Siddhartha Ahluwalia 1:02

Hi, this is Siddhartha Ahluwalia, your host at Neon Show and managing partner of Neon Fund, where we invest in the best of enterprise AI companies between the US and India corridor.

Today I have with me, Vasanth Namasivayam. Vasanth, welcome to the Neon Show. Thank you so much for having me, sir.

Vasanth, you’re the founder of Featurely. Neon Fund is very proud to be a recent investor and having co-led the seed round in Featurely.

And the outcome of that is that you are supercharging product development. You previously led AI initiatives at Meta, Dropbox, NVIDIA, you know. And I can very proudly say that, you know, you are one of the most intense and humble founders that I have ever met.

Vasanth Namasivayam 1:53

Thank you so much for your kind words, Sid. I mean, I’m truly excited at the opportunity both in building Featurely and in partnering with you all very closely as we look at the next generation of products out there.

Siddhartha Ahluwalia 2:07

So the first part of my conversation that we want to structure today is, you know, your experience working with these organizations, NVIDIA, Meta and Dropbox. They are some of the leading organizations today in the world, right? So what attracted you to join these companies?

Vasanth Namasivayam 2:28

Absolutely. I think different stages of my life, I think different things were more attractive. Starting off with NVIDIA, I think what I was really attracted to was just the vision of the founder, vision of Jensen and the reputation around NVIDIA.

Keep in mind that this happened back in 2015. So this was well before NVIDIA became what it is today. Back then, NVIDIA was extremely prestigious in the semiconductor world.

And it was starting to show a lot of possibilities, both for AI workloads and for Bitcoin mining. But I think what attracted me was how futuristic the company was, the kind of stuff they were doing and the vision of the founder. The second thing which attracted me was, I knew a bunch of folk who were already working at NVIDIA.

So the culture was something which was extremely attractive to me. NVIDIA is one of the few companies in the Bay where the average employee tenure is probably closer to double digits than single digits, at least when I was working. Primarily because of the loyalty which the leadership team and Jensen engenders in the entire workforce.

So that’s kind of what attracted me to NVIDIA. When I think about Meta, that was a bit later. And this opportunity came across where I would get to work on Integrity, which was something which was fascinating for me.

Because of everything which happened during 2016, all the misinformation which was coming out, and the very polarized world we were living in, the ability to kind of work on a problem like that was extremely enticing. And finally, when I think about Dropbox, the opportunity which was presented to me was fashioning or kind of rethinking how machine learning and AI is relevant to core Dropbox, and how we could think of using AI in the future of Dropbox. It was a very, you know, potentially very large scope.

And, you know, that attracted me to that opportunity. I was like, you know what, hey, I’m gonna give it a shot. And that’s what got me to each of these places.

Siddhartha Ahluwalia 4:50

And what have been your unique learnings across NVIDIA, Meta and Dropbox?

Vasanth Namasivayam 4:55

When I think about NVIDIA, I think NVIDIA is one of those few companies in the world, which builds for the future. I think that’s true for most semiconductor companies because of the lead time required to build things and the expense involved. But NVIDIA was amazing at that.

So they imagined a world where, you know, you kind of needed the complexity of semiconductors and the entire stack, even back in 2016, which is why NVIDIA is at the forefront of technology even today. AMD and others are just catching up. So big learning there was building for the future, you know, envisioning a future, building for it, and not worrying about just your day-to-day competition.

When I think of the second big learning I had at NVIDIA was cultural learnings. People just loved working for Jensen. We worked really hard, but the culture was a very intellectual culture where we were all proud to work with each other.

And we just loved the direction and the vision of the founder. When I think about Meta, I think the biggest learning was, so I worked in the integrity team, and I worked during a phase where we dealt with COVID, George Floyd and Black Lives Matter, and the US 2020 elections. We were also living in very polarized times.

So when I think about my tenure at Meta, I learned to move really, really fast because of everything which was happening in the outside world, but also learned that in spite of your best intentions, it is not always true that you will satisfy all stakeholders, you know, your customers, your advertisers, and whatnot, just because of how complex some of these problems are. And these problems have no right or wrong answer. And finally, I think about Dropbox, my biggest learning was, if you put yourself in a position where you are not leading the market, and rather you’re trying to reinvent yourself, you are reacting to the environment, you need to execute really, really well, else you end up, you could potentially end up being in a situation where there’s a lot of noise, and there’s a lot of chaos, and not the good kind of chaos. But yeah, those are the kind of learnings as I can think across those organizations.

Siddhartha Ahluwalia 7:35

And just to summarize, you know, NVIDIA people loved working there, because it was futuristic, they were building for the next 10 years. And Jensen could somehow predict that, you know, the kind of infrastructure that we would need in 2025, NVIDIA was ready for it.

Vasanth Namasivayam 7:53

Absolutely, absolutely. I think even things like communication speeds on chip interconnects, like we were going way ahead of what the rest of the industry was doing. It seems like a small little thing.

But that made all the difference when it comes to working with the kind of workloads we work with today.

Siddhartha Ahluwalia 8:14

What makes Meta so special? One is obviously a move fast and break things, speed of execution. You mentioned about the product obsession at Meta, but there needs to be something else that gives Meta its market leadership position, both in stock market as well as, you know, in the global markets.

Vasanth Namasivayam 8:37

Absolutely. Again, at Meta, I would say the brightest minds get attracted to Meta. Or back then it was Facebook.

Like, during that limited time I was at Facebook, I could honestly say that every day I’m learning something new from my colleagues who are all smarter than me. And I got to learn so much from them. So the, you know, the teams, the people, the intellectual power, and, you know, the cohesion was like really, really big at Facebook.

