Dan Siroker: Giving Humans the Superpower of Perfect Memory

23:02And if it's a feature nobody will use, then there's any amount of effort building it is wasted. I want to talk a little bit more about you leaving Optimizely. Do you think that that was the result of you having so much exposure to analytics between your role at Google, the Obama campaign, and then all the time at Optimizely? I think I definitely told myself I never wanted to do analytics again after the campaign, mostly because I conflated the painful experience in my body with the activity at the time. There was nothing inherent about analytics that I didn't like. In fact, I think that's still an area that I'm very excited about.

23:33I think data-driven decision-making is still a huge opportunity. So it wasn't the category. It was more the specifics of the mission and purpose. And I'm generally very, like I said, optimistic. I could get excited about a lot of things. And I just didn't. I actually probably made the mistake of thinking that my job, I got this advice probably from somebody. I thought my job as a CEO of a 40-person company in 2013 was to hire people better than me and then get out of the way. And what I did was actually hire really great product and engineering leaders. I wasn't that great at hiring sales, marketing, and customer success leaders because I just wasn't that good at that stuff.

24:08And so I ended up spending all my time on sales, marketing, customer success because I'd sort of hired the parts of the business I could do well, I knew. And I sort of didn't get a chance to be connected to the product. I didn't get a chance to be connected to users and customers as much. And so most of my day ended up being areas of the business I didn't love. And a lot of just as the company gets bigger, you deal with people issues all the time. And when we're 400 people, all of the issues were just people issues. And so So that also was very draining. And I think had I been more like Lou Cerny, who actually was somebody I met with, and Lou Cerny was then the CEO and co-founder of New Relic.

24:43And Lou was famously this CEO who codes. But I thought that wasn't my job. I thought I would be selfish. And in hindsight, that was probably one of the biggest mistakes I made was I was kind of pennywise pound foolish. I sort of did the things I thought the company wanted, and then I hated my job. And then ultimately, by hating my job, I wasn't as good as my job. And then that sort of, you know, that meant that we didn't reach our full potential. So you leave Optimizely and at some point you start Rewind. Talk to us about that journey between leaving this one amazing company that you built into now starting what you're working on today.

25:14Most people will tell you that if you're going to do two startups, you should take a break in the middle. Actually, a lot of people tell you they forced you to do that. I kind of did the opposite. I actually founded the company that now is Rewind before I finished selling Optimizely. Oh, wow. So it was something that I had, I was, and as I said, I had been called to do this from this experience of trying a hearing aid. But the time between that moment and actually starting the company, that was about seven years. So it was quite a bit of time before. And honestly, that also played a role in my day-to-day satisfaction at Optimizely.

25:45I had this grass is greener feeling about, boy, wouldn't it be amazing to use technology to augment human capabilities? Or do I make websites slightly better at converting? It's like when you have that choice in your mind. It's definitely hard to stay in love with what you're doing. I don't have any of the sort of feelings of being trapped. That's the other challenge I had at Optimize. I kind of felt trapped. I wanted to do this other thing, but the founder of the company, I can't just leave it and go do something else. So I felt very trapped and that also created a lot of resentment. So yeah, and ever since I started Rewind, I think my worst day at Rewind is still probably way, way, way better than the last any day my last five years at Optimizely.

26:27It definitely from afar and seeing sort of just the pace of shipping and new products that are coming out, it definitely seems like the team is in a zone now where you must you all must be having a blast right now. I just see all these announcements and these new things getting shipped. I mean, I recognize it from from when I built my company. It looks like you're having a blast. We are. We are absolutely having a blast. We're hiring, too. If you want to have a blast, and join a 20-person company. I mean, and we did so many better things in terms of setting up the team to have a blast. You know, we hire only really experienced people.

26:56We give them a lot of ownership and autonomy. So like today with a team of 20, we're doing far more than I think we could have, we did when I was optimizing the team of 400. So I'm a big believer in a small nimble team is able to do much better, ship much faster than a large bloated team of people. And so that's another big advantage is that also that minimizes the overhead as a manager, as a leader. I don't have to deal with all of the people problems that come with hundreds of people. I can focus on the things that matter. I can focus on problems, on customers, on product, and not on sort of man-made artificial politics and people jockeying for titles and positions and all that stuff.

27:31So let's talk about the product. I know what it is, and I'm sure some of the listeners do as well. But why don't you give an overview of what Rewind does and specifically about the product itself? Rewind is a personalized AI powered by everything you've seen, said, or heard. We started with a Mac app and actually launched it less than a year ago. It's kind of crazy to think that's gone so fast. We launched it on the Mac. It runs in the background, captures what you see. It does optical character recognition on what you see and turns that into text, stores it all locally. So we don't have access to it.

