Economic Chaos, More Pitch Madness & the Shopify AI Manifesto | E2108

26:37Yeah. Okay. So if we have it on YouTube and we have it on our server, then we have redundancy. I'll back it up to my machine. Editors back it up to their machine. We don't need to buy those SanDisk drives. We don't need to have network arrays. We don't need to have AAS3 and do all this stuff anymore. And the consumer surplus is huge for that. But there's also another side to this, because if we do see the major technology players, the people who own their own hyperscaled infra, come in and sweep the bottom two thirds, as you said, companies like Priya or, I don't know, Quora's Post Service and so forth.

27:09I mean, these could be essentially undercut by the already large. And it would be very disappointing to me as competition's number one fan to see that because it would just be essentially people with too much cash able to own part of the market, much as we've seen Spotify not have really any pricing ability to change their pricing because Apple has anchored everyone at$10 a month still, even with inflation where it is today. How much do you spend on storing your photos, Alex, right now? I think we pay a couple bucks a month to Google and a couple bucks a month to Apple because we're not efficient right now, much like you and your music services.

27:43You pay a little bit of storage for your overall phone and some percentage of that is not just your apps and music and video, it is your photos. So you're probably, it's abstracted that you pay an extra hundred bucks a year for your family plan. Sure, somewhere in there. It's so de minimis that you don't even know how much you're paying. I don't. In all likelihood. Same with me. I have Apple One. It includes Apple TV, Apple Music, Apple Arcade, and some amount of storage. People used to pay for Flickr. They used to pay for 500 pixels. Was that the company that everybody used? 500 PX. Yeah, I know what you're talking about.

28:17PX or something. Smug Mug. Oh, throwback? Yeah. And I think Smug Mug wound up buying Flickr. And I think the idea that a consumer would pay for those things, maybe you'd have to have a lot of photos and you'd have to really want like certain features so this is the problem with consumer subscriptions is they can get abstracted into you know a bundled offering the same way you know some spreadsheet can be bundled into an office just like how microsoft teams was bundled into office 365 subscriptions and that really did i think handicap um slack and zoom's growth to some degree i mean it's not trivial now they're both public companies are part of one but still I used to track the Slack weekly actives and monthly actives versus the Teams numbers when they were kind of competing to see who could grow faster.

29:07And now it's just Salesforce versus Microsoft, so who cares? But yeah, things can get eaten up pretty darn quickly. But korea.ai, k-r-e-a.ai, if you want to take a look, I think it's a pretty cool service. I hope they crush it. I hope that our concerns are wrong and they go to a pajillion dollars. We'll see. I mean, looking at it, I can tell you that this offering is much better than the free ones I've talked about. So just the ability to train, they just have, you know, a whole cornucopia of filters and styles and animation. And so you probably can do much better things much faster on it. So, yeah, I mean, it's definitely beautiful and people really eat this stuff up.

29:49I mean, I think that everyone remembers the time when they were a child and they just drew whatever they wanted without shame. AI image generation is quite a lot like that. You may not become the next famous painter, but you can have a lot of fun on your own and that's not to be diminished. All right. You didn't start your company to run payroll, did you? Of course not. We all know that. Gusto is here to help. Gusto is going to help you run your payroll and handle all your benefits onboarding and HR all in one place. The market agrees. 300 ,000 businesses trust Gusto today and you can too. As your startup scales, Gusto is going to grow with you.

30:27You got state and federal taxes handled for your staff around the country. Gusto does all that. And hey, maybe it's finally time for you to offer a 401k plan for your team, right? Gusto's got you on that. And you might need to get your compliance sorted, right? Well, three out of four employers say Gusto helps them be government compliant. And even better, Gusto is simple, easy to use software. So you can focus on what matters, building your startup. So here's your quick call to action. Do you want all Gusto has to offer with no hidden fees? Well, how about a discount? Try Gusto and get three months free.

31:04Gusto.com slash twist. That's G-U-S-T-O dot com slash twist. now jason there's a clip from mid journey you wanted to talk about which we have prepared here all about mid journey v7 and its new voice activation feature the reason i wanted to share this is people kind of forgot about mid journey and because it was taking too long to make images was taking too long and the prompting was too long what i thought was really interesting about this is um nick saint pair who saint pierre he's like on a twitter account he kind of releases a lot of stuff and keeps you up to date on it on x he just put this um image out where it explains how it works and the new model lets you in this draft mode just talk to it like your decker in blade runner saying zoom in two to the left two to the right do this do that and it's almost real time so let's show it mid-journey just released v7 and the killer feature is this draft mode that they have it lets you just have a conversation with mid-journey you don't have to worry about the prompting free flow the ideas let me show you how it works here so all you do is you hit draft mode and you can turn on voice now you can just talk to mid-journey hey can you uh make a photo of a man uh sitting in the park please there you go it's starting to it's writing the prompt for me it's giving me some images and look how fast it is can you put him on a like a picnic blanket and put him next to a picnic basket as well so good it's so fast uh can you bring it in a little bit tighter and then make him wearing a black t-shirt and i want him to be sitting next to a woman as well who's in a like a red dress it seems to be lightning quick it looks pretty promising uh there was a great scene in wag the dog if you've ever seen that with dustin hoffman he's basically working over the shoulder of a video editor and he's like okay add a dog add this and they're like making a war scene and you're like oh my god it's almost like conspiracy theory like they're faking the moon landing or whatever.

33:01And he's just over the shoulder working with an editor to make more and more interesting stuff. But they're doing it so quickly, it's kind of like shocking at how fast they could do it. This has the same feeling to it to me, which is you don't need another human. So when we talk about job destruction or making people bionic or static team size, here's like a perfect example. You would send your notes to a video editor 10 years ago, wait a week, they would send you some stuff back. And what we just saw might've been four or five iterations till people give up and say, okay, that's good enough. Now you can do all this stuff, then give it to a video editor to finalize them, polish with whatever tool they're using.

33:42Just like vibe coding, I might be able to build my annotated.com website and, you know, make it, make some wireframes and some designs and make it good enough and copy coda or notions design whatever and make it look like those products and then ship it to a dev team to polish it and finalize it that's where we're headed and you know this is truly impressive but this is version seven and this product how long has mid journey been out i wonder i'm thinking it's like a four-year project maybe they're in year five of the project so this is where they're at in year five version seven of this and i believe mid and he's an open source project?

