Raw Day Inside Wispr Flow

It's kind of unprecedented [music] the speed at which AI companies are able to scale right now.

30 40 50% month-over-month growth is the norm. Building the infrastructure to

norm. Building the infrastructure to keep up with that kind of growth is really challenging.

Early on in a startup's journey, [music] every founder is out there pitching, chasing investors, vendors, customers, [music] chasing anyone who listen. But there

comes a point where the tide turns, where suddenly everyone wants what you have. The thing you take is that AI is

have. The thing you take is that AI is only changing execution, writing the code itself. That's not true. It's

code itself. That's not true. It's

changing a bunch of different things.

In a world where we're all just exchanging slop, maybe flow [music] can help you sound like yourself while still helping you work faster and express your ideas more efficiently. Today I'm going

[music] behind the scenes with Whisper Flow, a startup that's raised a total of $81 million with their recent [music] round valuing them at $700 million to

revolutionize voice to text as a communication tool around the world.

We're not a destination. We're something

that puts adrenaline or gold dust into [music] your existing stack or into your existing workflows. That seems to be

existing workflows. That seems to be hard to communicate and we try in many different ways, but really seeing is believing. I strongly believe that a

believing. I strongly believe that a keyboard and a screen are just not a good way of interacting with a device and that voice is the most natural way that we communicate with another person.

Those two statements I can actually just believe no matter how much the world changes. [music] The question no longer

changes. [music] The question no longer remains whether the product works and whether the market has conviction. The

question now becomes how big can this truly get and can [music] they continue to deliver on the level of growth that they've maintained all this time.

Ah, here's a boy.

What's up, man?

How are you?

How are you? How you doing?

Just giving them a little tour of the office for Well, this tour is going to be irrelevant.

Yeah, because in one day we're moving. I

mean, today we're moving, but [laughter] later today.

Later today. Should we get the last glimpse before you guys leave? This

is our all hands meeting room. A bit

crazy. The team has grown a lot over the last two years. So before pretty empty and now you can see we have the bench over there, the chairs around here.

Man, you going to miss it?

Yeah, I guess a lot of memories, but I mean I'm excited. It's going to be cool.

The offices upstairs are sick. You know,

we started in hardware. Did you know that?

I had no idea.

But basically, you put this on like this and it could like read your movements and then like I don't know some crazy stuff. I'm not technical so I don't know

stuff. I'm not technical so I don't know exactly. But

exactly. But as in what like connect it up to like like electro pulses in your brain or something? Wow. So, yeah, some crazy

something? Wow. So, yeah, some crazy stuff.

[music] [music] Hector's just mentioned you guys going from hardware a little while ago. So,

could you tell me about that journey a little bit? So our goal has always been

little bit? So our goal has always been to build a voice interface that a billion people can use every day as their primary interaction with all their devices. And we started the company and

devices. And we started the company and we thought, hey, the thing that's going to block people is their ability to use voice around other people. Like I don't want to speak to my laptop when I'm sitting right next to somebody. And so

we started by building a wearable brain computer interface that could understand when you're thinking silently to yourself. We had a team of

yourself. We had a team of neuroscientists, hardware engineers, data collection technicians, all working on building towards that. And we

actually had a functioning prototype.

And what we realized as we started using it was we didn't even want to use the best voice interfaces that existed then when we were sitting in a closed room.

So what were we going to do? Well, we

had to build the voice interface, the software for the hardware. And that's

actually what led to the birth of whisper flow. And we saw our own

whisper flow. And we saw our own behavior shift. We saw the things that

behavior shift. We saw the things that we were doing changing. And that's what actually led us to put all of our energy into building that.

So I know we got a standup happening, right?

What are [snorts] we covering in the standup this morning? what's going on cuz we'll be in there with you guys.

There's a couple different things. So,

the first one is about how people are developing new workflows using AI. I

think everybody's doing lots of different experimentation about what they like and dislike and people take for granted the things that they're doing and don't always share it with

everybody. And so first half of it is

everybody. And so first half of it is about talking about that and the second part is actually just updating people on the different developments on the ML side, on the research side, on the product side for the new features that

we're building.

