Powering the AI-driven science revolution with Lilypad

GM GM. Welcome to what's going to be another great episode of DevNTell. So if you didn't know, DevNTell is a 30-minute podcast held every week to allow founders, hackers, and anyone in between to showcase what they've built. And today I'm excited to welcome back the Lilypad team, this time comprised of CEO Allison Haire and Head of Research Stanley Bishop. So if you didn't know, Lilypad is a decentralized compute network powering use cases from DeSci, AI agents, and everything in between. So if you stick around for today's episode, you'll see Allison and Stanley share their story behind Lilypad and how it's powering the AI-driven science revolution. Alright, let's get into it.

GM GM. Welcome to the show, gang. Pleasure to have you both on.

GM GM. Thanks, Narb. It's great to be back again talking with another OG of the Developer DAO community.

Indeed, indeed. And Stanley, welcome on the show. I know this is your first time, but you too are a Developer DAO member, as we both Ally and I have learned in the last five minutes.

One of the many treats from the pre-show.

Excellent. There you go, there you go. Indeed. Yeah, so like I mentioned, excited to have you both on. Really excited to hear how Lilypad has evolved since the last time, Ally, you were on the show. But for just in case folks weren't tuning in to that episode and missed you there, would you both like to give a brief introduction about yourself and then we can get into the meat of the content?

Yeah, absolutely. So, hey everyone. I'm Ally, CEO, founder of Lilypad Compute. I was working with Protocol Labs when we first founded Lilypad. So I was part of the launch team on the Filecoin Virtual Machine there, and also then moved over to a compute project called Bacalhau at the time, which is actually how I met Stanley. And then we founded Lilypad out of kind of a combination of both of those experiences for me. I previously worked at IBM, but you know, also just been riffing in this space for a few years now. And even met Narb a couple of years ago at what event was that at? Was that at ETH Denver? I can't remember now.

It was at a Graph Day, but it was actually a Developer DAO side event that we started to chat really.

But it was in San Francisco, wasn't it?

It was, yes.

Awesome. Yeah, amazing. Hey Stan.

Hey, I'm Stanley. I'm a data scientist. I work in machine learning and high-performance computing, and I particularly do a lot of work supporting the compute needs of different scientists and researchers, mostly in bio and medicine. I like to say I am not a James Bond, like all these incredible physicians and researchers, but I'm Q who gives them the gadgets they use to blow up cancer and other shit like that.

Yeah, and I think you're being humble there. I think you do much more in this space and we'll get into that. Just interested to kind of hear both your stories on like how you started in both your respective fields. So Ally, I know you're big in tech and you mentioned starting in IBM and whatnot. Like what was your inspiration for going into the tech field?

Yeah, definitely. It's a bit of a story actually, because I ran a cafe actually in my 20s. So I owned and operated a cafe in Australia in my 20s, which was kind of accidental. That's a different story though. But ran that for about eight years and when I sold it, I didn't want to be in hospitality forever. So I moved, I went back to uni to do, or back to college, depending on where you're from, and did, started in mechatronics engineering, so robotics engineering. I'd always been interested in science and math, you know, throughout my high school and schooling years. So I went back to do this kind of engineering degree. I thought this would be like kind of a practical application of those interests. And part of that was a coding course. So I did a coding course as part of engineering. And that's really where my kind of love of technology and this software space came from. So I did this coding course and I just wrote my first program at like 30 years old or something and I'm like, oh my god, I'm in charge of the internet. It wasn't quite like that. I'm sure everyone kind of has this moment where they remember like writing something like that for their first hello world and you know, feeling like pretty powerful. But yeah, so I decided to do computer science as a double major or double degree. And yeah, haven't looked back. I was part of a lot of hackathons. I even ran a startup when I was still at university. So I was doing my thesis. I founded a startup with a friend of mine. We won an accelerator program. So we went in a hackathon, we won it, and then we went through an accelerator and were going to go further. But we had a bit of a difference of opinion on taking external funding. So we decided to throw it in. But it was a great experience to have really early on. One of the reasons though, I think, and this has evolved over time, one of the reasons that I love technology so much, and it's not just because I feel powerful looking at it, but I feel like it is a tool for helping us. So, you know, if we especially, I did this presentation at IBM, I looked over the last 100 years of technology and how much it's changed our society. And you know, that's one of the things that I love about it. So things like appliances in the home really changed the game for women especially. You know, they could go and work then, they could get out of the house, things like that. And then there's obviously the printing press and the internet which gave education to the entire world. And now we're looking at, you know, a revolution with AI in my opinion, that is going to fundamentally change how we collaborate and communicate in the world as well, and some of the breakthroughs we could have in things like rare disease and science. And I'm sure I'll hand over to Stanley to kind of tell you how he got into it as well. But yeah, that's kind of my fundamental drive, and why I love technology so much. It changes how we do things as a society and can solve problems for us.