34:19The initial release was July 12th, 22. 22. I said five years. So this is three years. Holy cow, less than three. But that's seven years in AI. So. Yeah. Yeah. So 21 years in AI. But they're three years during version seven. So they're releasing a version every five, six months. Yeah. Man, if they get to version 14 in another three years, I can't imagine how fast and epic it's going to be. It's going to be 10 times. It's going to be essentially instantaneous and it's going to be three times as high quality. and you're going to be able to go from a still to a video very quickly and then talk to it live it's going to be insane i i can't wait and you're going to have an agent where you're going to say go get me every video of ashoka you know where i'm just going to upload you know 10 videos from the star wars uh genre i like and do my own training on my own machine with a you know a little tool and then i'm going to make my own trailer for the new sequels i'm cutting to do George Lucas's original script.

35:18And that's when we'll know we've crossed the Rubicon here is when you start seeing trailers that you get fooled by. You get fooled by a trailer, which I think is starting to happen. People can get fooled by a Star Wars trailer online, or you can make the entire episode of the next Mandalorian and people can't tell. Those two moments will be here in three, four years. Well, I mean, the whole world got confused today by a tweet and a misunderstood interview to the tune of several trillion dollars. So I believe we're going to have spoofed trailers very soon. But the point you made about doing things quickly to then give them to someone else so that they can finish it is happening in the business world.

35:57Toby Lutke from Shopify said today that his manifesto we sent around to Shopify employees was getting leaked. So he just put it on the internet. And point two is that AI must be part of your prototype phase. The prototype phase of any project should be dominated by AI exploration. prototypes are meant for learning and creating information and ai dramatically accelerates this process now jason i was going to bring up the toby uh memo because something else in fact um there's a final note in it that i think fits very neatly into our static team size or compressed team size trend and we pulled it out here into a quote the final item is before asking for more headcount and resources teams must demonstrate why they cannot get what they want done using AI.

36:43What would this area look like if autonomous AI agents were already part of the team? This question can lead to really fun discussions and projects. And increased earnings and faster speeds. This is exactly what I wrote, I think, two years ago when I wrote my little manifesto, my ADD manifesto, automate, delegate, deprecate. And here we are, you know, in our own company, when I add headcount, it is rare and it has to be somebody who's a bar raiser. No more, you know, throwing bodies at problems. And you know what? The team has stopped asking me for bodies, which is also very interesting. I forced everybody to put on their browsers, open new tab, and it shows ChatGPT.

37:32Open new tab, show ChatGPT. And what that did for the team was some number of people started using ChatGPT regularly. Now, as part of locking down our computers, because we have a lot of interesting information on it, I can see an aggregate. I can run a report and see what percentage of time are people in Notion versus Slack versus ChatGPT. And when I did this little beta test of the software, I'm not going to mention what we're using to lock down our computers, but we've got to lock everything down because we have partners and we have LP data. It's pretty serious stuff. And when you lock it all down, you get this productivity boost because I was like, wow, we use Notion and Coda and Slack and Zoom all day long.

38:17All day long. For certain people in our team, their entire work life is those three or four cohorts of products. And I'm like, where's the ChatGPT? I was like, oh, the investment team is using ChatGPT 3 % of the time. I'm like, oh, I wonder if that's low or high. People need to be doing a little more ChatGPT or maybe not. And then I realized, oh, a lot of the work we're doing in Coda and in Notion, the native AI in those products is what we're using. Yeah. Yeah. So it's actually really interesting when you start looking at the aggregate data of your teams and where you're spending your time. And that's the next phase of this.

38:52Man. Our computers are going to be studied what we do every day. This podcast, what you and I do every day, there'll be an AI producer in here by the end of this year saying, you know what? You guys don't show charts all that often. I'm going to start making charts from the docket. And the docket, I think we do in Notion. We've done it in Coda before. In this Notion docket, the docket is going to start suggesting, oh, there's an update to this new story. Here's a chart. Oh, there's a fact wrong. Like, why isn't the Notion page of the docket? There should be a button that says fact check every hour or every time we add something, fact check it.

39:29Now, Notion doesn't necessarily have to put that in, but somebody should clip this and send it to the Notion and Coda founders. Why not have fact checking built into these or why not do suggestions for additional facts, right? Instead of having to proactively highlight something and then ask the prompt to do it. And that's where a Gentic is going to come in. Right. So it's going to follow along and do things for you. Yes. And it should be following us here. And when I say something in this, you know, when I say pull up the jump seats from the Model S, it should have already, when I mentioned the Model S has jump seats, it should have shown that on the bottom and then given you the ability or I the ability to hit a checkbox to show.

40:04Yes, exactly. Why isn't Zoom doing that. I just think we're talking about the future six months away versus the future, sorry, the future we're living today. So I think you're dead on. You're just slightly early. But the fact that you're saying, why isn't this here to me implies either companies are going to be on it and they're already building this, they're going to win or they're not building it now and they're going to lose. But you have to have a really fast cadence. I do want to say though, your point about having chat GPT in our new tabs, going back to the Shopify memo, two quotes, Reflexive AI usage is now a baseline expectation at Shopify and quote, using AI effectively is now a fundamental expectation of everyone at Shopify.

40:40And by the way, folks, Shopify is a well-run company. It's done very well. Like this is not some random company with some random person. Tip of the spear company. And we're, I hope, a tip of the spear, you know, pre-seed accelerator. And so if Toby and I are thinking this way, I think it's probably going going to trickle its way down to every organization. We did our first round. Yes. Founder Friday. Here's the bracket as it now stands. So as you can see, a few companies have moved ahead. We've got Osprey from Houston. They've moved in to the next round. We've got Whisper from right here in Austin.

41:13And then we've got MedSimple from Florianopolis Brazil. They've all moved forward. They are going to be competing in the Elite Eight. As you can see, still a lot of these matchups to do. So we've got two fresh matchups. Right now, let's do it. We've got some decisions to make. So first up, from Portsmouth, New Hampshire, we have Pitchfire. Love getting cold calls and cold emails. In the history of the world, no one has ever said this. In fact, this year, there'll be 645 billion prospecting touches that will get ignored by prospects. It takes the average sales rep 220 pitches to generate a conversation with a prospect.

