Sterling Douglass (05:31.49)Yeah, only one I'm not super familiar with is the Brigger King AI assistant, which I'm like quickly reading about now. And that's all I, I think the most important thing whenever building like new AI stuff is what you name it. And I think Patty is just a 10 out of 10 name, you know, for Brigger King's AI project. So I like that part a lot. I don't know anything about it, but that part I like.
Joshua Sharkey (05:55.278)Here's what I think is interesting, more kind of holistically about this is like large enterprise groups. One, a lot of them are actually working on their own models. I think Yum, was it Yum that partnered with Nvidia? Like because they have so much data, there's some pretty interesting things that you can do there. The reason why this AI Patty, the AI assistant got a lot of coverage was it's kind of like a big brother.
Sterling Douglass (06:08.184)Yeah.
Sterling Douglass (06:20.822)Yeah, I'm friendliness scores.
Joshua Sharkey (06:24.135)Listening to everything you say and listening to the way that you answer. It's obviously picking up a of insights, but then it's also anything that you say is being recorded.
Sterling Douglass (06:34.286)Good way to get rich data. But yeah, this one's interesting. Yeah, I don't know a lot about that one, but all the other topics.
Joshua Sharkey (06:44.27)I just texted you so you have the list now. Does anybody know more about this Project Catalyst from Shake Shack?
Michael Jacober (06:53.326)Don't. This is the first time. Yeah.
Sterling Douglass (06:56.91)I mean, it's just kind of, I don't know, I don't know a V3 on their technology stack. they're just, they're improving tons of things around their whole technology stack. AI is definitely a big part of it, but you know, they also are, you know, redoing their point of sale. They're going to be rolling out Q. And they're, you know, trying to do a lot of internal loyalty and, you know, expanding the AI stuff. So.
It's just more of like a tech stack redo. That's AI forward. That was my, my read of it. I'm sure if you got them on, they'd make it sound way cooler.
Joshua Sharkey (07:32.698)Yeah, I mean, it's interesting that Q took this long to get rolled out in Shake Shack, given the EHI investment. Well, while you're looking at that list, Sterling, can you like, how's it going with podcasts, How many episodes you got live? What are you guys talking about? I listened to the first one, but I haven't heard since.
Sterling Douglass (07:50.498)Yeah, just the third one. It's really the only good one that we have, you know, the one about open claw. I would definitely perform the best. No, yeah, Aaron Newton, and I is the co-founder, chief data officer at thanks. Yeah. Aaron and I have been chatting, you know, informally for like a year and a half since like the somewhat earlier days of, AI and just trading notes and Hey, I built this. Hey, you built this doing a lot of show and tell we'd ask each other some questions. And each time we had a call.
We'd always walk away from the call with like, shit, I'm behind. Like I need to do more. Like I'm way behind Aaron. And Aaron would be like, I'm way behind, sir. Like I need to do more. And so we kind of were like, like basically like pushing because they were both fairly competitive. and w it kind of forced us to actually build some really cool things. And then what we realized is that half the stuff we were talking about was stuff that, you restaurants could do. He started doing these boot camps. and the boot camps were really popular. He's got a lot of C level executives at some of the top 500 brands.
And so for us, it was really just about like, Hey, what if we turn those conversations that we have every month or so, you know, into a podcast, we can get broader distribution and we can teach some more people and we kind of help push the space forward. Right. Like restaurants always have this, you know, what I think is an unfair, you know, perspective that they're slow to adopt technology. And I think that ever since COVID, I just don't think that's true anymore. And I think right now, especially in the AI front, they really want to, they just have no clue how. Um, and I've seen some do really well and some a little bit.
behind. And so this is a lot for them. It's also a lot for like the tech industry. Like we talk a lot about how we're trying to get our companies to be more AI first. What does AI first mean? What are some different tactics that we do? So it ends up covering a lot of those topics. We're on episode five, just dropped this morning. And we're going to do at least 10. And after 10, we'll kind of see where things stand and maybe we'll do some more or nobody listened to it and nobody wanted to it. And we won't do it anymore. You know, so far it's been a lot of fun.
Joshua Sharkey (09:47.224)What are the few of the good nuggets of things that restaurants can be doing with AI that you guys are helping them with that they might not be thinking about?
Sterling Douglass (09:58.294)man, we probably drop like two or three every episode. think that one of my favorite ones is there's a, thanks basically put together an MCP so that you could segment your customer data, you know, via Claude or chat to be here or whatever. And this group, it's like a 30, 40 location, Pokey shop. They plugged into that MCP and then they also had a weather MCP.
Michael Jacober (10:19.758)Oh yeah, I think I heard you talk about this. This is Pokeworks,
Sterling Douglass (10:22.286)I think so. Yeah. And, uh, and literally it was just the main operators, one guy basically just built a little script and all it does is basically when it's going to rain, you know, which he knows through the, through the weather MCP automatically like send a free miso soup, you know, to this customer segment. And he's basically just great. This little automation that's literally weather triggered. And if you think about like McDonald's bought a company called dynamic yield for like $400 million.
you years ago to basically do that. Like when it's hot outside, you know, adjust the digital, drive-through board to show ice cream. Right. Like they literally do. now like that has been so democratized and brought down that, a single person who's never written a line of code and I'll build that, you know, for specifically his restaurant, just the way that he wants it. I think that's probably one of my, favorite examples.
Michael Jacober (11:17.358)Do you know, was it Casper who, who made that? if it's Casper shoe, he's, he's like, he is slightly, he's technical. think he actually has a comp side degree. those, those poke works guys are, are like, they're slicker than, yeah.
Sterling Douglass (11:34.71)The name I remember, I'd have to go check back on the pod.
Joshua Sharkey (11:39.214)This is what I've been trying to sort of instill in my team and just, you now I'm meeting with just friends like CEO friends of mine and just helping them get started because I think we all should do that if someone hasn't. But that's a perfect example. Like when someone's like, I don't I'm not technical. I don't even know what MCP is. Yeah. Ask Claude. You know, like ask him.
Sterling Douglass (11:57.422)That's literally how we sign off every episode. like, if you didn't understand anything that we talked about and you want to know more, just ask Claude. every episode, that's how it ends.
Joshua Sharkey (12:08.398)It's funny because I remember I used to get upset with people when I needed something in Google Sheets or Excel and they're like, I don't know how to do those formulas. And I'm like, neither did I. I went to Google and this is before GPT and now it's even easier with GPT and now it's, you know, it does it for you.
Sterling Douglass (12:22.414)Stackoverflow was the original, you know, AI that you'd ask out on like fixed formulas and stuff.
Joshua Sharkey (12:28.106)And it's the same premise, you know, it's just like the way you learn these things, you just keep asking. And that's what I think is so beautiful about this is every time someone's like, yeah, but how do I do this? And I'm like, I could answer that for you. But I would, what I'd like you to do is think thoroughly about like what you need and then go ask, you know, go ask. I mean, I'm a, I've kind of went all in on Claude and my whole team is now on Claude and we kind of have a bit of a competition on who's, you know,
Sterling Douglass (12:52.834)Who's burning the most tokens? Anyone on a team token flexing? Anyone doing a billion a week?
Joshua Sharkey (12:58.038)My engineering team is all has been all on cursor. And so I'm trying to actually move them over to cloud only because of the, so we're all in one platform. You know, for me, like Carpathy did this note about a of weeks ago about basically implementing a Wiki in for like structure into your, into how you store all your data. I've always been like really bullish on that regardless, but then, you know, my, my developers were like, you know, you should, you should think about a vector database. So I started building this vector database and then.
