind of heads and shoulders above the rest to be able to break out. Um, so I, I don't love those. If you are going to invest in those, you really need to think about massive operational excellences. The equation changes, and you need to not just find the best product teams, but you need to find the best sort of company builders, maybe the best GTM builders on it. You need different perspectives.
- HSHarry Stebbings
How much have you raised now for Tassel?
- GPGuy Podjarny
Uh, 125.
- HSHarry Stebbings
125. When you have 125 million sitting in cash, the interest rate is, is quite high.
- GPGuy Podjarny
(laughs)
- HSHarry Stebbings
How do you approach that strategic decision of where you store cash?
- GPGuy Podjarny
... um, a- I think we're pretty, th- there's really high interest right now, so we don't need to get sort of too fancy around what we do. We get high interest rates, uh, as is. We don't need to invest it or, uh, or risk losing.
- HSHarry Stebbings
But with that, do you not have like $10 million a year in interest?
- GPGuy Podjarny
Uh, so you definitely get into somewhere between that sort of five and 10, right, if you're sort of massively, uh, massively, uh, um-
- HSHarry Stebbings
Which is basically-
- GPGuy Podjarny
... uh, conservative.
- HSHarry Stebbings
... your burn.
- GPGuy Podjarny
Which is, you know, definitely sort of at the burn level. I think the, the reason that we raised the large amount is that we, uh, we really have the ability to build for the long run, uh, and to invest when we get the thing that is working, alongside just having really, really good partners. Like I love Carlos, who now joined the round and Index is a great investor. And so I, I, I think the, the insulation from the craziness of the market, and as we discussed, you know, the potential bust of, uh, of, uh, valuations or realizations, uh, really allows us to think long term. I think eventually urgency comes from markets timing, not from runway. Uh, and so I would be concerned if, uh, if the-
- HSHarry Stebbings
What do you mean by that?
- GPGuy Podjarny
Well, so as an angel investor, and I've, I've made maybe like 100 angel investments over, uh, over the last decade.
- HSHarry Stebbings
Dude, you're literally the best fucking angel I've ever worked with.
- GPGuy Podjarny
(laughs)
Episode duration: 48:29
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode 3NROBOxVMLA
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome