This Week in Startups - Generative AI's potential, Spotify Wrapped + AMP Robotics Founder Matanya Horowitz | E1624

Episode Date: December 1, 2022

First up, Jason answers two great questions from the most recent cohort of Founder University! (1:34) Then, Molly and Producer Rachel discuss Spotify Wrapped's marketing genius and a potential new contemporary. (22:33) Finally, AMP Robotics Founder Matanya Horowitz joins Molly for another edition of The Next Unicorns! (40:02) (0:00) Molly tees up today's segments! (1:34) Ask Jason! Is the Generative AI boom for real? (11:42) OpenPhone - Get an extra 20% off any plan for your first 6 months at https://openphone.com/twist  (13:12) Ask Jason! How to survive text-to-image tools as a designer (15:00) Ask Jason! How to build relationships with investors before you start fundraising (21:07) LinkedIn Jobs - Post your first job for free at https://LinkedIn.com/unicorn (22:33) Molly and Producer Rachel break down this year's Spotify Wrapped (and a new contemporary!) (38:38) Crowdbotics - Get a free scoping session for your next big app idea at crowdbotics.com/twist (40:02) Molly is joined by AMP Robotics founder Matanya Horowitz to talk about building recycling-focused robots! (52:04) AMP's customers, target market, company mission, and how they train AI FOLLOW Matanya: https://twitter.com/mhorowit FOLLOW Jason: https://linktr.ee/calacanis FOLLOW Molly: https://twitter.com/mollywood Subscribe to our YouTube to watch all full episodes: https://www.youtube.com/channel/UCkkhmBWfS7pILYIk0izkc3A?sub_confirmation=1

Hey everybody, welcome to Wednesday. We've got a big show coming up for you today. First up, Jason answers a few really in-depth ask Jason questions from our founder university cohort. Then Rachel and I get together to break down Spotify wrapped the most wonderful time of the year. But actually, we talk about the viral marketing genius and a new player who's preempting wrapped building on top of this viral marketing trend. We have a good time with that, and it's Wednesday, so you know we're having a next unicorn founder on the show. This week, I interviewed Matanya Horowitz of AMP Robotics, which is doing something near and dear to my heart in climate tech and also robots. It's a great conversation. Stick with us. This week in

startups is brought to you by Open Phone. As a startup founder, a lot of mistakes are easy to roll back, but using your personal cell phone number as your company number isn't one of them. Open Phone makes it easy to get business phone numbers for you and your team right on top of your existing devices. Visit openphone.com slash twist to get 20% off your first six months. LinkedIn jobs. A business is only as strong as its people and every hire matters. Post your first job for free at LinkedIn.com slash unicorn. And crowdbotics. Great ideas can change the world. And crowdbotics is the fastest way to turn those ideas into code.

get a free scoping session for your next big app idea at crowdbotics.com slash twist. All right, everybody, if you haven't heard, I started teaching a free course called Founder University. Why am I doing this? Well, I love founders. I love hearing what they're building, maybe helping them with some inspiration. So every Monday, every Thursday, at 6 p.m. Pacific time, I spend an hour and a half with 350 founders or aspiring founders. and you can sign up for the next cohort at founder. It's just a lot of fun.

It's a way for me to spend time with founders right as they're starting their companies. And we will take about 10 of those 350 founders and put $25,000 into their company. We'll be their friends and family round if you think about it that way. So when I find somebody who's a real builder and building something interesting,

we'll put that $25,000 bet in at a $1 million valuation. That's a good deal for us. It's a great deal for founders. It's just like doing a friends and family. with your rich uncle or aunt or if you have rich parents and you're lucky enough, we put that 25K in, or we will invite them to come to our accelerator,

where we put $100,000 in for 6%, just like tech stars or Ycombinator. Anyway, we are having a great time inside of Founder University. We're using a cool tool. I'll give them a shout out. It's called Circle. Free shout out for the team at Circle, which is kind of like Slack plus message boards.

It's like a little educational community building tool. So shout out to the folks at Sucle. circle. In there, I have a channel, like a chat room called Ask Jason, and the founders are sending really, really well-vought-out questions. So I thought I would take some of their questions, and instead of just answering them there, which is what I do late at night, you know, I watch the Knicks lose and, you know, then I'll watch, you know, the Knicks fan TV show called Nix Fan TV on YouTube. And while I am watching the 25th year of my Knicks losing R.J. Barrett going three for 16. I will cry in my cogs hero and answer some questions. So here are some of them.

Here's Russell Palmer. He's from a company called Cyberfilm AI. And he says, what do you think about all the breakthroughs in innovation recently from generative AI? And for those of you don't know, generative AI is like GPT3. That's where when you start typing a sentence in Gmail or in other pieces of software, it fills out the next couple of words. So AI has studied large, data sets of writing. You can think Wikipedia, message boards, etc. And it then can do interesting things. You say you want to write a blog post about a topic like, I don't know, how to make Peking Duck. And it will go and start giving you suggestions based on what the AI has learned. Like, oh, here's the ingredients. It knows that ingredients are part of recipes. He goes, oh, here is, you know, the technique. Here is the temperature, right? So these tools are becoming very powerful. There's also text to image. So you say, hey, I want a picture of Jason Kalakanis, but I want him as a samurai from and mix it with a lightsaber from Star Wars. And this tool, you probably have heard it referred to as Dallie, D-A-L-L-L-E, or there's a meta tool that goes text to video. So you can do that same thing. Hey, Jason Kalakhanis as a samurai,

swinging a lightsaber and it makes a little five-second gif, you know, like a little short video. Anyway, Russell continues and he asks, with all these. powerful new APIs and models available to anyone, do you agree this opportunity can be the next big thing? Similar to the explosion of apps after the release of the iPhone store. Or do you think it's another hype cycle? Okay, so I will say that this is the real deal. Now, why will I say this is the real deal? And maybe cryptocurrency wasn't. Well, this collection of tools is available to everyone. You have multiple parties making really fast iterations on top of them. In fact, GPT3 just came out with a new version that I think is 10 times the corpus, you know, the data set than it was previously and people

are DMing me, oh man, this is a game changer. So when you have APIs, when you have tools available for free or low cost to the entrepreneurial community, they will come up with ideas that the people who made those tools and those platforms didn't even imagine. The smartphone is the perfect example. Nobody anticipated that a smartphone would be used for calling a taxi. But GPS and smartphones made that available and that became Uber and DoorDash. You know, stupid things like the flashlight or people using the accelerometer to know where you were in space and do things like look up at the clouds and be able to see stars and find Orion in the stars, right? There's that popular app. So this feels like the real deal to me because images, text and video are such a big part of consumers consumption.

And, you know, you're making things that can be 10 times better than what came before them. What's an example of that? Well, if you previously wanted to make an image and do a photo, well, if you did a photo and you wanted to make it look better, you would take a photo on your camera, you would upload that image to your computer, then you'd take it from your computer, you'd load it into Photoshop, and you'd do some stuff on it. So if you'd think about that journey of taking a photo on a digital camera, putting it on uploading it to your computer, it was like many steps.

Then iPhones came out. They had really fast processors. Some software got created. And Instagram allowed you to do these cool filters. And this made, you know, photo sharing on the internet explode. It really created the social media movement after the blogging, micro blogging movement. Then you had video. Okay, the phones became better at video and 5G and 4G and LTE and all the stuff emerged.