And then it turns out that like MD files are like gold for, especially for Claude and most LLMs. And I know that sounds like very technical to people, like what that means is you can build, you can build, you know, automated lint and page, you know, recompilation, all these things you can do without even needing to know what that is by the way, or knowing how to do it. And you can build.
Michael Jacober (13:38.542)natural language.
Joshua Sharkey (13:53.452)You know, like really powerful databases of all of your knowledge that you can then use to do all kinds of things easily, you know, like, and by the way, Claude will build it for you and we'll organize it for you. And like, it's funny, the same way I think about recipes and I wanted to put all my recipes in these. like, I want to put everything into this thing. so like 15 years of Apple notes, every possible meeting transcript, everything I can think of is going in this indexed and like, you know, with, this Wiki structure so that like I can, you know,
use it, know, brand, obviously brand assets and logos and all that jazz and you know, things that used to take me board decks that used to take me, know,
Sterling Douglass (14:32.11)Can give you my opinion on decks? Just in general board decks, but all decks. They're dead. totally dead. Decks are dead. They're dead. They've been murdered. I'm never building one again. I don't think anybody should build one again. If you want to present something, just have Claude or your AI build HTML.
Joshua Sharkey (14:37.302)I mean, we might have the same one.
Joshua Sharkey (14:54.286)That's what I do. It's 100 % HTML.
Sterling Douglass (14:57.006)It knows how to do it. It's much better at it. We're trying to fit a square peg in a round hole. We're trying to ask for faster horses by having AI like figure out how to do decks and, and do I just have a build HTML. It's more natural. It's more flexible and knows it better. It's just as easy to share. You can make it editable. It is all of the things that you need and way better and way faster. Our upcoming board meeting, my last board meeting I did AI generated slides, partly just to show.
the other board and it was a bit of a talking point, but also because it was way more efficient timing wise. this next one, that we're doing in a couple of weeks, like there will be no slides. will just be HTML.
Joshua Sharkey (15:35.714)Yes. Yeah, it's exact exact same. Some things people should know though is like you do need like I use Cloudfairy. You kind of need like an S3 bucket or something to host these things because otherwise they're local files. You know, you need to end if you want to gate them and you got to build a gate. But they're all easy things. Yes, cloud to do it for you. But 100 percent agree. Like these, you know, these decks that, you know, HTML, HTML is a pretty dumb app, but it can I mean, by default, you know, there's not a lot of back end, isn't that? But you can build, you know.
Sterling Douglass (16:04.63)Sure. Build a react app for all of your presentation. So you can, you can give it a little slide structure. You could template ties it. You can do anything you want in it. But I just think that like Google slides and PowerPoint, like what there's just no reason for them anymore. I think it'll take forever for that actually to be adopted across the masses. Like I think those applications will be used for a long time. There's a long tail. but I think that the it's shrinking now, not growing.
And I just think that shift over time is going to be there and I'm done making decks.
Michael Jacober (16:37.998)Yeah, but ultimately it depends on the source of capital that you're trying to apply to, right? Like if you are trying to raise, I don't know, 50, $100 million from an old school, you know, family office who's like, I need to print this thing out. need to be able to read this.
Sterling Douglass (16:59.31)I have a solution. I have a solution. Print to PDF. That's what half the people do with their slides anyway. That HTML, the PDF. Now they can print it out. Now they have something they can put on their tablet and use a stylus and draw notes on and annotate. Which by the way, I didn't, I did all the HTML stuff and I did not know the PDF. I was interviewing someone and they're like, oh, that's cool. And then you could just turn it into a PDF and make it easy to share. And I was like, shit, that's a good idea.
And then like immediately after the interview went on, I was like, dang, that's so good.
Joshua Sharkey (17:30.19)Everything I build in cloud that I present is HTML, but by default the skill now automatically creates a PDF. If you just save it to PDF, some of the page pagination gets messed up, but you can also create CTAs, a button that just says print PDF and it'll structure it for you. There's so much more you can do with these things when you dial them in. I have N8n connected to basically everything that gets me in HTML runs through a...
publish and like a published folder and then gets to you know this s3 bucket but then now like I got rid of my task app because I just want things I can go right to on my phone or my my computer yeah I have I have I haven't
Michael Jacober (18:09.838)You're no longer on coda. What do mean?
Joshua Sharkey (18:12.14)No, I have an HTML file that all the tasks from anything, any meeting, any email, any slack, any text, all gets sort of parsed and then based on my MITs and goals, it like prioritizes them. It's on a task board that I can go see if I want to and I can interact with it. It has, you know, you can press done, move things because it's got like it's connected to N8N to API calls. but I don't even touch it because it's it does most of the tasks for me. It's like it takes most of the
Sterling Douglass (18:39.392)How do you deal with the reviewing of it? Do you trust it just to go full autonomous all the way through and have it send to people?
Joshua Sharkey (18:45.046)Always review. For certain things and certain people, I will actually have a draft an email that I go and send because if you by default with the MCPs, if you send a slack via Claude, it will say sent by Claude. I don't love that.
Michael Jacober (19:05.678)Where draft things? Where do your drafts end up? Does it actually end up in your email client?
Joshua Sharkey (19:11.534)So I kind of want to show you guys because I have this daily brief and the daily brief it takes from every text message, email, Slack, meeting, world news, a bunch of news that I find important and a bunch of other things. It synthesizes it and then it only, you know, provides me what's, you know, what's important. And then it says, here's the things you do need to respond to. And then in the HTML file, there's a, in the UI,
It has, okay, here's the thing. Here's the email to send. Here's the calendar invite I'm going to do. Here's the rework of this doc I'm going to do. It's all like, let's say that there's eight action items and it says approve, edit or kill. If I press edit, I can say change this, do this. And then it has, it basically has all those tasks now teed up.
Michael Jacober (19:56.748)Are you just pressed on when you and when you edit, are you having like a context conversation with your agent within that edit flow or do you are you manually editing? Cause that's one thing I I'm working on trying to figure out now is.
Joshua Sharkey (20:10.767)I just explain I don't don't I don't like mainly edit I just say hey this needs to have include this part of it or like you know and then it
Michael Jacober (20:17.279)So you are, you are able to do that within the UI of your got it. Okay. So you built an interface for that specifically. didn't, you're not doing that telegram.
Sterling Douglass (20:27.406)Yeah, you're doing you on whatsapp. You have an open claw, know hack your whatsapp.
Joshua Sharkey (20:33.484)Talk about OpaClaw because, you know...
Sterling Douglass (20:36.098)We talk about it like every, every, every episode.
Joshua Sharkey (20:38.542)I remember bugging you about this for a while and like I've been every time I think yeah I'll use open I mean I'm using open claw for developing some for like code I do all the like strategy and very very specific spec docs and cloud and then I send it to open claw and it does some research for me but otherwise everything is in co-work because It's like for me. I don't know like you can schedule things. It's all connected to what I want more secure well
Michael Jacober (21:04.652)You're also, you're, you're, you're saving a lot of money doing that too. Right now, now that the.
Sterling Douglass (21:10.222)On Saturday on April 4th, they finally said you can't use max subscriptions for, uh, for open claw anymore. what I've been, since, uh, about like two, three weeks ago, uh, the cloud agent SDK came with a couple of major updates. The ones that got the headlines were, you know, scheduler for cron and dispatch to talk to Doug. I didn't really care about that stuff.