Second big thing, which I think helped them win was, they had kind of built the market leadership position in that entire network. So they understood humans better than most other people when it came to, you know, from an advertiser’s perspective. And so I could see a lot of value in what they were giving the advertisers compared to a lot of other companies out there.

So that’s kind of what I think. And then finally, again, the founder is obsessed. He is a one-man army.

He is crazy in terms of, like, what they’re able to accomplish. Because his vision is, like, you know, the driving force. When I think about the time they shifted from Facebook to Meta, the entire VR, the entire Metaverse was a lot of his driving force behind it.

It’s incredible how much, you know, energy this person has to kind of make those kind of company wide shifts. Final thing, which I would also love to point out about Meta, was we had this concept of town halls. And I think the founder’s willingness to be extremely transparent and answer all the hard questions, you know, it was really good.

In a lot of organizations, leadership tends to be sacred cows, where you can’t ask the tough questions. Not so for Meta. And I think that was also a crucial part of their success.

Siddhartha Ahluwalia 10:55

What I understand from you is, in Nvidia, Jensen decided what problems to solve. But in Meta, what problems to solve was, was it Mark deciding it, or was it more democratic?

Vasanth Namasivayam 11:09

I think it was a combination of the two things. I think there was a lot of democracy. And obviously, the leadership team as well, had, you know, certain directions, where they want to take the particular product.

So I think it was a combination, it was a healthy balance between the two things. We had wide guardrails. And as long as we were within these guardrails, I think, you know, the democracy worked very well.

Siddhartha Ahluwalia 11:37

And Dropbox was, at one point in time, the most viral product of its time, right? Why do you think the company was not able to reinvent itself when the world changed?

Vasanth Namasivayam 11:50

I think it wasn’t for a lack of trying. I think Dropbox has kind of worked on multiple products, has acquired a lot of companies to create, you know, multiple business lines. But, you know, I was having this conversation recently with someone, and they were like, there are very few companies in the world who have actually been able to stand up individual unicorn line of businesses, like Amazon is one of those exceptions.

I think in Dropbox’s case, they found the going tough. And they did the, they made a bunch of strategic bets with the right intentions. But those bets didn’t pay off.

It comes down to, you know, sometimes, you know, macroeconomic factors, all of your control. It also comes down to execution. And the amount of noise in the system kind of changes the probability of success.

Siddhartha Ahluwalia 12:49

The unique thing about all these three organizations is the founder still runs the organization.

Vasanth Namasivayam 12:55

Yeah, and I think that’s one of the reasons why, and perhaps, you know, you bring up a common theme on what attracts me. Each of these founders have very strong visions in their own way. And it’s a joy to work for something bigger than yourself.

And I think these founders were capable of delivering on it. As compared to professionally managed companies, I mean, they have their own value. But it’s very different from founder led companies, the vibe, the energy, everything.

Siddhartha Ahluwalia 13:27

Now you’re running your own startup. Do you think the intensity that you are running your own startup was the same intensity across these three organizations?

Vasanth Namasivayam 13:36

Yes, I think so. Maybe not in the same dimensions. I think intensity, multiple different dimensions.

But I would argue that the common thread among all these companies are incredibly ambitious, incredibly smart people, all trying to solve very challenging problems. And be the first at doing it. And so just that intrinsic competition would make sure that we were really, really driving ourselves to do the right thing and succeed.

In some ways, that’s common to, you know, what’s happening in our future. Where I would say that it’s a similar thing where you’re kind of driven to succeed. But I think there’s more dimensions beyond just that, when you’re trying to run your own startup.

Siddhartha Ahluwalia 14:25

And when you started to become an entrepreneur and, you know, go full time into it, what made you choose a problem? Let’s say, what made you choose that I want to work on synthetic users? And I want these synthetic users, the first problem they want to solve is market research and user research.

Vasanth Namasivayam 14:45

I think a couple of things drew me in this direction. I was very passionate about making sure the products we build are really customer focused. It’s a mantra, which is very common in the value.

But from my experience, it’s not always put in practice. My second big driving factor is, when I think about products, I think about technology. Even today, technology sometimes is elitist in the way it’s built.

It isn’t built for everyone, it’s built for the few. And I think, you know, that kind of reduces the adoption of what could be good for a lot of people. I think those are the two driving factors.

Going back to the first point, in most companies, you have separate market research and user research teams. And these teams run on their own cadence. Typically, they would work with, you know, third party companies to get a bunch of humans into the room, test out concepts and whatnot, summarize, deliver results.

And then you had the EPD teams, the engineering product design teams, which were running based on the sprint cycle. And in a perfect world, customer obsession is married, it’s kind of intertwined with product building. But organizationally, that was never the case.

So what ended up happening was, from my perspective, and not everybody will agree with this, I think in the plurality of companies, market research and user research becomes a checkbox, becomes a CYA, as opposed to what the customer actually needs. I could argue that this pans out even statistically, where I think it was Pendo, which found out that more than 50% of products or features built are not widely used by customers, or they’re not used at all. So you are kind of investing time, intellectual power, resources into solving problems, which are, I would almost call, made-up problems, or Silicon Valley problems, right?

It takes me back to Juicero. I’m sure you’ve heard about Juicero and what happened there. Well, that’s a made-up problem.

And if suppose the folk involved had done high-quality market research…

Siddhartha Ahluwalia 17:21

Can you tell our audience more about Juicero?

Vasanth Namasivayam 17:24

Yeah, so Juicero was an infamous company from the 2010s, which was building a hardware product and a service to deliver cold-pressed juice, you know, at your home.

And so they would provide this hardware, where you put this packet or sachet in, and it would squeeze out juice. And I think it was fairly highly priced. Back then, it was $450 or $500.