That allowed the next generation of social apps. Snapchat, doing very much. video, stories, right, Instagram video, and of course, TikTok. Well, now AI is going to add to this. So if you look at any practice that uses text, images, or video, this could be somebody writing code. And in fact, there's a lawsuit right now because Microsoft, GitHub, and a couple of other companies are getting sued by a group of developers in a class action suit because they're using their data set of open source code to help developers write code faster. So they start writing code and it says, oh, maybe this is the next line of code, or maybe you want to, uh, you know,

essentially copy, but it's suggesting. And that's really what the lawsuits debate is about. And so when you see lawsuits happening, when you see all these creative ideas happening, and the, uh, company that's being sued is called GitHub's co-pilot. And it is a very creative tool for finishing code for people. So these legal issues are going to keep compounding. But when you see legal issues like this, it's like a negotiation. We'll figure out some way for the training datasets to be compensated by the people who are building these tools. In other words, if GitHub knows that this individual wrote this code and that they want to train the AI on it, they should go to that individual and say, hey, we would like to license your code

for this amount of money one time per year for the next 50 years to train our AI dataset. Would you be okay with that? And that's the negotiation that's going to happen over the next couple of years. and it's going to be really exciting. I think it's not hype. And people are, you can prove it's not hype, because people are actually using these things in practice. Typically, these things, they start as toys. Hobbies use them, like people used to do. They would do Photoshop, they would edit something. But then they become a tool that everybody uses. That's the phase we're in now. And then eventually they become platforms. That's the, that's what will be coming next. They start as toys, and then they turn into tools, and then they turn into industries and platforms. So we're really in the tool phase. There's businesses to be made in the tool phase. I see. see many companies doing AI and GPT3, you know, text suggestions for bloggers, for email and salespeople. And we're going to see the next phase, which will be, you know, you'll open a product like Canva. And instead of it making you pick a template, you'll just talk to it. And you'll do text to image or text to video.

I want to make a birthday party invite. My daughter's name is Susan. and I wanted to be My Little Pony and I would like to take this photo and I would like to put my three children and their two friends on the My Little Pony faces and make them into My Little Pony and then let's put a date on it

and you'll be talking to it and it'll be making the invite and then you'll say, show me 10 different versions of that and you'll say, okay, I like two and seven. Give me three different versions of two and seven kind of put together and all of a sudden you'll have an invite that you would have paid $500 to a designer to make for free. And that's really what's happening here is these tools are coming fast and furious and they're really providing value.

That'll take the next two or three years. And then somebody will make these into platforms. Now, what will the platforms look like? Oh, my lord. I mean, who knows? But it's possible one of these things would allow somebody as but one example to make a short film based upon the IP and Star Wars, where they could just say, I would like to have a Jedi Knight fighting a Sith Lord. and I want a hand solo type character, but I want them to be, you know, female, and I want them to be a smuggler,

and I want them to have a Millennium Falcon ship, but it's slightly different in these ways. And then all of a sudden, you'll just have like a Millennium Falcon ship, and you'll have two people fighting in a lightsaber battle. And I think you're going to literally be able to create short movies that look pretty photorealistic and eventually look realistic and eventually look realistic and publish them. That's what the platform might look like. So, you know, this is usually the journey. and I think we're kind of in the second inning, third inning. There's a lot more to come. So if you are looking for a business to go into, I think studying what's happening with stable diffusion, Dolly, et cetera.

If you're a developer, going in there and contributing to the open source projects, forking the open source projects, making your own products. This reminds me of the early days of blogging. Matt Mullenwegg created WordPress. There was something called Typeform. There was Tumblr. There was Twitter, micro blogging. There were so many different creative things built during

that time period off of just the simple concept of blogging. So yeah, really double click on it. On the program today is Dorena Kulia. She is the founder of Open Phone. Welcome to the show, Dorena. Thank you so much, Jason. Great to be here. You know, I've read the ads a couple of times here. It seems like you're charging too little for this product. It's $10 a month per number. How are you able to do this so affordably? A hundred twenty bucks a year, 150 bucks a year per person is nothing. So we're a very self-served product, which is why many of our competitors offer much more expensive tiers is that they rely on like a customer success rep or someone help you out to set things up. We are very self-served. Now, we do have customer success managers who are amazing. A lot of our

customers are founders and startups. They like things to just work without instructions without. They'll read the manual. They'll watch the videos. They don't want to talk to a human. They just want to set it up and go. And you made the product so simple. Absolutely. That's where the cost savings comes. You don't have to have a sales team going out there selling it. And you know, the other big thing is that the way we also grow, and I think, you know, the way we get a lot of customers is that we have very strong word of mouth and people like tell others about us. And I mean, all of that contributes. Our business model kind of makes sense, it makes sense for us to be able to offer it at a very good price. All right, everybody, here's your CTA, the old call to action. Twist listeners, 20% off any plan for your first six months. Just sign up at openphone.com slash twist. And if you got an existing number, no problem. They'll put it right over.

openphone.com slash twist, O-P-E-N-P-H-O-N-E dot com slash twist today for 20% off. Now, if you're a designer, does that mean your job goes away? Are you going to be like a phone operator where you just don't exist anymore? I can see half of the designer jobs going away, but what will also happen is, as a designer, you might be able to just do more. So, you know, maybe the first level of iteration will occur with these tools, but then you'll want to refine it and polish it and make it look even,

better or be more thoughtful in the way you make things. In other words, you know, the actual drawing of an illustration might not be the skill anymore. It might be the actual conception, the creative process, talking to your customer a bit more, understanding what they want, not just showing them a bunch of designs that look great, but maybe understanding what their needs that is and really coming up with a better solution for them. So it's going to be a brave new world and there used to be a big business in people doing color correction on photos or putting filters on photos, and then consumers stop needing those. There used to be a huge business in hiring photographers to take pictures for magazines. And sure, that that still exists, but it's much smaller.

And people can take their own pictures and use them for their own invites. People used to hire photographers to take their family photo, to take their kids photos. Cameras are so good right now that arguably the photos being taken today versus one's taken 10 years ago by professionals, the photo's being. taken today by individuals are probably better. And certainly, you know, the photos I took in high school in the 80s are absolutely 10 times worse than photos I'm taking today of my children. Think about that. In, you know, 30 years or so, we went from professional photographers creating things that were much better than any consumer to today, consumers taking pictures that are much better than the majority of photographers just 30 years ago. So it doesn't take a genius to see where this is going. You're going to

need to add additional value if you're a photographer or a creative. Okay, our next question is from Cody from Diem. Everyone says to start building relationships with investors before you fundraise. Can you give some examples of what this actually looks like? I'm currently giving product updates every month, but it seems like investors don't have time for a 15-minute check-in, which is faster than getting coffee. You do want to build relationships with investors, but you do have to also be respectful of an investor's time. Let me give you the candid, behind-the-scenes conversation that occurs amongst VCs. When you become a VC, it's typically you're older, you're later in your career, you've made a little bit of money, and you don't have a lot of free time. And because of the

power law, the top one or two companies you've ever invested in represent 95% of the returns you'll ever have in your career. For me, that might be Robin Hood, Com, Grin, and of course, Uber. So what does that mean for VCs? That means while they are looking for new opportunities to invest, they know full well that the majority of founders fail, and you know that as well. They also know that their time is limited, so they become ruthless about their time. And they might also have families and other commitments, or maybe they're slowing down because they're tired and they're entitled and they want to ski 40 days a year, whatever it happens to be. So what you need to do is keep them up to date when you make significant progress, keep it short,