Joshua Sharkey (21:17.006)What do you
Sterling Douglass (21:34.478)The one thing that they didn't really announce that I cared about was they did, they launched a session manager and session persistence, which is really what I was using open call for, because I wanted to have kind of these long maintained sessions that are like per task. And so once they released that, I was able to then integrate that into my, you know, cause I have like my kind of personal to do application. And so I was able to kind of connect that I wasn't able to do that previously.
With Claude code, even though I wanted to open clause, basically the best way to do it plus open clause. mean, I had it back in like the first week of January, uh, back when it was before a mult bot, uh, I think it was just called like Claude bot and like, you know, the lobster clause. And, um, I wanted it for the crons and then I wanted it to be integrated into Slack. And then I wanted, you know, kind of their little API so I could build it into my app. I've moved, I've moved almost everything over now, uh, to, uh, basically the Claude agent SDK.
because now it has all of those things. But the reason that I really liked it is that I could customize it. Right. I can create so many different custom rules, custom skills, custom workflows, how it connects to different apps. I can sandbox it, you know, on the Mac mini, it can help me like manage infrastructure and I could do it for my phone. Right. That was the biggest thing. Like to be frank, like the number one reason I loved open claw is so that when I went upstairs in the bathroom or sitting on the toilet.
I could still use my to do things app and be interacting and having clothed sessions instead of hanging out on Twitter. And, that was really kind of like one of the biggest things for me.
Joshua Sharkey (23:04.066)Yeah, yeah. And think that's like the one, like dispatch, I don't really get, it's like a remote control for your computer. It's kind of like, fine. But I think that, I think cloud desktop will very soon be on your phone as well. So I'm like not investing.
Sterling Douglass (23:21.07)Yeah, they recently did that. Like that's part of remote. That was kind of the access they were giving to you is that like cowork and, um, Claude code were kind of available. That's the funny thing with AI, Aaron and I joke about this all the time. You'll build these cool little custom things to like make it do the thing that you want. And then like four weeks later, Claude's like, here's the native version of it. And you're like, well, this is better. So I'll just use this. So you have to be like very okay with just throwing away a ton of the work that you put into it, which I've always been fine with.
I'm never married to kind of like how much time I put into something.
Michael Jacober (23:54.862)Well, just think about the knowledge that you're, you're gaining with each of these, you know, tasks and projects that you're doing, you're learning how this stuff works. when they do drop it, I'm also here, my big concern is, I don't know if you guys have followed the subsidy. They actually announced, I'm not, it was Carpathy or someone announced the subsidies that Anthropic is actually, uh, so the $200 max plan.
is actually costing them around $5,000 per user.
Sterling Douglass (24:27.822)Only if you max it out each week. it's like, uh, when they did the math, there was 25 million output tokens, which would cost you like $8,000 in API costs or something. This is, I'm a couple months old on that data. But so what happens is if you truly maxed out the max subscription, they were, they were losing thousands of dollars a week on you. But the average person was only using like 10, 12 % of their max subscription.
But then once the autonomous agents and open clause started going in, now people could start running inference when they weren't in front of their computer, it just skyrocketed. Which is one of the reasons they've been trying to get people off of open claw is to basically just reduce the amount of people that are basically like, you know, making them lose even more than they thought they would on those max subscriptions.
Joshua Sharkey (25:19.586)That is wild, by the way. I'm like...
Michael Jacober (25:22.282)so then the question becomes how deeply in bed do we want to be with these businesses who, mean, Anthropic is planning on going public this year. OpenAI is planning on filing their S1 and Q4. Like these companies are all of a sudden going to have to become real companies. So then the question is like, where do you, as the consumer want to be? Do you want to be, you know, completely leveraged on these tools, and a hundred percent reliant on these, you know, these frontier models?
Or is it time to start thinking about local? Right? So that's kind of where I'm, I'm excited right now is like, you know, are we, I don't know if you saw this Chinese model that just dropped. think it was, it was either earlier this week or late last week. It actually is performing better than Opus 4.6 in code, in terms of coding. and then there's, you know, a ton of these smaller, these smaller models that you can run on free, like forget about getting a Mac, forget about getting a Mac studio.
like six month wait list, like it's not happening. Yeah.
Sterling Douglass (26:23.95)10 billion parameter models are only about four or five months behind the trillion parameter models right now.
Michael Jacober (26:31.448)So like, you, are you, are you looking for, you know, Nvidia GPUs right now? Are you trying to think about building your own little server? Cause I am, which I'm kind of excited about. I'd love to be able to just like burn tokens constantly and not even think twice about it.
Sterling Douglass (26:47.534)boys be burning. I mean, the look it's been really interesting. I dabble with them. I check in on them kind of every so often as of right now, the best like, you know, there's like, there's lots of different tests that you can do. And then there's more like holistic ones, like, okay, taking a weighted average across like the 10 key categories that you actually use. And for the most part, like as of right now, you know, on April 8th, the local models are only about five, six months behind, you know, the, the best.
you know, cloud models. So Opus four six or the newer Grock models or some of the open AI ones. Uh, you can get kind of those Quinn reasoning focused models. Uh, they're performing about as good as sauna, you know, four or five was, which came out in like December. So like they're only, you know, a few months behind. And then if you even go a level deeper on the models that you can actually run on your phone, I can use even less compute. Like those are about like 12 to 18 months behind.
So if you think about like how technology usually works, like the time ranges, they're exponential, right? They get closer and closer and closer. So that gap isn't going to get wider. It's actually going to get smaller. And I think that you'll probably see a lot more local models being run. I know folks that have switched over to it and you know, they've designed kind of their workflows and stuff where Sonnet four five is enough to be able to do a lot of the autonomous work that they need done. And yeah, I'm right there. You know, I've got a server here.
I've had for a while, you know, I dabble in the local models right now. I'm not super focused on them. I don't really care about sending spending a few hundred bucks a month right now.
Joshua Sharkey (28:22.382)It's just all like, I mean honestly, it's just so, I mean, I don't know about you guys and I know this is funny because my therapist was like, hey, you know that there's a reason why you're doing this and I was like, every night I put my kids down and literally 8 p.m. to like 1 a.m. am just working. Like I haven't seen a TV show or a movie in like five months. And by the neither my wife because she's got this new project she's doing and she's all, I got her all in on this thing. You know, I'm being careful not to, I'm is this an addiction?
I mean, this is all stuff from my work, you know, obviously, so it's like very, very helpful things that would take me a month now take me, you know, a couple minutes, things that would take me six months, take me, you know, a day, you know, the output is pretty incredible. So it's, it is a bit addictive, but it's
Michael Jacober (29:04.594)I have feelings about this and I'm experiencing really some, what I explained to my wife on a regular basis, like I'm drunk on this stuff. And I think what it, to me, like I've tried to understand why and I've been held hostage and I'm sure you guys have and I should maybe held hostage isn't the right term. We've been managing developers and engineers for years and wanting.
constantly wanting more and more and more and pushing, pushing these teams to be producing more because, you know, we, some of us understand are more technical than others, but, know, you know, to say like, listen, that's just one line of code. Like, why can't you just do that? And in some instances, you're wrong. And in some instances, you're right. And some instances, just like someone isn't really that engaged and they're not working very hard. Now we have like,
a team of the most badass expert developers to do whatever we want, whenever we want. And for me, it's like Christmas every day. It's like, man, if I just had this a year ago, if I just had this two years ago, how much more productive I could be, how much more I could build. And now we can just build anything. And it's just this unbelievable euphoria that I'm just jiving on and vibing on on a daily basis. The question is like, what is it gonna, are we gonna crash out? Are we gonna, we're gonna burn out? Like what's.
what's eventually going to happen. I don't know. I have no idea, but that's kind of my two cents on like why we're so like.