This was in the mid-2010s. And then I think it turned out that one day, one of the folk involved realized that he could take one of the Juicero juice packets and just squeeze it with his hand. And, well, he would get the same juice out.

And, you know, shortly after the company went off. But it kind of talks about solving problems, which are made-up. You know, we think they’re important.

But do our audience, do the actual users, think that they are valuable, hair-on-fire problems, which is worth solving? Or are we just building feature factories, where, you know, we just want to ship stuff? What I also realized was, for a lot of big tech companies, a lot of the product builders, it’s a politically incorrect thing to say.

But for a lot of product builders, their focus is as much about getting promoted at the end of the appraisal cycle, as it is about building a product which really resonates with the audience. And when that is your incentive, then it is in your best interest to push out feature after feature to kind of, you know, put things on your annual appraisal. But that doesn’t mean that, you know, the actual customer finds value in what you’re building.

You know, and so I saw this happen repeatedly again and again. And as I looked at it, I figured the bottleneck was really in an ideal world. I would have a magic wand.

I have a question. I can invoke a group of humans, ask them what I want, check if am I making the right decision. And then, you know, move on to making the decision and moving fast.

In a real world, getting access to humans to ask so many product and market questions along the way takes so long that we just go with very limited market research, very limited user research, a lot of gut instinct, a lot of debates, a lot of highest paid person in the room’s opinion to kind of decide what to build, and which is why products, you know, kind of fail. And I kind of came to a point where I was asking myself, do I want to take up one more product role? Or do I want to solve what I think is an inefficient way of building what humans and consumers need?

And I figured I would do the latter. So that was one big aspect of it. The second aspect was about democratization of technology.

You take an iPhone today, and most people would call this to be the pinnacle of design. But I’ve seen my dad use it. I’ve seen other people who are not that tech savvy use it.

And if you ask them to change something in settings, it’s going to take for them forever. Right. And I’m picking on Apple, not because I want to pick on it, but I think it’s one of the greatest products.

And even products like that face this issue where it’s not very understandable for the average person. And my question is, why are we not building products so that every single user can just get it? It just works.

You know, this is one of the things I learned from Dropbox. Dropbox’s motto is, it just works. Right.

And when I work with most of the products around me, it works with an asterisk. It works for people who are extremely proficient in a certain skill, but it doesn’t work for all the users. And, you know, that kind of, you know, we talk about personalized user experiences.

Why don’t we have it still? You know, these are the kind of questions which kind of drove me towards working on how do I make it very easy to really read a user’s mind, to read a human’s mind, to understand what is important to them, what is not, what is easy for them to understand so that they can, I’m solving the right problem, and I’m solving it in a way that the user just gets it. It just works.

Siddhartha Ahluwalia 22:13

And what is the difference between synthetic user and AI agent and like, what are they?

Vasanth Namasivayam 22:17

So when I think about AI agents, AI agents for me is hyper, like, though there’s an overlap, there’s an intersection between the two sets. AI agents is hyper focused on efficiency. I have a workflow, which probably in today’s world for knowledge workers spans across multiple products.

And I want to automate this workflow, make a workflow more efficient. And this was obviously being done even in the previous decade, with, you know, workflow automation and whatnot. But with LLM powered AI agents, it’s becoming more flexible, more easy to create these kind of workflow, these kind of automations, so that the agents can kind of, you know, do things for you.

When I think about synthetic humans, it’s not about efficiency. It’s rather about mimicking the cognitive processes of humans, mimicking all human weaknesses and foibles. Because we as humans, we’re not focused on efficiency, not efficient.

But when you’re building a product for a human, you are not building a product for this magical, efficient human being. Rather, you’re building it for, you know, blood and flesh with all the things which come along with it. And I think that’s what a synthetic human is.

So I think the use case or the goal is kind of very different. With synthetic humans, there might be cases where synthetic humans do not complete an onboarding workflow. And you know what, that’s very realistic, because that’s what real humans do as well.

Whereas in the case of agents, agents always get it right, because it’s all focused on hyper-efficiency. The analogy, the pattern which I draw is, back during the Adam Smith era, or during the classical economics era, we always believed humans to be rational economic agents, who made decisions in a very rational manner. Now, obviously, with behavioral economics, we know that’s not the case.

We know that humans are humans. And therefore, you know, as you think about economics, you need to worry about the real human, and not this hyper-efficient logical human. You know, and when I think about synthetic humans, that’s what I’m focused on, you know, can I mimic the real human?

Can I bring that human on demand? So people who are doing stuff for these humans, are building stuff which are more relevant, which are more usable, which are more attractive to these folk.

Siddhartha Ahluwalia 24:55

And what are the tasks that these synthetic humans can do that an AI agent cannot do?

Vasanth Namasivayam 25:01

The synthetic human, the best thing which synthetic humans can do, which the AI agents cannot do is, they can fail. They can fail at doing a task, which is a reflection of how a real human would fail in using a product. An AI agent will always figure out, or the goal is to always figure out how to solve the problem.

But synthetic humans will constantly fail, they will complain, they will tell you why this is bad, all before you launch a product.

Siddhartha Ahluwalia 25:31

So how do you build a world where, let’s say maybe 10 years from now, both synthetic agents and AI agents, synthetic humans and AI agents coexist?

Vasanth Namasivayam 25:41

I would imagine there could be a world in the future where, even when I think about future right now, there are aspects which I want it to happen from a workflow automation perspective constantly. And AI agents fit in very well there. And then there are aspects where I want to mimic human behavior, where synthetic humans fit in.

I could also see a world where synthetic humans kind of use AI agents to automate some of their work, you know, going forward. So I would always see the two things are kind of intersecting with each other, parallel with each other. From a technological perspective, there’s a lot of overlap.

But applications is of course this.