tight is right. Don't ask for coffee. Asking for coffee with a busy person is really a bad idea. Now, if it's a new VC, it's an associate, maybe they need to fill their schedule. So this is where contacting Bill Gurley or Chamath to, you know, hash out your ideas, well, I mean, they're super busy or David Sachs. Now, there might be an associate working for those folks who's looking to meet people. They might be really interested in the 15 minute check-in. So feel free to go, you know, meet with them because those people are looking to have their calendar

fill up so they can say to their boss, hey, look, I did seven meetings with founders. None of them are ready, but these three are very promising and I've identified this trend. So there's, who are you targeting? And there's other ways to build relationships. Social media is perhaps the best one. If you go on Twitter, VCs are all on Twitter, talking all day long with each other. If you make a beautiful social media profile and your profile is focused on you and your company and your vertical and nothing else, you're just incredibly focused on if you're

an education startup, educational topics, your company, if your company is called, you know, unlimited education. You have your domain name there. You have a description of the company you're building. You have a description of what you're passionate about. And you don't talk about the Knicks. You don't talk about what you're having for lunch. You don't talk about politics. You strictly stay focused on your online persona, Instagram, TikTok, Twitter, your own personal blog. If you start a podcast, you just make all of your content focused on your vertical. So when they hear your name, Cody, they think about the topic and your company name. And there's no question about who you are and what you're focused on. And if you seem relentlessly focused on one thing, that makes it easier to place a bet

or to take that 15-minute check-in with you. Of course, things that matter are when you launch a new product or feature, when you have customers, when you have revenue, when you've done a funding round, those are the things that people see as truly worthy of maybe jumping on a Zoom call to figure out what happened. So if you're growing 5% week over a week for 10 weeks and you can put a chart, A chart is worth a million words. It really is. I see a chart. I see a table.

I see a list of customers. And all of a sudden, I'm like, okay, this is real. So you're trying to move from your building a project to your building a business. Projects have features. Projects have ideas. Projects have a bunch of words. Projects have mock-ups. Businesses have customers.

Businesses have revenue. Businesses track key metrics that are going up and to the right or sideways or very business is hire people. So remember, when you are signaling people, you don't want to signal things that put you in the project, lost in the words, meandering bucket. You want the things that signal a business, a going concern, a viable enterprise. So just use common sense. If you hire somebody from Google to be your new CTO, okay, that's a sign that you have the ability to recruit top talent, and it's a business. You sign a lighthouse customer. You have Twitter or Uber start doing a trial of your company. You grow 7% week over week, but then one week you grew 47%

and the chart looks awesome. Those are the things that are going to put you in the right bucket. So until you have those things, I would not waste a Visi's time. I would get those things. Remember, long emails with lots of words, your life story, nobody cares. TLDR, too long didn't read. Most founders just start writing and writing and writing and it's actually working against their best interest because they don't look credible. You're trying to look credible to investors. How do you look credible to investors? You do business things. Business things include hiring great people, landing customers, tracking your metrics, launching new features, having lighthouse customers, having revenue. Those are the things that will get you the meeting.

So until you have those things, focus, don't waste the VC's time. And the social media presence cannot be discounted. I have met countless founders over social media, and the ones I tend to connect with are the ones who are consistently talking about one thing and one thing only on social media, and they're writing blog posts about one thing and one thing only. They are ultra-focused and obsessed about one thing. So make sure you signal that you're focused because most founders in 2022, and for the last five years, have been not focused. And that goes for everybody, if I'm being honest, a lot of VCs were not focused. So get focused and start moving the metrics that matter to investors, and everything will work out just fine.

Okay, now let's hand it over to Molly and producer Rachel, and they're going to talk about the new Spotify wrapped and producing your own music festival based upon your Spotify consumption. All right, everybody, LinkedIn jobs. We have to talk about the downturn right now. there is one silver lining for all of us running businesses, especially small businesses or nascent companies. You know, when you got that half dozen dozen people, the talent pool is getting stronger and stronger. And people are looking for interesting companies to go work for. And that's where you come in. And that's where LinkedIn's going to solve all your problems. When you run a startup, you run a small business, you know every single new hire is high stakes.

So you have to ask yourself, what if I hire the wrong person? What is that going to do to my team? My team dynamics. Well, that's why you have to check out LinkedIn Jobs. There's so much great talent out there right now, and LinkedIn Jobs helps you find the right people for your team, and they help you do it faster. Now, you want to be 100% certain. You have the right candidate pool that you're looking at.

Well, that's LinkedIn. You're talking about over 800 million people are there. Use screening questions. Ask people thoughtful questions. Hey, if you were a podcasting company, what do you love about podcasting? What are your top three podcasters and why? Hey, if you're an app company, have you downloaded our app and what do you think of it? Then maybe give me some feedback.

So that will help you filter the people who really want to come work for you, who have that passion. LinkedIn jobs helps you find the qualified candidates you want to talk to faster. Post your job for free at LinkedIn.com slash unicorn. That's LinkedIn.com slash unicorn to post your job for free. Terms and conditions do apply. Okay, everybody, it's Wednesday. It's the middle of the week. We don't want to talk about layoffs.

We don't want to talk about crypto. We don't want to talk about anything serious because it's Internet Culture Day. We're going to have some fun on the Internet because it is that special time of year when we're all sharing or hiding due to our shame, our Spotify wrapped. And I am excited to bring on our special guest, Rachel, reporting, to have some fun with me to talk about our music listening. I'm so excited. This is, I'm pumped. I waited all year for this. I kind of did.

I mean, it's really become like, it's a thing. It's just, it's like the Mariah Carey, you know, all I want for Christmas of, but, but it's been interesting this year because there was like a wrinkle. There was a, an upstart that sort of preempted the Spotify Rapped. But let's start by, let's start with RAPT. Spotify Rapped, this annual digital experience where it lists, you know, your individual data to create these charts and these visuals about your music taste for the year. they started doing these recaps in 2016, made them shareable, and it's become this kind of like can't miss end of the year internet event. It sums up your top artists, your genres, your songs. Now your podcast listening this year, apparently that's new and your total minutes listen to on Spotify. Are we ready, Rachel, to share? I'm ready. I'm ready. It's funny, 2016 was the year I graduated high school and I got in a college and Spotify is an amazing like college discount. So,

I've been using Spotify for as long as they've had these. So this is like, this is my Christmas. I'm excited. Let's compare and contrast our artists. Okay. Let's bring up yours first. I know we're going to go from like, your top artist.

Gen X to Gen Z in one conversation. I'm my top artist. Okay. Okay. Childish Gambino is cool. I listen to a lot of Gambino. I've been on a little Bon Jovi kick lately. I've been going to the Bon Jovi B sides.

Fish. I'm just going to say that one. and five were hijacks. Steely Dan and Kanye West do not sound like things I can imagine you listening to all the time. So I can imagine that. I've come around to Steely Dan and I like Steely Dan more than I thought I would, but that was a relationship influence. Okay. And then my kid has been all about Kanye.

I don't like Steely Dan. And then that's really complicated, obviously. Steely Dan being your number one. Probably good that he's your ex. Sorry, I don't like Steely Dan. My ex-boyfriend likes Steely Dan too, so it's a sign. I wish I could go. I wish I could have filtered by the last six months of the year

because like Steely Dan dropped off dramatically to zero. That would be good if they could do that. If they did it by like time intervention. I would like to know also like this month. You know what I mean? Because my music taste totally changes during the holidays. Exactly. So this is like this is a time capsule in lots of ways.