Sterling Douglass (30:37.951)it
Joshua Sharkey (30:38.574)It makes total sense. It's also a little bit like it is a little bit scary because of this sort of like rug pulled up from under you thing You know I was on a call a couple days ago with Eli who was on the show a couple times in this guy Evan Armstrong I don't know you know Evan he's got a this sub stack. He used to be a tide mark, and he's got this really cool sub stack about AI and He was saying yeah, you know you're just at some point the cost
of what you're doing is going to be exponentially higher. They could just completely change the cost of you. And I'm like, I would be fucked. Imagine if like all of a sudden this is all like 10x more expensive.
Michael Jacober (31:20.514)Local, you have to go local, baby. You gotta go local at some point.
Sterling Douglass (31:24.374)I would be, I would be surprised if it got exponentially more expensive because usually in the case of technology, like these subsidations, that, that come, like they are helping like the initial build and then the costs come down a ton. Like, I still think that long-term the cloud costs will actually be cheaper. Like it's the same thing when, like when Dropbox and all these ones are trying to do cloud storage in like the late nineties and early two thousands, it was just so expensive.
There's billions of dollars of capital poured in there to try to subsidize it, build the data centers, and they just didn't quite get it right. But then by 2008, 2010, 2012, like we don't even think about it anymore. Like AWS and these S3 buckets, like they are nothing. They cost nothing. They're, they're free. And they have all these tools that then actually make it better than hosting local. I think you'll see the same pattern, you know, on, on the token use inside, for AI.
So, I mean, I'm not super worried about it. don't, even if the cost went up, I don't know, even if it went up 10X, I don't think I'd really changed much because I'm getting way more than a 10X ROI on it. it's like you're getting, yeah, it's I try to make sure like I tell all of our, team and employees, like, don't want to hear anybody talking about tokens. I do not care about tokens. Burn as many tokens as you'd like learn, do it because the lift that we're getting is so much more than the actual token costs that you burn.
Joshua Sharkey (32:33.346)So I-
Sterling Douglass (32:50.894)As long as you're being reasonable and somewhat responsible, don't literally light them on fire. Like you don't have to worry about being super efficient right now. Um, because once you figure it out and once it clicks for people and now all of a sudden they're like locked in and they're like, Oh my God, think about all the things this could do. They start making their own output, like five and 10 and 20 and 50 acts. literally produced 20 times more code than we did a year ago. And with 20 % smaller of a team.
Like, and the quality is better. There's, less outages. There's less bugs. Like it's literally better and faster and everyone's more happy because they're not having to do some of the things they didn't want to do. And they just feel like they're doing so much more.
Joshua Sharkey (33:35.074)I do think that there's got to be a, there's a period where I noticed with some of the team where they actually were behind on things because they were spending so much time trying to figure out a, you know, some let's say sales, trying to figure something out as opposed to like, Hey, there's like four deals that you actually need to follow up on. And I, you know, I think there's this balance between yes, dig in, do your thing, but like there needs to be an ROI on token. Right? So like if you, if
You know, there's one thing to say, blow out as many tokens as you want. You know, but I do think that there's, there's, there can be this sense of like, yes, just keep doing these things. But like, you got to look, you know, probably even at a more frequent cadence than monthly. Like, okay, this week you just, we spent now twice as many tokens. You just spent 30 grand or 20 grand or 10 grand. What's the Delta on like the difference between like actual return? Like are we more productive or are we actually.
Sterling Douglass (34:25.845)I would say
Sterling Douglass (34:29.88)For me, I don't care. I don't care for more productive last week or not. I care more that the team is adopting. I think that I am heavily biased towards adoption. So I don't care that they're going to waste, you know, half a million tokens this week. If it's in the pursuit of them figuring out how it actually works, making mistakes and learning, it's like a free training. It's like a training course. So I don't, I don't care how productive or what the dollar per token output was or what the projects were.
I care much more about adoption.
Joshua Sharkey (35:01.518)Yeah, but there's a ceiling to that. At some point, everybody's adopted.
Sterling Douglass (35:05.314)And then I'm not like, once every single person in the company starts every single task with Claude code, I'm really only worried about adoption because they end everyone who goes on this path, figures it out. I mean, we, have, we have, you know, technical people who haven't written code, but you know, can go, you know, troubleshoot five different point of sale systems. Like I don't care that they wasted a bunch of tokens building some skill and MCP that never worked last week. I care that they just like went through the process and like thought about how it could help.
Because now they're going to be two steps closer to actually figuring out how they can, you know, better automate and get better results and more. So I'm much, we're in the mode right now where I think optimizing for ROI is the wrong approach. I think that right now you need to optimize for adoption and learning because the faster that you enable and upskill your team, the faster you'll be able to reap those ROI benefits for the whole company.
So I am like hyper focused on, I do not care about tokens. I don't want to talk about tokens. I don't want to talk about that. can burn as many as you want. As long as you're in there and you're learning and you're, and you're making progress. Like that's really all I'm focused on.
Joshua Sharkey (36:17.794)This show is brought to you by, you guessed it, Me's. Me's helps thousands of restaurants and food service businesses all over the world build profitable menus and scale their business success.
If you're looking to organize your recipe IP and train your team to put out consistent product every day in less time than ever before, then Miii's is just for you. And you can transform all those old Google Docs and Word Docs and PDFs and spreadsheets and Google Sheets into dynamic actual recipes in Miii's in lightning speed. Plus, stop all that manual work of processing invoices because Miii's will digitize all your purchases automatically.
And there's a built-in database of ingredient yields, prep yields, and unit of measure conversions for every ingredient, which means you're going to get laser accurate food costs in a fraction of the time. Visit www.getmees.com. That's G-E-T-M-E-E-Z dot com to learn more. And check out the show notes moving forward because we're going to be adding promotions and discount codes so that all of you lovely and brilliant Mees Podcast listeners get a sweet deal on Mees.
Sterling Douglass (37:21.294)you
Joshua Sharkey (37:22.19)I don't disagree, but I do think there's a... I don't agree with the premise. would say, especially because you're servicing like, hey, restaurants, you can do more. There are companies where payroll is next week and they have X amount of dollars and they're like, oh yeah, I could spend more tokens or I could pay my crew. And I think in tech, we tend to like say, yes, blow through all this dollars if you have venture, if you're...
Sterling Douglass (37:25.358)You're allowed.
Joshua Sharkey (37:52.187)you know, lot of us aren't like necessarily profitable, we're just growing and that, and that does make sense.
Sterling Douglass (37:57.547)If you're in crisis mode, right, and you've got, you're short on runway, then yeah, absolutely.
Michael Jacober (38:05.134)What is crisis mode mean? What does runway mean? going back to we are very much benefiting from massive subsidies from the dollars that are behind these companies. Sterling, you mentioned that the cost is just inevitably it's going to come down. It just has to, because that's always what's happened. We've never had an infrastructure build like this in the history of the world. This competes with...
any massive infrastructure build, you know, the railroads, like, it is such a critical time. And also depressions have come after massive infrastructure builds. This is like kind of unprecedented with the amount of spend that's going in this bet on AI usage and data centers. I think we as people who are, you know, responsible business owners have to be thinking about
where, you know, at some point the debt needs to get paid. At some point, you know, this is going to come due. It does say a little bit like if the methodology is like burn, burn, burn. I think it's, you know, it's, definitely one way to look at it. I guess the other question is, you know, how can you think like a, like a truly responsible business owner where you are actually needed to think about if I'm investing in this today, what am I getting out of it tomorrow?