Siddhartha Ahluwalia 26:26

And why do you think, you know, companies like OpenAI, which are really good at AI agents, right, cannot do user research or market research well?

Vasanth Namasivayam 26:40

I would not use the word cannot. So when you say cannot do user research or market research, themselves as a company are providing it as a service.

Siddhartha Ahluwalia 26:48

Providing it, let’s say a type of interface into ChatGPT that tell me this Figma file or this user workflow works or does not work. Will ChatGPT be successful in it or not?

Vasanth Namasivayam 27:01

Yeah. So a couple of things to consider. One, when I think about OpenAI, their vision is very different from trying to mimic human behavior.

So they are going after, you know, a different part of, like, I think Altman keeps talking about AGI. So I think they’re going after a potentially far larger fish than synthetic humans. But that said, we have tried prompt engineering ChatGPT a lot to act as synthetic humans.

And what we learned is transformer models, OpenAI, is great at providing feedback on the average. But it doesn’t cover the edge cases or the edge cases of your opinions. So that’s a problem.

So ChatGPT is good at persona feedback. It kind of averages it across the population. But if you really want to understand how does the population react, how does the population interact, not only in the Gaussian middle, but also on the edges, then, you know, ChatGPT is not built for, you know, that kind of application at all.

And so it would require kind of not just using base transformer models, but more hybrid architectures to be able to pull that off. And I don’t see that being a part of OpenAI’s vision right now.

Siddhartha Ahluwalia 28:30

Why do you think, you know, that ChatGPT is such a successful product? Like, if you have to think about all the products, right, right from iPhone to anything else that you can imagine, is the first product to reach 100 million users the fastest, right? And I think right now it’s about to reach a billion users.

Vasanth Namasivayam 28:55

Yeah, I think it’s the wow factor and the time to value. When I first used ChatGPT, I think it took me around 30 seconds to kind of realize that, oh, wow, this can do so many things, depending upon how I prompt it or speak to it. So the time to value was very, very small.

And in today’s attention deficit world, ChatGPT got that perfectly right. The second was a wow factor. After decades and decades of really annoying chatbots, starting with Microsoft’s, I don’t know what it was called, I forgot the name, the clip thing, paperclip, which used to keep coming out, right?

That was back in 97, 98, I think something of that sort. After all those decades of annoying chatbots, you finally had something which just worked. And I think, you know, that wow factor was just something special, along with the fact that around the same time as ChatGPT, you had DALL·E coming out from OpenAI as well. And people being able to generate images, right? All of a sudden, the entire concept of generative AI took off, you know, and it became like people could understand the potential, the possibilities, and they could understand that very, very fast. Even though this is a hyper complex technical product behind the scenes, for the average user, I would argue that it’s easier to find information on ChatGPT than it was on Google when Google first came out, and that’s saying something.

Siddhartha Ahluwalia 30:33

Yeah, I think if you go back in history, why Google search was so successful is it was the fastest search which gave as possible to accurate results as possible. I think ChatGPT would have taken it to in terms of at least the fast or what the user wants, not in terms of accuracy, but to at the next level of what a Google search can give you.

Vasanth Namasivayam 30:56

And again, like, thanks to everything which happened socially over the last two decades, our attention spans are falling more and more. So, we don’t want to trawl through, we don’t want to do the work. We just want knowledge to come to us in information size, like bite-sized pellets, which you can quickly consume and move on.

Because we live in the Instagram generation, we don’t want to spend more than six seconds trying to understand anything. And I think GPT is greater than that.

Siddhartha Ahluwalia 31:29

And coming back to synthetic users, how big do you think the market for synthetic users is?

Vasanth Namasivayam 31:38

So, let’s think about from the perspective of synthetic humans, as opposed to just users, because I want to cover not just people who are currently using your product and mimicking their behavior, but also potentially future, right? And I think I was recently seeing this podcast by, or this video by a A16Z partner, where he referred to this as social simulation. And he talked about this being as the next frontier in AI and machine learning.

We’ve been able to predict images, we’ve been able to predict text. Now, can we predict human behavior or the way humans think? The applications are enormous.

We obviously started with market research, but even the word market research or user research is so loaded. You can think about something as narrow as usability, but you can also think about which ad copy will work well with this kind of human. You can think about what kind of user interface flows work well with this particular human at this point of time.

And when you think about it in that way, the applications can range from product building all the way to government policymaking and politics. As an example, how do I, if I’m in an election, how do I ensure that I am giving the right message to the plurality of my constituents so that they are more impelled to vote for me? Or if I’m in policymaking, how do I ensure that the policies we are crafting benefit the most number of people?

The ability to simulate humans, the way they think, the way they behave, individually and at a population level, kind of opens up all these applications which were not possible earlier. So, I think, when I think about synthetic humans and synthetic human simulations, I think the market is tremendously large. I don’t think it’s well quantified today because it’s opening up applications which didn’t exist before the advent of this.

Siddhartha Ahluwalia 33:48

So, just to reiterate, market research, user research is one application, product building is another. What are the others that you can think of?

Vasanth Namasivayam 33:59

So, I can think of gaming. So, as an example, you want richer games. Can you make your NPCs actually behave like actual humans?

That’s a fascinating application. Today, I think I was speaking to some mentor. He was talking about how many people are constantly trying to build new games.

So, then the question is, can you build more and more games which actually work? Because people who play them would kind of like it. Another thing would be, like we just talked about, what do you say, politics, like winning elections, policymaking would be one more thing.

I think any place where technology interfaces with humans, being able to simulate humans would help you build an interface which is easiest to use for that particular human. One more application I can think of is trade war simulations, country simulations, because you’re trying to simulate populations and you want to see how populations would react to anything, any kind of decision you’re making at a macro level. The one common thread in all this is you are making a decision.