27:59Your employer doesn't have access to it. It's very private. And does the same thing with audio. So meetings, you can transcribe everything you say, everything you hear, stores it all locally. Then with all of that data, you can do simple things like rewind time. That's why it's called rewind. You can just rewind back saying, what was that thing? I just saw or I read, and you can copy and paste anything from your past. You can search through anything you've seen or heard, and you can also ask Rewind. We have this integration with a large language model where you can actually ask any question. A good example of that is you can ask Rewind to write you an email.

28:29There's a demo on our site I show where it actually asks it to write me an email to Sam Altman asking to catch up. And it uses all of the context of my past. It starts with, hey, I remember fondly being introduced in 2010 by Paul Graham. It's great to have you as an investor in the company. I want to give you an update on some of our latest progress. All of those things are all possible with a context of everything I've seen, said, or heard. So that's what we did on Mac. We've launched on iPhone not too long ago, and we're working on Windows today. And also, more recently, we launched the pendant.

28:59We launched it into pre-order. It's not available yet, which is a wearable, an AI wearable that captures not just what you experience on your digital devices, but out in the real world as well. In a very privacy-first way, we capture what you say and you hear, store that again all locally and respect the privacy of other people you meet. We don't record anyone without their consent. And that builds a much better context to really give you the superpower of both perfect memory and a much more productive, much more efficient way to live your life. There's so much to unpack here. You mentioned that the application stores everything locally.

29:35But also now you're cross-platform, right? you're on my desktop, you're on my iPhone. How are these things communicating with one another and making sure the context travels across device while storing everything locally and securely? Yeah, so today it actually does not travel across devices, which is one of the top requested features we get. We're actively at work on end-to-end encryption across devices. So it is able to give you the full context of everything you've seen, said, or heard in any device. That's a really hard problem to get done well. So it's gonna be a while before we have that available, but that's something that we're definitely very keen on making possible because that's really the promise of our product is how comprehensive.

30:10On the Mac right now, it's not like an integration with these three silos. It's not integrating with these different tools. It's literally everything you see, any app, any experience. And when you extend that metaphor to your iPhone, that massively improves the comprehensiveness. And then you extend that metaphor to the real world. And you finally get so much context that you ultimately could build a clone, a digital clone of you to do the things you do every day and save you time. There's so many things a bot can do for you in your day-to-day life that will save you time. And we think we can do that in a way that respects your privacy.

30:41That's a huge win. It feels like a no-brainer. When I first heard about Rewind, I was like, oh, wow, amazing. If I could have perfect memory and I could remember everything, that would be incredible. But it also feels like there's a little bit of a difference between sort of capturing and storing and then me as a user kind of tapping into that memory. And it seems like you're coming at it from multiple angles. How do you think about creating that shift in behavior for a user where they have to almost remember that they can remember now to get value out of the product? That's a question we've thought a lot about.

31:16And it's hard. Behavior change is hard. Regardless of what the behavior is, people are ingrained in their habits. There's a hook model, which is this idea that you need sort of an external stimulus to get you to remember to use the product and then sort of variable reward. And that sort of builds this feedback loop. One of our most successful features, which is meeting summarization, does this beautifully. Instead of having to remember to use Rewind, it just pops up after the meeting, asks you in one click, do you want to summarize and draft an email to all the participants in the meeting of the meeting?

31:47And that is like a really magical experience because for me, I'm in back-to-back meetings. I don't have time to take notes. I'm mostly trying to be more present. If after every meeting, I could look like a genius and just send you like to hear the 10 bullet points that we talked about here, the next steps and action items. That's a huge win for me. It's a huge win for everyone I meet with. So those are the kinds of features. We have several other things like that in store. I don't want to give any secrets away. Some of the things I'm really, really excited about that don't require you to remember to use Rewind.

32:14It just interjects itself into your existing workflow in a way that just feels delightful. And that's what we're really focused on is how do we take the power of everything we've captured, but now make it so easy and frictionless for you as a user to not have to remember to use Rewind. Right. So it's really about the products or taking the initiative to sort of prove the value rather than me saying, oh, Dan and I were having this conversation a year ago. What was it we said? Is that how you think about it? Yeah, I think so. I mean, I think that's helpful, but it's much more rare. And it goes back to my point around people go around their life with blinders.

32:47They don't know what they don't know, and they don't realize what they forget. So many things we capture are so valuable to somebody, but they don't have the trigger to go back and ask or think to ask or even know what to ask to go back to that. And so that's where we feel like if we can avoid needing somebody to remember to do that, instead show up elegantly because we know exactly the opportunities to do that. If we show up elegantly in their existing workflow, that's a much more compelling way to create value for them. What are you finding to be the most common ways that people do actually engage with the product beyond it, obviously capturing and storing and encrypting?