Did you have any of these like artists on your rap last year? I think Steely Dan was on my rap last year. year. I think Childish Gambino has been for a long time because I actually, and I'm not trying to be that girl when I say this, but have been a fan for a long time. So I think those two were for sure on there. I think Bon Jovi, Fish. Fish is a good one. That's a good one to have. Honestly, Fish is like great to just have on. It is. It is. When you're just doing stuff. I got, my family got rid of our, what is that? It's satellite radio or whatever that is in your car. Fish played all the time and it's the best driving music.

It's the best driving music. Yeah, exactly. I think that I put on fish a lot for like cleaning and like chores and things like that. And then obviously our Kanye listening is considering my son's Jewish has dropped off dramatically as well. So we got a first half of the year and a second half of the year situation happening here. All right. Your turn. Your turn.

All right. It's myself long enough. No, mine's like pretty predictable. Zach Bryan is my favorite artist of this year. And he is super active on Twitter. And he's not my favorite artist as in like. like music-wise, I guess, but he's my favorite artist as in writer. And then Taylor Swift.

Like, if I could switch any of these around, I think Taylor Swift should be number two. Taylor Swift has always been on my Spotify wrapped. She will probably forever be on a huge Swifty. But Zach Bryan actually has been tweeting a lot about how mad he is about the whole Ticketmaster situation with her. So I think it's interesting that they're both on my wrapped. And then Rainbow Kitten surprised. I like them.

What is that even? Well, my favorite song is called Cocaine Jesus. And they sing it. And it's a really good song by them. I like them. They're cool. I actually was going to go watch. The reason their number two is I had tickets to go see a concert of theirs in Central Park.

And the concert got canceled. So it's on there. I like them. But because their song is called Cocaine Jesus, I can't necessarily put that up on my Instagram story without my mother having a heart attack. Right. And then also, you know, it's not your fault. That's not your fault. That's true.

And then Ian Muncic is good. I don't listen to a lot of his music, but everything's kind of folky, country poppy over there. And then Tyler Childers is another great country artist. Cut it in that same bucket as Zach Bryan. I don't like, I like country music, but I love like the back, like, I think they called it like backroads country. There was like a word they used it, like stomp and holler on my Spotify rap. They like, like that was a genre of music. That's awesome.

Yeah, you've got a real, you've got a real like soundtrack. to Yellowstone kind of vibe here. I think maybe Zach Brian actually even had a couple of songs in Yellowstone featured in Yellowstone. But yeah, so what's fun about this in addition to just comparing our relative ages in picture form is this just overall virality, right? Like, I feel like there aren't that many things anymore that go viral in this like real water cooler way. Yeah.

And so what happened this year that's kind of awesome. is that this very clever, fantastic hustler came along, and Shea Sabu, a developer who created this free web app called InstaFest, which was basically a way to tie this app into your listening platform of choice, like it works best on Spotify, and create your own kind of version of a festival poster.

So like Spotify puts all this energy into RAP, every year. And it's wildly free marketing. And then this guy comes along, creates this InstaFest app that's free like a month before Spotify Rapps drops. Yeah. He swooped dead. Swooped. Yeah. Like just completely. And so then everybody had so much, I saw this everywhere on Instagram. I made one, which you can see if you're watching our video version at YouTube.com slash this weekend where it made a festival poster for me out of all my listening that frankly is a lot more, I think maybe it's not more flattering, but it shows more layers than my top five artists that are heavily influenced by outsiders.

One reason why I really like, I loved InstaFest. I actually didn't realize that it wasn't by Spotify until a few weeks after I made my first one. Like it came out a bit ago. And then I saw all my friends have have it up on their thing. And I went through my Spotify and I was like, how do I make this? And I had to Google it. So I think this is really cool. But But I also think that it's breaking into this, like, this person is obviously somebody that listens to music a lot and obviously enjoys Spotify. And I think it's funny because I hate when people post their music on their Instagram story. I'm like, who cares? Unless it's Spotify wrapped, which I think is cool that Spotify kind of, I feel like Spotify took the hint.

Nobody cares if you post a song on your Instagram story. If it's like a sad song or like, okay, you got dumped. Like, whatever. Need some. Need something else. So I like this because it feels like this is like a lot less cringy to have. up on a story in my opinion. It actually really is. And it was the right time of year. It was like right before Spotify Rept. And I agree with you. It looks so pro that I didn't realize it. For somebody posted it in

one of my group chats. And I was like, oh, Rapt is out already. And I went to Spotify and was like, oh, I guess I haven't gotten mine yet. You know, I was all confused. So I think it was, it was, it was perfectly designed to pick up on that kind of yearly virality that we're all super used to. And then it was so, And then you've got to admit it, like, perfectly dialed into the festival vibe. Yeah. What he said, what Sabout told TechCrunch is that he got the idea for the app thinking about what Coachella's lineup would look like if he picked the artist. He's a student at University of Southern California, USC in Los Angeles.

The cool thing that they did, that I wish Spotify would do, is they actually did do time intervals. So you could have chosen your music from the last four weeks, six months. or all time. And I'm like, you know what, Spotify RAP should do that. Spotify Rapp should definitely do that. And the basic score is also really cool. Sterecical. That's also cool. For those who have not done this or don't know what we're talking about, go to instafest.com. You can do this online or on your phone. You connect your listening platform of choice. The Apple music one, I think, is still in beta, so it's not as full featured. But if you do it on Spotify,

yes, you can choose your time periods and then you get a literal list of artists by day as though it's a three-day festival. There's Rachel's. Okay, so when's, so day one, you got Taylor Swift headlining. And then it puts this kind of, it puts this like related, but still different enough collection of artists underneath disclosure, Tyler Childers, Illumaneers, Chilled Pig, Jack Kay's, all these other things I haven't heard of. Sooky Waterhouse, that's amazing. Day two, Zach Bryan featuring Ian Muncick, Frank Ocean. I'm on a new, Frank Ocean's going to show up high next year for me because I'm on a new kick. Right. He's a good one. And then day three is you're kind of easy listening, right?

Kat Hasty, Casey Musgraves, Michael Boubley. That other- Day two has Machine Gun Kelly and Kenny Chesney? I was about to say, also Friday, there's, um, so Kat is a wonderful artist. I actually used to DM her because I listened to her like a while ago and I was like, I'm obsessed with your music. I think she came viral on like TikTok. She's a country music.

But Jaden spelled JXDN is kind of like Machine Gun Kelly music. Like I love, I listen to a lot of punk music, I'm not going to lie. And then Fred again is also in that. And Fred again is like a DJ. He does these things called boiler room sets, which is like a DJ kind of set up where you can go stand behind the DJ booth and like with him when he's DJs. And he blew up this year. So day three for me is a little, I don't know if Michael Boubley would necessarily be out there.

Like in the thick of it with like punk pop music and like a DJ booth. I like it, though. I don't know. What was your basic score? Oh, it has right there in the bottom right. I actually let it be public. Oh, yeah, 64. Niche.

You have a niche festival lineup. This just made us all laugh, like this idea of a basic score. So the lower your score, the more niche it is. Mine was 71, which honestly, for given my age, I'll take it. I'll take it. I appreciate that. I'm super basic with Taylor Swift. I'm sorry.

