Finding Real PMF in the AI Era with Tanay Kothari (Wispr Flow) + Hans Tung (Notable Capital)

Renee, good to see you today. It's, uh, I still remember the first time Chelsea connected us, and on the Zoom call when we first met you, you asked me to speak in Mandarin and see what happens, what does it do to your app. And I was astounded by how easy it was to use. We're not the only ones. You are highly recommended by Marc Andreessen and Reid Hoffman, even though both are not investors in your company. They talk about you on their podcast. What is Whisperflow and why so many people like it so much?

So for anybody here who hasn't used Whisper, I saw I think 30% of people here are Whisperflow users. Is really simple. It takes you from doing this on your phones and this on your computers all day long to just speaking. And that's the way we talk to everybody else. That's natural. And so we do this just incredibly well. You speak naturally, and Whisper writes for you in a structured way, cleans it up for you in exactly the way you like across every single application on every single device. No integrations, no nothing else. You just press a button and it works the exact same way your keyboard does. And so it was really funny, 'cause when we started working on it about 2 years ago, everybody thought we were crazy, 'cause hey, there's a dictation button on every single thing that already exists. And the thing we always told people was, but do you love it? And the answer was always a blanket no. That meant there was still a problem that needed to be solved, and that's what we've been working on for the last, 2 years now.

And how fast, how accurate is WhisperFlow?

So the metric that we measure is not what a lot of otherwise companies measure, like word error rate, the number of words you get right, because that 96%, 97% sounds great. It's terrible because that means in a 20-word sentence you're going to make a mistake. The real thing that matters is how often do you have to edit The bar we hold ourselves to is if in a 3-paragraph email you dictated, if one comma is off, that's a zero for Whisper. We have failed you. That's the metric we measure. 93% of the messages that people write with Whisper do not require any edits at all. It's perfect first time.

I remember when on that call you asked me to say something in Mandarin. I played a little trick. 'Cause in Chinese you can have something phonetically sounds the same, but the text actually are quite different. And I remember the output in English caught that nuance and translated correctly in different words, even though phonetically Mandarin is the same word, sound the same. So how is that your system just so much more intelligent than what people used to in alternative?

So the thing we realized is speech is very nuanced. When you're listening to somebody, somebody has a hard accent, let's say, it takes you a minute or so to figure out what the person's talking about. And a lot of times you need to know the context to just better understand, hey, this— like, what did the person actually mean? Because a lot of words sound the same. The first thing we built with Whisper was the ability for it to just have the context. So for example, if I'm messaging Marco or Hans and their name is there in my recipient field, be it in WhatsApp or email, of course it should get their name right. The name is right there. And all of the little nuances of language, if you train with real-world human data, you're gonna get that right. And so all of this was built in the product from day zero because there is no other way we're gonna give the people perfection until you can at the very least emulate how humans work before making it better.

Yeah. And by the time Chelsea found you and you, you already have the product pretty well done even though it was, the revenue was very small at the point in time. But this was not the first, you didn't get it right the first time. You guys iterated, pivoted. Initially you guys started with something, a hardware product. So how do you find product-market fit and when do you hold onto your original vision? When do you make what we call in-game adjustments as you see more feedback from the life, from the field? How do you make those tough calls?

So I'm not going to answer the question of how do we find product market fit, because that is— we could go on for hours. But I'll talk more about the in-game adjustments. So for anybody who doesn't know, Whisperflow software product been around for about 18 months, but the company's been around for 5 years. And for the first 3 years, we were building a wearable device very similar to the one I'm wearing right now that could pick up silent speech. You didn't have to say a word out loud. And we had a team of 40 researchers, primarily PhDs in neuroscience, signal processing, electrical, mechanical, that built the world's first thought-to-text device. June of 2024, we finally got this device to work after 3 years of working on it. Our chief scientist used to manage 5 Nobel Prize winners before. There were two other people on our team who had left their tenureships at universities. They were professors, and they joined Whisper full-time. And we got this device to finally work. And we started to use it with different things, and we realized that ChatGPT, Siri, Alexa, they all still suck. So we built this little thing called Flow that went from your mental rambles into something that was structured and ready to send. And I remember coming to the office one day and open office, 40 people all sitting there, and everybody was whispering to their computers. And that to me was a slap in the face because the reason we built a hardware device in the first place was for people to be able to use voice around others. That was a problem that still needed to be solved. To see people just have the hardware device lying on the side, the thing that we spent tens of millions of dollars building, 3 years of just blood, sweat, and tears, people just using a simple $10 microphone with our software product. To me, I think about startups a lot as truth-seeking endeavors. Right. And for us, it was really clear that the problem that we wanted to solve was real. The solution was not. And so a few weeks after that, we make the decision to pivot the company. We still had half the money in the bank. Our investors were really happy. We were 6 months ahead of schedule, but we still went from 40 people to 4 people overnight to put all our eggs in the one basket of this product that had 100 users, not 100 daily actives, just 100 users. But these people loved the product like nothing else. And that was the thing that we placed the bet on. And this is August of 2024. And then next morning we woke up, we had transitioned from a deep tech brain-computer interface hardware company to now building consumer software. And I told the team, gun to our head, we need to launch this product in 8 weeks, and that's what we did.