41:47It'll cost their company over$3 ,500 to produce a meeting. 80 % of B2B companies are missing their pipeline number today. There simply isn't enough incentive for a buyer to get back to a prospect. This is why we built Pitchfire. Pitchfire is a paid prospecting platform that allows companies to build pipeline by rewarding their prospects to engage with their sales team. On Pitchfire, businesses get answers, buyers get respect. A business can find a prospect on Pitchfire, write a quick thousand character sales pitch, and offer them money to review and respond to that sales pitch. Buyers get paid to answer these pitches even if they're not interested, as long as they get back to that person in seven days.

42:18When they do answer, 40 % of the money goes to them, 40 % goes to their company so they can use it on Pitchfire, and the other 20 % goes to us. That's how we make money. That's our business model. Businesses only pay money when a buyer responds. Our whole go-to-market motion is a flywheel. We convince businesses to deploy capital on Pitchfire, and we run paid prospecting campaigns at their target accounts. Last month, we sent over 50 ,000 paid prospecting pitches out for companies. All right, so what do we think of Pitchfire's approach to building a circular economy of pitching for money? At first, you would think this is smarmy, icky.

42:49I don't like it. And I didn't at first. When I get one of these things, hey, if you do a call, we'll give you a$25 Starbucks card. I'm immediately like, my time is worth more, which means the sales team with that$25 gift card is probably wasting their time with somebody who just wants the$25. However, if you do this with enough prospects, I always think about those timeshare scams where you go and buy a timeshare in Jamaica or whatever town. Free trip, but you got to listen to our pitch. Right. And then they hard sell you and then people record it and share it on TikTok. It's a whole genre. For some reason, I got presented with one of those and I liked it.

43:30And then I was just like, it gave me 20 more. And I was like, oh God, don't like these again. Super uncomfortable. But if people are doing them, it must work, right? What I like about this one is the reciprocity. if you take my pitch i'm going to put half the money into your account so if i pay you 50 bucks to do this you get 20 lawn or i should say your company gets wait lawn gets 20 lawn's company gets 20 and this pitch fire gets 10 exactly so i like this concept and i would like to pull the string and see if it actually works what the devil's in the details what you want to make sure of is that the people who are on the other side are not doing it just to get the money and And that, hey, if some number of them actually use those dollars, it would be awesome.

44:15If they don't use the dollars, then that's found money. I wonder if you get the money back or it sits in that other person's account for a year and they have a year to use it. And then who does it? Who does the breakage go to is my question. But this is weird. And when I hear weird, I get intrigued. Juan, your thoughts quickly. You kind of said what I was thinking too, which is this, it's a good idea, but I think it's really all about the execution. I could easily see this turning into mostly spam and junk, like people using it just to collect the money who don't really care about the pitches and aren't going to really take your product seriously.

44:52And like then how do you weed through all of the garbage accounts to find the 10 percent of people who are genuinely using it like that? I feel like a lot of these kinds of systems that they eventually degrade into that. But if you did a really good job of managing the community and moderating, and like, I think you could actually turn this into something pretty intriguing. Some people like to, you know, get paid to do surveys. Who's next? Up next, we have a AI-powered workspace platform for creatives from San Francisco. Here's Creatant. We're building a creative super app, an intelligent visual workspace for teams to focus on creativity on Qatar.

45:29Meet James, a fashion designer from New York. To take his projects from idea to production, he was juggling between multiple fragmented tools, wasting immense time and losing track of his content. But James is not alone. 86 % of content creators want a single unified platform to handle all the creative needs. This is exactly what Created provides your entire creative world in one place. Since James started UserCreat in May last year, he reduced his project time from 6 months to just 2 weeks. Now let me show you how Created works. You just need a browser. Now you have a beautiful user interface.

45:57Your images, your videos, your music, your online content, your articles, your designs, and real-time collaboration. This is how you can reorganize folders. Just drag and drop. It's intuitive. It's made for efficiency and to work very fast with your content. You can create mood boards. One click from the folder, you go to mood board view. Here's an example, interior design and architecture. It's also used in fashion. We have visual bookmarking. Here you can see Pinterest, drag and drop, into our browser extension, Instagram, drag and drop, Spotify. You can play instead of creating, and YouTube.

46:26And here are the articles we use AI on top of it. Created Box in VR, augmented reality for Apple Vision Pro and Quest. Just bigger screen. We have AI Instant Search on top of it. You can search hundreds of gigabytes of data in seconds. No manual tagging anymore. You can select content like this in one click, and the content goes to the correct folder. There's no manual organization anymore. And we support high resolution 3D content, and this environment is also real-time collaborative. And that's how Creatant helps creators, designers, and teams to work smarter, faster, and more efficiently. All right, so that was pretty quick.

47:00But Jason, first thoughts on Creatant? I've seen it before. I've met the founder before. It's a clever, like, Hail Mary, crazy idea. Anybody can do anything and organize this information. And whenever you're trying to do something that expansive, it's kind of hard to get an ideal customer profile. When he started talking about creating mood boards, I was like, well, that's interesting. And I think this is what Pinterest learned. You got to have a there, like a very specific customer you're going after. All creatives are not created equal. I don't want to use a tool like this. I'm too busy. But I might like to see if I was, you know, redesigning my shelves back here and, you know, asking you to redo yours and asking Lon to redo his.

47:44If there was a tool where, I don't know, Chris or Lon or all of us could be throwing together mood boards of, hey, this is what we want. this studio set to look like this studio set, and it was just a quicker way to share and present ideas, I could see the person we hired to redo the studio using the tool like this, right? Yeah. So that's where I like it to be more specific. Houzz does something similar for interior design as an example. H-O-U-Z-Z. Yes. Really great company. And then what they did was when they nailed that, then they were able to put all the providers on. And then they made money by having profiles of the providers.

48:23So it's like a crazy idea with a crazy interface. It's like being presented for everybody. I would like to see it be presented for a more narrow group. Yeah. Like the first one is extremely focused on salespeople and even SDRs to get the first call to go from email to meeting to book a meeting, not even close the sale, just get a meeting. So interesting. What do you think, Alex, of this one? Well, first of all, I'm an absolute sucker for a new interface. I love that they're building in VR. I'm a very visual person, so I really enjoyed watching them blast through it. My thought was, are the people they're targeting having the budget needed to make this a big company?