100%, and couldn't agree with you more. And also found out today that you were a part of the restaurant scene before tech scene. That's a cool story to hear.

Yeah, I did it backwards. I was supposed to retire to the cafe after the startup, right? Did that in reverse. Mixed that up.

Sorry to be unconventional. We like that, we like that on the show.

You know, being able to organize a busy coffee shop I think is up there with an aircraft carrier, you know?

Maybe, but but less consequences if people don't well, maybe if people don't get their coffee.

I don't know Ally, have you seen me when I don't get my coffee? It's...

Yeah, me too.

Yep, yep, all three of us are quite avid coffee drinkers. I'll also take a sip to join you. But yes, jokes aside. Stanley, I know that you are regarded as one of the OGs in the DeSci space. I'd love to hear kind of your origin story of how you got or what your motivation was to going into that field, and how you mentioned before where you're the provider of gadgets for folks looking to cure cancer and other rare diseases. Would you like to get into that a little bit?

Yeah, I'm happy to, and I wish I could say I did it all on purpose. But my original intention was to be a quantum field theorist and I was actually in grad going through grad school working towards a PhD when I actually, I actually got very sick about a year and a half into grad school. And this was before we had the Affordable Care Act in this country, so I did not qualify for care because I had a pre-existing condition because it was diagnosed right between undergrad and grad school. Wonderful system, really wonderful system. Ended up needing to take a role and I I got this really cool role, I got kind of lucky. I was a principal machine learning scientist doing language localization for Google's codebase and a number of other projects at Google through a system called Falcon. And Falcon was a system that used some of the very early what we call multi-internode GPU compute topologies. So these are compute topologies that have multiple GPU nodes and then the the speed and the data logistics across all the nodes is sort of the most non-trivial part of the engineering. And yeah, we were building these systems to actually support the work of about 100,000 language translators all over the world. So you might describe the platform as Uber for translation or localization. So yeah, just really fun, juicy project. It made me very passionate about these kind of systems and how to build them. But more so, it kind of gave me my the big passion for my career, which is building robust systems containing both human and machine intelligences. And then that's definitely the thread I've been following and then particularly when it comes to scientists whose work I find very fascinating. I like to build, you know, systems that support and accelerate scientific work without being frustrating for the scientist.

Amazing. Amazing. Yeah, and and the work you do in in in your field is amazing and and we all applaud you for it. But this is actually a great little segue here into how both of you kind of came together at Lilypad because your your stories are have the synergy together where you're both really into tech and looking to use it for for good. So I guess either of you can take this, but how did you both come together to found Lilypad?

Yeah, I'll talk a a little bit about the origins. So I think I I kind of mentioned it briefly in my intro but I was working with Protocol Labs, I'd just one of the projects I was doing was helping to launch the Filecoin Virtual Machine and that included you know, you know, getting it to market, getting it across developers, trying to make it accessible to developers and building out the programs to onboard developers there. And then I kind of hired a team and finalized that and moved over to a project, really interesting project called the Bacalhau project. And that was run by one of our advisors still called David Aronchick who was one of the leads on Kubernetes at the time. Under the hood, Lilypad still actually uses the Bacalhau project, but I might be putting the cart before the horse there so I'll just I'll park that one. But was working for the Bacalhau project and this it was all about compute. Bacalhau is a peer-to-peer compute, distributed compute network, fully open source compute network, check it out if you haven't heard of it already, super powerful. But it didn't have any incentives. And so I'd just come over from this, you know, really incentive-driven like you know, chain into this peer-to-peer compute project and I'm like these two should be together. We should be able to, you know, run this kind of thing with incentives based on blockchain, with kind of these guarantees that blockchain can provide us. And that's kind of where the Lilypad project was born. What's interesting was Stan was in this Bacalhau space as well. So we had been working on, you know, building out projects together in the Bacalhau world as well. So we were were talking a bit and then when I when I did go and found Lilypad, like we started out at Protocol Labs, but then when I did we did go and found it, I was like, I couldn't think of a better person for Chief Scientist than Stanley and I'll let him tell his version of the story of how we met at Bacalhau as well. But basically this is how it happened.