[music] [music] As a reminder, as always, we are focused on current devs. And the reason why is because we want to be building the first voice interface that a billion people can use every day as their primary

interaction. Are there things that you

interaction. Are there things that you do that you don't know that other people are doing that you think might make them more effective when using AI?

The way that I organize my thoughts around this topic is that I generally bucket things into one of three things.

So, first bucket is quality. So how this is where AI helps me get to a better answer than I would have without it or on my own. The second one is

compression. So AI gets me to the same

compression. So AI gets me to the same answer but it closes the gap or makes it much quicker things that took a week.

And so that's increasing the velocity of the uh important decision- making I need to make every day and so there the focus then comes well what decisions should I focus on? And then the third bucket is

focus on? And then the third bucket is uh project management. So, how is AI making sure that all the things I committed to in my day across a milestones and timelines actually get

done the next day? And I found that that's been helpful then because to deconstruct the ways that AI can help you rather than thinking across the board. This is a little weird or

board. This is a little weird or uncomfortable thing where I kind of believe road maps are kind of dead outside of high level direction. So,

it's like this is the direction we're heading. But if I can scope it down to

heading. But if I can scope it down to we're going to do this, then this, then this, then this, then this, and I can scope it to any detail, then it's almost done. And so it's kind of like, okay,

done. And so it's kind of like, okay, great. We're like able to rapidly

great. We're like able to rapidly execute, and then once we do a thing, we learn the next thing to do. Like, we

couldn't have known the next thing to do until we did the first thing. I know

where we're headed. I know that we're going to have to solve many problems along that way. And how we solve those problems and the specificity of that road map is going to be a lot bigger than it probably used to be. And one of the things that we've learned in content

centers this week is that how you're prototyping should also match the problem you're trying to solve. And so

for example, now that it's easy to jump to a prototype in cloud code or any other tools, like maybe a really good skill or a really good prompt where you're learning what the actual functionality is that you want, the

output that you want is actually what you need to learn. Um, and so it's both finding a problem and figuring out the right type of prototype or the right level for what you're trying to learn because for me personally, sometimes

I've jumped thinking about the UIUX when actually I just needed a damn good prompt to teach me what I wanted it to do and then get to that.

[music] [music] How was the morning stand out this morning?

I had a good walk about yesterday with Sahes. This notion of like what is the

Sahes. This notion of like what is the road map when the road map can change completely in a week or two just based on the velocity of what you can ship.

It's a really interesting challenge. So

increasingly we're really focusing on like how do we build a brand that resonates and is distinct not get lost in the fast growing tech bubble. Right?

I use the product every day for four or five months. And so I know that once

five months. And so I know that once people get into the product, things work beautifully from there. Meaning we

activate very well. People build a new habit after a while they're pretty comfortable paying for it and then they stick around for a very long time. That

whole part of the funnel is bestin-class of any software company I've ever worked at or seen. So the biggest challenge is no one's heard of us, right? And so a disproportionate amount of my time is

thinking about well how do we change that? We currently have EU campaigns

that? We currently have EU campaigns which are reaching the UK with generic content. So are you able to see

content. So are you able to see benchmarks from those campaigns and then compare the performance of this UK?

We can do the country breakdown within EU. We haven't done the baseline

EU. We haven't done the baseline comparison because that for me is really the real we're trying to understand which is should we bother doing specific stuff at a country level. My inclination just

because we're bucketing multiple markets at the same time is metas in like team efficiency right so it's just going to serve to the most likely candidates

sometimes that means you sometimes that means p Germany where wherever so it's not necessarily like a clean onetoone comparison but that is a good question in that like maybe there's some insight

to help us build the baseline for this this market speaking of building baselines for this market we also have India specific creative that has been running with Namisha. Um we have two

separate um well we've had our Android uh campaign going since launch but we now have a desktop specific testing campaign within India um with quite a

bit of variety of um UTC based [music] ads.