33:21Yeah, there's two classes. There's the, you know, find something from your past. And I gave you some examples of that. And then there's the turning your past into action. You know, finding something from your past, very often it's this feeling. If you've ever had the feeling of this tip of the tongue feeling, like, what was that thing I saw? Where did I see that? Very often, most people will then go into their email, search for that thing. They can't find it. And it's because it was an iMessage. It wasn't an email. Or it was a Slack message, not an email. Or they'll be searching Google, trying to find it.

33:48But they can't find it because it was a tweet, not a website. So that sourcing problem is very common. Because we capture everything you see and you just search for the word or the thing that was related to the thing you're looking for, people are able to get that. That's kind of a class one of you's case, which is find things from your past. Class two is turn your past into action. And that's where having the context of how you know somebody is really, really powerful. That's where you can sort of figure out, you know, you can, right before every meeting, by the way, we have a, how do I know this person?

34:13You can click it and it sort of goes back and figures out all the references to that person. You can, you know, you have a draft emails for you. Anything that ChatGPT can do, with the exception of the more recent multimodal stuff, that you can do with Rewind and is powered with the context of the things you've seen, said, or heard. You mentioned ChatGPT. You said you started this company, I think a couple of years ago. So did you always think of this as an AI company? How are you using AI exactly? Talk a little bit about that. So it was always an AI company, which is funny because I'll just draw an analogy to Optimizely.

34:46So in 2009, 2010, when we started Optimizely, we were in this sad category of companies that people didn't really think were sexy and nobody wanted to work on called SaaS, software as a service. Everyone wanted to be the next Facebook. It's kind of crazy to think about this. So everyone was into social consumer. That was like, and we were this like weird, like you want to do like SaaS, B2B? And then all of a sudden, it was the hottest category ever. Two years later, everything was SaaS, SaaS-ter. All these venture firms pretend like SaaS-ter. It was so weird to have been doing the thing for a while.

35:14And then the world around us conspiring to say, that thing that you're doing is now a good idea. And the same thing happened with Rewind. We've been doing this since March 2020. We had no idea, even though Sam Altman was actually our first real investor, we had no idea about what the power of GPT-4 could offer. But we started with a product. It actually was a different version of, or it was a different product. It was called Scribe. It was an online meeting bot. It would join meetings. It would transcribe the meeting. It would use AI to do speech transcription, taking what was said and turning that into text.

35:47And it was growing well, doing okay. We had this realization that, wow, Apple Silicon is really powerful. What if we did the magic of what Scribe does for meetings, but for your whole life and all locally? And that's what finally connected the dots. So it was a key enabling technology was Apple Silicon, you know, M1, M2 chips. That's what enabled it to happen. And then while we're working on that, all of a sudden GPT-4 launches and that becomes this amazing, like totally by accident. I had no prescience or no sort of vision that this would occur. We just happened to be collecting along the way the most valuable data set that you could use as context for GPT-4 to be able to actually make GPT-4 personalized to you.

36:21That's fascinating. And, you know, I think like other AI companies, it's been, the timing has been very, very beneficial For the listeners that maybe don't know about fundraising and raising venture capital, you know, the traditional fundraising process is, you know, you go either to a meeting or over Zoom and you pitch to venture capitalists. You had the benefit of so much excitement around AI that you were able just to post a video and say, hey, venture capitalists out there, if you want to invest in Rewind, send me a note, send me a video, and maybe we can do something together. Tell that story because I thought that was pretty fascinating.

36:57First of all, we didn't need to raise money. We actually had just raised from Andreessen Horowitz a seed round. We had, I think, well over two years of runway. So it wasn't the fact that we had a budget of hiring that we needed to fill. It was actually two things. First, we felt like there would be an opportunity to tell our story and explain more about the company that would build a ton of trust with users. And we felt like a lot of people were concerned. They said, I love Product Rewind, but what happens if you go out of business? Or what if Google comes by and buys you? So we really wanted to be clear that we are a big, strong, independent company, financially stable.

37:30We're not going anywhere. And then we had the thought, well, if we share that publicly, why not just put out the pitch deck that we would show investors and customers at the same time? Sort of pitch our customers the same way we would pitch our users. I did like a seven-minute video of our pitch deck, which is basically the same pitch deck I gave to Andreessen Horowitz a few months ago, both updated numbers. And it took off. millions of views on social media, over a thousand offers to invest, a bunch of term sheets. I highly recommend asking investors to put in two numbers, the amount they want to invest and their value and at what valuation.