Once launched, somehow you had created a word-of-mouth impact from early on, and you seem to have very good traction with certain users, mostly around marketers and product managers, and later on developers as well. How did you know who's the sort of the right ICP to focus on? And how did that word of mouth effect with that group somehow materialize?

So I'll get to them one by one. So the first question is, how do we figure out which ICP? Because the product is you speak and it writes for you. It's like, well, it can work with really anybody who speaks. So our audience is the 5.5 to 6 billion people in the world who have a phone or a laptop. But if we chose that our ICP, we would probably be dead by now. And so we had to filter down. And we found 26 different personas like founders, developers, lawyers, consultants, all of whom were using Whisper and they loved it. And we had to pick which ones. If you just look at the primary metrics, which is the NPS score, the conversion to paid, and so on, they all kind of look the same. But the core metric that we had, and so it all comes back to it, like what are we really trying to solve for? The thing we were really trying to solve for is we want to get Whisper in the hands of as many people as possible as quickly as possible. The two personas that we chose to go after were founders/business owners and developers. Why? Because they're very high K-factor audiences. When developers love something, they talk about it everywhere, on Twitter, on Hacker News, on Reddit, on every single Discord channel that you can find. And that is why now you usually hear cloud code and Whisper Flow in the same sentence. Founders, again, when they love the product, they deploy to their teams. And at this point, we have almost 8,000 enterprises using Whisper.

Mm-hmm.

We add 1,000 more every single month. This includes a lot of the Fortune 5s, uh, a lot of the Fortune 500s, public companies, and your other favorite startups like Notion, Clay, Lovable, and so on. And so that was the beginning of how we chose our ICPs, and now we're expanding out to more, but it's always about both the first-order effects on how they impact your business directly, but also indirectly how that matters. The word-of-mouth question that I get asked a lot, and you guys saw this when Margot was raving about it. That was completely organic, right? And this is what— the usual WhisperFlow user does. And the question always is why? How do you create such raving fans? And the answer is just really simple. It all comes down to why do people refer anything to anybody else? When you found great food at a good restaurant, why do you go text your best friend that you found this? And whenever you meet somebody, you tell them something that you're recently really, really excited by. It all comes down to you had an incredible experience, and I'll talk a bit more about what makes an incredible experience. And it is just human behavior that you want to share that with somebody for them to have that same level of happiness, delight, pain relief that you had. And there's a part of you, there's a selfish part of you that wants that when this person feels this, they want— you want them to attribute a little bit of that to you. That's because of Hans or because of Margot, I found this tool and it's incredible and this is fantastic. And that is what matters the most. And so if you just think about it that way, it becomes really simple. How do you create that delight? Well, if you think about what has surprised you that way, it was ChatGPT about 3 years ago. It was Whisper Flow for many people. It was Claude Code. In January of this year. And all of them, very simple philosophy. Number 1, they surprised you. Number 2, they changed your worldview because they showed you something was possible that you did not know about before. And the third point, and this is the most important one, is that for the amount of value they added to you, the friction was extremely minimal. This third point a lot of people miss out on. So you can have the first and second one, but if your experience is bad, if you're asking a lot from the user, if it's not as seamless as it really should be, you don't get the compounding word-of-mouth effect that happens. And this is basically now in the company something that we just treat as a science instead of an art, where in every single product, in every single feature, in every single thing that we do, This is the first thing, or after it hits PMF, this is the first thing that we optimize for, is how do we make every single one of our users a champion and an evangelist for the product?

