This Week in Startups - Stop ghosting your friends with Nox’s RPLY, plus Alloy Automation and a Shopify flashback | E2209

Episode Date: November 14, 2025

👉 Register here to join Founder University Japan’s kickoff: https://luma.com/cm0x90mkToday’s show:Nox’s new app RPLY reminds you to text your friends back, getting you to iMessage inbox zero.On today’s TWiST, we’ve got Molly Cantillon of Nox, telling us how she plans to make lackadaisical texters (like our co-host Alex) more effective, even if it means breaking into Apple’s walled garden. Hear about how RPLY maps your relationships so that it can understand your messages contextually, and help you stay organized.Then, Gregg Mojica of Alloy Automation swings by the show to dispel some common myths about AI agents. For the most part, these little machine learning helps aren’t truly autonomous… Hear how Alloy’s agents decide when it’s time to bring in a human for help. PLUS is AI in a “lull” right now? Hear why Gregg thinks there’s still a LOT of room for these apps to improve.Finally, it’s another TWiST Flashback. We’re jumping back to 2013 for an eye-opening chat between Jason and Shopify founder/CEO Tobi Lutke. Hear some fascinatingly prescient takes on the rise of SaaS, how early Shopify differentiated itself from other ecommerce and website building solutions, AND why Tobi dreamed of making Ottawa a true tech hub.Timestamps:(02:57) Molly Cantillon of Nox is trying to help you get better at texting(04:01) How RPLY gets even chaotic texters like Alex more organized(05:10) The complexities of entering into Apple’s “walled garden”(07:20) Keeping RPLY safe, in case you’re using it for work!(09:24) Miro - Help your teams get great done with Miro. Check out miro.com to find out how!(11:59) How RPLY uses relationship mapping to understand and prioritize your messages(19:23) Perspective AI - Real insights, straight from your customers, and your first two months are on us. Just go to getperspective.ai/twist.(21:08) Why Nox wants its AI to become “the Invisible OS.”(29:30) Pilot - Visit https://www.pilot.com/twist and get $1,200 off your first year.(30:40) We’ve got Gregg Mojica of AI agent makers Alloy Automation(31:42) How the rise of AI agents changed the core of Gregg’s business(34:03) Why most AI agents still aren’t truly autonomous(35:00) How Alloy’s agents decide when to bring in a human for help(37:02) Why the rapid acceleration of AI development has made some enterprises pump the brakes(39:00) Is AI in a “lull” right now? Why Gregg thinks there’s still room to improve.(40:31) How close are we to AI apps going truly mainstream?(43:41) Why Alloy Automation didn’t raise a massive round like so many of its AI peers? (Or are they?!)(46:48) The difficulty of hiring in today’s AI landscape(50:21) In today’s TWiST Flashback, we’re revisiting Tobi Lutke of Shopify’s appearance from way back in 2013!(52:11) Why Shopify switched from selling snowboards to storefronts(59:37) Turning Ottawa into a tech hub, and the “secondary market” theory(1:05:32)The constant tension between growing your global footprint and expanding your features(1:08:52)Who were Shopify’s actual rivals?

I have more than 800 unread I messages on my photo. 800 is maybe even the lower end of some of the inboxes that we've had to deal with. It works with two services today, IMessage and WhatsApp. I'm curious about the privacy element of... What you want to do is not send your messages or at least have them anonymized and never personally identifiable if they're ever going to be sent to a cloud server. When I saw you guys built this, my first thought was, hell yes. My second thought was, will my friends yell at me for using it? AI reply generations sound like you.

I feel like we're like a year or two away, though, from having enough RAM and every single new Mac computer that we're going to be able to run local models without it being a question. People are not as excited as I think they should be about local models. And the reason why I'm excited is because you can have 300 requests in a second if your computer is gig a chat. And then you also talked about mapping relationships, kind of using people's text messages as a way to sort out who's in their life and when. What's interesting is if you take all of this hardcore statistic and you try to weave a story of the tapestry of your life over it.

This weekend startups is brought to you by Pilot. Focus on your product. Let Pilot handle your bookkeeping. Pilot provides the most reliable accounting, CFO, and tax services for startups. Head to pilot.com slash twist and get $1,200 off your first year. Perspective AI. Surveys, they never capture what customers are really thinking. That's why we use prospective AI. Real insights straight from your customers. And the first two months are on us.

Just go to getprrospective.AI slash twist. And Miro. Help your teams get great done with Miro. Check out Mero.com to find out how. Hey, everybody. Welcome back to this week in startups. This is Alex, and we have an absolute banger of a show today. We have not one but two Twist 500 interviews.

and then at the end, we're going to rewind that clock all the way to 2013 when Jason sat down with then much younger and much smaller company CEO, Toby Lutkey. Of course, Shopify has grown tremendously since then, but it's really good fun to go back and look at where the company was and what we were asking Shopify back in its relative infancy. All right, so Knox up first, great company. They're building kind of like superhuman but not email instead for iMessage. It helps folks like myself who are chronically and constantly behind on text messages. stay up to date and the company has a big vision for a feature AI-driven, invisible OS.

You're going to want to know what that is. Then we're going to talk to Alloy Automation, a company that I've covered since its earliest funding rounds. It has matured though. It has gone from being a e-commerce automation platform with a no-code twist to instead today being something akin to the data and integration platform for building AI agents. Trust me, there is a through line there that we explore. Greg's great. I love both of these companies.

So Knox, then Alloy, then Shopify, let's go. I'm not very good at texting. I'm actually pretty bad at managing all my inboxes, Twitter DMs and emails and LinkedIn messages and all that. But my SMS folder is always a mess. Happily, though, there's a company called Knox. They built a service called Reply that just might save my bacon. So please join me in welcoming to the show.

It's Molly Kintyon, founder of Knox. Hey, Molly, how you doing? Hey, Alex. I'm good. How are you? I'm good. It turns out I have more than 800 unread. eye messages on my phone according to my just checking that. Am I like the target for what you're

building? Because I feel like I'm always behind on text and I feel bad about it. And then I put it off because I feel bad about it. And then I feel worse. And, you know, it just goes on and on. Oh, absolutely. I think 800 is maybe even the lower end of some of the inboxes that we've had to deal with. It's been, yeah, really serious amounts of people there. But it's a self-selecting thing, right? There's already this bias of people who come towards our solution and they find it and they have 10,000 on read messages or in my case, I just read the messages and I don't know how many I have to respond to, but I haven't yet. Oh, so you mark them as red, but don't respond. So you have an invisible number of people that are mad at you. Yeah, yeah, yeah, yeah. Let's talk about reply.

It's spelled RPLY and you guys call it kind of a unified messaging platform to help people survive their text message inbox. How does it work? And really, apart from us, complete freaks, for whom is it for? So as you said, it's this place where there's one home for all of your messages. So it's iMessage, WhatsApp. We're working on the roadmap towards doing Slack, Discord, telegram, email. And essentially, we've just embedded this concept of inbox zero into your text messages. And so the idea of you having to respond to a person, what it does is it filters on exclusively the conversations where you were the second to last person to reply. Right. So, I mean, you know, once a week, every Sunday, I would go through literally my text box, my, you know,

iMessage and spam control tab and just spam that tab until I would see a gray bubble in the bottom left corner, which would mean I need to respond. And I realize, you know, there's an easier way to do this. What if I could just have this, you know, one filter, this one query, where I could find all of the people that I haven't responded to where I have an obligation to respond and just have one home for that. And so that's how it came to be. It's a desktop app because it's actually not distributed through Apple's App Store. a direct sort of DMG downloaded app, which is how most desktop files work. Can I ask about that? Is that because you couldn't release this through the App Store on iOS? Because I presume that there's privacy rules or Apple not wanting you to mess with IMessage.

So a few weeks ago, we actually launched our reply for iOS app. And it is not right now on the App Store. We're on test flight, which is Apple's beta software. And the way that it works is you register on your computer, and then we relay those messages to your phone. And so, yeah, it's essentially what you're saying. If you tried to do this, Apple would block it. So you have to do it in different, yeah, different engineered way. Do you think the Apple restrictions there are reasonable? Are they actually enforcing intelligent privacy protections for people?

Or are they just walled gardening and rinseaking their way to the bank? You know what? I think it's the thing that differentiates them and is a large reason why people are buying iPhones. Eye message. Yeah, I message. not having a green bubble, having a blue bubble instead. And so it's completely within their gate to reserve this right. Wow, that's the most positive thing I've heard about someone say about Apple in a really long time.