38:03And with that, I had such an amazing data set on the actual market price of our company. A lot of people thought, oh, this is, you know, you know, you shouldn't, you know, we were also very early on. We had$707 ,000 in annual recurring revenue. Most typical investors like to tell you, you shouldn't over get across your skis on valuation And like, you know, but like I had market data. I can tell you, I have a thousand data points of what people think we're worth today. So like, I'm not going to go crazy and choose like, you know, we had several people offer us over a billion dollar valuation. We ended up settling on a$350 million valuation, which is still very high given our revenue.

38:31And I basically cherry picked the very best investors. I also had a little field, like anything else I should know. And one of the things I loved about this was like, this was like the opposite of how venture capitalists, how ego-driven venture capitalists like to interact with founders. What they would love is for you to drive to their office in Sand Hill Road, kiss their ring, wait half an hour in their lobby. This is the opposite. I turned that experience into them having to humble themselves to submit into a little form, just like everyone else. And they have to put a little blurb out. And the investor who I ended up taking as the lead wrote an amazing letter or sort of paragraph of why they wanted to invest.

39:07They actually sent me a, I think, 30 or 40 page PDF about why they were the best firm for us and what they believed about our category. And they truly wowed me in that process of clearly, showing their commitment. And so they were one of the finalists I got to meet. And then I ultimately chose that investor as NEA because we were so aligned on values. They're very long-term oriented investors. They invest their fund life cycles 15 years, not 10, which is typical. Oftentimes for companies like CloudFare and others, they're actually buyers at the IPO. So because I want to be doing this company for the rest of my life, I view this as something I want to do for decades.

39:40I want to work with people who I'm going to work with for decades. And that's why this process was so amazing. I could cast a wide net, find a good market price for the company, and pick the best investor who really aligned on values. Another analogy I like to draw is to online dating. Most venture capitalists, most entrepreneurs venture capitalists meet at the proverbial bar. You go to a bar down the street, if there's a person there that you happen to like, you hit it off. Maybe that's a way to date. Maybe that's a way to get to know people. But there's so many other people than the people who are at the proverbial bar.

40:10And by casting a wide it's kind of like online dating. You go out there, you see who's a good fit. It's a numbers game on both sides. For some investors, we're a great fit. For others, we're not. Similarly for us, there's some investors, great fits, some aren't. And so by casting a white net, you get this sort of match made in heaven that you wouldn't get if you just went to the local bar down the street. Yeah. Let's dive back into the product a bit. Talk to us how it actually works. How is Rewind seeing and understanding everything on my screen, on my phone? And then we're going to get into the pendant, of course, as well.

40:37But maybe let's start with the iOS and macOS products. It captures passively the things you see. And so basically the way you can think about it is it sort of takes a screenshot of your screen every two seconds, then it compresses that. So it can take that data, which is actually quite large. You know, each of these screenshots of your full resolution monitor is quite large. It compresses it and compresses it exceptionally well.

41:03So it way we can do that, by the way, is there's a ton of redundancy. Even I'm sure as you're watching in this meeting, you can see me gesticulating, but I don't know if I'm full screen on your screen. And there's a lot of pixels that aren't really changing on your screen. So like the most part, most knowledge workers interact with their computer and most of the data that appears, most of the pixels that appear don't change frame to frame. So step one, compress that, effectively compressing it so that you could have indefinite storage. You know, it's about 14 gigabytes a month. Most hard drives you buy today, that's multiple years of data that you could store.

41:32We're also working on cloud and an encrypted cloud backup. So eventually you can put it into the cloud if you want. But that's step one is compression. Step two is OCR. So then we take what's on your screen and we do optical character recognition. Every word that appears that's pixels, we turn that into the actual English word. We store that in a local database that's encrypted. Then we can query later for those words. And then we do a sort of analogous process to the things you hear and say. So we take the audio that comes out of your system audio, the audio that goes into your system microphone, we transcribe it.

42:03We use a model that runs locally on your machine called Whisper to do that. That turns into words that, again, you can search later, source it all in the local database, and you can go back to and play back that audio. That's the sort of foundation. And then on top of that, we have an optional capability, which is asking rewind questions. That uses GPT-4. That's the only thing that sort of everything I've said so far that connects out to the internet. It goes out there. It sends only the snippets of your past that are related to the query. None of that data is used for training. It's only retained for 30 days by OpenAI for spam and quality purposes.

42:37And that data then gets taken from OpenAI, from ChatGPT4, and turned into a response. It sends the context to GPT when I'm asking the question. So help me write an email to Sam Altman. I, as the user, don't see it. But under the hood, it's taking this package of information that Rewind has built and it's sending it to ChatGPT to help form an answer. Yes, a very, very, very small portion of all your data, not your whole data, just the portion needed to actually answer that question. Got it. OK, so I understand how that can work on macOS. iOS, I'm guessing, is very different. Yeah, so iOS is much more limited, mostly because of the operating system limitations.