That decision affects a human or a group of humans and you want to de-risk or optimize your decision as best possible before you spend resources in implementing that decision. To allow that to take place, you need to authentically simulate humans, both at an individual and at a population.

Siddhartha Ahluwalia 35:48

Vasanth, I now want to cover, what is that one thing that surprised you the most while building a startup?

Vasanth Namasivayam 35:56

I think I did expect a lot of intensity going in when I was going to build a startup. The one thing which I think surprised me was how complex or challenging it can be to hire the right people and figure out who’s the right person and who’s not. I’ve done a lot of hiring in these larger companies, but they obviously have very structured processes.

The cost of failure, while it’s big, is not earth shattering. In the case of a startup, I now believe that the first 20 people can individually make or break a company because they have outsized impact on the future of the startup. It’s not that I didn’t anticipate the size or the impact or the intensity.

Siddhartha Ahluwalia 37:52

One mistake every beginner product guy makes.

Vasanth Namasivayam 37:56

A lot of beginner product folk obsess about how innovative their technology is.

We’re all technology obsessed. I think the focus should be on the value and the experience. Technology for me, technology modes in most cases don’t matter.

I think it’s the and most initial product folk make that mistake. A second big mistake which everyone makes is they hyper focus on coming in and coming up with amazing ideas. Ideas for me are valuable, but they are the very, very, very tip of the iceberg.

There’s a lot more which goes into building a good product.

Siddhartha Ahluwalia 36:54

One mistake that every experienced product guy makes.

That they know everything, that they don’t like a lot of experienced folk, and this might not be limited in product. We tend to not approach questions with the same humility as is warranted because we think that we have been here, we’re done.

Siddhartha Ahluwalia 38:17

One thing about startups that you had an intellectual understanding, but now have developed an emotional understanding as well.

Vasanth Namasivayam 38:25

Every founder would talk about how the startup is their baby. I intellectually got it. It kind of made logical sense.

The kind of emotions I go through on a daily basis, the moment I wake up is not dissimilar to when I had my first kid. Your entire world is this particular thing you’re creating. You’re trying to make sure it lives and bring it to life.

Now, it’s personal, before it was intellectual.

Siddhartha Ahluwalia 38:57

Now, I want to spend more time on discussing Featurely. What does Featurely do? What does Featurely solve?

I want to hear it from you.

Vasanth Namasivayam 39:06

Absolutely. Let’s go to the problem of understanding humans in general. One of the insights I had was election campaigns in the US.

I think 2024, correct me if I’m wrong, a few billion dollars is spent in every election campaign. Each of these campaigns run their own user research panels. You have a lot of third-party companies doing exit polls, all that kind of stuff, doing a lot of representative sampling to kind of figure out which direction the election goes and how to make sure you win the election.

Sid, let me ask you this. All that money and they generally get it wrong. And my thought was, well, if in spite of so much money and expertise, these guys can’t get it right, then what’s the chance a small little product builder sitting somewhere can actually really understand their humans and figure out what is required to build?

I would argue that the political folks have everything they need to get this answer right every single time. True for big tech companies as well. So the problem which I’m solving is when you are building anything, and when I say building, I’m talking about content, it could be a design, it could be a user interface, it could be a workflow, anything you’re building for a human on the other side, you need to really empathize with the human.

You need to really understand what they’re doing to make sure that you are solving a real problem and you’re solving it in the right way. Featurely helps you do that by simulating the way humans think, both at an individual level and at a population level. Let me give you a very, very crisp example.

Let’s say you’re building an ERP software for a chief marketing officer. Okay. So you do a lot of user testing.

You do things of that sort. But all this is done expecting that your chief marketing officer is sitting in a vacuum using only your product. But let’s think about a real CMO.

She might be having 25 tabs open, multiple Slack messages coming in, flipping in and out of your product. Now think about that experience compared to in an ideal environment, testing it, making sure that your product works. Two very, very different experiences, two very different products.

That’s the kind of user empathy and insight needed as you’re building successful products. You can’t build in a vacuum. That’s what Featurely hopes to democratize.

Now, furthermore, my belief, my hypothesis is that product building or generating products is going to get more and more democratized, thanks to all the advances in generative AI recently. But what is not getting democratized is user understanding. So it’s not just about building products.

It’s about, am I building the right thing? And am I building it in a way such that the person on the other side can use the product? That is going to be the bottleneck.

And that’s what I want Featurely to solve. So at the core, it’s about human behavioral simulation, but I’m hyper-focused on this entire space of human understanding for markets and products.

Siddhartha Ahluwalia 42:45

What change do you want to see in, for example, in the companies that you say, this makes Featurely successful? So for example, an example of success can be if a company was previously doing user research in 10 days and building an X product with Featurely, they can do it in one day and the product, the outcome X that they were showing, it’s not X, it’s much better than X, 2X.

Vasanth Namasivayam 43:12

Yeah. I think the signals which I would like to see from an outcome perspective is conversion rates at every part of the funnel. So let’s say you’re building a landing page.

Does Featurely improve your conversion rate by an order of magnitude? You’re talking about retention. Does Featurely increase your Sean Ellis score or your retention metrics by multiple orders?

The time to generate a product and solve a problem, can we go from months to minutes in terms of doing user research, understanding everything? I think those are the factors. If we are able to achieve all that, I think Featurely would be very, very successful.

Siddhartha Ahluwalia 44:07

And what are the things that you are able to make difference in your life of your initial customers right now?

Vasanth Namasivayam 44:13

Right. So I think two things. Straight off the bat, speed.

Our initial customers are going at multiple orders of magnitude faster because of being able to work with Featurely users. Previously, our customers either would take months to understand, schedule user interviews, versus, or they would not do it at all and take a bet that something would work and then just go to market. So now with Featurely, they get access really fast and they get access to actual synthetic units.