Just like these companies who are making this big infrastructure bets are thinking of.
Sterling Douglass (39:34.702)I think that like, we got to make sure that we right size the dollars here. Claude max subscriptions, $200 a month. And it took me two months of using it before I ever hit any session maximum on it. And so when I say burn, burn, burn, like we're not even getting $200 a month. Like where the burn? Like, it's not like I'm sitting here and saying like, we need to be burning hundreds of thousands of dollars a month. I'm saying that.
Yeah. Spend $200 a month. That investment has such a crazy ROI. Totally agree. I care more about usage. You know, even, even Aaron, you know, who's one of the heaviest users I know, like the most he's ever done was a billion tokens, in a week. And it was like a couple of grand. Like that was, that was it. And so like the dollar amounts are not crazy here. I don't think you need, you know, we haven't raised as much venture capital as.
Uh, you know, a lot of these other companies, don't need a hundred million dollars of venture capital to do it. Uh, but I am saying that like, go get a hundred or a $200 a month subscription because, try to max it out. Like you'll struggle until you really know what you're doing. Um, you'll struggle to use it unless you're just like, literally just like, well, write a hundred thousand word essay, you know, on, uh, you know, on the financial crisis of 2008. Like, all right, that's a, you can go.
Joshua Sharkey (40:53.742)I mean, that's a fair point, Sterling. You're not talking about... I was thinking more like everyone running OpenClaw and... That's when my dollar started spiking up a bunch, when I had a whole bunch of deployments going every day on OpenClaw. I was like, wait a second, what just happened? 10x my spend overnight.
Sterling Douglass (41:14.126)Yeah, that'll do it. I mean, I've burned, you know, think I've burned 400,000 tokens since we started recording.
Joshua Sharkey (41:21.262)I mean, good for you. Okay. Alright, well, okay, so you got this podcast. I guess what I would ask is, you independent of, obviously we're all digging into AI here, like, how's it going with the show? Like, how does it feel to start something? I always have people ask me like, like, I want to start a podcast, you know, what should I do? And I'm always just like, start and don't stop. But like, I love to hear how it's going.
Sterling Douglass (41:45.438)second attempt at starting one. I won't name names but a couple years ago I got a pretty
Michael Jacober (41:51.713)I heard
Joshua Sharkey (41:52.046)The recording of the one you tried before.
Sterling Douglass (41:53.774)Yeah, I got a pretty sweet group of, uh, you know, four of us. Um, you know, some, some of the most interesting, in my opinion, some of the most interesting people in restaurant tech. Uh, I mean, we, we put together kind of a pilot episode and, you know, we weren't able to like really get it produced and get the cadence going. Um, so I learned a lot going through that process. So I definitely use some of that. Um, I'm, I'm always big on, you I don't mind failing as long as I learned from it. Uh, and so, yeah, we started, I, it's been fun, honestly, like.
AI makes it easy. Um, you know, it takes me. But 30, they're only 30 minute episodes. And then it takes about 30 minutes of, you know, all the AI stuff to get everything posted and manage, um, the run books like we put together, you know, it takes Aaron and I like five to eight minutes, uh, we like a little brainstorming thing. So it hasn't been a great, it hasn't been a huge time commitment. Um, it's a heavily automated, like the cover art and stuff is obviously AI. Like we leaned into that.
kind of did that on purpose. And so yeah, I think it's been a lot of fun. Reception has been pretty good. It's always fun when you get kind of messages from people that are like, oh, I learned this thing, you know, from, listening to you guys. Like, that's cool. But yeah, I mean, so far so good. Aaron's awesome to work with, super easy. You know, both of us are pretty, pretty low on the ego side. So like it's really easy just to kind of give straight feedback, critical feedback, improve back and forth. So it's been a lot of fun.
Joshua Sharkey (43:17.55)Charlie going, we haven't talked about that in a while. Yeah.
Sterling Douglass (43:20.724)yeah, Chali went through a hell of a transformation, in the last 18 months. So everybody knows Chali for the third party marketplace integration, integrating the Grubbob and Uber orders. That's how we got our start. but since January of last year, so we're about 15 months in, we've really only been selling our new platform product. And the new platform product is the actual like full marketing website, online ordering, mobile app, email marketing, Google ads, Google business profile.
surveys. And then of course, all the stuff we do at the marketplaces. And it's all bundled into one. It's meant to be everything outside of your four walls. that theoretically you could manage a whole restaurant with like Jolly Square and restaurant 365. You've got your out of house, you've got your POS, and then you've got your back house. so, sorry, me's. So we launched that last January. It's all we've been selling for the last
15 months and in April, we'll actually, it'll take over and we'll actually do more revenue through our new platform product than all of our legacy and historical stuff combined. So been really fun. Lots of growth, lots of AI, you know, helps us build the product in the product. The whole go-to-market motions, heavily automated with AI and you know, it's been a lot of fun. It's been busy, but really fun. The whole team.
It's like 50, five zero. we've, we've been trying to get, you know, more efficient right over over the past year or two. So, we've been, you know, keeping the team small and mighty and nice and efficient.
Michael Jacober (44:59.576)What have you guys seen? You guys are both still, you know, venture backed.
Joshua Sharkey (45:05.23)Hmm.
Michael Jacober (45:07.106)businesses. I don't know much about your cap table, Sterling. I know a little bit more about Josh's, but you've obviously been following what's going on in the in the public markets. How has that been kind of how does that hit your boardrooms? How does that, you know, affect how you guys are?
Sterling Douglass (45:23.854)Is that what we're getting at?
Michael Jacober (45:26.67)You know, we're still in the midst of it, you know, the private credit crunch. Yeah, I'd love to hear how that's affecting both your guys' and, you know, how it's making you think about M &A on both the, is there anyone for us to buy versus, you know, is anyone interested in buying us? Yeah, I would love to hear a little about what's going on there.
Sterling Douglass (45:45.624)Yeah, I think on the, on the fundraising side, I think that what AI has done is it's just created kind of two, you know, it's created two categories for SaaS. it's just the haves and the have-nots. And it's not necessarily AI native. Like there's tons of articles written about this, but there's going to be like five companies that win AI native. Like the Anthropics, know, Proboxity, OpenAI, you know,
There's really only going to be a couple of companies that win like true AI native. Everything else that's going to be a good company is just going to be like AI enabled, AI first, AI like, and it's much easier to do that when you're starting your company from fresh, from the start, right? Cause everything you do get centered around it. It's much harder for these legacy SaaS companies. And I'll put myself in that category to try to lead and transition and like become AI enabled. which is, you know, one of the things that I.
I mean, think the last February or last March, I just declared like war. You know, if we do not become an AI enabled company, we'll die. Like we have to, we have to, we have to go as aggressive as we possibly can. Hence some of my stances on burn as many tokens, right? Because we have to transition this company. And so then what ends up happening is you just have two categories. You have the companies that are AI enabled and their metrics show it. So they have these great metrics. And then the reason is because of AI.