I mean, I remember back in Airbnb back in 2014 and Brian Chesky had this philosophy is that, and I'm paraphrasing here, let's not have a million people like you, focus on getting the first 10,000, 100,000 people just love you. And that definitely is the user experience that you strive for here. When we made the investment the first time, from the very initial time, I remember one of the reporter from TechCrunch in India called us and he said something was interesting. He said, He knows that, you know, the US market is our number one market. You grow very quickly here, but what he sees on the ground in India is that it's just Whisperful provides great access to good business English. And for anybody who, like what you asked me to do, to say something in a non-English language, it translates so well into English, so it's easy to share that information within your team or within your organization, that secondary impact is actually could be as great, if not greater, than original value proposition. Was that something that you guys noticed early on? Is that something you're trying to capitalize even more? How do you think about that secondary effect?

I think you hit the nail on the head. So the way you get 10,000 people or the first 100 people, which kind of, it can keep scaling up beyond, to absolutely love you is you need to make the product feel, you need to make the people feel like this is a product made for people like me. And when you do that, that starts to become a part of their identity. So for example, when we were launching in India, So this is basically like— this has been something that we've been working on for the last 2 months. And there were a couple of things we noticed. The first one is India is an Android-heavy market. OK, check. We need to do that. Second is India is very value-conscious. So you can't have the same pricing mentality that you have in the US apply to India, right? Netflix costs 99 rupees, which is about $1.05 per month. And then— The third thing is a lot of people— not a lot, basically everybody is multilingual, right? I grew up learning 4 languages, and so did most of my friends. At home, you speak a combination of Hindi and English. You don't speak one language fully in any situation. But the thing is every single other product makes you speak in American English. And if you are anywhere else in the world, you have to adapt to the tool. In India, our promise was we're meeting you exactly where you are. You don't have to change a single thing to use Whisper. And that messaging just struck the hearts of people in India. And we did a big launch yesterday. Today is Tuesday. Yesterday. And so if you go to my LinkedIn, just see the video. It's 1.5 minutes long. And that's what we made to really make people feel like, hey, this is not an American brand that is coming to India and trying to shove their product down our throat. Like give us money. But it's really a brand that understands the country, that understands our life. It had me running around in an auto in a noisy, crowded street using Whisper on my phone while talking to the auto driver, which is just how life looks like. And that has gotten India to basically scale in both revenue and user base at over 100% month over month over the last 3 months. The key thing is when you hear the Airbnb story, you use that for the first 10,000 users, I don't think you should ever stop. 'Cause for me, we did that with the initial founder group, the initial developer group. Now we're doing it for India. Then we're gonna do it for every single country that's out there. Whether we go to UK and tell them how we're the only AI tool that supports British English, or how when we go to Australia, we're the only product that supports the very unique Australian slang that only exists there and the very unique ways of speaking. Or when we go to a salesperson or a lawyer and tell them, like, if you say never, wherever, therefore, wherever, it just writes that as one word in your complex legal jargon. That is just something that is, again, not a one-time thing. It's an engine to just use repeatedly. And at the end of the day, it is all these second-degree factors because these second-degree factors are really what hit the emotions of every single person because at the end of the day you are building something for people.

Yeah, a lot of people initially look at you just as a voice dictation app because that's what's there today. If you fast forward say 5 years from now or even beyond that, I don't know, in the world of AI, 1 year, 2 years makes a huge difference, but if you, you know, really squint and think far ahead, what kind of company do you think Whisperflow will end up becoming?