49:02Because when I think of creatives, no offense, I'm not thinking about people in the iBanking income bracket. So I was just concerned about their ability to sell this for a price point that will allow them to run the expensive AI in the background. but I will say my vote between Pitchfire and Creatent is going to go to Creatent because I don't like people sending me meeting invites. Okay. I'm going to go with Pitchfire just because as a business to invest in, I think they have a quicker path to getting to a million dollars in revenue would be a better investment right now. The other one's a little bit of a Hail Mary, like I said, and it's going to take a lot to get there.

49:36I wonder if the other one should, I love Cretan should be more AI based where it just builds you the mood boards and it posts them based on a prompt as opposed to me doing all the work, which it seems like they're adding those features. Lon, you are the deciding vote. I thought creating is it is interesting and I think visually it's cool, but I think the pitch is it's too focused on the drag and drop. And actually, I think the better or the thing that's more intriguing to me is the organizational. organizational like when you're trying to edit something that's complicated and you've got like a ton of assets that you're putting in that gets out of hand immediately and that's one of those use cases for ai that i think is so strong like ai can identify all the assets keep them organized sorry so that that that was kind of missing from them so i'm gonna go with pitchfire oh okay pitch fire by nose okay let's do our next pairing all right next up we're now gonna do kippy which is As a personal language tutor, let's hear from Kipi.

50:33Okay. Hi, my name is Steve Walshansky and I'm the founder of Kipi. There are is your personal language tutor. There are over 2 billion people who are currently learning a foreign language and many of them are struggling. The traditional methods don't really work well because those are designed as a one size fits all. And at today's world, people are expecting more personalized approach. If you opt for a personal approach to tutor, then it can actually get really expensive, which is pricing out many people around the world. So what people do is they are trying to find their own speaking partners.

51:08There are hundreds of groups where people get together and trying to speak with each other. The problem with that is that if a beginner or kind of intermediate is trying to teach another intermediate, they don't really make any progress. And I've joined a couple of these calls just to find out how difficult it is for these people. While the language learning market is growing at 20 % a year, it's expected to reach about over 300 billion in 2032. Languages like English, French, Japanese and Korean are gaining traction as well. So, KPIs here to help people learn languages and advance to a conversational fluency.

51:44Because many people don't necessarily need to reach full fluency unless they want to go for a job or a university. For most people, it really means knowing enough that they can travel and connect with local people. Kipi helps you practice your pronunciation, practice real-life scenarios, or simply chat casually about the things that you care about. Kipi launched in December 2023 and pretty much grew tenfold in last year. Kipi already gained quite the traction of superfans and continues to grow. Our competition doesn't sleep, but our competition is focusing mostly on English learners, where Kipi found its stronghold with people who are learning languages in English.

52:23Jason, first thoughts on Kipi? I mean, I think this is the best way to do it because the AI will learn your style very quickly. So if it knows you're good at just learning words, but you're not good at grammar and putting a sentence together, you know, it'll give you enough positive reinforcement, make you learn a bunch of words and then gently get you to do sentences. If you're great at sentence structure, but your short-term memory, you have a hard time with remembering colors and food, maybe it will help you with that. Or maybe you're like me and you have pronunciation issues because of your dyslexia.

52:59I like adaptive learning. And we used to do that through machine learning or through like forked trees where we'd ask a series of questions and then try to adapt the learning style to you. AI is just going to do this naturally and it's going to work. I also like the idea that you could customize it a bit if it's ai you could say hey i would like to learn about um this topic through reading my nix news and so show me my nix scores and show me my tech news in german and force me to learn you know german that way so i kind of like it i think it's i think it's great i love the focus on non-english speakers and you know it hit me as i was watching that I pitch that I would use this because I've gotten to know a couple of people who are of Nepalese descent.

53:46And so whenever I go to where I see them, I learn a new phrase in Nepali and then I try it on them and they giggle and we have a good time about it. And if I could show up with like a couple of sentences, the line's blown. Anyways, Lon, who's next? From Arlington, we've got Vidala Labs. They're creating generic versions of high cost drugs. Let's take a look. Hi, my name is Isaac Yonemoto. I am the CEO and co-founder of Vidala Labs. I'd like to tell you the story of Ben, who is my friend and a member of the board. He's a pediatrician, and early in his medical career, he did a medical mission to Central America, where he unfortunately contracted a pretty severe gastrointestinal bug.

54:27Now, one of the side effects of this bug is that you can get an autoimmune disease called rheumatoid arthritis, which is, as it sounds, a debilitating joint inflammation. To treat this, he takes a drug called Humira, which is a biologic antibody drug that costs upwards of$150 ,000 per patient per year, and sometimes can even be more depending on how frequently you need to self-administer the drug. So this is well out of the range of affordability for most people around the world, and also a burden on our healthcare system, since we all subsidize each other via either private insurance or public options such as Medicare or the VA.

55:05So the goal of Vidala Labs is to bring down the cost of the drug by reimagining the manufacturing, which allows us to make good margins and charge way, way less. This will increase the accessibility of the drug around the world and also at the same time lessen the burden for patients all around the world. Because not every patient is lucky enough to be a health care provider and have fantastic health care insurance like Ben. Thank you very much. Jason and Lon, what do we think? Hard to know. Not my, you know, vertical. So for me, easy choice. I'm going to go with the language KIPA because I don't know enough about this space, but it does seem promising.

55:50And so I do know that doing these drugs is, you know, really effective and generics change everything. I think the whole Roman and HIMSS business model is based on generic and compounding pharmacies. And instead of even bothering with your insurance to just buy these things for pennies on the dollar. So I do think this is like a big part of the future. I'm sort of with Jason on this one. Obviously, a cheaper version of Humira that people with rheumatoid arthritis could take would be great. I'm very for it. I wish them all the best. I think that sounds like a great concept. But in terms of a pitchable startup that sounds like something that people would invest in, that is like a Founder Friday sort of friendly company, I think Kippy is a very cool idea.

56:34Another great use for AI, just helping people run their languages and train. You don't need a live partner for that. An AI partner is perfect for that. So yeah, I'm going with Sydney's Kippy. Yeah, same. I just want to say though, Vidala Labs, your website's a little bit bare, but what you're saying is awesome. So when you have more to show us, let us know, because if you can actually bring the prices of those drugs down, you're a legend. But in the meantime, we can play and touch Kippy. So I think that's going to be my pick as well. So three to zero on that one. But shout out to the biotech people for pushing ahead on trying to make the world a better place.