43:15So when it comes to your screen, there's two things that we capture. We do everything I said before, but just for Safari. So when you install Rewind for your phone, we have a Safari extension that takes what you see in Safari, stores it all locally, and you can rewind time back and forth on all your Safari browsing, so any web browsing. And for native apps, unfortunately, the operating system doesn't allow that capability. Well, it does, but it's a little bit kludgy. I can explain if you're interested. But instead, what we leverage is screenshots. People already take screenshots to remember things.

43:43I'm sure many people listening, you know, take a screenshot of their confirmation number or there's times you take a screenshot that you think, I should probably save this for later. We automatically, if you choose to, you can import all of that into Rewind combined with all of your Safari browsing data. And then you can search and ask any question of any of that. And by the way, that was really exciting because when we did that for iOS, unlike Mac, on iOS, people had already been capturing screenshots for many decades. And for Mac, you start at the beginning of Rewind, it's kind of this cold start problem.

44:11You start, nothing we have in there. So on the very first day we launched on iPhone, we saw all of these examples of people asking their iPhone Rewind app things that we never imagined. Like one person asked Rewind, when did I have COVID? And basically, when did I have COVID? And because they took screenshots of their kind of COVID test results, it could tell them when they had COVID. And like, I never thought that that was a thing that Rewind could help you remember to do or help you remember to answer. But because people use screenshots, remember that actually worked really well. Yeah, it's so true.

44:40I mean, we all take these screenshots all day, every day. And let's be honest, we barely, if ever, look at them ever again. And now you've unlocked all this new value for them. That's really, really cool. Okay. So, so Mac OS, iOS sounds like windows is coming soon, but obviously now there's another vector as well. And that is the physical device, the pendant talk to us about the pendant, or at least your vision for the pendant. I know it hasn't shipped yet. So the vision for the pendant is to go beyond just the experiences we have on our digital devices, beyond Mac, beyond your phone, beyond, you know, windows and really help you capture and remember the things that happen in the real world.

45:16And, you know, there's a bunch of use cases for that. Most people sort of go through life expecting that those things will be forgotten or, you know, if they need to remember it, they'll take a note, which is kind of funny if you think about it, you know, like dependent to some extent competes with note-taking. But note-taking, if you really boil it down, is this idea of taking a tree, chopping it down, turning it into a rectangle called paper, taking a stick with ink at the end of it and scribbling things down. And that's the way we think, most people think, is the best way to like remember stuff.

45:42And there's, of course, digital versions of that. But it's basically the same idea. It's like you have to actively think, okay, this is a thing I want to remember. I have to capture it actively. I have to be able to actually understand what somebody said, write it down, and maybe, just maybe, I might go back to it later if I did a good job of having foresight to know that this is the thing I might want later. So the pendant and even rewind on the Mac changes that model. It's entirely passive. You don't need to have foresight. It works in the background. You don't have to actually actively tell it what you might want to remember.

46:09And the joy of that is you can be much more present. You don't actually have to force yourself to try and remember things you want. It can be as simple for the pendant, for example. It can be as simple and trivial as you're at the shopping, you're at Target or at Whole Foods, or you think to yourself, what did my wife say I should pick up? And she might have told you that 15 minutes ago and you just forgot. You could text her, sure, but wouldn't it be great to just go back and like, what did she just say? Okay, got it. Or it could be you're at a conference and you're trying to, you're back-to-back sessions, and you want to send summaries back to your team.

46:41You know, it'd be great to just be able to passively listen and then have technology be what like a perfect chief of staff would do for you. Sit there and take notes, summarize it and provide it for you. So how does that work? Obviously, it's recording to this device. Is it then, we talked earlier about syncing and how you're not syncing between platforms and devices. I'm guessing this has to sync in some way, right? Is it sending something to your phone? I'm a big believer that all products should do a few things well and not a thousand things poorly, especially hardware products, I think should do only a few things well.

47:09And so in some ways, you could think of the pendant as really just a Bluetooth microphone. It's capturing what it's hearing. It's doing it in a way that has great battery life. And most of the heavy lifting happens on your phone. So I have a belief that we all carry a supercomputer in our pocket everywhere we go. I don't think we need two supercomputers everywhere we go. So there's a lot more flexibility. So for a lot of reasons, having a purpose-built great product, it'll be cheaper. It'll be more reliable. It does one thing really well, which is gives you perfect memory of everything you've said and heard.