A second big thing which I’ve seen which has changed is the confidence with which they ship stuff. They’re able to catch a lot of edge cases even before they ship. In hindsight, when they look at the synthetic human opinion, they’re like, yes, makes absolute sense.

I should have thought of this. But in the day-to-day of products, there’s so many things you don’t think about. These humans catch them.

So then they’re just building a higher quality experience, whatever that is, whether it’s your landing page or your actual product. And they ship with more confidence.

Siddhartha Ahluwalia 45:26

One thesis that I have is, in today’s world, the best of enterprise products will have the time to value equal to the best of consumer products. Like, for example, OpenAI, 30 seconds time to value. And that’s why it’s so successful even in enterprises.

What do you think your time to value is in Featurely?

Vasanth Namasivayam 45:47

So right now, our time to value is multiple minutes. But I believe that over the course of this year, I aim to have a time to value of 40 to 45 seconds. And as I mentioned earlier, we are living in an increasingly attention deficit world.

It’s so important to have that really, really small time to value. From my perspective, you should build a product like it’s going through PLG motions to an end consumer. And if you’re able to crack that level, then it kind of follows on that for an enterprise setting, you should be able to hit the TTMs.

Siddhartha Ahluwalia 46:30

And where do you think then modes of distribution with companies like Microsoft would stand?

Because Microsoft today wins, or Salesforce wins, or Oracle wins, majorly because their distribution is in, if you talk about Fortune 500, entire 500, right?

Vasanth Namasivayam 46:50

100%. 100%. So this is interesting.

And this is just a hypothesis. I don’t know how this is all going to play out. I suspect that we increasingly see a world of more and more hyper-customized, on-demand, ephemeral software, which is kind of spun up on demand, used for a certain application for a certain person, and then it kind of disappears.

Siddhartha Ahluwalia 47:13

What do you mean? Can you explain it in layman language?

Vasanth Namasivayam 47:15

Yeah.

So right now, let us say, as an example, let’s say I need to create a document. Well, I would use one of the primary software, Word, Notion, Paper, Google Docs, things of that sort. But I’m not creating a document.

I’m actually trying to perform, like achieve something, maybe get called for interviews, or maybe create a letter which makes a lot of sense for my team, or for the board, right? Today, we are forced to use general-purpose tools, and then use our own time and intellect to make sure they work for our specific application. But with the rise of all generative technology, I hypothesize a future where software is more on-demand for that particular application or goal you want to achieve.

Because it’s on-demand, and it’s kind of hyper-personalized for you, the person, it also needs to be ephemeral because you probably will not use it again. So you need this, the software will come up, it will help you achieve that goal, and then it will kind of disappear. That’s kind of what I imagine is a potential pathway to products and software in the future.

And when you think of that kind of world, distribution channels kind of change a lot. You’re still going to have the classical channels, but I think it kind of changes more for the non-large companies. When you think about distribution, Sid, back in the 2000s, for all rich media content, you had these big production and distribution companies, and that was the only way to win.

And then thanks to the rise of social media, the mobile phone, and all that, distribution fundamentally changed. I imagine that that’s potentially a possibility, even with software and software products. I don’t think that the current distribution strategy is going to be the only or the primary distribution strategy 10 years from now.

So that was kind of the analogy which I was trying to…

Siddhartha Ahluwalia 49:36

Yeah, it’s strange that you pointed it out, because right now, software is built online, but sold offline, if you imagine that. You don’t win enterprise software contracts because the team liked the software the best. It might be because Microsoft has the best sales team, that’s why they won.

Vasanth Namasivayam 49:58

Exactly.

And as you’re looking at enterprises becoming smaller, becoming more efficient, then the focus is going to be on value, value creation, as opposed to just who do you know, just your relation and your distribution network. And I think that is probably going to erode over time.

Siddhartha Ahluwalia 50:17

And apart from the above companies that we discussed, name three companies that you respect the most in the AI world. And how do you think that they win in distribution? And they can be Cursor, Windsor or anybody else.

Vasanth Namasivayam 50:33

When I think about AI companies I admire, I still think of NVIDIA as one of the biggest AI companies, primarily because without a lot of what they do on the bare metal, a lot cannot be achieved on the top layers. Second big company which I admire, and I think the reason why they win is because they’re literally building for the future. They are defining the future market as opposed to reacting to the market.

A second big company which I think is amazing in the AI world is Aldi, like DALL·E, the specific product from OpenAI. Why do I talk about DALL·E? It’s the wow factor of creating images.

It made AI so easy to access for the average person. That I think it’s brought on a lot of more people into using these kind of tools. So I think the way they went about doing that was just amazing.

If I think about a third company which is really, really good at this, I don’t think this company exists, but I think there is going to be a company in the future which generates media content, hyper-personalized to the person, hyper-dynamic based upon the context of the viewer. I don’t think that company is there today, but maybe somebody is already building upon it. I think that would be a company which would be very interesting in terms of its potential impact.

Yeah, those would be the three things I would think of.

Siddhartha Ahluwalia 52:12

And imagine we are living in 2035. What do you think in our daily lives, the work is being performed by AI?

Vasanth Namasivayam 52:23

I think one of the big things which I’m seeing now is we have leaped forward in the software or the pure software layers, but we are starting to integrate it into hardware layers. I find that fascinating because I would imagine there is more embodied AI in 2035 as compared to today. If that is true, then that would lead to a lot of automation or having AI power in day-to-day work like folding your laundry or doing things which are considered chores.

That’s kind of where I would see the world. In an ideal world, AI should not replace the creator. It should replace the mundane.