It is up. Taylor Swift alone just puts me in the category. Like, I'm exactly like every other girl. when it comes to that. It gave you a couple points. I'm not going to lie. So one other cool thing they didn't wrap that they haven't done before is they do what kind of music like you listen to depending on the time of day.

And I believe I send you guys like an image of that as well. And in the middle, there it is pumpkin spice. That is my afternoon. I hate it. I hate it. I hate that it said that. And then the one above that, it's like one of the things for my nights. It says hopeless.

And I was like, I don't know. I don't know. Evening listening is bittersweet, energy, and hopeless. Like, it sounds. Yeah, thanks for nothing, Spotify. I know. And I was like, I mean. But at least my morning.

My mornings are off to a good start. Our lighthearted, joy and funny. One thing that sucks about Spotify and the thing that doesn't suck, I guess, about InstaFest is in Spotify, you have to watch a video for each freaking social card. I think that sucks. And I get it. Like, they want you to watch the whole thing. But oh, my gosh. I don't know. I don't know. I wish they sent to be like in an email.

Right. Like this is, it shouldn't be impressions based. Like this is all about, you know, to get to the kind of the startup takeaway, if you will. I think Jason phrased it as, you know, micro apps for marketing are really powerful. Yeah. And that's really true. And anybody can do this at any signs. I mean, Spotify has completely nailed this with making it super easy to share. And then for whatever reason, this year, they made it hard. harder to share. Yeah. They really have. And it's like this forced march through every single card to get to the information that you want that you want to share, which I don't understand. Like, is it generating, because there's music set behind it, is it generating a play for each artist? Like, it doesn't, it doesn't make sense. It feels like they've undone, they've gotten beat in two ways here. Like, first they got beat by Instapest. Yeah. By several weeks. And everybody thought it already was the Spotify Rapped. And then second, they're kind of getting beat in the

like micro-marketing part of this, which is like, we're about the speed. Yeah. And if you think any other app should have a RAPT, what do you think that should be? If another app could have a RAPT. Doesn't, does Instagram already do, they do some kind of end of the year like your most popular posts or something, right? If you have a public-facing Insta. They might. I think they do. Like, they show your, oh, your most, I know Snapchat. Snapchat did something like that. Yeah. I feel like Instagram could do that, and that would be pretty cool. Yeah, I think it's true.

I mean, they should all, like, honestly, I think the lesson we're taking here is they should all do it. Like, TikTok should have a little carousel of, you know, whatever, the types that you watch the most or the creator that you like the best. Uber could make an awesome rap, like the fun places that you went. Airbnb should totally do it, like all the places you sit. Like, everybody should do nostalgia economy for the win. Everybody loves the look back. I was thinking, be real, should have done this. Like, I know it's dead, but I liked it.

And one thing that I think would be really cool to do for any other app is maybe have, like, if you were somebody that reacted to, like, a bunch of people's other things on Be Real, like, you could have your own video saying like, hey, I was the person that liked a bunch of your content, happy that we're friends or something like that. Because Spotify had like a really cool thing this year that they haven't done. before and they were basically like little cameos. They were like artist messages. And Taylor Swift did mine, which was cool.

And I think that'd be cool to have on other platforms too, but from your friends. That would be amazing. But wait, they have little, did I miss this in my Spotify? A little video? I had a little video from Taylor Swift was the first one. Camp was the next. I don't listen to a lot of music from camp, but. Oh, here it is.

Your artist messages. Is that the video you mean, Nick? This is it. I just screen recorded this. So you're seeing, you're seeing my phone. right now. Hi, it's Taylor. So it turns out you may have listened to a lot of my music this year because I'm one of your Spotify rap. So it was cool. And I listened to all too well, the 10 minute version a million times. So that was cool that I got a little blurb from Taylor. A little cameo. Right. But yeah,

I'm already excited to see next years. Super clever and adorable. All right. That's our little OK boomer slash what's happening internet culture segment for today. All right. And it's when. So we've got a next unicorn interview. I'm talking with Metanya Horowitz from AMP Robotics, because robots are awesome. And we just keep open that there's going to be a great unicorn there. This could be the one. I'm excited for that combo. If you want to build an app, you want to build a startup, you want to build a website, a service, a marketplace, any of these things, well, you're going to need to have a really good plan. And if you don't have a good plan, well, your idea is probably going to fail. Well, I've got a solution. for you. CrowdBotics gives you access to the best practices for your specific app. What this means is, CrowdBotics has all the pre-built app templates ready to build your startup faster. You're not starting from scratch. Your developers would have to start from zero. These developers at CrowdBotics and their team, they've got that technology ready to go. And so the architecture is going to come

together very quickly. You can think of CrowdBotics as a CTO as a service, right? A chief technology officer who you can just drop in and boom, they are going to build your app for you quick and they're going to speck it out perfectly. So if you're not sure where to start, CrowdBotics also offers professional scoping. Now this is going to help you flesh out your project. They're going to ask you the important questions. They're going to help you flesh out the features maybe you haven't thought of, right? I want you to talk to the folks at CrowdBotics and let them show you how good they are at what they do. They will schedule a free scoping session with you and get you a detailed build plan at CrowdBotics.com slash Twist.

That's crowd, B-O-T-I-C-S dot com slash twist. Try Robotics and see what they can do for you in their scoping session. You got nothing to lose. Today on Next Unicorns, I have Metanya Horowitz, founder and CEO of AMP Robotics, which is right up my alley as a climate tech investor. Not only is it robotics, which is awesome. It's recycling. Welcome to the show. Yeah, thanks, Molly.

Glad to be talking about AMP robotics. Yeah. So tell me, you know, in your words, I don't want to mess it up. What are you building? What is AMP robotics and how did you come to start this business? Yeah. Well, so in AMP, we're enabling a world without waste. So we are using robotics and artificial intelligence to improve the fundamental unit economics

of harvesting different commodities out of the waste stream. Something that I think a lot of people don't appreciate is our garbage stream or our recycling stream has a lot of value. There's lots of good stuff in the bottles, in the metals, in the plastics, and the paper. And the challenge the industry has had is the cost of extracting these commodities, so basically getting the metals in a pile or the plastics by themselves rivals the cost of those materials. And so you kind of erode away all the value by separating it. And it's left the recycling business not as exciting as it could be.

But you can imagine with robotics, with artificial intelligence, and the ability to create a sensor that can identify all these. materials, you can really see a path to changing those economics and turning recycling into a phenomenal business. My background is actually not in recycling. My background is in robotics. I got PhD at Caltech studying robotic path planning, but saw some of the early results in deep learning and thought it would be transformative and went out to kind of find places where

I thought it could be useful. And as soon as you visit recycling facilities, you see there's a strong need for automation. Yeah. Okay, so tell me more about, before we dive more into the robotics part of it, tell me more about the recycling part of it. Like, dig a little deeper into that, no pun intended, into that waste stream and how, you know, the parts that are inefficient, the parts that aren't being recycled at all, like, what is the problem that you saw that you realized immediately needed to be solved?