Okay, but it does, it works with two services today, iMessage and WhatsApp, probably the two largest non-straight SMS platforms out there. What would it take to get traditional non-IMS text messages in there too? Yeah, so you're talking about what Slack, Discord, Telegrams, or what types or SMS? Maybe this is me about to show my ignorance, but like, I message is a form of SMS, but not all messages that I send via text message on my iPhone are I messages.

Like, for example, if I text them with Android. So is all that brought in as well? It all works. Yeah, yeah. It's all in one. So it is all my text messages. Okay. Exactly.

I'm curious about the privacy element of this, because if you have got my Slack messages from work, I really don't care. They're coordination. They're boring. They're us talking about what we're working on. My text messages, though, often include stuff from my friends that, is not really mine to share. And so when I saw you guys built this,

my first thought was, hell yes. And my second thought was, will my friends yell at me for using it? So how do I keep my friends and loved ones information secure while using reply? There's kind of two separate parts to this answer. The first one is obviously just the privacy policy that we have and the compliance we have.

So we're working towards SOC2 right now because a lot of people are using this, you know, just so happens to be for their work messages, for messages where they're coordinating things, for their actual job, and a lot of their occupation sits inside of their text messaging inbox. But then the tech answer is kind of interesting because what you want to do is not send your messages or at least have them anonymized and never personally identifiable if they're ever going to be sent to a

cloud server. And so the first thing that we completely make apparent is that no data has ever sent sort of personally identifiable from our servers to any other servers stored anywhere along that process. So you're the only place where these things go. So as long as you secure at rest encryption on your end and then in transit, we're pretty much good. Yes. So we have zero data retention policies with all of the cloud models.

And we also support these local models. And so what that means is you could literally turn off your internet and you would be able to still have the AI reply generations sound like you, albeit, you know, might be it. turn your battery up, you know, go lower and it might have more CPU usage and there might be some other performance issues, but all this would be possible without having to send things to any sort of server. What local models can I run, let's say on my MacBook Pro here, to use in reply to handle the offline processing? Is it just stuff from meta or if you guys open the aperture also, you know, models from Moonshot and so forth? So we're using Apple's MLX framework, which does

open the ability to have a few different models and to support a few different ones. The best one we found so far is the Lama, the 7B model. And so that's the one that we're supporting. It works best on obviously the later stage. So M3s, M4s, but also M1s, M2s it works on. It's just going to be slower. Obviously, lots of people are worried about AI coming for their jobs. And, you know, it's not entirely unreasonable, but that's not the only story.

For some people, hey, AI is going to make them a lot better at their jobs. these individuals are the ones who are going to thrive in the next wave. And they 10x their productivity. Enter Miro's innovation workspace. From a company that's been helping teams collaborate and brainstorm for over a decade comes an AI-powered workspace. That is not just going to take your ideas to the next level. It's going to help you progress from unstructured data and a bunch of random notes to product briefs and prototypes in just minutes. It's a solution that will keep us all organized end. It isn't just moving these things around. It's turning them into polished charts, slick diagrams, well-designed slides.

It's a massive time saver. Quickly start developing your next project without spending hours perfecting the ideal question or prompt. No, your whiteboard is Miro's prompt. So help your teams get great done with Mero. That's Mero.com to find out how. 7B is the smaller of the Lama 4 model family. I forget how much bigger the other one is, but how much RAM do I need to actually run the 7B?

I don't have the exact numbers, but it's a manageable amount. And if anything, we have a sort of disclaimer at the top when you don't have enough RAM to be able to support local models and then we just let you know, hey, this won't work right now. I feel like we're like a year or two away, though, from having enough RAM in every single new Mac computer

that we're going to be able to run local models without it being a question. It's weird that we're not quite there yet because RAM isn't that expensive. The great thing about local models that we all forget, and I had this tweet that I don't, no, I just, I was expressing this frustration because obviously local models are, you know, not

at the place right now where they're widely adopted and even widely discussed or talked about. There's not really just, I guess, potential people are not as excited as I think they should be about local models. And the reason why I'm excited is because people forget they are fast. They work without internet. And above all, they're free. Like, you can, if you have a strong enough hardware machine, you can spam as many models as many requests as you'd like. So you can have 300 requests in a second if your computer is gig a job. So you can run this locally. You have a good privacy setup. That seems to be good. Working towards doctor compliance makes a lot of sense for me. You also have a feature inside of reply that lets you kind of search through

all your messages, which I thought was quite smart because search on iOS is kind of garbage. And then you also talked about mapping relationships, kind of using people's text messages as a way to sort out who's in their life and when. Tell me about that and why that's an important feature. It's really interesting because the app started with just this one sequel query, which was, hey, here are the people that you have to respond to. And you should respond to these people. And here's the draft I've created for you.

And then we realized, oh, wait, we're sitting on this treasure trove of information. And there's so much context here about just understanding the relationships you have, you know, the people that motivate you, the people that serve different purposes, the people that are work related, that are personal related, that are maybe even service related, like my cleaner and, you know, different people. And what's interesting is if you take all of this hardcore statistic and you try to weave a story of the tapestry of your life over it. So, you know, in 2020, you were really close to Sam. And then in 2024, that relationship plot. toad. Then in 2025, you got super close to Lola. That's cool. And so the last page of our onboarding, we have this really, I would say, hard-hitting and then sort of just world-bending delineation of who you are through your messages. So it's a closeness graph of the top 10 contacts that you have, all time, your messages, and then a graph of how up you're going and sort of how down,

you know, certain months or certain periods of time. So you see, during COVID, you got your best friends, you see your family, maybe your sisters, you had some valleys and peaks with. And I think the coolest part about having all this, as well as the ability to run LLM requests, is that you can annotate your life. And so, obviously, you know, 2021, I moved to Stanford and it says, oh, this was the big move to California. Your friends became, you know, closer in this way. And you guys started talking about this. And then this was when you had this huge internship. How does it know that I moved or you moved to California? Does it know that just because you started talking about Stanford quite a lot? Or is there any geographic information attached to the

messages that you can pick up on? There is no geographic information, unfortunately, but it's all just predicated on the relationships that you have and what you're texting those people. So. And that's enough to put to put me into a geographic place. Interesting. I mean, a lot of it is, hey, you know, we'll go to the ice cream shop or we'll do this thing that has no geographic context. But when there is geographic context, we have in the prompts to make sure we pick up on that, as those are big life changes. And we look for other things like graduation events and job changes and moves and things like this. Wouldn't this graph you're describing also provide like a look at how long your personal romantic relationships lasted and when they feel out? I don't know if I want this

in a way. Like I've been on my message for a long time. And, you know, across different states, different stages of my life, education, careers, relationships before I got married. I got, I don't how far back I want to go. Can I like time bound this? Like just give me like the last six months. I don't want to go. I don't want to go too far back here. You can skip the slide in onboarding that will tell you this thing that we call Ebb, which is people that are just not as close in your life that used to be close. And yeah, a lot of people have somewhat somber experience going through and seeing the people that they sort of lost, either symbolically or like literally as as, you know, partners. And you could just skip ahead and not like at this slide, not take it too seriously. But it is a lot. It's,

is, yeah, sort of a part of the onboarding. This is when I think applying intelligence to our personal lives gets unsettling in a way. I don't mean to bring up like black mirror too much. But like, there's certain things that I just don't mind that I've faded, you know, that I've kind of let go of. But I think as we have increasing digital lives and better search tools and better surfacing tools, we're going to look back more. And I think that's going to be a pretty interesting cultural motif.

Okay. Clearly, the reply is not aimed at people who want to go back and see when they stop texting their third ago partner. You're charging $30 a month for it for the paid plan. So I presume this is currently aimed at your busy executive, your parents who has seven children that have to keep track of, people that are just high volume by nature. What's interesting is, you know,

like a lot of people who are using it are simply people like me and you who are in group chats all the time where they're getting introed to someone. And imagine being the person in the group chat who doesn't respond, right? So it's, I'm so excited to introduce Molly to Sam. And Sam responds, hey, Molly, I'm so excited,

and heard amazing things about you. And then me, just because I have this bog in my brain, haven't responded. And then two weeks later, I'm like, oh, my God, this was crazy. And so we're solving that problem, which is just making sure that you're on top of your messages, even from a scheduling standpoint and just from a managing relationship standpoint. And, yeah, there's a lot of customer service and there's a lot of sales enablement. There's recruiting. There's consultants.