The barrier to that is you don’t have a good physical hardware layer which can interact with the environment. Now, I think that’s starting to get built. And so, a combination of that along with the brains of whatever you have on the software side, on the model side should open up a lot of those applications.

Siddhartha Ahluwalia 53:39

And I now want to touch on the transformers and LLM. So, if you want to explain transformers and LLM to a layman, how would you do it?

Vasanth Namasivayam 53:50

So, when I think about the evolution of AI, you first had stuff like Bayesian networks, which was basically probabilistic reasoning. If the grass is wet, what is the probability that it rained? Can you predict that?

So, that was the very initial form like a very simplistic way of thinking about intelligence. Then you had neural networks where you were trying to look at tons and tons of data and predict human neural patterns to kind of make sense of this data and make predictions going forward like extrapolation or interpolation. I think the great thing about transformer models is it’s almost like transformer models can listen to and see all your inputs simultaneously in a completely parallelizable way.

That was not possible earlier. So, imagine if you didn’t have one pair of ears, but you had an infinite pair of ears, where if you were at a dinner table, you were listening to every single piece of the conversation at the same time. And you were able to instantaneously put it all together to understand the entire story.

I think that’s kind of what transformers unlock. A great application is large language models. And when I think of large language models, I basically, at a very simplistic level, it’s a next word predictor.

This is the first word, what’s going to be the word after that, and so on. But how does it get it right so much? It’s because these models have been trained on all human knowledge across all time.

And so, it’s been trained on so much knowledge, it’s kind of able to see the patterns on what should be the next word based upon this word, based upon the context. And it’s a transformer. And so, it’s able to better understand your input queries as compared to more traditional models.

The closest real world example I can give you is, think about there was an art on how to search Google back in 2005. You had to enter your query in a certain way. But today with LLMs, that’s not really the case.

It can understand most things, almost everything. And I think that’s kind of what I would think of large language models. It looks very human, it sounds very human, but I almost think of it as a parrot which has memorized all of human knowledge, and is able to just tell things which make a lot of sense.

But sometimes it doesn’t make sense. But yeah, that’s kind of the way I think about, or easy way to think about large language models.

Siddhartha Ahluwalia 56:48

So, Vasanth, I’m grateful to my partner Ribhu in NEON that, because of him, you, Ribhu, NEON, me and the entire NEON team started working together. But to go back in time, what made you partner with NEON? And how has the experience been since then?

Vasanth Namasivayam 57:10

I think partnering with the right VCs is a very strategic decision. It’s more than just the money. I was looking for a partner who could kind of be very hands-on with me, who could help me on the ground, and who’s invested enough that they have the time to spare with me on a weekly basis.

I think NEON fit that bill excellently. Second, I think your reputation proceeds both through the NEON show, and through other folk whom you have already invested in. And I think the overall consensus was someone who is very founder-friendly, who has a lot of skin in the game, and who feels almost more of a partner than someone I’m kind of reporting to on a weekly basis to kind of talk about where the company is going, things of that sort.

And the final thing was, I was planning to be more capital efficient by building quite some of the work in India. It’s been 20 years since I left India, and I really wanted one amazing VC to partner with, who could kind of help me navigate the system, and build over here in as efficient a manner as possible. So those are my three, what do you say, hypotheses.

Now, in terms of the way it’s played out, I think it’s exactly played out that way. I would say that I love our weekly conversations, because it feels like we’re brainstorming. We’re constantly seeing how we can push the boundary.

I love how generous NEON has been, both with their time and with their network. It’s not something you see often, and I think it creates a lot of trust in the system. Personally, I’m not a person who feels comfortable very easily, or who kind of, you know, like, forms, like, it takes time for me to trust people.

And I think with this particular relationship, it’s been a kind of an exception. Just because of how much skin you guys put in the game. I respect that like crazy, because I know you have a big portfolio.

And I’m assuming you’re doing this for every single person you invest with. And I really respect that. Because it’s more than just writing a check and walking away.

It’s about putting your reputation, putting your blood and sweat behind this, constantly thinking about how you can help this team, this group of founders to be the best version of themselves and make the company successful. I think Neon does that very well.

Siddhartha Ahluwalia 1:00:18

And what do you think, if I were to ask, what could make Neon even better?

Vasanth Namasivayam 1:00:30

I don’t think like off the bat, I can’t think of much. You know, I think Neon is still a growing fund. And right now, I think the focus is hyper focused on very early stage bets.

But I imagine over time, you will have a healthy growth fund to kind of, you know, double down on the bets which are paying off. I also think that Neon, it’s great your thesis, the Indo US thesis. But I think Neon would benefit from investing in companies which have, you know, more of a Bay Area presence over time, just in terms of exposure, you know, and things of that sort.

But yeah, off the bat, I think there’s really like you guys are doing the right thing.

Siddhartha Ahluwalia 1:01:17

And what do you think, you know, can help Neon partnering with the best of founders in Bay Area, like we are still building that reputation right now?

Vasanth Namasivayam 1:01:27

I think there’s like, definitely the sustainable way is building the reputation over time. You know, I think that there is no replacement for that. But I think having a presence, a big presence in the Bay is going to help.

Right now, the presence is kind of more remote-ish.

Siddhartha Ahluwalia 1:01:53

You know, that we have one partner in Neon in the Bay Area.

Vasanth Namasivayam 1:01:57

I think that would help a lot in kind of just having that presence. And second is, there is a lot of, like, even within the boundaries of your thesis, there are a lot of folk who are trying to get into the space. But, you know, they’re not had the ability or the courage or whatnot.

Potentially looking at stuff like incubator models in the future, at least focused on Bay Area, might help you kind of, you know, get access to higher quality talent very early in the process, as compared to what you all are doing now.

Siddhartha Ahluwalia 1:02:35

And why do you think Bay Area is so important for founders?