And then you have everybody else who's, you know, still, you know, has like SaaS metrics that might've been good two years ago, but aren't good enough anymore. And so I think what that does is it just concentrates capital on the haves. So it used to be, you wanted to be like top quartile in a lot of your SaaS metrics. That's how you'd get funded. Now it's top decile and maybe it's top 5%. And so like, you have to have those, like, you have to be even better. Like the bar is just higher to get capital, but I still believe great companies, you know,
Great companies will get like capital finds great companies. It's just how it's how the markets work. And so I think that as long as you accept that, like the rules have changed, the benchmarks have changed and you start changing your targets accordingly. Like that's how you're able to kind of like, you know, survive this asspocalypse. Um, you know, that's been our theory and that's what we've been going after. And, know, so far, I think we've done a really good job of that.
Joshua Sharkey (48:02.744)transformation of a legacy team into being AI native is definitely, that's a tough one. We had our summit last month and I had a similar like, you know, hey, this is no longer a buzzword. This is a mandate. It's no longer okay to just be dabbling and, know, really like, but also taught everybody more about like what it looks like. But there's clearly folks that are like, well, you know, I wasn't hired to know AI and even on the inside there was, you know, a little bit of that. And I'm curious, like,
How did that transition work for your team, especially on the tech side? And are you just deploying a lot more product now or is that you're just deploying fixes and refactors and things faster?
Sterling Douglass (48:44.206)We're not done. We're definitely, you we have not like finished that transformation. You know, Aaron and I talk, talk about that a lot. And I'd say thanks is actually probably a further along than we are. I think we're, we're going to catch up, but I do think that we still have a ways to go. We're using it a lot. Our customers are benefiting from it a lot. We wouldn't have been able to launch and manage our platform product without it. We wouldn't have been able to grow as fast as.
as we grew with it, you know, without it. So I think that we still have a ways to go on that. And in terms of like the difficulty, mean, I think we hired our last person that wasn't AI. I think a few weeks ago we hired someone who hadn't really spent that much time, you know, in any of these. And I think that was the last person we're going to hire that does not have experience using AI. And I think that we hired people three months ago that we probably wouldn't hire today.
Michael Jacober (49:27.843)Slow.
Sterling Douglass (49:42.542)And that's okay. We're going to take them on it. We're going to train them and we're going to do all those things. Uh, but as a business, like if you want to be AI first, like all of your people have to be AI first. And so, you we've got to bring those people along and we're making good progress on that front. I'm really happy with it and you can see it in a lot of the numbers, uh, but we still got a ways to go.
Joshua Sharkey (49:51.021)Yeah.
Joshua Sharkey (50:00.43)I resonate with that. had a meeting with my product managers last week and one of them is newer. And I had a tough conversation where I said, just to be clear, the job that I hired you for is not the same job anymore. I know that's tough to hear, but here's the expectations now. it does, I understand, I want to be apathetic to folks, but it's just, it's a very different world.
Sterling Douglass (50:26.958)Yeah, but like, even though, like, so like, like creators are probably the ones with the most pressure being put on them right now from, from the AI front. We have a graphic designer, right? She's never written any code. Like she's graphic designer, right? She is now managing like deployments, you know, through cloud code. Right. So we set her up. We, set up the MCPs, we trained her on it. She got into it. She like loved it. And now she can actually do.
parts of the funnel that she could never do before. She always had to depend on someone else for, and sometimes they would make changes or sometimes it'd be difficult to communicate. And now she gets to do it all of herself. So even in that, you know, kind of, in that kind of, in a vacuum, that smaller world, she's actually more empowered than she ever has been in her entire career. And that's good for us too. Right. Because now you have one person and the same person who has kind of like,
The design brain is also getting to see the rest of the funnel. So when she goes to the website and someone doesn't look quite right or she notices something, she can just go through and change it herself. She doesn't have to wait for someone. She doesn't have to do a ticket. She doesn't have to try to explain it to them. Right? Like she gets to do it all on herself. And I think like that is like the best. So like, yeah, your job has changed and it's actually better.
Michael Jacober (51:29.773)Yeah.
Joshua Sharkey (51:42.774)It's way better. Yeah, it's way better. it's also, you know, one of things I was sharing was you used to have, we would have all of these ideas and things we wanted to test and make sure that we built it the right way, it serves the customers right way and that they, you know, that the UX was right, the UI was right. And I was telling my product managers like, you don't need to guess anymore. Go build it, get it in front of customers. By the way, it's likely products that we can just, you know, already, you know, commercialize.
You know, we have a pretty complex app in terms of like the database and all the recipes and all the the calculations that happen. So it is it is a little and the permission access of all the different ways you can view it. So it's a little bit tougher to to like implement into the app. You know, certainly doable. But, you know, the idea that you can if you have a new a new feature that customers are requesting and you want to build. Go build it, literally go build a version of it, get it in front of them, have them use it.
You're going to way better feedback, way faster so that you have a lot more confidence in what you're going to send the engineers to go to go to go build. And that was impossible. That was impossible even, you know, six months ago. To your point, Mike, of like we're kind of like kids in a candy shop. I kind of think about it like if I wasn't if I wasn't a chef, I was like, I want to build a restaurant. I want to build like the next like, you know, Noma or whatever. And I don't know anything about cooking, but somehow I was able to actually produce like incredible dishes.
with my ideas. know, then great. just made, yeah, you know, just made food. I was like, wait, now I can actually like start a restaurant and I can create the kind of food that I really want. And maybe even the service style I want, even though I don't know anything about cooking. And that's that's how I feel is like this sort of.
Sterling Douglass (53:22.702)I imagine like a chef that had, you know, could basically like fully automate and manage exactly how, you know, all of the sous, all of the cooks and all the servers actually did right. The person with the creative, the subject matter expert now has the ability to get closer to what they envision to actually happen in real life. It's one of the things where I think that, you know, AI is going to shift, you know, it's going to shift power from people doing the tasks to the subject matter experts.
Yeah. Right. So if you are an architect on the engineering side and you have a really good mind for how systems are supposed to work together, that's not going anywhere. But now instead of relying on three people that might be twice removed from you to actually go execute it, you get to have a much closer front row seat to making sure it's done the right way. And so I think that the power really goes into the subject matter experts. And so the people who do understand exactly how things work from a ground level, like
They are actually more valuable, not less.
Joshua Sharkey (54:23.854)Yeah, 100 % agree.
Michael Jacober (54:26.51)I a question here because this is I wouldn't say this is concern, but the fact is all these all these things that we're developing right now, they're all they're on the digital world. Right. And. The subject matter experts are right now the ones who are able to be supercharged and to build things really fast, really efficiently. What's your guys thought process on? Because these are digital products and because this technology is getting democratized at a
really, really rapid rate. At what point is there going to the market be flooded with the product that you as the domain expert architected, built, developed, first, were first mover on, where now someone across the world has the same tools can literally just rip out, recreate your product in a day. I mean, these models are going to get even more and more advanced.
you know, what's your thought process on when that, that's not an if that will happen. Who, how do you win there? Like how does, how do you maintain to be the one like that you have the trusted product that like when the, when the price could just get driven and driven, driven down.
Sterling Douglass (55:39.818)I think about that in two ways. think one is the idea of moats have changed dramatically. What was a moat for your business, what like helped protect or insulate your business before is dramatically changed. So things like your partnership ecosystem or your ability to distribute, right? How good your sales are. I think that really complicated, deeply integrated workflows, your ability to contract. These things are all much more important than they used to be. think the second.
thought I have on that is that product market fit used to last three to five years. Right. When I, when I came up in tech, like every kind of five years, you really needed to kind of re-invite your, your product, right? You needed to really do a big expansion and, and there's lots of, you know, saster and, and Jason Lumpkin and a lot of others have talked about that. That I think right now, I think we've gone from three to five years to one to two years. I think roughly every two years, you've got to make a pretty big change in the product.
to be able to kind of keep that product market fit that you have. Just because the competition comes a lot faster. And it's not just competition from brand new companies or someone around the world. It's your competitors, but it's also your partners and anyone. Anyone who's adjacent to what you're doing, like it's much easier for them to jump into that because the walls are lower. And so I think that companies like Toast and DoorDash
Michael Jacober (56:48.27)It's your competitor.