There's freelance people. So it's people who live in their inboxes and are texting hundreds of people a day. 30 bucks a month is an interesting price point. It's a bit higher than most things that I see. Not in a bad sense, but I'm kind of curious if there is a lot of processing costs in the background that are leading to this costing as much as it does. Are you essentially covering cogs there? Or are you just charging that much because people are willing to pay for it?

So ha-zah, great margins. It's probably both. It does require quite a bit of compute, especially so there's a toggle of how long you want to go back. If you go back to the past month or the past six months, there may be hundreds, maybe even thousands of people of threads that you've left on red. And then we're sending all of these requests in parallel. And so that's thousands of requests, you know, in a second, which gets quite expensive. But it's also, I think, just a comparison benchmark, Superhuman, who does a great job of this on email. We're sort of, you know, mimicking an inbox your concept from, and they charge the exact same.

I mean, Superhuman's done quite well. just sold itself into the Grammarly Confab that now has Coda, Grammarly, and Superhuman. We've talked to, Shere, the CEO, we're going to put the episode number right here, so you can go find that interview if you want. It was a lot of fun. One more thing before we talk about the future in February. You told TechRunch you had about 1,000 paid users. We're now a couple quarters later.

How's the company doing? Yeah, it's been good. So we had a sort of smaller launch in February where we announced, hey, we're here. We're doing this really exciting thing, sort of 9 to 5 Mac, and a lot of Apple enthusiasts picked up. on it. And then, yeah, we launched another thing three weeks ago, which was the iOS app and then announcing that we're supporting WhatsApp and aim to support a few more. And it's just been explosive since then. I think people really like the idea of having everything all in one place where you can see and you can take longitudinal context, right? So I responded to Maria about the Wi-Fi password.

And I should take that to the conversation I have with Masha when she asked me about the Wi-Fi password. And so there's just a lot of really interesting overlap dynamic-wise that is going on here. And yeah, we always, I'm a big fan of these launch videos. So that was a ton of fun and just got a bunch of interest. How was growth been on the either free or paid user site? What's nice is that every time we have one of these big launches, we have, like my audience is like quite power user heavy. So we have a few thousand people who go and spike. I think that day we had maybe just short of 10,000 people go and at least download it, try it, check it out. In terms of conversions, I don't have the exact number.

You can't make your product better without listening to your customers, but how are you supposed to actually figure out what your customers want? You can send out a survey, of course, but people just want to give the right answer in that case. They're not really giving you their honest opinion. Well, at this week in startups, we actually found the perfect answer. It's called Perspective AI. Their expertly trained AI conducts one-on-one interviews with your users based on your prompts and questions. Just tell their system whatever you want feedback on in simple language.

And they take care of the rest. We've learned so much about our audience since teaming up with Perspective AI. Reginald, for example, is a founder who listens and wants more quick-hit, faster-paced content. And an anonymous guy from Canada gave us tons of helpful feedback about some audio-syncing issues we've been having and we fixed them. This is the kind of feedback that's invaluable when you're a founder who is product-obsessed like you should be.

Sign up today and get started in just a few minutes at getperspective. AI slash twist and you'll get your first two months free. That's get perspective.a.ai slash twist. Thousands of pan users now. Yeah, yeah, yeah. So what's sort of the most spectacular thing about this product is not actually the growth number because it's, we're not really looking to be this super consumerist, massive adopted thing. I think it's for the loyalist power user that is using it every day. What are their attention numbers? And those numbers are phenomenal because every single person who struggles with this problem comes to this and they're like, this is my huge grail. Like, this saves my life. Where have you been? And so that's been something that we take a lot

of pride in. Again, when I was prepping for this, I was like, oh, this is designed just for me. And it's rare that our product feels so directly tuned into exactly what I need, frankly. All right. Now, while all this is cool and replies a lovely product, I'm glad it's doing well. You guys have some longer term plans. You actually put out a white paper of sorts entitled The Invisible OS. And you said that the Trojan Horse is building. the iMessage assistant that texts like you. I feel like you guys have built that and you've taken it to market and you've shown that you can monetize it very effectively. You also say that you believe in the world where A, it doesn't replace you. It helps you show up. The next wave of a software

be proactive yet invisible, seamlessly layering into the tools you already know and love, becoming the unconscious default. Tell me more about the future we're working our way towards because it sounds pretty fluid in a really nice way, but I'm curious what that really kind of means. I think starting with text messages is so interesting because there is this uniquely deep, rich, unfiltered, sort of stream of information about a user that you wouldn't get anywhere else that determine what their intentions are, you know, just the manner in which they're responding to people, the frequency in which they're responding. It's really good, at least even a base level, like, here's what my product does. It's almost one of these mini apps, right? Like,

one of these rappers almost like a Cal AI would be or a, I don't know, there's a bunch of these like random AI apps that sort of print because this is a universal problem and very easily understood. And so that has always been the goal. Like we have to start somewhere where people download the app. They think it's a phenomenal experience. They find real utility in it. And it doesn't even matter if it's an AI company that's building it or their neighbor. Like it's providing use to them. But I think I've always been really interested in how we can add proactive personal intelligence into large language models and how we can create this experience that actually 10x is our lives that we live on and we depend on every single day.

We started the company really looking at the voice assistant space and building things on iOS and thinking, oh, well, if we aggregate all these different streams of context, right, your emails, your calendar events, your random visits, your random photos, the places that you go, all of these intricate little details into one place, then we can create a legible stream of insight. Okay, I should help Molly find an Uber to the next destination she has, or I should help her create some, you know, dossier doc for the next conversation she's going to have and whatever it is. And through that process, we realize that I think building on iOS is a lot more associated with entertainment. And building on Mac is where real utility or value where we're sort of people live and create things. And so I've always been more interested in building the utility thing.

And I think once I had reply out in the hands of people, I don't know if you've used it or seen videos of it, but we do quite interesting things behind the hood just inside of the iMessage app. Right. So we layer a little logo in the text field, always accessible in every single conversation so that if you click on it, it starts magically streaming and typing in a response where you can just send. If you don't like it, you delete it,

you press it again, it streams another one. And then similarly, if there's a calendar invite, if there's a mention of, hey, me and you are planning something, it's going to be at this restaurant at noon, there's actually a pop-up and it says, hey, you want me to add it to your calendar? And you can send the ICS file to both parties right then and there by just clicking on that once.

And so this idea of just in time help, of implicit help of, I don't even have to trigger anything, but it just watching my screen at the same time that I'm doing things and I'm learning and I'm going through my routine and my workflows. It should understand how do I format the calendar event? How do I like to send it? All of these things are implicit decisions. And so that's where this idea of invisible OS came from.

So I feel there's three ways to go about building an AI first OS. You can take, let's say, Windows 11 and just jammed AI into it until it bursts. You could delete Windows and start from the ground up or Mac OS, pick your OS, and build something that was AI native, or you could, as a different company, build the invisible intelligence layer on top of macOS and Windows and so forth. And I feel like you're going for option three.

I've been thinking a lot about this because it seems very silly to me that my personal chat GPT instance, GPD 5.1 now, knows me, but only exists in this little box and can't touch anything else. It's so siloed. It's almost lobotomized in a way.

Yeah. But I feel like you're describing a much more personal, integrated intelligence layer that I take with me that knows me and probably plugs into a whole bunch of things. So you know how we have model context protocol from Anthropic? Yeah, of course. I feel like you're building like the personal context protocol. PCP is not a great acronym. So if you have all this context, are you going to allow other companies to kind of tap into

what you've collected about individual people or do you see yourself always being the company that uses that information in a product context? The thing about building an AI company is that you could kind of make a decision about whether you want to be the infra company, or do you want to be the model layer, the data company, the training company, or you want to be the application layer company. At the end of the day, all of them converge and you kind of do everything, right, like cursor and maybe cognition winser for like meat in the middle. Making their own models and now they're powering their own stuff and cutting their cogs and so forth. Exactly, yeah. But I think with the background that we have and what we've already launched,

it makes sense to start with the application. It also is something that you can be incredibly tasteful and it pinated about. And that's something that has always been, you know, top of mind, like, how can we create the implicit assumption? And I talk a lot about this, actually, in the memo, that us as humans, a lot of times are not asking for choices, right? AI gives us, hey, which response is the best? Or, hey, do you like this one? Do you want a negative one? Do you want one in the middle? It's like, no, there's one right answer here. And whether I accept or reject the suggestion should just qualify and give you the right momentum to tell you, you know, whether the further suggestion should be accepted or not.