So the answer to that question has not changed in, at least for me, the last 18 years. I've been working on different kinds of voice systems for many, many years. And to me, couple of years later when I walk down, go down on the street, I don't want to see anybody hunched over their phone doing this all day long. Like, this to me is depressing. It's not the world I want my kids to grow up in. For me, the ideal is you go downstairs and you see everybody looking up. They might have something in their ear that they're talking to, but for the most part, technology fades into the background so you as a person can be a lot more present. That to me is the vision for Whisper. That is why We decided to do this when we could have done literally any other thing in the world. And that to me is what I had as the dream when I was 10 and I saw Iron Man for the first time and I wanted to build Jarvis. Because to me Jarvis is less of what it looks like in the movies. It is this vision of the system that just gets you. It's with you 24/7. And you trust it to do things on your behalf. And we basically boil down from that, OK, if that is the vision, that is the kind of world we want to see, what are the tactical steps to actually get there? How do you get there? Well, number 1, you need to build voice input that just works. It's reliable. People trust it blindly. If you don't have that, you can't build anything else on top of it. Phase 2 is Whisper doesn't just write for you. It's able to do things on your behalf. It becomes your confidant who you want to tell everything to. It helps you on a day-to-day basis in small ways without being too intrusive. And then finally, gets to a point where I can say, Hans, you know this thing you love using on your phone and laptop? Imagine if you didn't even have to take your phone out of your pocket because Whisper lives in everything else around you, be it your smartwatches, smart rings, smart glasses, and everything else. And while everything else is trying to be a system of record, all I want Whisper to be is a system of you.

You talk about we, talk about doing this, want to do this for a long time. You and Sahaj were initially dormmates and then roommates at Stanford. How do you guys decide, how do you guys meet? How do you guys decide to partner together and what has that journey been like?

Yeah, so it's really funny. Saj and I met first day of undergrad in our freshman dorm and we started bonding over standup comedy and then we roomed together for the next 3 years and we both gave each other imposter syndrome for all of this time. He was a prodigy. He was one of the youngest researchers in Stefano's lab, who some of you might have heard. I think his talk was a couple of firesides ago. And Sehej was one of the people who invented diffusion models in a lab filled with PhDs. And I, on the other side, was extremely focused on product and scaling things. And I'd been writing code since I was 9 years old. We never thought we would work together because I wanted to build a number of software companies. He wanted to go do AI research.

Right.

And beginning of 2021, Saad gave me a call and said, I kind of want to build something. And then we started talking about what were some of the biggest problems in the world that we saw that nobody else wanted to work on. And the biggest one that came up was computers can understand natural language. And so the way we interface with them is going to change. And everybody's building the capability layer, but nobody's building the right interaction layer, what it looks like. And that is the problem we wanted to solve. And we actually spent 6 months, because he was my best friend, right? I was the best man at his wedding. He's going to be the same at mine.

You finish each other's sentences.

Yeah. And so I was like, I want to make sure we preserve the relationship. So we actually spent 6 months talking about the key things that matter to us, how we would navigate different situations if we got X dollars of an acquisition offer, like what matters to us really. The thing that we both aligned on was life is short. We could focus on building a business, get it acquired, and make some money along the way, but I don't know, that doesn't feel that exciting enough when the alternative is like, hey, you have your 20s and 30s to do the biggest and best things in your life. We wanted to spend that time taking the biggest swings that we can.

I mean, seeing you guys grow 7x in 6 months is, I mean, I've been involved with over 10 unicorns, but 7x in 6 months is still right up there as among the fastest ever. But you guys, the chemistry that you guys working with each other and the ability you have finishing off each other's sentences, was what really draw us to wanna bet on you, the two of you, when we met you. For, if you have to give people in the audience one advice, like how do you fight through the labels people give you? Initially a hardware company, people think that you're X, then you turn into Y, and then people say, well, are you really up for doing Y? How do you fight the labels and morph your skill set to become this great go-to-market expert for an AI tool versus a hardware company when people first find you?

When we pivoted over to Whisper, most people told us that we don't have the DNA to build a consumer company. Most people told me that I don't have any idea what marketing is going to be like and this is not going to work because of that because you were very technical founders. Like, I've been an engineer my whole life. But I think the biggest thing is that most things in life are simple, not easy, but really simple when you boil it down to the basics. And so whenever I'm at least trying to do anything new, just boil it down to like, what is the absolute core essence of it? And can you just do that really simple thing? And in my case for marketing, it's— marketing's really simple. You need to give a very clear and crisp message, deliver it to the right person at the right channel. And if you're building something meaningful, then that is going to work. And that is the simplest thing that I keep telling my team to cut everything else out. Because if you just keep doing the simple thing repeatedly, you're already 90% of the way there.

So reducing a hard problem, complex problem, to something simpler based on first principles is what has served you guys well through different iterations over the last 5, 6 years of doing this.

Yeah, every complex problem is just a bunch of simple problems.

On that note, thank you very much for your time today.

This is great, Hans. Thank you.