Yeah. So it's kind of stepping back a little bit. The way the recycling industry works is, you know, we all put our recycling in a recycling bin or something similar. Maybe we go to a drop-off center. It'll go to one of these recycling facilities. They're called material recovery facilities or MRFs in the industry lingo. But they're really what you're doing is you're separating out these. Just going to be saying that all day, by the way, MRFRA. Yeah, exactly. You get used to it. And then people are like, what in the world are you talking about? But yeah, so the material goes to the MRFs and they separated out into

the individual commodities. These will be things like aluminum cans, number one, plastics, which are like Pepsi and Gatorade bottles, cardboard, paper. And once you have those separated, those get sold on to paper mills that will process it into do paper, smelters that will melt down the metals, or what are called plastics, reclaimers, that will melt down the plastics into new plastic pellet that can then be used for different materials. Like I was saying, there really is value in this stuff. When you look at the typical kind of material coming from a, you know, a community, a set of households. It'll vary. It depends on the commodity market, like this, this world,

this recycling world lives in a commodity market that's fluctuating all the time, but you may be able to find $100 per ton of value in that material stream. The problem is that running a recycling facility can cost $80, $100, $120 per ton. And so you're kind of left with, it can be pretty unattractive economics. The recycling facilities will often charge to accept the material. They charge what's called the tip fee. So they still have a business, but what you can see is it doesn't have to cost that much to do that. There's a path with automation, with robotics to bring that cost down to something more like $50 per ton or even $30 a ton and also raise the value of material. Just to put a fine point on it, it costs so much because you have to have people doing the sorting.

The sorting is time consuming. You have to jettison a bunch of stuff that has like food on it, that kind of thing. Yeah, sorry to be a little long winded there. But yes, that's right. In these facilities, what you have is some big pieces of equipment that are sort of screening the material separating by size and other things and magnets and stuff like this. But largely, you have people standing around conveyor belts separating out material by hand. Not a very thrilling job. It can be very hazardous. And many recycling facilities are run understaff because the turnover is quite high. You can imagine it doesn't smell great.

And so our robots basically plug into those existing facilities and automate that sort of. task with very minimal retrofit. Most of our systems are installed over the course of a weekend and provide them a path to automation without really changing their way of operating. Now we get to do the fun part. Tell me about the robots. So are the hands? Are they pickers?

There's vision built in. Like, what are we talking here? I think we actually have a video that when we post this, people will be able to see a video. There it is. Oh, great. Yes. Oh, this is a bit of another one. Yeah.

So our main robo. We call it the cortex, but it uses a suction system. So we have a suction cup and a vacuum that's plugged into that suction cup. The robot reaches out, engages the vacuum, kind of sucks onto the material, sort of like using a vacuum at home, and it just picks it off the belt, throws it into a bin. It would seem simple. In many ways it is. We're using a robot that's been around for decades.

What is enabling this is a vision system built on AI, where we've shown it millions of examples of all of these materials. examples of all these materials. And it's learned to identify this stuff, even though it's smashed, folded, and dirty, and it's generally inconsistent. With that capability, you can now use a lot of these sort of tools from the industrial automation world in manufacturing. You can bring those to recycling. So there are, as I understand it, four parts to the surface, right?

Neuron, the vision system, cortex, and clarity? Yeah. So, yeah, at this core, we, we, we, We call it neuron, but it's sort of our vision system platform. So it gathers data, it records. It is a common way of, we have a neural network that we, or we have a common way of updating the neural networks across all of our systems. Then it does record pretty much everything it sees, even if you're only sorting

aluminum cans, it still sees all the cardboard and everything else. That gets reported into a cloud dashboard that we call clarity. And there you can see it's called material characterization in the industry, but sort of what is going past the robot? What is the robot separated out? How pure is a material stream? All this sort of data analytics you might want. And then you also have, of course, the robots that does the physical sorting.

Many of our systems now are starting to be used just for data collection, like even beyond the sorting system. But we've also started to deploy other robots that we've built from the ground up for the recycling industry. And we're announcing a lot more of those this upcoming year, but we are going to beyond kind of using an off-the-shelf manufacturing robot to custom robots for the industry as well, but all using the same common vision system, which is our neural platform. And then if someone just implements the vision system for data collection, what do they get out of that?

Why do they want that data? So for a couple of reasons, you know, interestingly in these facilities, because there hasn't really been like a garbage sensor, they're not able to really measure what is happening on a continuous basis. The way they have to measure is they do what's called a waste characterization study where someone goes through the material by hand. You can imagine. But just like one or two times, I would imagine. In addition to how crappy a job that is. Yeah, it's not super pleasant. It ends up costing thousands of dollars per ton when you sort of imagine, if you want to sort a whole ton of garbage, it's a lot of sort of labor that goes into it. So for our vision system, you actually see it ends up

costing like dollars per ton or even in some cases under a dollar per ton. Or even in some cases, under a dollar per ton to scan the material. So in multiple orders of magnitude reduction in that measurement. In terms of what people use that for, it's you can get higher value for these commodity streams if you hit certain purity levels. So they want to understand how they're tracking to those purity targets. They also want to understand sources of loss in their system, so material that'll go to the landfill that they want it to recover. And then also things like equipment breakdowns or when people aren't doing the sorting that they're supposed to in real time detecting those issues. one of the things we're very excited about too is adding the capability of sensing hazards.

So propane canisters, hypodermic needles, knives, this sort of thing that can hurt people. Batteries in particular are actually burning down recycling facilities left and right in the industry. And people throw their batteries in recycling. They're like, oh, great, it'll get recycled. It's electronic waste. And what happens is the batteries crack and start fires in these facilities. And so detecting that for the facilities is a very high priority. I had no idea.

Yeah, so a number of different use cases. And then tell me about this, at what point did you realize or see the opportunity in building custom robots? Because it sounds like what you have is hardware that doesn't sound that expensive if it's off the shelf and its existing technology. But then you decided to make more expensive, more custom hardware. Well, you know, what's interesting about what we're doing is the material we get is already damaged. Like, it's already been, you know, through your trash bin, and it's been through this hauling truck that compacts it. And it ends up, if you say, I'm going to make a sorting device that lets me damage the material more, there's a whole realm of more efficient, lower cost, higher throughput systems you can imagine building.

A lot of industrial robots are built to handle packages or handle manufactured goods and you really don't want to damage. You know, you don't want to damage the car, you're building and things like this. And so we're finding there are sort of physical mechanisms for separation we can build that are an order of magnitude, higher in throughput and lower cost. And, you know, we look at a couple different metrics to kind of measure our progress. And one of them is kind of the cost per pick that you're effectively paying with some of these technologies, we think we'll be able to bring that cost per pick down by an order of magnitude. Because you can smash it.

Did I understand that correctly? Yeah. Basically, if you can smash it. Are you going to call it Hulk? I mean, I hope you're going to have such a great. opportunity for branding here. Okay, you'll get sued into oblivion, but still. Yes.

Yeah. Internally, at least, call it Hulk. Yes. I would say, people don't let me name things anymore. I tend to, the names I come up with aren't too popular. Yeah. All right, well, I see you. I'm clearly where birds of a feather here.

All right, so then let's talk about sort of throughput, like at that sort of current robot and AI, how much material can your technology process and how accurate is it? So for our vision system, we can see hundreds of objects a minute, most of our, I believe even thousands. Our robot, the one that you saw the video of, sorts at about 80 picks per minute. A person will do roughly 40 picks a minute. They can get up to 80, but most people will get tired. So 80 gives you kind of a two sorter replacement for some of these other devices.