And so that's where I've been thinking about. One last thing before I let you go. We're going to allow AI to increasingly intermediate human relationships. If it's just nudging us to respond to something, if it's helping us draft a message, people use AI to write cards and letters and all sorts of things. And we're really kind of allowing AI to sit in between us as humans. Do you have any qualms about that? I'm not asking because I'm trying to set you up here.

I'm legitimately not sure how I feel about it. I'm just curious how you think about AI taking kind of a third wheel in human relationships. It's really interesting. I think with emails it already has. So I am an extreme anti-email person. If it isn't clear from me just building an I message app, like I hate to be in email. I don't email anyone. And it's because of these formalities. It's like, this is so unnecessary. I'm texting best regards. And I have to say, you know, hey, dear this person or hey, how are you, hope all as well. right, these things that you're just wasting your breath on. And I like the idea of having a extremely curt, straight to the point, this is what I want to say, and just help me get there

one step closer. And that's what reply is supposed to be, is just come up with exactly the line that you think I would say in the manner and the voice that I would say it in. And I can edit it. I can say where it's wrong. But really, I'm just there to press approve or send, interpret the message. And that's, and that's pretty much all. In my sci-fi optimistic mind, I think it's a really great thing because now, you know, this Dunbar's number, which is this metric, like you can only have 150 friends and your, or 150 people that you know and maintain a relationship with. I actually do think, since building reply, I've been having double, triple, you know, just maintaining the same bar, but with more and more people, which has been phenomenal because now all of the lagger, the logistical, the, you know, brain fog is intermediated by the AI. And I get to do all the interesting work, which is, okay, What are the conversations in which I need to focus, right? What are the ones in which I actually have to give back to this person in a meaningful way? And then just, you know, write those myself.

I'm excited about your future. And as a sci-fi optimist myself, huzzah, Godspeed. And when you hit, I don't know, 10,000 paid users or something, come back on the show and tell me about it. I'm really curious about what you guys are going to build next. For folks who want to learn more, it's hey knox.com. And reply is spelled R-P-L-Y. And Molly is also over on Twitter. Molly, a treat.

Thank you very much. Thank you for having me. Probably the best part of being a founder is watching a project that started as just an idea blossom into a full-fledged startup. The worst part, it's bookkeeping, it's boring, it's detail-oriented, and it takes time away from building your product and, hey, growing your team. It's time for you to let Pilot take care of all this bookkeeping so you can get back to running your company. Pilots the largest accounting farm, and it was built just for startups. They understand the technology industry and the landscape, they know the stakes, and they're already trusted by amazing companies

we love, like Open AI, Scale, Airtable, and Beyond. And you know what, Pilot is about so much more than just getting your taxes done on time. Pilot has a smart dashboard that puts all the essential metrics your startup is tracking at your fingertip. And they have helpful advisors who can walk you through crucial financial decisions you're going to need to make in the coming years. Plus, when it's time to raise your next round, Pilot's COO and CFO services. are there to help. So it's time to focus on your company and let Pilot handle the books and Twist listeners get $1,200 off their first year. Just go to pilot.com slash twist to get started. Alloy Automation, a company that I've known about since October of 2019 when they launched

on Product Hunt. I had the pleasure of talking to the founders a couple of times as they raised a seed round in an even larger $20 million round. But now down the road, the company has evolved and I want to learn more about what they are doing. I love these founders. I've loved their space. And now we're going to catch up and figure out what they've been working on for the last couple of years.

So please join me, welcoming to the show. It's Greg Mojica, CEO and co-founder. Greg, how you doing? Great to be here. Thanks for having me. I feel like I'm just going to keep getting new jobs every couple of years in the podcast game and I'll just keep bringing you on the show

wherever I am. How does that sound? Sounds great. That seems like the theme here. I love it. So I wasn't kidding. Back in 2019, you guys launched on Product Hunt, and the pitch is pretty simple,

complex automation made easy, with no coat. And as I talked to you around that time, you guys were talking about the e-commerce market and how that was really pulling you in. And so I thought about you guys as a place where I could go to connect applications, get workflows done, but really it's about selling stuff on the internet. Now, clearly, I'm dramatically out of date. So why don't you explain to me how MCP comes into this and now we're talking about agents. So catch me up. So we start our life, like you said, in end of 2019, kind of beginning of 2020, crazy time with COVID, of course. focused really on the commerce space.

It was a very fragmented space at the time, right? So there were a whole bunch different platforms out there, Shopify, Adobe Magento, WooCommerce, and many, many others are very fragmented ecosystem. And our initial vision was how do we connect all those different platforms together, right? Orchestration, helping them to integrate those systems together. This is really the pre-agentic era, right?

So this is the pre-AI era where AI wasn't really a thing, and so APIs were all the rage. As we've expanded as a business, you know, Commerce has ultimately become a little more centralized. And so we've expanded to not just commerce, but also accounting, ERP integrations, which is also a very fragmented space, payroll integrations, and many others. And what's top of mine for a lot of folks these days is, well, how, of course, do we inject AI into this? Right.

So our business has actually evolved in many ways from being just an integration platform to be more of an orchestration layer for AI. And so some of the things we've done recently is we've expanded to have MCP, add an MCP layer on top of all of our integration. So if you are looking to integrate with a whole bunch of applications like NetSuite or QuickBooks, we have an MCP layer. We're allowing folks to essentially take our technology and build agents quicker.

I think I now understand the progression pretty clearly. So you had a couple of products. You had alloy embedded in alloy flow. One was your kind of white label of integration layer. Flow is helping people run workflows using those integrations. And then today, if we think about Anthropics MCP structure, it's essentially another layer on top top of that that allows AI agents to tap into information that you guys are already bringing together

in your integration product. So really what you've done is you've made like a series of hooks on top of your previous work so that way it can itself fit into the agenda. Completely. It's exactly correct. Is this something that you guys went out to do because you saw, look, this is where the market's going or were you just kind of plowing ahead and then your customers were just like telling you, hey guys, we need this, we need this, we need this. A little bit of both. You know, we certainly saw a poll from the market. Folks were saying, hey, we want to to have integrations to things like GPT or like Anthropic Claw and whatnot. And so we added those connectors naturally in our product because the customers were asking for

that. And then we also kind of realized that in many ways, agents do require a certain degree of what we call determinism, right? So there's a lot of talk these days about completely autonomous agents in the space. But there's not that many truly autonomous agents that are actually being deployed in production. And the reasons for that are privacy, security, compliance, it's hard to just say to an agent, here you go, go do something, and we're not really going to have much oversight. And so we realize that blending the kind of the determinism of our products that we've

historically built our workflow-based products with agents and kind of adding AI and infusing it into a workflow creates what we call kind of semi-determinism, right? It essentially creates an agentic workflow that can think, that can reason, but still works within kind of the constructs of a deterministic flow. and it behaves relatively consistently across the board. So you're actually able to essentially get the best of both worlds. This is why you guys wrote with your, quote, AI connectivity platform, agents handle the busy work and humans step in only when needed.

So how do you tell when you want to step off of the probabilistic AI side of things and go back into the more deterministic, perhaps even human-led part of work? Is there a trigger that comes into play? Is it just a complexity point? It's basically a human-lil-loop connector, right? how our AI workflows work is, you know, the, someone will define the flow, they'll build it out, and it'll have various different agentic tasks in it. And then if the agent doesn't have, if the LLM doesn't have, let's say, confidence, right,

to address something, maybe the confidence score is less than 80% or 75% or so. It'll actually know then to escalate that to a human. It'll send an email or a WhatsApp message or a text message and say, hey, you know, can you please approve this? When you think about the percentage of confidence, I'm a human and I'm terrible at that. I think I'm either 100% or 0%, but I'm usually more like 46. How doesn't an agent know when it doesn't have enough confidence to proceed and should call in something? Because 75% is a great special, but if it was a little synthetic.