Vasanth Namasivayam 1:02:42

I want to say it’s the climate in California. But I think in, when I look at the Bay, failure, like it’s almost sounds like a cliche, but failure is celebrated. You’re not afraid to fail.

And since you’re not afraid to fail, some of the best ideas come out there. People dream stuff which should not happen. Like, as an example, let’s talk about future, right?

You’re trying to model human behavior. Man, that sounds like science fiction. There is more reasons why they should not work than it should work.

But you know what, like two decades of living there, I’ve kind of internalized that anything is possible.

Siddhartha Ahluwalia 1:03:26

You would not be so courageous in your pursuit of this idea if you were not living in Bay for the two decades?

Vasanth Namasivayam 1:03:33

I think that I would be more derivative in terms of what I would want to pursue, as opposed to something of this.

Siddhartha Ahluwalia 1:03:43

What do you mean by that?

Vasanth Namasivayam 1:03:46

Like, let’s say I was living in Bangalore.

I mean, I grew up in Bangalore. And let’s say I deem to go to the US in a different universe in a different world. I would think of building startups, which kind of follow startups, which are companies which have been very successful, and kind of replicate those business models in the ecosystem, which I work in, so that I can be capital efficient, without having to, with a high chance of success, right, where a lot of the hypothesis has already been tested out.

And here, it’s more about refashioning it for the Indian context, and then building it efficiently. That’s kind of what I would potentially do, suppose I was not in the Bay. I think being the Bay has given some level of latitude, which I think allows me to explore this.

Siddhartha Ahluwalia 1:04:39

And this is for other founders, right? What helped you in your capital raise journey, your pre-seed was oversubscribed, you get regular VC inbounds, which you’re not able to cope with?

Vasanth Namasivayam 1:04:52

I think I was just, you know, writing a letter to my team the other day. The focus should be on building a business, building a product, and solving a value. It should not be on purely what your valuation is, you know, and raising money.

Then, I think VCs are a really smart group of people in pattern recognition. And when they see something working, they will come to you. But on the flip side, if suppose you’re focused on just attracting the money, I think you’re kind of getting distracted from the actual art of building a business, of solving a real customer problem, which is monetizable.

And that means that, you know, you are less attractive, you know, to a lot of investors. That’s my thesis. Again, I’m so early in this game, maybe my thesis is completely wrong.

It’s helped me the first time. Let’s see whether it helps me, you know, in the future realms. But like, I have a lot of conviction that if I can build something, which is very valuable to customers, if I can build a solid business, then the money will come, you know, there will be VC interest, because there are a lot of proof points.

I’ve de-risked a lot of things for VCs. Whereas if I’m just focused on a narrator, and it’s a hollow shell, VCs are smart enough to look hollow shell. And so I think that’s where you’ll churn out like a lot of stuff.

So I would always say focus on the problem.

Siddhartha Ahluwalia 1:06:34

And if time, resources and money was not a constraint, how big you want to be built Featurely?

Vasanth Namasivayam 1:06:49

I think even today, I think I’m at a place where I want to be category defining this particular market. I keep, you know, kind of semi joking with my team, that I want billboards all over San Francisco, which say that Featurely, our humans are the best in the world. I don’t want to use the word monopoly.

But I basically want to create a completely new market of synthetic humans, which is embedded in workflows across every single decision, no matter what it is. I’m starting with market research, user research, but I expect to expand vertical by vertical over time.

Siddhartha Ahluwalia 1:07:34

And what do you think will make Featurely so easy to use that I can deploy synthetic human for whatever I’m doing?

Vasanth Namasivayam 1:07:43

It’s not here now. Okay. But I think the best, most usable products are products which are aligned with our pattern recognition as a human.

So we don’t need to think about how to use a product. So an example would be the swipe on an iPhone. People didn’t have to think about how to use it because flipping a page like this was kind of embedded in human consciousness over all these centuries.

So when I think about Featurely, I think that’s the level I want to achieve. Can I look at how synthetic humans are interacting with each other through a glass door virtually, you know, mimicking human behaviors in the digital world so that you don’t need to think about how to use a product. That’s the gold standard.

Siddhartha Ahluwalia 1:08:39

I think it’s analogous to what Jeff Bezos said that whenever they’re taking a decision in a room, they always keep like one seat empty, for a customer.

Vasanth Namasivayam 1:08:50

Exactly. And can that seat be filled with a synthetic human?

Siddhartha Ahluwalia 1:08:54

Yeah.

Vasanth Namasivayam 1:08:56

I think that’s the notch which I want to go through. Even for this year, it’s part of our roadmap. Obviously, it’s one thing to plan it, it’s another thing to execute on it.

But that’s the standard we need to achieve. There are, like you saw the A16 report, there are probably 10 companies in this market. I think a lot of the companies are going to get it wrong, because they are going to be hyper focused on the technology modes.

Technology modes are interesting, but I don’t think they are long lasting. I think what is more interesting is, does it just work? You know, is it so simple to use that anybody gets it?

Siddhartha Ahluwalia 1:09:40

Yeah.

Vasanth Namasivayam 1:09:40

Right. And that is magic. Right?

All those debates that product managers and designers debate over, or should this button be orange or yellow or whatever? No worries, invite 10 people and then ask them and they will tell you on the spot.

Siddhartha Ahluwalia 1:09:55

Yeah.

Vasanth Namasivayam 1:09:55

Right. That is the company which is going to win this race.

Siddhartha Ahluwalia 1:09:59

Thank you so much, Vasanth. It’s been an incredible conversation. I enjoyed it a lot and learned a lot.

Vasanth Namasivayam 1:10:04

Thank you so much. Thanks for the opportunity. This was a lot of fun.