57:05So there we go. So now we're going to have these next companies, if they want to do a new pitch and tighten it up, I'm OK with that. They just have to get it in within the next week or so. We've got six companies, three matchups left to do before we're before we're at the Elite Eight. Great. Let's do it in the next episode. Do we have a founder question, Lon? We have a quick founder question from Reddit. Oh, I like this one from Sean Guitarius on r slash startups. What was your hardest learned lesson as a first time founder? He's got his own example. He read the books, knew that he was supposed to keep 18 months of runway.

57:43Nonetheless, at one point it dipped down to more like 12 months, but he had so much faith in his team, didn't want to let people go just in order to shore up the financials. But as a result, conventional wisdom in this case was right. Things got worse. Eventually, about 50 % of the team had to be let go. That was this person's hardest learned lesson. So his question to the others is, what were tough lessons that you learned from your first startup? And what's the hardest classic principle of running startups to put into your actual practice? So I have learned this lesson twice, maybe like to, you know, write the ship and, you know, get the expenses correct.

58:22and that is very hard. So I empathize with him. Once you've been through it once or twice, you understand that you want to preserve the altitude you have to find runways to land and get more gas. So if you're flying between islands, you don't want to be out at sea and have to ditch the plane on the ocean. So instead of ditching the plane, you ditch some baggage from the plane to lighten it. It's a pretty dark metaphor, but it is what it is. Humans are sandbags. Yeah. If you were transferring iPhones from China to San Francisco, you might throw some iPhones overboard rather than all of them going down.

59:02There you go. So I think the other one is with talent, how to manage talent is always a hard lesson for folks. If you want to do a training program, do a training program. Invest in the training program. That was a choice we made here at our investment firm. If you don't want to do training, then hire people who have the skills and just get out of their way. Both of these strategies can work and they can work in the same company. You might hire very elite people to run finance, let's say. And you might hire one really great sales executive. And then, because you can't find great sales executives or they don't match your philosophy, you decide with your top salesperson, I'm going to create a training program for salespeople and I'm going to do professional development.

59:53What doesn't work is hiring people who don't have the expertise and then expecting them to be able to perform like an expert. So when we hire researchers here, we use the term researcher, not associate or not partner or principal or managing director, the other venture terms. This is a hard lesson. You can't be upset at people who you hire to be experts if they're not experts. You need to train them to be experts and you have to invest in professional development. And if you have experts, then you have to demand excellence from them and not be intimidated by them and demand they hit targets and that, you know, they are self-starters, et cetera.

1:00:31And it's really about crisp, clear communication when you're interviewing people. I find almost universally when a founder tells me or when I've had this situation, somebody's not hitting expectations. It's because either one, two, or three things happened. One, they oversold themselves in the hiring process and you got duped and you weren't able to get reality from their references, et cetera. That can happen. It does happen. I see it happen all the time when people hire like this sales executive with press shirts and cufflinks and they come across as incredible, but then they're terrible at actually getting on the phone and closing sales.

1:01:09The other is you saw the signs that this person wasn't going to work out and you just ignored them and you just hoped they would, you know, kind of get there. And you probably saw that, Alex, many times at TechCrunch. You hire a writer and you don't, they sold themselves as something and they actually don't file on time. And, okay, they misrepresent themselves. But then there's other times you enabled it. You didn't, you know, cut bait. And so if there's any doubt, there's no doubt in hiring and firing fire fast, almost universally. When I knew something wasn't going to work out, I took too long to cut bait.

1:01:46Yeah, people stay too long in relationships, jobs, and employee situations. Can I just throw in a thing that I've learned at startups? Because I've worked at a couple of days. Don't tolerate toxicity. That's a good one. Like this is a thing that will bring down your, even if it's a high performer, if they're just miserable to be around, they will scare away other talent and they will end up by themselves with a lot of mediocre people around them. Again, Jason, it's kind of one of those things you can detect early, but you might ignore for the sake of a deadline or whatever, but it's just, it's never worth it.

1:02:16Don't tolerate it. I call this my no jerks rule. I've got one too, actually, from years of working. Not a founder, but just like as in screenwriting, they use the expression, kill your darlings, which means sometimes you may love a character, a scene. It may be the reason you wrote that thing in the the first place, but it doesn't end up serving the story or the plot or the moment when you're done and you've got to cut it anyway. They call that killing your darlings. I feel like startups are the same way. There may be a feature or an idea or a concept that it's been with you since the very beginning.

1:02:51It was part of the pitch. It's a beloved part of the product for you. But if it doesn't fit where you're going now, you might have to lose it. And I've seen that happen multiple times in different startups where it's always a hard decision like, Well, but I love this feature. This was what I thought power users would be into, but it's not where we're going this quarter. Some refer to it as sunken cost fallacy, where you've put a year into a feature. It's got modest success. And modest success is almost universally the blocker to breakout success, I tell founders. So if you have a product or an aspect of your product, let's say you've got this week in startups, it's sold out, it's doing incredible, Then you do some other thing and it's not.

1:03:35But it's not growing, but it's growing 10 % a year. And the other product's growing 200 % a year. That thing that's growing 10 % a year, you could just reallocate those resources to the thing that's really growing quickly. You just have to understand the sunken cost fallacy, sometimes referred to as throwing good money, throwing good money after bad. So you made a bad decision. You invested a million dollars in a product that only makes 100 ,000. And you decide, oh, I'll put another million into it. We should put a million into that one. And it's like, yeah, nope, better to do something else. This is a great Reddit rapid response.

1:04:05Let's get this. We'll fly up there as quick as possible to our friend. I'll add it on Reddit for sure. And well done. He is x.com slash Lanz. I am x.com slash Jason. He's x.com slash Alex. Go do a search for This Week in Tech to see Alex and I guesting on Leo Laporte's podcast, which This Week in Tech was the predecessor to This Week in Startups, which was a predecessor to all in fingerprints along the whole way. And if Leo hadn't taught me how to do all this and Alex and giving us a shot, you, we wouldn't have created this week in startups without this week in startups. There'd be no all in without all in Trump would have won this presidency.