Well, we're always evolving, right? So we're always learning more here. We have some built-in e-vails that allow you to kind of test in the product and to help the LLM to kind of generate that confidence. So that's certainly one way. And ultimately, it is essentially a judgment called the LLM is making, right? So we are in a certain, to a certain extent, relying on. on the LM to have that judgment. You can still also use just good old-fashioned conditionals, right? So if you want to be very much more deterministic and say, okay, based on the output that

the LLM provides, if it literally just is black and white does not meet the certain criteria you're looking for, you can build a conditional flow into the workflow that just says, okay, if this, then that, and that's not really agentic, so to speak, but that is a kind of an off-ramp to be a little more deterministic in an agentic flow. I'm glad we're here because I was really here is about just how durable, non-fragile and useful today AI agents are. Because if you go back to last year, Sierra was talking about, you know, this was going to be the year of agents and so forth. And I think they've made progress, but I don't see that many companies putting them out there and using them in a way that seems to be entirely free from human intervention. It sounds like, Greg,

that that's going to be the case for at least another year or so. It feels like we need a step function change and AI model quality to remove the need for humans in the load. Absolutely. I think there's a number of reasons why. In general, I think what we're seeing is that the technology is still so new, and enterprises, of course, take a little bit of time to adapt to this new technology, so they're being a little bit more careful in adopting it. And just in general, the models are also evolving so fast and such a rapid pace that because of all this change, I think that is driving that little caution in the enterprise saying, hey, we need to take a breather. and make sure we're doing this in a safe way.

But even with that note of caution, it still seems like people really want to move towards a more agentic future. And so I presume that even though the technology may not be as mature as we would like, you guys are probably still seeing quite a lot of demand for this side of your work because folks really want to start at least getting their feet wet in agents. Yeah, absolutely. Well, I mean, there's kind of, there's two, you know, two, I guess, broad groupings of agents, right? there's the more semi-deterministic agents that are, you know, requiring human oversight and

advice. And then there's completely autonomous agents. And I think what a lot of folks are starting with is that, as the former, they're starting with these more semi-deterministic workflows, these AI workflows, that they are agents, but they may require a little bit more oversight. But I think the gradual shift over the next, you know, probably, let's say, 12 months will be primarily towards the, you know, the truly agentic, a completely autonomous future. I'm trying to figure I have 12 months is a lot of time or not very much time. Because on one hand, it's an entire year, Greg, and AI, but also like 12 months is, you know, that's four quarters.

It's four earnings calls. I mean, it's not that bad. Are you content with the pace at which we're seeing the underlying AI models improve? To me, speaking just kind of crassily, it does feel like right now we're in a bit of a lull between major releases and we're not seeing this in kind of like gains and intelligence. So do we need to see an acceleration there to hit your 12 month window? Yeah, I mean, I think that we absolutely do need to see an acceleration. I completely agree with you.

There's been a lull, right? I think that there's been a lot of talk about how GPT5 was not as impressive as everyone thought it would be. I've come around to that. I'm kind of not on that page. No, absolutely. It's more like in many ways I still see myself actually going back sometimes to the previous models and actually using them because they're frankly better sometimes. So I think we are in a low period.

I'm pretty confident that technology is going to continue to evolve pretty rapidly. I mean, there's such investment, obviously, in this space. So I would be surprised if it didn't significantly increase. over the next 12 months. But that being said, even I think what we're going to see, even if we don't fully get to autonomous agents, there's so much efficiency gains that you can get from these semi-deterministic agents, right?

These agents that are not 100% autonomous, but they're, you know, 50% or 60% autonomous. That's still 50% time saving that you didn't have before. Yeah, and then it's 50, 55, 65, 62, 70, 75. It goes up from that. Okay. So one thing you guys talked about recently that I really like was the idea of building intelligence systems

as simple as prompting a, a workflow. And to me, that kind of feels like the North Star of where you guys are going, kind of the combination of what you were doing before, and then on top of that adding agents. So how long until companies that don't have internal tech teams that don't have the same level of resources to bring to bear that your larger customers, your Amazon's have, can really take full advantage of what Al-A is offering? And I'm asking us, I'm curious about the state of the market. Yeah, well, I think there's still a barrier to entering the market, right? Like, I think the people who are still developing agents or who are primarily deploying agents right

now, it's other folks who are, you know, smaller startups who are on the cutting edge and they're, you know, they're, they're, they have teams internally where they're building agents. And then you obviously conversely, you also see that in the, in the, you know, the big hypers, like the Amazon's and the Googles and whatnot of the world. But I think that the long-term vision, of course, for agents, the dream of agents is that it becomes mainstream, right? That, you know, the mom-pop store on Main Street is also able to, to leverage, you know, AI agents.

And so I think that's coming pretty soon, frankly. The barrier to entry still, we believe, obviously, we're biased. we believe is by producing a really elegant experience where you can build that, you have all the tools, you have all the technologies in a visual kind of builder. So as opposed to requiring someone to write code, you can do it yourself. It's almost like you'd want to have a company that had an initial foundation in no code automation as the place to build your future AA agents for mom and pop. So this brings us to customers because we were joking before we started recording about your current logo list. And it includes a lot of very impressive large names. And you're not going to put Bob's

pizza store on there. But when do you think that happens? I mean, when do the restaurants near my house that are run by folks who, you know, might have an iPad at most can take advantage of this? Is that five years? I think it's less than five years personally. I don't know. I'm not sure that it's one year, but I think it's less than five years by all means. I think what's going to happen, like just like the internet happened, of course, is, you know, it starts, of course, with people in tech and big companies that have the investment budget and the R&D budget to invest in this. But it's going to expand pretty rapidly. And I think what we're seeing is we're already seeing kind of pilots of, you know, of companies that are doing this and who are smaller.

They're not just enterprises. So there's certainly a lot of interest. It just will continue to trickle down. And there probably will be this wave crashing moment where there's just this massive, massive splurge of people who are saying, like all these mom and pop surf on saying we have to do this. For me, that was when my mom asked me if she'd help me put something on eBay for her. I'm like, oh, okay, this is now become a real thing if my mom's going to ask me about it. And the same thing happened with, like, social media, too, right? I mean, like if you look at like Facebook story, for example, the same thing happened, right?

So like, I think it's inevitable. And if you draw comparisons to the past, like it didn't take that long. I mean, it certainly took time, but it wasn't like it took a decade or so. It was much faster. Talk to me about the business itself, because the last time I think we spoke was two or three years ago. You guys had raised money or two kind of quick rounds. Things were looking pretty good. How has business performance been since people last spoke?

Yeah. So we've been going more off market, right, over the past few years, right? I think with the integration space that we were kind of previously in, we realized that the primary folks who were benefiting from that were these larger companies. So today we have companies like Amazon that uses us, Best Buy, Zero, and many others that use our platform. And I think, you know, AI is interesting because now AI presents an opportunity to go not just a market, but like almost go even more horizontal, right?

Like we're still obviously focused on the middle market and then ultimately the enterprise. But again, to your point before, like, it is going to go more mainstream. there's going to be just this endless, I think, opportunity with everyone just raising to LAI agents. So business has been good. It's been really exciting. I'm trying to chip away. I'm confused why you haven't done like everyone else has done and go out and raise a $100 million round to, you know, reinvent the world of agents. Like, I don't think you guys have raised money in years. And, you know, just giving your focus, it feels like you're in a position where if you wanted to, you could go out and raise a pretty large amount of money. So what's your thoughts about

raising capital in this market cycle when feast or famine? But you're probably in the the feast side of the divide. Well, I didn't say we're not going to raise capital, right? It's certainly top of mind to a certain extent. But yeah, I think that we are, you know, we're aggressively just growing this AI platform and I think we'll look to a fundraise in the not so distant future, frankly. Are you counting down to a specific like ARR milestone before you do that? I think a mix of ARA milestones and also just, you know, certain feature we want to see.

But I mean, we're in very healthy position as a business. You know, we've been very capital efficient. So we've not needed to raise capital, which I think is obviously exciting because, you know, we don't want more dilution than we don't have to take on. That's actually been a theme of the company since you raised your last round was just not spending too much of it. So I did not presume you weren't raising because you couldn't. I just presumed it was more of a choice. What are you going to invest in when you do raise more? What are the things that you can't do right now that you would like to unlock?