47:37Everything else, you can use something else for. So when you announced this, obviously, it caused a lot of discussion. It caused a lot of pre-orders, too, I think. I think you've already sold a bunch of these, which is super cool. But I know there was definitely a debate. There's some people claiming this is the most groundbreaking thing ever. Some people were concerned about the privacy. Talk a little bit about that discussion and sort of what's happened since. Yeah. So we are now at 6 ,000 pre-orders, still every day, like 100 pre-orders, which is kind of crazy. How much does it cost? $59. So that's$354 ,000 in pre-orders, which is great.

48:11I mean, one of the reasons we did this in full transparency is to know, is there even demand? Like, if we built this, will they come? And so it is in that sense, you could describe this in my optimizing days, the way this is described as a painted door test. Like, put it out there, see if people want it. If they do, then put a lot of resources behind it. If they don't, then just accept that, okay, maybe this isn't something that's top priority. So that's, you know, now is sort of number one thing we learned. the reaction from a privacy perspective was exactly what we're hoping for. We really wanted the category of these devices to have a privacy-first mandate from the public.

48:44I think that's going to be critical to making this a mainstream, socially acceptable device. And you just have to go back and look at Google Glass to see the opposite of that. Google Glass is a classic example of technology in search of a problem, a fancy device that they didn't really have a clear understanding of what the problem is. And they put it out there and hoped people would figure it out for them. But instead of being, and because they were so cavalier on privacy, it ended up having a backlash. And the story isn't yet about privacy. We really got to get ahead of this and make sure the narrative is focused on how do we make this a socially acceptable privacy first device, not a cool fashion gizmo that pairs fashion week, but something that ordinary people can look at and feel comfortable that that person isn't invading their privacy.

49:23So that's, that was the goal in this announcement. And we achieved that goal. Well, I certainly got a lot of hard questions. I'm really glad I did because those are the questions that I think are going to be important to answer for this category of devices, not just our product, but the entire category of AI wearable devices to really be a success. Through these questions, what have you learned that people are demanding from a privacy perspective? And how has that impacted how you plan to build the product? It's very similar to the announcement of the telephone. You know, people think today, like telephone, you know, the most obvious or the most common, one of the very common complaint of the telephone when it first emerged was privacy.

49:58And a big reason why was the very first version of the telephone required you to go to the general store to make a phone call. It wasn't in your house. So you people worried about, oh, this is going to eavesdrop on my call at the general store. Even when it came to your home, you had this issue of the eavesdropping switchboard operator. And it wasn't until the technology could catch up to solving those two problems. It was cheap enough to have in your home without a switchboard operator. Then it went mainstream. And I see the exact same things happening with the pendant, which is everyone is criticizing the equivalent of the pendant or the AI wearable as the telephone in the general store.

50:30Yeah, there are issues around privacy that we need to solve and fix, but those are solvable and fixable through technology. I'm happy to share several examples of that. And that was the biggest thing I learned. It's like, wow, history is repeating itself. This isn't new. Every new technology goes through this kind of awe, skepticism, scorn, hate. And ultimately, those are all steps along the way to mass market adoption and normalization. The thing I thought about was the Alexa and the Google Home and the fact that, you know, only a few years ago, we put these devices in our homes and that are effectively listening.

51:02Right. And I think the way we all got comfortable was we trusted the fact that until we said a certain, you know, wake word or trigger trigger phrase, it wasn't actually listening. Right. The listening was just stored very locally and very, very briefly. but in the case of that and also the telephone you think about the fact that these things for for sort of like all of the participants are kind of like opt-in right is there an element of that with the rewind for for for people will there be like a red light or or something where hey if i'm sitting with you dan like i know i'm being recorded but without having to ask first of all visual indicators are very important so we will have that it will not work under your shirt it will only work i'm sure that's why the form factor How do you do that?

51:45And part of it is just how the microphones are designed. It'll be awful, yeah. Yeah, so it is by design. We don't want it to be used in a covert. You can go on Amazon, buy some spy glasses or whatever. We don't want this to even be used in that way because I think that will damage this category writ large. There's so many uses and utilities for this product when it's out in the public. It would be silly to sort of just because it's possible make it possible to use it as a covert device. So that is number one, the goal. Definitely, we don't want people to use this in a covert way. One idea that we're really excited about from an opt-in perspective is voice fingerprinting and speaker diarization.

52:22And what that simply means is by the audio that gets captured by the microphone, the smartphone that's attached to the device can determine who is speaking and has that person previously actually verbally opted in to being recorded. So what that would mean effectively is if you and I were talking, I had the pendant on, I, as the owner of the pendant, would have opted in beforehand. At some point during our conversation, maybe I'll say, hey, Michael, are you okay if we capture this, record it with the pendant? And if you said, sure, yeah, I'm fine with being recorded, then the pendant hears your voice, recognizes the voice, fingerprints your voice.