We're in the kind of 200 picks per minute to over 500 picks per minute. And it's pretty exciting to see some of these things. You're just sort of seeing a whole cascade of garbage. And then like something you want is just sort of like coming out of that. But you can imagine when you're talking about sorting through an entire city's way stream, like you sort of need these three puts to be viable. Yeah, definitely. Okay, so what's the status now?

How many facilities are using your technology? Who are your customers? Yeah, it's, so our customers will be large waste haulers, groups like waste management, waste connections, smaller businesses, and then also municipalities that run their own recycling facilities. I believe we're present in a bit over 50 different facilities around the world. Most of our systems are here in the U.S., but we have some in Japan. We have some in Europe and we're getting more systems out all the time.

And this is hundreds of robots across all of these different locations. Most of them are in single streams. So plastics, metals, paper. We do have some in electronic waste and also what's called construction and demolition material. So basically broken down buildings. And you find it's pretty much the same technology stack. It's just you're pointing the data set of one of these different kind of types of material. How many robots would one facility need?

I mean, I suppose that's highly dependent. It is, but you're usually looking at it. Do they buy five? Yeah. So one of the key kind of elements of our growth go-to-market is allowing customers to get just one or two, get comfortable with the technology, then get more. And then once they're comfortable with that, get more and never have to really bet their whole business. And that, you know, eases the path to adoption for them.

But a typical facility can kind of take between 12 and 14 robots to really take advantage of all of it. I'd say at this stage, we do have some facilities at that level. Most customers are kind of in the like three to five robot range. And then at this point, most of the most quarters, we sell more than half of our systems to repeat customers. But they're sort of in that adoption phase of getting more and more systems and fully getting close to full automation. And then how does the pricing work?

How much is a robot? Do they pay on a monthly basis? Do they just buy the hardware outright? Yeah, they buy the, typically buy the hardware outright. And you're looking at, there's a lot of specifics that go into it, but a price point that might be over $300,000 for a system. This will compare favorably against kind of two full-time employees replaced per system. They'll run two shifts a day. So you're replacing four FTEs tends to lead to as roughly a two-year payback.

There's lots of specifics there if more robots are installed or depending on ease of installation. It may be lower cost and maybe higher cost. But you're talking about something in that range. We do then have different lease arrangements that bring the cost in under. Typically, when they get a lease, the lease payments will actually be under their existing labor spend. So there is an ROI pay, you know, on day one. And, you know, we found that's very popular. And for all of our systems, we tend to offer this kind of KAPX focus thing, a lease.

And then there's different subscription services to either increase the capabilities of the robots or provide different updates, software updates and services. Got it. that's the unicorn part of the business plan. It's some kind of stuff attached to it. Yes, yes. Tends to be important, yes. And then how hard is it to sell into, it sounds like once people try it,

they want to buy it. You've got a real Pringles thing happening here. But like, how hard is it to sell into municipalities or waste management or, you know, is it a tough sales cycle? Just because it's hard to change things? Or is the value pretty obvious?

I think we really engineered the solution to require little change to their existing operations. And so that's made it easier. I'd say, you know, what's really important is you're plugging in a system and really asking someone to trust a part of their business to your technology. And that technology is new, has different maintenance requirements, sort of behaves differently than what they're used to. And so there's most of the sales cycle involves having people trust that your technology is really going to work. That means visiting reference sites, you know, visiting us at headquarters and really kind of like seeing the technology and action and understanding in my setting, how is it really going to work? And so kind of travel in person visits ends up kind of dominating the sales cycle. But, you know, with a strong reference base and strong repeat customers, you know, that, you know, helps sell these systems quite a bit.

The labor issues in the industry are so acute that, you know, when you start to have macro trends like what's happening in the labor market. right now, it also tends to push people towards this type of solution. Right. And it sounds like right now this is a solution. You said most of the installations are in the U.S. It's a solution that sort of enables improvement at existing recycling facilities. What about places where some of this sorting doesn't yet exist? Like, do you see this as an enabler of value in places it does not currently exist?

Yeah. Absolutely. Most of our focus is on upgrading the existing infrastructure and existing. facilities, but it is, what you start to see is building a recycling facility around this technology rather than deploy, kind of retrofitting into an existing one, allows you to build much lower cost recycling facilities that can operate 24-7, which means you require less sort of hourly throughput. You start to see that there is one of the challenges of the recycling businesses today when

you build a facility, it sort of has a certain price tag to get a certain returns to scale, And what that means is you end up having a minimum population that you need to serve. So you see recycling facilities in bigger cities where it is a better business and fewer systems out in rural communities. And this is part of why actually a large fraction of the country doesn't have access to curbside recycling. It's just not viable to build smaller facilities. When you start to look at lowering the capex of the facility, now you can have a payback on a smaller population and you see more rural recycling or in more developing sort of areas. the recycling starts to be viable. I would say we're only beginning to make an impact there now,

but that's the real sort of long-term vision is really expanding the recycling industry, not just sort of helping the industry that already exists. Right. And that's how we get to the world without waste. Yes. Yeah, exactly. Yeah. Thank you.

Yep. And then I guess talk to me about the, before we go back to economics, like, tell me more about the benefits, right? Like, as we recapture more of this waste stream, what is the overall impact? What's the, you know, the climate impact to the mission? Yeah. So, you know, I think I think people are maybe sometimes skeptical about recycling. They sort of like hear different stories and things like this.

But when you really look at the numbers and the EPA actually has some great numbers in what they call their warm model. But what you can see is for every ton of material diverted, it's going to have some, it's going to offset the production of some virgin material. And that almost all cases has a positive. carbon impact. It's especially high for metals. It's still high for plastics and it's still there for paper, but on a per ton basis, it's usually a little less, but the paper volumes are so high. But you have this immediate offset on carbon from the recycling process. So of course, that's helpful. When you're dealing with waste, you also, it's not just about what goes into the landfill or outside of the Western world sometimes going into the oceans. It's also about the transportation

costs and then the transportation emissions to kind of get material to these destinations. The ability to kind of localize the reprocessing of material ends up having some of these auxiliary benefits as well. But I'd say, you know, the biggest thing that I'm excited about is if you can make recycling and the diversion of material a strong business, you start to create this economic incentive where people want to bring material to somewhere that's going to create value out of it. And so you start to see a stronger incentive to building waste management infrastructure in places that don't currently have it. A lot of the sort of waste pollution and sort of plastics in the ocean are actually coming

from the developing world where there's a lack of waste management infrastructure. And it's my hope that by creating a stronger economic incentive to diversion, you start to see a buildup of infrastructure in those places. Right. So if the waste is valuable, it is, it fundamentally has more value and you know that and it becomes a sort of a global market for glass or aluminum or paper that is predictable and much more valuable than throwing it in the ocean. Yeah, exactly.

Right. And, you know, one of the things I remind people of is, you know, you see these videos of sort of scrap pickers going in, in the developing world, kind of going over landfills. is trying to pick out like little plastic bottles and things like this. It obviously looks horrendous and unsafe and all of these things, but like fundamentally there is value in that material. Like people are trying to capture that value and working in those kinds of conditions, I think speaks to just how valuable these materials are. But yeah, the sort of techno-optimist in me and the capitalist in me is kind of like sees this and says, okay, if we can create an incentive,

the world will align itself behind the capture of these materials. And that would be very good. How are those materials priced now? Like, do you see a future in which your data collection, for example, and your data processing can contribute to that kind of pricing? It's a really good question. So, recycled materials have historically, usually traded as a discount to their version counterparts because of different quality issues. Like, the material itself is just not as good.