Is it acquisitions or? I think honestly, an acquisition might be interesting, but frankly, just scaling up our forward-deployed engineering, team, right? So I think with AI, what we're seeing is we're seeing a lot of just, you know, like these implementations are very heavy, right? So it's not like you can just press a button and turn on a template and, you know, you're off to the races super quickly. Like there's a lot of a lot of configuration required. And so our implementations today are very much like we have to, you know, where we're working really kind of in tandem, almost an extension of one of these companies, basically. And I think we want to grow that team specifically. So it's just more

investment probably in product and frankly in the forward deploy engineering, kind of go to market motion, like how we get deeper penetration. There's probably a seesaw effect here because on one hand, forward deployed engineers is one of the most popular in growing jobs in technology because I think a lot of companies are realizing, like alloy, you're going to want to have people to help make everything kind of work. But if we go back to the mainstream conversation, you can't require a small company to need forward to deploy engineers because if that's

the is requirement, it's price, they're priced out. So invested a lot in forward to plate engineers now, go after the, enterprise, but over time, as technology gets more advanced, you can probably have a lower weight sales motion. Completely. I think we're already kind of seeing hints of this with like what Replit is doing, for example, right? Like, you know, the Replit agent, like, they've reduced the barrier to entry. So like it's primarily builders now that are able to build website. But like, and formerly that was engineers. I think that's going to go even more and more down market, right?

Like, they're going to have anyone who can build a website pretty soon. And like, I think the same technology applies with, exactly with here, right? Like, I think that there will be a world where you can just say, hey, here's my process, or maybe you have a JD of some sort. You say, you know, I need a bookkeeper agent. Here's my JD, and here's a bunch of the tasks I want to go do. You just put that into alloy, and then all of a sudden, you just get an agent with very, very little configuration. I don't think we're quite there just yet, but we're certainly marching in that direction.

One thing we have heard, though, is that everyone does want to hire these forward-deployed engineers. So I'm curious how hard is it to hire the right talent in the market today? Because it feels like on one hand, you have meta spending entire European football clubs with a capital on individual nerds. And then on the other hand, I read all the computer science, you know, subredits and forums. And everyone's like, I can't get a job for one dollar. So I'm just curious from your startup CEO chair. How tough is it out there to hire? Well, it's not easy to hire. You know, I mean, there's, I think that there's exactly what you described. There's these two kind of very wise spectrums, right? You have a lot of folks who, who are, you know, just being

gabbled up by these big labs and who have infinite capital. And then simultaneously, you have, you have folks who maybe don't have as much work experience and therefore are not the right candidates to hire. And we're also trying to be picky because we started out really during COVID as a business. And the first year of the business, we were remote, right? So the whole business was just, you know, was not in person, was not an in person culture. We're trying to change that. So we're like being very, very intentional of like hiring more in person. Now, we're a hybrid company. So like we have people, of course, who are remote and we will continue to hire remotely. But we are trying to be more intentional while hiring in person. So that actually makes it even

harder, right? Because you've got to find people within a, you know, certain geographical location. What's the maximum commute you think is reasonable for someone to take on? Because I've lived in San Francisco. I'm super familiar with the traffic there. And let me tell you, dear God. Man, probably 30 minutes. But that means that you can't even live north of Market Street on bad days. Like, if it rains in San Francisco, at 30 minute, it commute six blocks.

It depends. We're close to Kaltrain. Our office is close to Kailetra. So you're down in Selma. Yeah. So if you're, if you're in South Bay, you can hop on Kale Trane pretty quickly. Kaltrain, if you're not familiar with Kauklai-Port system, is the American Bullet Train. Faster than Tokyo's trains, faster than, oh,

Wait, no, it's not. No, it's slow as hell. We're still waiting for that upgrade, you know, one of these days. Can't some one of the billionaires in Silicon Valley just buy a new train? Like, why is that? Why can't we just get faster questions? Like, come on, people. I've been asking myself that question for years, but hey, you know.

I want to throw one more at you, Greg. Last time we spoke, you were the CTO of the company. Now you're the CEO. I have seen CTO take on the CEO mantle before. Not unheard of. Not super common either. So for folks out there who are either a CTO or a CEO today, Tips, tricks, things you can kind of share about swapping the roles, handing off certain reins, picking up other ones.

I'm just curious about how much you rocked your world to make that pretty dramatic shift in position of the company. One of the unique things about at least our business was that because the product historically has been very technical, I was doing a lot of implementations myself, right? So obviously, we talked about forward to plane engineers. That was you at the start. Yeah, of course. I mean, that's got to be somebody, right? And so I was doing a lot of that work myself. And it was really exciting was I was also working directly with, you know, many of these customers, implement.

implementing it as the FDA. And I got to see that kind of sales motion firsthand. So I think it was unique in that I had that experience and was able to quickly transition that over to, obviously, the CEO role, which of course, oversees, you know, more of the go-to-market org historically. But I'm still, you know, I'm still, I have my hands dirty in the product by all means. So I think that's been a been a important thing. And just in general, hiring a leadership team is been really great. We've been focused very heavily on growing our leadership team over the past few months. So we have a CRO now. We have a head of marketing. We have VP engineering. So we've been really building up the leadership team and, you know, kind of bringing in obviously domain experts

in all the categories needed. And that's been just super helpful as a CEO now. You guys have been really busy. So I'm really curious how big the business will become. But Greg, a treat as always, when you do hit the next major milestone, whatever it is, come back and tell us about it. We appreciate it, man. Keep the Bay Area cool. And we'll talk to you soon. Thanks, Alex. I absolutely love my job because I get to talk to people when they're absolutely starting off. when they've raised their first money and even when it begins to hit scale. It's an absolute treat. And speaking of founders, we've done that with.

It's time to turn to our interview with Shopify's Toby Ludke. And I want to bring Lon Harris up. Yeah, I think it's so funny to go back into these archive clips and see those are, those are always some of the most fun moments where they're talking about like, yeah, well, we just hired 40 people. It brings our total to almost 200. It's like, wow, like to think about Shopify with, well, you know, like a hundred-some people trying to make this thing happen in the early days.

It's like it's such a huge shift from how we think about them today. You can tell in this interview, we're going to play a couple of short clips from it here in a second how on the ball Toby is, how serious he is, how deep into the problem he is, and how non kind of buzzworthy the whole chat is. It's so focused on helping people sell stuff on the internet, make companies. I don't know, it just seems very serious compared to the current launch video cycle that we're in. Yeah, for sure. And it's also he's got such a, this is the thing Jason is telling founders like all the time

and founder you is like, you really have to have this incredibly deep, nuanced understanding of like who your customers are and what the market is and who your competitors are. And you could really see that Toby was super drilled into that. Like Jason throws a lot of other platforms, many of which are still around. Adam, like Etsy, what about Etsy? Are they your competitor? What about Kickstarter? Are they your competitor? What about Amazon marketplace? Are they your competitor? And Toby has like every one of those. He's like, well, they do this and we do this. And they're aiming for these kind of people. Like he's just got it all. very clearly laid out, like, what the whole landscape is like in his head. And I think that's probably

what made him such a, you know, dangerous competitor in a lot of ways that he was able to sort of, he saw so much of what was going to happen over the next decade plus of e-commerce coming. But every company starts with an idea and a start. So here is the Shopify origin story. I'm Toby Luckie. In 2004, I co-founded a business. And initially, we were trying to do an online store ourselves we were setting snowboards online. So you were selling snowboards? Yeah, right. So we were actually, we were using Yahoo stores back then. So yeah, Yahoo had a store product. I forgot. That's right. That's right. Programmed, Speer web. We sold that to Yahoo. Yahoo, we wrote it a couple of times and then became Yahoo stores. My background is I'm a programmer. I

apprenticed as a programmer, which is something you can do when you're from Germany. And I wanted to get sort of out of that a bit. I wanted to start a retail business. business. This is something else I was really interested in. We worked with manufacturers, got some snowboards, wanted to sell them, we wanted to use something off the shelf, and realized there was no freaking way we could make, build the business we wanted based on off the shelf software that was available back in 2004. So over the course of this first season of sales, we replaced it with software we built ourselves. Which 2004 was right the time when Ruby and Rails came out. So that was fun to play around with and I got involved

very much in the Ruby and Rails community. We had a lot of fun building the technology, so like reinvigorated my kind of interest in programming back then. People were saying, you know, it's really, really cool that you guys have this, you know, Snow Devil, the Snowboard store, but would you license the software?