52:54And then when you in the future speak, you then get captured. And if you don't say yes, it's as if you never were in the conversation. It's just a one-sided conversation. I'm talking. I can capture myself. There's no laws against that. There's no privacy concerns around that. But when you speak, you only get captured if you opt in. So that's one of the ideas we're playing with. That's a hard problem to solve. But I think that's the equivalent of the eavesdropping switchboard operators to the telephone. It's like, let's make it so that people feel comfortable. Those are the things that I think will make a big difference in people's perception of this product.

53:21Going back to something we talked about earlier, sort of startups versus incumbents, who has the advantage, empathy versus opportunity. It feels like there are incumbents that can also do this really, really well, right? If they wanted to. We talked about Alexa's in my home. We talked about, you know, well, we didn't talk about it, but I'm wearing an Apple watch. Maybe you are as well. Do you think that the incumbents will try to attempt something as well? And how will their distribution factor into this problem space? I absolutely think if the incumbents chose to do this, they could crush it.

53:54I don't think they will. And there's a basic reason why. The innovator's dilemma. OK. The last thing, and I can guarantee you, there are people at Apple today who've seen Rewind on Mac and think, this is amazing. This is going to give us$10 billion in subscription service revenue. Services is the future. Let's build this app. And for every one of them, there's 100 people at Apple who are either lawyers or privacy zealots or paranoid, conservative, brand-oriented people who say, don't do that. No, this is scary. This is taboo. This is new. They would much rather guys like me blaze the path and sort of take all the punches for them and then slowly come in behind after we've proven the market.

54:31Then they're going to be the ones that innovate. And maybe I'm proven wrong. I'm not trying to be cavalier here, but I would be shocked and, in fact, very impressed if Apple had the courage to take the risk of creating a new category in this space, especially given what Tim Cook has said publicly around privacy, what he said to Congress around privacy. I think the last thing Apple is going to do is take a risk around a new product category that we're defining. And, you know, they have yet to see that. And just look at Dropbox is a good example. You know, they wanted to do Dropbox, what Dropbox did for a long time.

54:59They tried to buy Dropbox. They couldn't. And eventually they launched iCloud many, many years later. And the big difference is, you know, yeah, I'm sure iCloud is doing great compared to Dropbox. But the main difference is we aren't just files. You know, what we're storing is capturing everything you see stay in here, an encrypted proprietary database on your device. You can't just copy and paste your memories from Rewind into the Apple free equivalent. So I believe we're building a huge advantage. I hope Tim's not listening to this. But I believe that we're building a huge lead many years, hopefully, before Apple does this.

55:26And I think, frankly, it's inevitable that they will because it's a good idea. it's just a question of when will they, when will the innovators dilemma, you know, get so small compared to the opportunity they see in this category. Maybe zooming out a little bit, staying on this privacy topic a little bit, zooming out, not super specific to Rewind, it feels like given everything that's happening with AI, data is kind of like the new gold, right? Everyone, just every company that is building these personalized experiences just wants to kind of like suck up as much of this data as possible to provide a superior experience.

56:00And frankly, it feels like it's also a little bit of the Wild West out there, right? How do you see this all playing out, this sort of data gold rush for AI? What does this look like a few years from now? I think data is super critical to making LLMs work well. We certainly would never sell our customer data. We don't have access to it. It's all stored locally. So like, you know, in that sense, we have a deep commitment to the business model we have, which is a subscription. It's free product, generally very, you know, very futureful free product. And if you want to pay for some of the bells and whistles, you pay us.

56:29And that's how we make money. And that's how we're sustainable. And that's how we will grow. We're not like an ads-driven business. We're not a data-swapping business. But I do think data is critical. I mean, data and context in particular, that subset of data that is useful for a large language model to produce an answer, I do think is pretty compelling. I have some secret ideas around what actually will work and won't. I won't share those because I think they're easy to copy once you see that they work. I saw a long time ago a talk by Diane Greene and I think Bill Gates around China and the data gold rush and the United States and the data gold rush.

57:01I don't think it's as simple as who has more data wins the way who has more gold wins. It's not a more is better. It's a better is better. Better is better. If you have the right data, if you know how to take the pixels on your screen and turn it into words, turn that data into knowledge and take that knowledge or into information, then information to knowledge and that knowledge into wisdom, that's the part that gets valuable, not just the sheer number of terabytes of data you have. Okay, now zooming back to Rewind. What's the plan for the pendant? When is it coming out? Maybe also, what's it like building a hardware product after only ever building software?