[music] [music] So basically the idea is that when you're in office setting like if you're dictating and like it can pick it up well with like your MacBook but number

one if your MacBook is closed the mic isn't isn't enabled so won't pick up the audio and then secondly you also feel kind of awkward because you might be like speaking a bit louder than than usual. So with the mic, if you have it

usual. So with the mic, if you have it like up close here, I'll show you guys how good whisper flow is. Actually

whispered into the product. So

hey, let me know if you're running me like wild. And I was running behind.

wild. And I was running behind.

[laughter] But yeah, so if with the mic then if if you put that into consideration, then people are like working and they can just be like whispering without creating like a lot of background noise in an office or in like an open space like

this where everyone has mics and we're all talking with Whisper. Um it creates for like a less noisy environment. And

the idea is that if we send these to offices, um then Whisper will be adapted faster and more people will start using it and they'll be like, "Oh, why do you have a mic on your desk?" And it's like, "Oh, I'm using Whisper." And they're like, "Oh, like what is that?" And then

it creates like that open conversation to people for people to get introduced into whisper.

So as a linguist my main role is to kind of bridge the gap between the ways in which we speak with all its inherent messiness. So we put in a lot of filler

messiness. So we put in a lot of filler words like um and like. We also speak in long run on sentences in general. And

what we expect when we write is to see not that right. We want precise sentences. We want punctuation. Things

sentences. We want punctuation. Things

that are implied by what we do when we speak but really aren't there. And so my job is to kind of bridge that gap so that we're still honoring what people said in the tone that they said it in the words that they chose by meeting

their expectations as if we are writing exactly as if they would have typed it themselves.

[music] I saw you writing a lot in the morning stand up this morning. Tell me about that.

So, I think what's really cool about flow is it helps you clarify your thoughts. Um, but there's no substitute

thoughts. Um, but there's no substitute for pen and paper to get down certain types of thoughts. We're saying that voice is one way to share your thoughts and it's fast and it helps you be very

conversational and natural and that helps me get out more things than I would have if I were just writing all day. So each has its place. But yeah, we

day. So each has its place. But yeah, we write to think, we talk to think, we talk to express ourselves. And flow

[clears throat] is just one way of capturing that, making sense of it, and helping me do it a lot faster. And every

afternoon people would come and do some follow-ups. I would take notes. Um, and

follow-ups. I would take notes. Um, and

then we also did a hanging contest.

That's where I made the leaderboard. And

this actually came from on the day of our launch we like initial desktop launch we had a product hunt out and at every hundred up votes I was like let's

all like someone has to do something. So

we had people do like pull-ups like at every hundred we had to do an activity.

I've got a question here. Where are you on this list?

Well I did a challenge with Tobin and loser had to buy sushi and let's just say I didn't pay for sushi so I won. But

I'm not I I did I never wrote my name on here but I have I have video. Oh, I'm

sure. I'm sure. I have video evidence. I

do. I do.

The audience is going to have to see that. It's like It's like picture it

that. It's like It's like picture it didn't happen. [laughter]

didn't happen. [laughter] I saw you got mogged in uh one of the YouTube videos where the the founder started doing muscle ups and then you got called out. You're like, "No, no,

I'm not on peptides yet."

[music] Right now, you speak and Whisper Flow will write for you. But we also wanted to do things for you. And so we have a team spun up that is running 20 experiments a day of how we want those

actions to work.

For a long time in in different environments saying meant not getting stuff done. [music] And so that paradigm

stuff done. [music] And so that paradigm shift we are actually saying is reducing the time from idea to capture and capture to execute is a really positive thing.

That is something that is just so special because what we sound like is what makes us individually us, right?

preserving people's voice [music] and authenticity when every other tool seems to just be saying like do it one way do it faster and that standardization is just something that like I worry about as an industry [music]

if you know that people want exactly what you're building on the research side makes a ton of sense to spend a bunch of time and energy building one big bet that you believe that you hold

and if it's an observation about the world then it's not even just like a thing that you believe that could be false that's [music] the only thing that's like really kept me grounded

personally with all of the the ways in which things change. [music]

Yeah. So, you're like dictating like tasks to employees or whatever. Let's

just say our product is pretty damn