1:04:44So it's all dominoes. Thanks to Leo Laporte for getting Trump elected. That was the joke we made yesterday. He loved it. We had a lot of fun. So we went for like three hours or something crazy like that. It was a lot of fun. Give it a listen, everybody. And we'll see you back here on Wednesday. Bye-bye. Bye-bye. All right. So today we're going to talk to a company, a startup called Menu Data. But before I get into what the company does and I bring up the founder, I want to tell you a little story. So I live in Rhode Island, which is a very small state here in the United States. And actually, my family has some history here.

1:05:14My father got his advanced degree here in the city. So why does that matter? Well, back when my parents lived here in Rhode Island, in Providence, they told me that at the time, and this was back, I think, in the 70s, that there was essentially no place in the entire state to get Mexican food. And the one place where you could wasn't very good. By the time that I moved here, there was an absolute plethora of food options. We've got Uyghur cuisine, Thai cuisine, stuff from all over the world. And I'll just say some pretty banging Mexican food too. That's now the norm. I kind of go to a city and I expect there to be a great number of food options out there.

1:05:46Anything that I want should be probably within five or 10 blocks or five or 10 minutes if it's a driving city versus a walking city. This is the new world. People eat more variously, more diversely, and I would say probably eat out a bit more frequently than they used to. What this means is that there's a lot more data and information out there about what people are eating across menus and other sorts of things like that. So what do we do with all that data? Well, this startup, Menu Data, has an idea about just that. So please, welcome to the show with me. It's Sonny, the CEO of Menu Data. Hey, how you doing?

1:06:19Nice to meet you, Alex. Good to be on. Oh, absolute pleasure. So first of all, before we break into the company, you were telling me in the pre-show that menu data is now actually based not in Silicon Valley, not in Austin, not even in New York City, but instead in a place that I mostly associate with whiskey. So tell me about that. Yeah, so we're actually in Chattanooga, Tennessee. Part of this really awesome kind of accelerator program called Brickyard. So 12 teams come out here from all different parts of the country and just grind on their startups. You know, they have a sauna in there, they have like a cold plunge, they have a gym, and it's just a really awesome place to build.

1:06:56So little known fact about Chattanooga, it is the home of Coca-Cola bottling. So as you know, it's a famous railway station endpoint. There's just a lot of Coca-Cola and freight and 3PL going on here. I just pulled up the BrickHair website and it says, for the maniacs. I love that. I love that. It's always good to know that I'm talking to a fellow maniac, Sonny. It's my favorite type of person. All right. So I wanted to bring up just a little context about why I care about what you're building. But in this case, I'm a little bit curious what the starting spark was. Why did you choose to build a company around data from menus?

1:07:32And we'll get into why that matters in the business model in a second. But I'm so curious as to why you picked this problem to really put your back into. Yeah, it's very interesting. So kind of coming out of grad school, did a lot of research in human computation and training data for AI before AI was cool back in 2012. I worked for a startup called Crowdflower, where we'd solve these awesome big data problems for large companies. But when I was living in the mission, I would see kind of chefs with their hair on fire. They couldn't read what they were buying, 30 ,000 different vendor templates.

1:08:04So there's this kind of big data problem that always existed, especially for the independent restaurants all the way up the supply chain. And I wanted to focus on that. Is that because the hospitality business has always been pretty hands-on, people-oriented and dare I say it, low tech? I mean, a gas range doesn't change that much, I think, over time. And so did that lead to just the restaurant business being relatively fractured and kind of stuck in the paper and PDF era, if you will, longer than it should have? Yeah, I like to call it old school. But yeah, so it's a very kind of relationships oriented business.

1:08:37There isn't a lot of data transparency that's going on. And in our opinion, you know, food and beverage is one of the last sectors to kind of modernize. So it's undergoing lots of change all across the industry. You can see it from the way consumers are eating at restaurants to trends around Ozempic, GLP-1. The need for data is just astonishing. As a person who has young kids and therefore I'm time strapped all the time, I've become a real connoisseur of finding menus online to figure out what they have and what's in it. So that way I can check against allergies and so forth. It does seem that since COVID, a lot of information was taken off of menus that I can hold in my hand and turned into QR codes and digital formats.

1:09:20Did that lead to an explosion in the amount of information that was actually available in the market for you guys to kind of hoover up, if that makes sense? Yeah, absolutely. After COVID, almost every single restaurant and grocery retailer has put their information online. So through advanced scraping techniques, whether it's PDF, JavaScript, HTML, we're able to not only extract that data, but make sense of it using large language models. So we're able to convert the menu item and description, look up the recipe online, and ultimately translate that into ingredients. But okay, so let's say that you and I are sitting down and we're both going to make a red sauce.

1:09:58I would go about it with my favorite methods, my favorite olive oils, spices, and so forth. You would do it differently. but if you're going to be assigning ingredients to recipes or to menu items, isn't there a risk of having the wrong actual recipe for a chicken parm or a chicken piccata or whatever and therefore missing something that might be important? Yeah, I mean, so everything is based on confidence intervals. So for the large, high confidence items, if it's exactly explicitly mentioned there, then of course it's going to be 100%. But if it is something obscure like a sauce, it might have a different sort of percentage.

1:10:31So we're able to wait those out and give that to manufacturers or distributors or, you know, even menu strategy teams to sort of execute and either create new products or make sales. I'm sorry, menu strategy teams is a business or a profession? Yeah. Why do I work for a living? I should be eating for a living. Yeah, why not? I quit. Anyway, I want to talk about the scraping point because when I was just thinking about the business, I was thinking, okay, there's a lot more data out there. So that makes a lot of good sense. But we've also talked a lot here on Twist, and I think out in the broader technology scene, about information rights, intellectual property, the ethics of scraping, and so forth.

1:11:08Do those concerns apply in this space? Or is menu data just kind of treated as something that everyone can collect if they want, and kind of no harm, no foul? Yeah, we kind of view it in the no harm, no foul point, because this is kind of restaurants that on their own website have put their menus for public consumption. We're just simply using it for trend analysis. And eventually our goal is to have an independent restaurant version, almost like a freemium model, where actual restaurants can find localized trends and they can start becoming more data-driven. So we sort of view this as an overall improvement to the ecosystem.