57:37Yeah, it's so early to say on all those questions. I would say I've been, people have sat me down and tried to convince me. Don't do hardware. Don't do hardware. Hardware is hard. And I think they're right. I'm not, I'm not, I'm not being cavalier about it. But I go back to the point I made earlier. I do think hardware is hard when you don't know the problem you're solving. It's not magic. It's hard. Physical atoms are much harder to move than digital bits. But if that's what it takes to build a truly personalized AI, I think it's worth the schlep. So that's my view on it. but I'll just be very transparent.

58:11We're so early that I can't even give you a date. My guess is it will be much later than all of our competitors, in part because we would much rather build a better product. We hope by announcing it now, we force those competitors to really think about privacy. But even if they don't, we're gonna come back. We're gonna come back after them with a much more privacy-focused product that will ultimately win the market. It sounds like you're building this company in ways that you didn't build optimizely, maybe from some of the lessons you learned about your path there. What are some of the other ways that you're building this company in a different, new, fresh way for you?

58:42Boy, if I had to go back, I could do so many things better. You know, if I think about it, if I took myself and put myself in Dan from Optimize-y 10 years ago, so many things I did wrong back then that I'm doing right now. First is a small, nimble team is much better than a large junior team. And I think is I couldn't have hired the people I've hired at Rewind. I couldn't have hired them back at Optimize-y. I wasn't good enough. The company wasn't good enough. That is, I think, a huge, huge advantage is the people. The second is culture. You know, I was actually very explicit at Optimizing about our culture.

59:12We had cultural values that we codified. And I think we did a good job of walking the walk. But I think I took that to a whole other level with Rewind. You know, we publish our culture memo. It's out there. You can see all our values, why we think culture is important. Every important decision we make, we root it in the cultural values. Every quarter, we do 360s, very lightweight 360s on each other. And that's where we give each other feedback on how we're living up to our cultural values. So all of those things, I think, have made it so much better this time around compared to last time. Well, it really sounds like you're having an awesome experience.

59:43We covered so much ground. Is there anything I didn't ask that I should have or any stories you feel like need to be told? I think I'll add, you know, I am actually having a lot of fun. And it kind of, as you mentioned that, it reminded me of something I saw yesterday from Jensen Huang, who's the co-founder and CEO of NVIDIA. And he was asked, you know, if you're starting a startup today, what would you do? And he said, I wouldn't do it. But he said that running NVIDIA and creating NVIDIA was like a thousand or I think maybe he said a million times harder than he thought. And look, I haven't had the privilege of building an NVIDIA.

1:00:13Once I have – and I don't think I can even question him. Like once I've built a trillion-dollar market cap company, maybe then I can have a talk with him about it. You can quit then. Yeah. But I guess my experience has been different. Like it's been hard, but the hard part is the part I love. And I think I just, maybe my brain is just wired in the wrong way or the right way, where like the pain that he and maybe Elon Musk writes about is the pleasure that I feel, you know, like the uncertainty, the effort, the work, the sacrifice are all the things that I love about the job. So I guess if you have my one last thing is if you're, some of your listeners are thinking out there, boy, maybe I should be an entrepreneur, maybe I shouldn't.

1:00:49I don't want you to think that the only path to success or the only path as an entrepreneur, success or failure is pain. It's certainly, there are hard days, but I can tell you unequivocally, there's nothing I'd rather be doing than I'm doing right now. And I think it's the best job in the world is to be an entrepreneur. I know you said it earlier, but please remind us again, are you hiring? Where can people find out about opportunities to work at Rewind? Yeah, we are hiring across the board. Go to rewind.ai slash careers. We've got a lot of amazing roles. You join us, you'll have a huge amount of autonomy and ownership and trust, and you can help us build a superpower.

1:01:21And then last but not least, tell us where we can try the product. Rewind.ai is the best place to go. You can download the product today for Mac. We have a very generous free product on Mac. You can go to the App Store, download Rewind there as well. And soon you'll be able to download it on Windows. Awesome. Dan, thank you so much. This was so much fun. Really, really appreciate you giving us all this time. My pleasure. Thank you so much for having me. Thank you so much for listening to Generative Now. Now, if you liked what you heard, please do us a favor and rate and review the show on Spotify and Apple Podcasts.

1:01:54That actually really helps. And if you want to follow me, you can find me at Magnano on X and LinkedIn and anywhere else. And if you want to follow Lightspeed, you can find us at LightspeedVP on all the same platforms. Generative Now is produced by Lightspeed in partnership with Pod People. Big thanks to our production team. The show wouldn't be possible without them. And I'm Michael Magnano. Thanks again. We'll be back next week with another fascinating conversation.