And then we sort of realized, well, maybe helping other people go through this and building their stores might actually be a better business than selling snowboards. And this is sort of how we pivoted. In 2006, we launched Shopify. by taking lessons from base camp. This was again, even 2006, software the service wasn't a term yet.

Even just putting a price on the website was innovative in 2004, even though that sounds crazy. That screen where you have like, here's your three options and this is the best one for you. That was a true innovation by 37 signals. Absolutely. And it's hard to remember that at this point. But yeah, now this was all really, really impressive. And it was so clear that that was the way software should be sold on the internet. And then we wanted to do this for online stores.

So I love that they were selling snowboards, which is such a specific thing. You know, like of all the world of products that Shopify could have started with, snowboards would not have been on my bingo card. And that they were using Yahoo stores, which is, I mean, you could see even in the clip, Jason, at the time, is like, oh, right, Yahoo has a stores. That's right. Like, I totally forgot that Yahoo used to have a like buy and sell stuff, marketplace, facing kind of Yahoo stores.

So that I thought was really funny. And the other thing that really stood out to me is that when they launched Shopify, SaaS was not even like a term, like was not even a concept yet. We think of that now, now if you think of software as a service, it's like such a, that's what companies are and that's how you build a big startup is. You know, you sell software over to go. Like to think of that as being like a new fresh on the scene concept. Like I wouldn't have even thought that was during my lifetime, let alone.

you know, like just 20 years ago now where that was like, and they're talking about, you know, the origin, like 37 singles kind of invented this whole thing with Basecamp. And that was the first piece of software. People were like subscribing to and re-upping on. And it's like, wow, I think of that as like really distant history, but it's not. And the sounds point brings us to the clip that I want to bring up first lawn about pricing. One thing that we saw them discuss is how do you charge for this?

Do you charge too little? And Jason runs through the exact sort of mental calculation he still does on the show today, which is, all right, so if you're going to replace and build this yourself, what would it cost to construct this? Can you charge that amount? And it's funny to see how Toby's kind of not thinking along the same lines of replacing the cost with their pricing. It's very interesting. Take a listen to this. The one thing I've heard from people is that you charge too little. How do you respond to that? Like people say like you're charging 10 bucks or 20 bucks a month, you know, $250 a year to do all your e-commerce in a dedicated site is absurd. It's insane. And people would pay a lot.

more if you asked your clients if this software didn't exist how much it would cost you it would they would say five or ten thousand dollars to hire somebody to set up a competing service what do you answer to that I usually just tell them be a better pricing and that's a that's a truth of it it's maybe I regularly talk with people who are converting from like not just five thousand dollars people are you often paying fifty thousand dollars a month for their systems that's a that's a fairly regular price in our industry because you're going for third parties who are building the systems and they are hosting

for you and you being charged and all these kind of things. So converting to that on a $170 plan and their service much better with that. The nice thing about our businesses, we really want to just make it a lot simpler for people to start these kind of businesses because it's hard. You know, like when we started, there was so much to learn. There was, you know, just how you're dealing with manufacturers, supply chain. How do you get the word out? There was a lot of complexity around technology.

too like we had to to get approved for credit card gateway back then we had to post like a ten thousand dollars down payment or bond with the banks to do this kind of stuff they needed to be you needed to engage for many many thousands of dollars a year a company doing PCI scanning for your site which meant that the site had to be up but then you couldn't get access to the to the gateway credentials before you had the tests so it was a huge tick-neck problem doing development for us and it was just hard so there's a lot of these kind of problems we wanted to make go go away so our business is focused on hey let's let's make

it so that everyone who has products to sell like people with interesting stuff but there's just nothing in their way to get to get this out and get them in the hands of people that's sort of what we care about and then for us there's opportunities to monetize of course to ask people to pay money for this but there's also opportunities for us to work with the payment gateways and the shipping companies because economies of scale matter in this business. And if you can go to FedEx and say, hey, we have 60,000 people who want to probably ship with you. That means we can get better rates for everyone. And that means

there's potentially some opportunity for us to make money to. I still think we're trying to figure out how to price products today. I mean, we mentioned now SaaS is kind of a dated term. Sure. And now we talk about usage-based pricing. But at the time, we are two generations in pricing from where we are today. So it's really interesting to see how far back we're still trying to sort out. What should software cost and are we undercharging? Questions we're still asking today? twist even in the present day, the idea of like, isn't your product absurdly cheap? Like, Jason asks Toby that directly. Like, isn't $10 to $20 a month for everything that you're doing for people absurdly low? Wouldn't people pay a lot more? And this is a topic that still comes up

to this day. It's like, well, you do want to, you know, that that is always this sort of like really fundamental tension is like you do want to delight your customers. You want people to be so excited to use your product and making it extra cheap as a way to do that, but then, like, are you sacrificing where you need to go and the ambition and the scope of your project by making things super, super cheap instead of, you know, charging more so that you can get further? Let's move on to talk about Ottawa, talk about building in a secondary market. I think right now people are very focused on San Francisco.

Again, it is once again become the hub of, you might say, the cutting edge of technology. A lot of AI house parties, I'm told, are in San Francisco. And yet, Shopify built a simply world straddling company somewhere far away. What we think is one of the things that Shopify became really, really good at, is making what we call secondary markets work. Like secondary markets is, you know, every city which isn't, everything other than Silicon Valley, maybe New York City. So how a great, big, meaningful, impact for companies being made.

It's like there's a geographical region that somehow realizes there's this one company that all the best people go to and spend a couple of really exciting years of their career together. And then they disperse again to do many other things. And we would like to be this company. So to us, the most important thing is you want people to think of us when they plan their careers. And if they are signing up for lots of interesting work, but are willing to work really hard with, you know, really caring and with a lot of passion. And hopefully you come to us. And it turns out there's a lot of these kind of people around in this city.

It is a topic that we're constantly talking about. It's like, you know, like that's maybe the number one question that we get from founders when they submit. make questions for twist is like, I have this great idea for company, do I need to move to San Francisco to start my company? And, you know, like, it's always kind of a nuanced answer and it shifts and stuff over time. But I feel like Toby really plugs into this very sort of like romance, what I would consider like almost like a romantic idea of startups where it's like, it's like that PayPal mafia idea. Like you're going to bring a group of really brilliant people together under one roof. They're all going to make this great product together. And then they're going to

splinter off and become like, you know, like the all stars. Like they're going to all shoot off and start their own great companies. And that's what makes this sort of city a startup hub. And I'm like, I love that. I don't know how practical that is or how many companies have been able to like set out on that course and actually achieve that. But he and I can tell that he and I share a lot of ideas about like what the ultimate kind of collaboration or the ultimate kind of workplaces. Because that's what I feel like I'm always trying to find too. It's like, where is there a group of just really excited, like cool people digging in and working on something interesting? And then they're all going to go off and do their own things after that.

Well, you know, just for fun, people do talk about the Shopify Mafia. I mean, Business Insider in 2023 were a story entitled, Meet 38 members of the Shopify Mafia. So it really does seem to have worked out. But that's how you build entrepreneurial hubs. There has just been enough people, you know, building, exiting, recycling capital in the Bay Area for so long that it is what we call Silicon Valley. But I think you can pull it off in Ottawa. I think you can pull it off in New York City. I think Austin's got a good shot.

I think we've seen a lot of attempts, you know, like, and sometimes it takes. And sometimes, I mean, I'm sitting here in Austin. So obviously, there was a sort of focus, like, let's see if we can go move our companies to Texas and get going there. And it worked in Miami we've seen, too. But I also have recalled, like, I've had a few of these conversations about places that where it did not take. Like, I started working in tech companies in Silicon Beach in Los Angeles, if you'll recall. And still today, not the most amazing startup hub in Santa Monica, if you went there today. And then I also recall Vegas.

That was going to be a big thing. It was like downtown Vegas is right for the picking. We're going to put all of our tech companies there. Tony She's vision. And I came of age, if you will, in Chicago technology scene back in the era of Groupon. And you know what? Chicago is still about where it was in terms of the ranking of the global startup charts. Did they have a silicon nickname for Chicago?

Silicon freaking cold as hell too much of the year. dot com is how I described Chicago. Chicago is the best city in America except for the weather. Like that literally just ruins it. I did. There was one more thought from that discussion that I thought was interesting.