1:11:43I was going to take us into the product before customer, but what you just said is very interesting. It sounds like then, Minudata, by the way, Minudata.ai, if you want to take a look, is not aimed today at the single store that might be looking at trends, but instead more at chains. Is that fair? Yeah. I mean, our initial customer base is going to be that large-scale manufacturer. They're typically spending millions of dollars on R &D and innovation, and they're frequently coming out with products like McDonald's McPlants or Spicy Coca-Cola that completely missed the mark. So there's 85... McDonald's McPlants?

1:12:18Yeah. What? So plant-based burgers are really popular on kind of the coastal regions, but not so much in the Midwest. And they just didn't have a lot of data around it, right? So when you can kind of compile the right sort of data and sort of create products that people actually want to eat, you're starting to really use this data to provide insights for business outcomes. Okay, so the tech stack here is a crawling mechanism, a data cleaning mechanism on top of that, and then AI on top of that to parse through it and find things and also answer questions for people and you're selling to more Panera than to, you know, Percy's single person coffee shop.

1:12:56Okay. All of that makes sense to me. Can you show it to me? Can I see it? Because I would love to get my eyes on this. Enterprise tech is always a little nebulous until you see, you know, the actual thing. So let's take a look. So imagine I'm an analyst at a consumer insights team at a food and beverage company. And I just simply want to know what's out there or what's innovative in the market. You can simply search for something, you know, like Mocha, for example. And you could simply see the amount of data that you have to deal with, right? 64 ,000 results. That's a lot. So we're actually running AI through every single menu item, looking at all of the ingredients and identifying information.

1:13:34So here you could see the cinnamon cayenne mocha is very interesting. Maybe a charcoal mocha, Alex, if you've ever heard of that. No, Sonny, all of these are blowing my mind as a black Americano drinker. But just for folks who are listening or watching, what the system has done here is pull up what appeared to be the more exotic mocha variants that allows people to then see what other people are trying. But you can't today tie this to sales data, for example. So you can't probably tell me if the cayenne mocha is doing better than the black walnut white chocolate mocha or whatever. Yeah. So what we do as an approximation is for every single restaurant location, we have kind of geolocation data.

1:14:17So that geolocation data will say how often consumers are going to that restaurant, the overall sentiment analysis based on reviews, and then for top, you know, within kind of, you know, food delivery websites, we're able to kind of grab those popular menu items and then tie that back to these items as well. So if you are a sales rep, you could understand that, hey, is this really one of their top sellers or are consumers mentioning it a lot? Okay, I want to try to put this into Dunkin' Donuts terms, if that makes sense. Because I live in Rhode Island, so the Dunk is a big deal out here. Let's say that Dunkin' was going to roll out a nationwide promo.

1:14:57Let's say they were doing their equivalent of the McDonald's McPlant. And it doesn't do well in certain markets. Let's say this new item only does well on the coast to stick to that point. Would Menudita be able to say that, hey, you know, the Duncans that are in the middle of the country have pulled this menu item from their menus, but it's being kept up on the coasts? Can you get that granular today? Yeah. So one reason why clients really love our platform is we are able to grab not only what are the top most popular items per location, but we're actually able to grab pricing per location. So we're able to say, whether you're on the coast or whether you're on the Midwest, how are those prices changing, whether it's for their own business or against competitors?

1:15:40And I'll give you an interesting example, especially for chains like Panera Bread, where they're not looking at conventional competitors anymore. Because if you've kind of seen in the news and headlines, Chili's is really exploding because they have this$9.99 triple dipper package and a$9.99 burger where you can kind of sit down, have a meal just for$10. So the idea of competition is changing, especially as consumers are looking for things that are definitely not as expensive. I love it. So what does it cost? So if I'm a big company, I've got big dollars. You've mentioned Teams. And if I'm spending a million dollars on this, my impression is that you probably have some pricing flexibility, just given that you're working with companies that have budget.

1:16:24Yeah, exactly. So, you know, all of our models, you know, start at probably around$5 ,000 to$10 ,000 a month. But if you want access to kind of different countries' worth of data, or, you know, if you want the retail module or, you know, the social media module, of course, it would cost more. I'd love to hear that. I mean, I love to hear startups charging a lot of money because it means they're doing quite a lot of good. Sonny, clearly having a lot of MoCas in the database is good, but I'm curious what other types of data you're hoping to add into the overall database, because I presume that there's quite a lot of information out there that you're going to accrete over time.

1:16:57Yeah. As we were mentioning earlier, we kind of are also storing thousands and thousands of ingredients. So within our database, it's segmented by different categories. Let's just for fun, we'll look at top fruits. So acai is a mainstream ingredient growing at 47 % in retail. So we were talking about new data sets. So menu data has just recently added in retail information. So we're grabbing data from new retailers, showing growth numbers, the top categories it's being used in, brand information, average pricing. So, you know, one of my favorites is kind of Dole, right? Dole, they have this acai, Dole line, and they're focusing on protein, right?

1:17:42So these are interesting trends and tidbits. Competitors would want to know, new product development teams would want to be interested in, consumer insights teams, as well as marketing. and eventually sell. This is awesome. So really, the thing that I've learned most in this second demo is that you're really not just targeting restaurants and chains, but also grocers and buyers and a much broader demographic of people than I had initially planned, which is good because in startup terms, that means your TAM is bigger. So that makes a lot of sense. Do you think you'll have marginal pricing power on adding more data in the future?

1:18:15Will that allow you to basically up your prices? Yeah, I believe there's this immense opportunity to become kind of the leader in food and beverage AI. So what I mean by that is to be able to continually add more and more data sets and by using AI and eventually adding in chat functionality, create this sort of product that anyone on the team can use. You don't have to be an expert on food. You don't have to be an expert on innovation, R &D. Anyone within the food and beverage company can just simply log in, ask a question, and get their answer. Can I give you a tip slash request at this juncture?

1:18:54Yeah. One thing that I've gotten a lot in my history of writing about startups for so long is, you know, how do I make my company stand out? How do I get more press and attention? But I always tell people is, if you have interesting data, anonymize it, and then publish it in some way. Turn it into blog posts and headlines and share that data with the masses because me, the journalist, doesn't have the same data sets that you do. You guys are like the best possible example of this, like especially in a high inflation environment with everyone watching. I've seen more charts of egg prices than I ever want to admit in the last couple of weeks.