It's all super- yeah, yeah, yeah.
Sterling Douglass (57:05.322)are now much bigger threats to any restaurant tech company in the space. They were always threats. That was the number one question I always got in every investor meeting is, what happens when Toast decides to do this themselves? happens when Aloha decides to do this themselves? And then there's a whole fun story around that. But like now that question is now even more threatening. And I think that that is where you're seeing a lot of the convergence on the technology side.
So I think that what it does is it kind of forces these larger products and it forces much faster product velocity. If you want to stay relevant and you want to keep that product market fit.
Michael Jacober (57:44.024)Well, then a follow-up, how can you defend against your, you know, your partners just cutting you out? Like what are some, what are some things, what are some tactics that, that, like, how are you thinking about this, about these potential threats of toast? Just saying, yeah, now we do everything that, that Chally does and it's, you know, another 20 bucks a month.
Sterling Douglass (58:07.086)So I think there's a couple of things I think that integrations and deep technical workflows, those are harder to replicate. Like even, even for someone like Toast. So the fact that we integrate with 64 different point of sale systems, right? Like that is like a very deep set of integrations and deep partnerships. It's very hard to replicate that. And then you've kind of got in the case of Toast, like they are locked into certain menu structure. So like they have a competing product to our marketplace integration.
They still refer us deals because virtual restaurants, for example, like their menu infrastructure makes it really hard to support those properly. Whereas we kind of say one layer, we were super early in the virtual restaurant game. Virtual restaurants are definitely not as big as they used to be, but like we can support those. And so you have like these kind of like more complicated workflows. So those are ways that you can defend against that. And then honestly, I personally think the biggest note you can have as a business is distribution. If you have a super efficient go to market motion.
I think that's one of the biggest most, if you can create a sales motion where you have like a CAC payback of like three months, or you have like the LTV CAC ratio, like seven, eight, nine, then I think like that is actually a moat because that is really hard to do. You can't just ask Claude build me a super efficient go to market motion, right? That takes iterations. It takes time. It takes a lot of subject matter experts, takes training. There's people involved. There's certain tasks that I cannot do yet. And so.
Having a really strong go to market and having that distribution, I actually think that is the single best mode you can have as a tech company right now.
Joshua Sharkey (59:44.686)I think it's probably always been that way. you know, a great product with poor distribution will always lose to a good product with great distribution. I think that's really always been the case and more so now than ever. I still think, well, two things. One, I believe that like if you're not the system of record or the rails, like you're just, you know, ultimately it all can be replaced because we're just seeing how quickly we can build these things. Integrations are tough. I don't know what the half-life is on like these
every one of these systems not having, you know, APIs and MCPs that are really good. I think that's good. There'll be enough inertia where they just have to, all of us. But I also think that there's, restaurants now have a decision that they never had before, which was build or buy. You know, like, that wasn't an option before, but now restaurants can decide to build instead of buy. you know, the SMB probably won't have this. They're not going to build.
Who knows, they might have their scrap you found or something, like, you know, that option was not on the table before, but now it is. You know, like there's, you know, reporting BI, things like this, or like, I don't know if I'd pay for that.
Sterling Douglass (01:00:53.518)I think that's exactly what this unlocks. think, uh, big restaurant groups have always had the build the buy decision. You know, there've been, you know, Wingstop was really famous for the stickies barbecue. Uh, did this, um, you know, there's a lot of, think enterprise groups that have picked different pieces, whether it was point of sale or online ordering, you know, all the big pizza guys, right. They all had their own online ordering. So I think that the really big restaurant groups have had that decision. You haven't seen as much of it, but it was still happening. Whereas independence, they did not, like I had no frigging choice.
And now they do. I know a single location group that actually built a really damn good website. Not just probably the look and feel, which I don't really care about, but like everything they did on the SEO and the GEO side, like I got really close and they even built like a little deployment pipeline. And then we gave them some tips and helped them out a little bit, but like that was never a thing.
And it's not going to be a hundred percent of independence, but even if it's 5 % of independence, went from zero. Went from zero to 5%, they are so much more enabled. So we have groups that we're working on where they want to build their own little dashboards to go track their Google ads. You know, we actually showed them where they could get a Google ads MCP, and they could be able to ask questions and do a little tinkering. And now we're working with them to see if we can give them some data to like help, help kind of make those decisions. And that's a three location group. And so like, I'm seeing more and more of it. We talked about the Pokey.
place example where they're doing it, you know, that's not a huge enterprise. So I actually think that it's AI is going to do a lot of democratization of tech for independent operators. They shouldn't manage everything themselves. Like supporting this stuff is, is still a lot of work. AI makes it easier, but it's still a lot of work. But I think that you're going to see a lot more independent operators have a lot more homegrown stuff or demand from their vendors. You're going to be my, my source of, my record holder.
Right. Like what you were saying, Josh, but I demand I get an MCP or a skill, but I can actually do what I want with that data. and I think we're going to see more of that. and I think that's going to be a key part of every single restaurant tech product is what do you have for skills? What do you have for MCPs or whatever the next version that comes out next week that makes both of them obsolete is that basically happens once a month.
Joshua Sharkey (01:03:08.706)Yeah.
Michael Jacober (01:03:09.538)Yeah. My methodology here is, if you, if you're an enterprise that's large enough to have like a director of technology or a head of IT, you should, you should consistently thinking about, you should be thinking about builder. Absolutely. Yeah. Yeah. I think the days of just like the, the, the technology director being the person who is, you know, procuring all these tools and, and dealing with the contracts and maybe overseeing, you
the physical IT director, like you gotta be building stuff now. And if you're not, think your CEO is not being the best fiduciary to the business.
Joshua Sharkey (01:03:50.414)This was honestly like a bet, I thought about from very early on, before AI was it like, know, ERPs, reporting systems, know, visualizations of data, things like this are like, just, are going to be commoditized and those things are very easy to replicate. And which is why we focus on like, okay, we want to replace Google Drive, Google Docs, Google Sheets, the things where like, you no longer store your IP there, you're going to store them in...
in these so that no matter what, like, fine, go use whatever reporting tool you want, build your own. But the, you know, but the, but the system of record, I think becomes really important and, even, you know, Stripe, you know, like you can't replace MasterCard or Visa, but like you could probably, you could build Stripe.
Michael Jacober (01:04:35.95)You could build Stripe. Yeah, absolutely. Yeah.
Sterling Douglass (01:04:38.702)I also think too that the system of record is important, but I also think there's a ton of value in the system of action. Like I think that there's a lot of nuance and building kind of the orchestration layer or the agent layer to be able to actually do stuff with that system of record. I've been saying this since 2016, but I believe a point of sale system at its heart is literally just a series of databases. Like it's really all it is, right? It's a menu database, an order database. Yeah.
you know, prep station configuration, right? Like it's literally, you could boil it down to nothing, but you know, a couple of Postgres tables with some JSON payloads. And then, you know, they use payments, but that's payments is just a way for them to make money. Right. Like payments isn't really like a value add thing. It's how they make money. And then they wrap all this other stuff around it. And then they want to try to get more SaaS. So then they start out, try to add, you know, KDS and this and that. But I think at a heart, that's all it is. But I think what it's going to become.
you know, as the years go by, it needs to become the system of action. Like it needs to become like, how do we grow from just like, all right, well now I can ask my, you know, square IQ about a report. So how can the point of sale actually just go do stuff for me? You know, it knows that I'm out of tomatoes, so just 86 it yourself. Like it knows that like this particular item has a bad margin, take it off the menu. Like how do we make it stop ordering this one from the supply chain side? Because.