He talks about how even great the top people, the A-list people, the people that you desperately want to come with you on your startup journey and help you build your company. Like it's not, it's not so easy to find those people

as just scanning a resume and looking for like work experience. You know, like he notes that a lot of people have been fired or have been like poor performance at one job and they just needed to sort of figure that out. He specifically discusses it while working with a seaman. So let's take a look at that clip. If you poll 10 people, high performing people about their careers, three, four, five of them

will tell you, oh, I got fired once. Like people just, people's careers are never these like meteoric raises that people think about. They're never perfect. They are like, I think Cheryl in a book called it sort of a jungle gym. Like that's much better metaphor for what a real career looks like. And you know, sometimes this is the stuff that has to happen. I worked for Steam Man's as an apprentice and I got one of those letters from them saying,

your performance isn't up to par. And I was like terribly shocked by that because I thought about myself as a really good programmer. But then I realized, you know, like I probably haven't slept in the entire week because I'm playing video games at night or programming. Probably my performance wasn't very good. I made some changes and these kind of things might end up turning out to be really, really, really, beneficial things. And I think that's worth reminding people off. And I thought this is such a great point to make that we think of, you know, your career path as being this like straight line,

like, you know, up into the right, like you start at this job and then you move up and then you move up and then one day you're Paul Graham or whatever, or your one day you're Toby Ludke. And I think that, you know, it's more, especially in the early days, it's a little bit more chaotic than that. And everybody tries jobs that don't work for them or career paths that work. the right path and then corrects. And so when you're evaluating somebody where you're looking at maybe hiring someone, you can't necessarily be that rigid.

You've got to be able to like, does this feel like the right person and does their experience line up with what we're doing? But it's more nuanced than a lot of people think. Yeah, you can't really be reduced or boiled down to just your LinkedIn page because that's not going to capture anything like the full context of yourself, your career, and what you have learned.

All right, Lon, the next segment in this interview we're going to talk about is the question about expanding one's global footprint versus adding features. This is the sort of trade-off that founders have to deal with and they don't have a current-day Shopify's employee base. Yeah, I mean, this is, again, it's just fascinating to go way back into the archives and find these amazing companies from years and years ago. And they were struggling with the exact same problems. We hear from founders week in, week out. Even people working on their, you know, MVPs are already sweating this.

Like, well, how do I, do I make it more appealing to more people? Or do I make it like extremely robust and like there's tons of useful stuff in here and it takes a while to sort of figure it out. And I just always think it's interesting that, you know, like the same problems persist, even at such different levels of scale. How do you decide between going global and adding like killer fun features? We are a very small company given the size of a problem. We had 200 something people, I think 16 at the latest count. You do that all with just 200 people?

I mean... Last month we added like 40, so it's like... You added 40 people in a month? Yeah. Wow. That's more than that's two per business day. I give a monthly newbie session. I had to move it into the park because it was nice weather.

There's a lot of shop-if-if-I get-togethers and I just recently went to one in New York City. They had a hundred people there. People are exchanging ideas about how to, you know, take their products to the next level and finding new market channels. But it's powerful. You go around there and you talk to people and say, where were you when you got your first order? And everyone would be able to answer this. because it's a seminar, it's a live event. You have an app.

Yeah. And so you get a push notification. I saw somebody who had it and they showed me their Shopify notifications and they were watching sales come in all day long and they were absolutely raving about the product. People are addicted to that, the metrics, huh? Yes. Yeah, I mean, it's metric business, right? The thing that sucks about online businesses is that you can't watch the person walking into your front door

and looking at the various product and taking a path through your store. And then leaving without buying. And then leaving without buying. There's so much information in this that, like, Like, online stores have, like, or just online business have trouble surfacing and creating an equivalent for that. You know, it's just, those are our challenges. If you can figure out what that equivalent is for online, then we would do a massive job of educating. Eye tracking, where their mouse is, whatever, just what images they swiped on.

Right. And then try to do that while being in 100 countries. Not easy. Try to deal with the privacy concerns of various different, like, content and so on. So one question that I had is, what's a lot of? like to be the CEO of a hyper-growth company because there's not that many people actually out there long who have done this because a really, really fast-growing company is rare. So by definition, there aren't that many people that have experienced it. And so I loved how Toby

described his biggest challenge, viz the market. The hardest part of the job? Is it the regulation or hiring? No, it's not. It's the hardest part of a job is how to, it's that I don't think like there's no books for how to build a company that, fundamentally needs to be able to, while following a very ambitious roadmap, at the same time, the core competency of our business needs to be how to thrive in chaos and how to react quicker than anyone else. And we are competing on that ground with companies which are very good at this kind of thing, like Amazon, right? So, because... That is the big competitor, though. Well, yeah, really. So if I am, if I've made 10,000 iPad cases,

what is should why should I put them on Shopify instead of just selling them all to Amazon and seeing what happens well Amazon people have to make that decision right yeah no it's not really a decision so usually the way it works out it's like people people need their online shop which is where they capture all the margin value and then they have various sales channels Amazon marketplace is going to be one of them so you you from from your Shopify store you federate out of you know eBay Amazon all these other channels If you are in retail stores, you integrate Shopify into that part of the supply chain. So like we're not really competing on that.

But if you scale up the ambition of the business, like what Shopify is doing fundamentally is trying to make commerce better, right, and trying to disintermediate further, trying to connect the... Because what, you know, Amazon has a monopoly on all the price that have barcodes, and what we would like is to have a monopoly on all the products that are actually interesting. So... Because the people who actually care about these... the products and make them themselves come to us with them.

Got it. So we would like to disemediate further and get them directly in front of people, and that's what we really try to accomplish in the long run. What's funny, Lon, is now there's a bunch of startups out there who are trying to out Shopify, Shopify, and now view Toby Luckie, like I'm sure Toby used to view Bezos. I think it was interesting too that Jason, you know, tries to give him like, well, what about Amazon?

They're your chief competitor. And the way he thought about it, I think, was pretty productive, where it wasn't like, I'm going to try to steal people away from the Amazon. Amazon Marketplace, which like, obviously today maybe they could have that idea. But back then was like a David versus Goliath scenario. So he was already even thinking about it as like, well, how can I like work side by side with them so we don't have to beat them? Like, well, maybe you can sell things through Shopify on Amazon Marketplace so you could

use both services. And I think that's like, you know, the sort of clever approach. Like I don't, you don't have to kill Goliath. you can figure out how to like hang out with Goliath for a few years until you grow and then maybe you can take on Goliath. Yeah, you go hang out with Goliath, learn his workout schedule, his macros, his peptide routines, and then once you're also enormous, you go kick him in the face. Lon, one last segment here about competition and the broad array of it.

He was mentioning names like Etsy and Kickstarter as well. Who's a bigger competitor, Etsy or Amazon for you guys? I don't think of... A Kickstarter. I don't think of... I know. They all facets of the same... market having slightly different approaches. I mean if you if you really will

for like for shopping cards specifically our competitors like you know big commerce maybe which does something similar magenta or maybe but again we have like I think fundamentally Shopify is trying to climb a bigger mountain than than the others and yes like there are similarities in this but this is I mean like Shopify's market is making websites that make more money when they cost what's the you know, what's the market size for that, right? Yeah.

So this is a big place. And there needs to be lots of different approaches. Toby had a great quote in here. Fundamentally, Shopify is trying to climb a bigger mountain than the others. And then he reduces it. You know, we do one simple sentences all the time, like describe your product as quickly and effectively as you can. And he gives a great one. Shopify's business is making websites that make more money than they cost.

So when you think about it that way, it like really reframes the whole business. Like, oh, that's not what Amazon is doing. Amazon is like, we're going to get you that toothbrush cheaper and faster than if you went to CVS. That's like the core promise. This almost make it sounds more like Squarespace or something. Like it's about making this website and making this online business and we're going to give you all the tools to like make that happen. So I thought that was really interesting.

Also very focused on individual businesses, because if you're thinking only about the cost of the website to build and run, you're not thinking about a large staff. So he's really saying, how can we enable people to build small companies? I think that's really great because everyone else wants to serve the enterprise. I love seeing someone go after the small guy. And we'll just close with this. Shopify today, Lon, worth $190 billion. So it all worked out. A nice happy ending.

All right, everybody, that has been Twist for this lovely day. We'll see you all soon. Lon, you're a peach. We'll see everyone next week. Bye-bye.