1:19:27So I just really feel like you guys with your data sets and your ability to already extract insights out of it. My God, do you have a publishing campaign like set up to just show off what you have and also explain the world to people? Because it seems like a big opportunity. Yeah, so basically our marketing right now is basically done for free because we have these two amazing partnerships, one with Kantar, one with the Food Institute, where we're able to kind of partner on new ideas, new trends that are happening in 2025. One example of that is definitely what's happening in probiotic and prebiotic sodas, right?

1:20:01Oh, because Poppy just sold them. Yeah, yeah, yeah. Yeah, so we had this really cool slide showing how many SKUs were added at Olipop and Poppy and how they were able to differentiate and win the market. Because when you sort of look at the flavors that they have, they always have flavor combinations. So it feels more premium. You know, they have a raspberry rosé or they have an orange cream, which feels more unique and they're able to charge more. So we were able to kind of have this information out there. And then people from Coca-Cola were reading it, Pepsi, obviously. and it just kind of brings this new wave of brand awareness for us.

1:20:37Okay, that's great. But have you had Olipop and Poppy before? Have you drank them yourself? Yeah, I've tried Olipop. It's really, really interesting. Oh, I knew it. I knew it. I knew you were going to say interesting. What you're reaching for there is it turns out that humans find high fructose corn syrup delicious. And when you swap it out for something healthy, it's not as good. so shout out to poppy for making the bank but that's a trend that i'm not entirely sure i believe in which makes my point that i would love to see charts from menu data kind of breaking down like our actual grocery stores adding more skews to this trend is this real and if i was a like a cpg investor i would kill to have more than information and then i would go share it and spread the word and make you guys look like kind of the uh the crunch base or pitch book of food yeah that's our goal.

1:21:27And the thing is, Gen Z, for example, they don't, you know, they're increasingly interested in dry January, the sober curious movement, you know, healthier lifestyle. So every generation is going to have those fundamental differences when it comes to food and beverage. And yeah, I mean, it's really leading to more and more innovation in the market. Sonny, tell me about growth. I know you guys founded the company and I think it was early 2024. So how big is it today? And there's been a lot of talk about startups needing to grow. frankly, faster than ever. So I'm just curious about how quickly you're going today and kind of your expectations for 2025.

1:22:03Yeah, I mean, you know, towards the end of the year, we're already a little bit under half a million, you know, of ARR. We have new clients really lined up that are willing to pay more if we're able to add more functionality, and especially this new AI feature we're working on, where essentially we're going to have the first of its kind where it's going to be a chat GPT AI for food and beverage, essentially. Pulling in all sorts of data, right? Whether it's retail, social media, menu data. And the first step is to create a system where consumer insights folks, folks in food and beverage, they can easily search, get an answer on external sources of food and beverage.

1:22:42The next step, though, is to connect in internal data. So maybe it's supply chain information, maybe it's information about their existing products and their sales data. and the AI could give personalized recommendations on how their company could be doing better, whether it's a personalized sales script or just even what sorts of products they should be pushing to the market. So we're actually building this AI tool to allow insights teams to actually create products. So it's going to be like an AI product generator. It's going to bring in trends and ingredients and flavors and consumer preferences and create new ideas on the fly.

1:23:16And it might yield actually a world in which we have more menu diversity because if people see more trends, they can be more creative. So maybe we're going to end up with better food overall. As an eater, I'm a fan. But Sonny, I would be remiss not to ask you this. So there's been a lot of questions in the startup world about where value is going to accrue in the AI game. Is it going to accrue to the data centers, the AI model companies, the data owners, or apps? Or is it going to be vertical AI? Are you concerned at all? And I'm asking for the founders who are tuned in today. Are you worried at all about the AI model companies taking on ingesting the same data that you are and offering a similar competing service at what might be a more consumer-friendly price point?

1:24:01Yeah, no, I'm a firm believer in the vertical AI kind of model because there's just so much data ingestion, getting it in the right formats, normalization that has to be done along with fine tuning your own RAG pipeline. So I don't really see a world where, you know, the AI is literally going to do every single business. Although, you know, there's been so much talk about it on X where people are like, hey, you could just spin up Salesforce in a day. Yeah, but Salesforce is still in business. Exactly, right? Not to throw shade, but I think they're doing fine. Yeah, exactly. So our belief is that because these data pipelines are really hard to create and the data normalization is very, very unique to the industry and food data is so messy inherently, it turns out there's actually 82 ways to have an ex-Benedict on the menu.

1:24:54And just even normalizing their data takes millions and millions of LLM calls. We really feel like we're in this unique space where we're able to leverage what's already out there and have our own proprietary data set and then, you know, bring on existing models like Cloud or OpenAI or even Mistral and then have the ability to create these curated insights that never have been seen before. Okay, last couple of questions then, Sonny. How big is the company and staff today? Are you hiring and what is a difficult role you're trying to fill? We're primarily a team of three, you know, core team. Yeah, so really, really small.

1:25:31We have two outsourced developers, though, so actually a team of five. We just hired our first AI engineer. So, yeah, it was a very interesting process. So we actually had, you know, ChadGPT01 create the job description for the AI engineer. And then it also created a challenge. So we actually got 50 ,000 rows of our menu data and then issued that out where people could actually build their own version of this bot, send it back to us. And it really allowed kind of a real life example to see how people were approaching it. Were they using RAG? Were they using TFIDF algorithms? And it was just really fun overall.

1:26:07Okay, last word for folks out there who are building and are not in a major technology hub setting. Can you just give them some props that you can build in Chattanooga, you can build in Wyoming, you can build anywhere? Yeah, I mean, I would just say, you know, get out of the house, go to your local AI meetup. If there isn't one, just create it. But just try to find like-minded individuals. They're literally everywhere. I was so shocked to see the amount of great startups coming out of Tennessee and just the amount of innovation that people are doing, especially kind of the younger folks out there.

1:26:37It's just incredible. You know, they just taught me Cursor two weeks ago, so I'm actually pushing code. So yeah, it's just a wonderful time to be an entrepreneur. It's also a very funny time if they're letting people like you and I write code. I don't know if that's incredibly bullish or incredibly bearish. It's got to be one of them. All right, Sonny. And just to make sure the URL is menudata.ai. People want to take a look and we'll have you back on when you hit the million in ARR. So let us know. Yeah, looking forward to it. All right. Thanks, man. See you. See you.