Like you can get a better price at this place. that, those actions, that's where the real value is long-term. And so the databases and stuff, I think that gets commoditized. I think it's just easier to build off of it today. It's why not that many people are trying to vibe code a CRM and replace Salesforce. They're more interested in doing the actions and just using that data, because I think that's where there's a much bigger ROI. But I think that's where the POS will, the the successful POS is in the future. Will we try to be the, of the system of record.
They're really going to want to be the system of action because the actions are what actually produces true value for the restaurants.
Joshua Sharkey (01:06:40.27)Yeah, I think that's, I think that's, um, that's a. Yeah. lot of the orchestration is going to be a while before that gets replaced. think the first thing you replaced is like, give me the report of XYZ. One, you don't really even need a report anymore. You just want to set this as of what's important. It's nice to have a report, but reports are very easy to generate now. Um, but the synthesis is even more important, but those, those you don't need tech for, you know, you just need the data. You need a good, you need a good data source that's structured well and that's really it.
Michael Jacober (01:06:42.252)I it's interesting.
Sterling Douglass (01:07:08.654)You guys wanna talk about a non-AI topic?
Joshua Sharkey (01:07:11.15)You know, can we end on a non-native topic? It's funny you say that because like you know what lately one like therapy has been helping a lot more for a couple things but like I find myself cooking more now because I I miss cooking as a chef but like also I find just getting more and my my wife has this really cool nature thing that she's building for our our school for our kids to get kids out and you know out off of off of tech and like basically like planting
plants and being in nature as a restorative process. And even if they get trouble, instead of like going to the principal's office, they go to this nature. And it's really cool. And like every weekend we're there with our kids and the community just building, hauling, know, dirt and mulch and carrying things. I find that like, I want more physical things because so much is going on digitally.
I don't know about you guys, but I find I'm actually leaning more into those things now because of how much is happening on the digital side.
Michael Jacober (01:08:13.134)First of all, I want to mention that I made those sausage burgers this weekend. They were great. They were great. I think it was a hard sell over cheeseburger. Yummy.
Joshua Sharkey (01:08:18.734)So how did they turn out?
Joshua Sharkey (01:08:27.854)If you position it, Sterling, we were talking about, think that like sausage meat is the great vehicle for burger because you can make any kind of, you can make a Vietnamese sausage, you can make a pork cheddar sausage, you can make one. And you don't have worry about temp, it's perfectly cooked, perfectly seasoned. And I sent Mike a Vietnamese sausage that I make at home. I think I sent you the pork and shrimp one, right? Yeah. But if you say, hey, do you want a cheeseburger or do you want this?
Did you? But it sounds like you did.
Michael Jacober (01:08:58.254)No, I didn't ask. didn't ask. We all sat around and had these pork burgers and they were delicious, but they were enjoyed. They were just not enjoyed nearly the same way that my kids would have crushed.
Joshua Sharkey (01:09:11.179)Well yeah, for the kids. you, was it spicy? it like cilantro, mayo, that kind of stuff?
Michael Jacober (01:09:16.73)I left, I left several of the ingredients out. Yeah. Just from a, from a complexity perspective. I know my, I know my customer.
Joshua Sharkey (01:09:31.402)so i will buy you man
Sterling Douglass (01:09:33.144)I've always, I'm a fair, like I love to obsess over things. I love to go down rabbit holes. So from a tech perspective, I don't actually think I've gone, and I don't think it's been that different for me on the AI side than it was for, know, APIs and the restaurant tech industry and, you know, virtual restaurants when I was a business model. was like, I typically always throw myself at those things. And it's one of the reasons I have always, like ever since then, I've been very outdoorsy.
so that's like always been, you know, part of, part of me is like, I'm sitting in front of a computer. I'm doing research. I'm, I'm, trying to understand this. and so I'm always trying to escape like big snowboarder love hiking, is one of the reasons I moved to Denver. and so like, I've, always am obsessing over tech and then escaping. and so I, this hasn't been, this hasn't been that different. Honestly, I, I'm still going super deep and. Yeah. I'm trying, you know, lay literally laying in bed doing flawed code on my phone.
and then falling asleep and then the next morning be like, my God, I need to get out of the house and go on a bike ride. so I think that for me, it's kind of, it's been a bit similar. I think the only, the real big difference has been I probably had been a little bit more excited to share it with others and because it's so, I just think it's so cool. No one's really interested in once you kind of like really figure out like how to get menu scheme is right. You know, in different points sale systems, like nobody really wants to hear that. but I think so I'm like, yeah, I can make this like.
really sweet bird watching apps since you like to watch birds and you can like do all this custom stuff. Like they get excited over it. And so I think that for me, I've been, that's been really, that's given me a lot of joy, uh, is just being able to actually share it with others.
Joshua Sharkey (01:11:15.574)I agree. It's funny, like I feel bad. It's actually even you too, because Mike and I talk about this a lot and like basically every day I'm like, I want to text Mike another thing about something I'm working on. you too, so I don't bug you nearly as much because you don't always.
Sterling Douglass (01:11:29.934)Look, I don't want any texts from you about AI, but I do want texts about cheeseburgers. Those are, if you look at the group texts with my friends, it is half of what we send each other is cheap pictures of cheeseburgers. So anyone can send me a good smash burger anytime they want.
Joshua Sharkey (01:11:46.222)Well, you have been helpful on that so I will send you some cheeseburger stuff, but like every day I'm like working on something and I want to be like, you know, I have like four or five, you know, friends that like are in the weeds as much and I just at all times want to be like, dude, check this out. Because it's so like there's just so much to share so.
Michael Jacober (01:12:04.238)Going back to the beginning, it's interesting. You talked about your compadre on, I'm sorry, I'm blanking on Aaron and how sort of you guys would, it wasn't a competition, in talking to each other, you're like, my God, I'm behind. I need to catch up. I have that relationship with Josh and I've got like maybe one or two other buddies. It's such, I think it's such a healthy way to communicate about this, this, you know.
Sterling Douglass (01:12:30.958)There's always a bigger boat. There's always someone who's done more with AI and is further along with you, no matter who you are. So the quicker you just accept that, get over it, you know, and, and just start sharing and getting feedback, like the better.
Joshua Sharkey (01:12:45.742)Well this was fun. Thank you. I actually learned a lot today. a few levels. I appreciate it. Mike, thank you as always as well.
Michael Jacober (01:12:57.89)Great. Yeah, thanks for having me, Always good to see you, Starling. Yeah.
Sterling Douglass (01:13:00.728)This is fun.
Joshua Sharkey (01:13:02.734)Thanks so much for listening to the show. If you liked this episode or any other ones, you can actually check out more of this at getmes.com slash Josh. That's G-E-T-M-E-E-Z slash J-O-S-H. I have my podcast there, The Mese Podcast, plus some other shows and interviews. Starting to write some stories and blog posts, some recipes, recaps, things like that. So I think you'll enjoy it. Again, it's getmes.com slash Josh. G-E-T-M-E-Z.
Thank you very much. Very grateful for all of you.
.png)
.png){kind=icon}