Two reasons for that. One is there's actually impurities in there, so you don't want your plastics to be. like one part per million diapers and one part per million, you know, whatever shoes. So making it more pure makes it more valuable. But also, there's a lack of a consistent chemistry. So when you separate out something like number one plastics, you get bottles, but you often also get what are called clamshells, but like think strawberry container type things. You'll get, you'll get stuff with different colors in it, creates an inconsistent chemistry that then degrades it. One of the really interesting things about artificial intelligence in this industry is not only can you make

these lower cost sorting solutions, that's kind of phase one. Phase two is you actually can identify material more specifically and say, oh, this is a Pepsi bottle. This is a strawberry container. I can now separate them and deliver more precise chemistries, which significantly increases the value of these materials. And so not only do you bring the cost down, but you make the stuff more valuable, recycling becomes an even more exciting business. And the the ultimate hope here is we can basically create bespoke chemistries for different buyers and materials and make the recyclable materials much more valuable than they are today. And be superior to the sort of virgin stuff.

What does your timeline look like overall? I feel like we're at a moment where recycling as an industry is pretty moribund. It feels like you're coming in with a cape on at a moment when consumers are starting to lose faith. you know, we finally like sort of got the message out about recycling and now people are starting to lose faith in it as a concept. And it feels like technologies like these end economics that you're describing are pretty clutch. Like we sort of need need that message out there ASAP. Yeah. Well said. No, very, and hopefully all the VC's out there here, our economics are very clutch. Yeah. But, um, you can just clip that part, just that part, TikTok.

Yeah. So recycling is actually at an interesting point where what's happened is people are really genuinely concerned about carbon emissions. They're concerned about plastics in the ocean. You know, these figures about like there's going to be more plastic than fish in the ocean, like really bother people, like as it should. What's happening is consumers are actually making purchasing decisions based on their perception of the sort of sustainability of the packaging.

and what's happening is brands are starting to use more recycle content in their material as a way to create this better narrative. Like they're sort of doing the right thing and they're being pushed by consumers to do the right thing. So again, like the opt-of-mist to me is like, oh, the system is kind of working, but it's really consumer-driven. Right. And it's also now being legislated. So California recently passed what's called an extended producer responsibility set of laws Colorado has as well.

Recycled content use is being mandated into different types of packaging. So what's happening is actually the demand for recycled material is rising. And what you find is if you look at the commitments that the brands have made and what's being legislated, it actually exceeds what the industry can produce. So it's almost flipping over where you don't have enough supply to meet the demand. And so the ability of technology like ours to sort of grow the industry is really a necessity. So I'd say for consumers, they should, when you start to get into some of these dynamics, like you start to get very optimistic about recycling.

I think the other thing people should realize is it's a commodity market that they're selling into. Like you're selling plastics, you're selling paper. So you're competing against oil and natural gas. Like sometimes oil is a great business. Sometimes these guys get crushed. Same thing happens to recycling. But I think because everybody has this personal connection to recycling, like, you know, you're putting stuff in your recycling bin, anytime it's not going,

well, people feel it personally and get frustrated. Yeah. But so it goes through cycles. That's a really valuable insight, I think, for people. Like, just keep doing it. Keep doing it. Oh, yes. Absolutely.

And then the commodities part of that is fascinating. I had not thought of it that way at all. I would just add to that and also continue to make purchasing decisions that drive sustainability because it's actually having a really major impact in the industry right now. Yeah. I'm so glad that you said that because sometimes I feel like the last VC in the world who still believes in the consumer. the power of the consumer like people.

A collective problem. A collective is made up of individuals. Yes. Yes, absolutely. And then tell me, because we're calling, because you're here on our next unicorn segment, how do you scale? How do you become a unicorn? Like, what is the, what does the road ahead look like?

Well, there's kind of two main avenues for us. So one is just making excellent products, like having our robots just be exactly what these recycling facilities need. And we have some really excited. innovations that are landing soon, just having the robots be incredibly intelligent about how they pick, incredible hardware improvements that kind of get us to the next level in performance and making them ubiquitous in the existing infrastructure. What we're also doing is then pioneering the design of next generation recycling facilities

built from the ground up around our tech. And as we prove out these facilities and start to get them adopted across the industry, you start to see these two sides to the business where, you know, our technology is helping the existing infrastructure. We're growing new infrastructure that expands the market. And if we're successful deploying both of those, then hopefully we have a very large and successful company pretty quickly. Unicorn galore.

Yeah. Yes. How, just as a quick kind of technology follow-up, tell me a little bit more about the AI and the training and how long it's taken you to develop the software part. I mean, the really the brain part of this. Um, yeah, uh, too long. Um, it's, uh, when I started, uh, people are listening, well, see, I don't exactly have a lustrous head of hair, but I did, uh, before starting all this. Um, but, uh, the, so, um, so yeah, we're, I think, uh, not, not very dissimilar from like autonomous cars or,

other things. We are doing basically in the industry you call a semantic segmentation problem. We're looking at images and we're identifying different objects within those images. And then with that, we're getting sort of gathering characteristics and then making decisions like targeting on those items. So we get RGB color images, not that's different from what you have from a cell phone camera. It goes into our neural networks. They do identification of the material, I'll sort of spin out these targets. That data, that's done on the edge, meaning we actually have basically high-end gaming machines

with each of our robots doing that identification. They'll also store data that'll be uploaded to cloud infrastructure we've built. So we built our own annotation toolkit, our own databases and everything. Then we have annotators go through that and say, this is number one plastics, this is a piece of newspaper. That ends up being a really tough part is not training the neural nets, training the people, because you have to get people to be experts in, you know, like if I were to say this is colored H.D.P.E. Or this is, you know, a corrugated

container or, you know, all of these different subtypes of paper that everybody in the industry cares about. The average person doesn't know this stuff. You have to train them to become recycled materials experts. Anyway, they annotate it. That goes into a training set. We train a neural net. Those will get deployed around the world. We have our robots, you know, in the United States, It's learning from our robots in Japan and vice versa. It's really a pooled data set and pooled neural network. What's interesting is, so I started the company really, I started working on it in 2014, and then the company really opened its doors in 2015.

We started to have neural nets to do this within six to 12 months, even back then when the deep learning technology was a little bit more basic. But to have it work well, it took another two years. Like we needed datasets from dozens of recycling facilities. we had to do tons of like banging our heads against the wall on the right, like data augmentations, and this is the right algorithm. And here's our custom tweak on the algorithm. There's a, it's quick to get a lab demo.

It's very hard to have something that works well in the real world reliably. Yeah. Wow. It sounds like it's a solution, like many things that rely on this kind of level of computer vision. It's a solution that like couldn't have existed before. Yeah, that's at least our belief. The robots, a lot of the hardware,

You could have done this 20 years ago. And we're not the first to think of robots and recycling. I actually found a research paper, I think, from the 60s talking about this. But having a vision system that could identify inconsistent material is really what's kept this from happening already. We should quickly note that you have raised a total of about $162 million. Your last valuation was $571 million. And hopefully you're well on your way to unicorn status.

This is a great conversation. Thank you so much. I appreciate it. Yeah, I really appreciate being highlighted in this way. And this is my cheap joke, but always happy to talk trash. Amazing. Bravo. I'm going to give you a golf clap on the way out. Matanya Horowitz, his founder and CEO of Amp Robotics, I see why they don't like you name stuff. Thank you, Molly.