Oh, I have to say too, it's so funny the way, you know, our stories play out and always kind of did remember the first project with Ally as very very special and of course so excited to have been able to, you know, grow that energy into such a special and important project. And 100% like I am very very passionate about decentralized computing and citizen science. Grew up a big folder at home, had a big nasty cluster that took up almost my whole wall in my childhood bedroom. And and yeah, being in the decentralized science space and then also having a background in HPC, I get brought a lot of interesting projects and a lot of interesting ideas come across my desk. And a lot of them are things that haven't happened for a reason. They kind of suck, like not having enough compute or, you know, sometimes it's just the people who want to use the compute not knowing how to access it, you know, there's sometimes little twists and turns that can prevent a scientific computing project from happening. And nowhere I think is that problem more critical than in marine science. And that might sound like a little odd, but if you're kind of familiar with what we get from the ocean and how critical the ocean is to the whole planet, you you might really sincerely consider the lungs the ocean to be the lungs of the planet. So the things that are happening in the ocean are quite negative related to climate change right now and we we don't even know a lot of what's going on because, and here's the wonderful tagline, there is nothing more decentralized than the ocean, right? It it's on every every laps up on every shore. So the problem of generating data and biomonitoring in the ocean incredibly challenging. And I had been the technical leader for a project called New Atlantis that was actually building what we call metagenomic technology for bio-surveillance and Narb, I think you'll love this metagenomics is so crazy. You you literally take a cup of water, that's not what this is, I'm not drinking salt water right now. But but you actually sequence every fragment of DNA present in the water and and you can just imagine how many different animals have DNA present, just floating around from plankton to DNA fragments from fish and other larger animals. And it turns out that actually just with this little water sample, you can get almost a fingerprint of the local ecology that can tell you way more than you would ever expect because the life in the ocean is so interwoven. But it's rather a lot of compute. I think cost to process one sample right now is about $5 and to monitor just something like the the Gulf of Mexico for a year you could be talking a million samples. So the problem of finding the compute, you know, not just in an affordable way but in a way where it's close to the data, very challenging. And just when we started working on this problem, we realized, you know, Bacalhau and the decentralized data paradigm was like a perfect fit.

That's amazing. Yeah, I mean 5 million, 5 million for for like one sample, that's that's that's crazy. Or a million samples, that's that's crazy. No no small change.

You know if I can add, sorry to just jump in, but tell you one other just kind of fun thing because we are talking about the the patterns of technology a little bit in in this convo. The cost of performing that kind of analysis actually falling faster than Moore's Law, if you can believe it. Yeah, it's Ally it's actually a double exponential. It's it's double Moore's Law.

Yeah, it's amazing, isn't it? I think it is like when storage came out and that was super expensive but over time it's become like much cheaper to do. But it's great that we're bringing access because those kind of walls prevent kind of science from being able to to happen really. They prevent science from being able to be innovative, they you know, put up walls so you only have proprietary information or proprietary models happening, which is not what we want with science. We need need the world to have access to it like that's the dream of the internet, right? Sharing kind of data and information around the world to be able to collaborate and make that better, that's the dream of open source. So yeah, it's it's it's frustrating when I when I definitely I'm frustrated when I see blockers like that happening for projects like this, incredibly important projects in the world. And you know, that's you know, I'm back on my soapbox but this is like you know another reason we need to give tools to to people, right? We need to give open access not just to the compute but to other resources that they need, to the tooling they need to be able to create these things and reduce the walls and pain points for doing so while still, you know, enabling them to kind of make a living as well because you know, scientists are kind of the starving artists of the world I think sometimes, which is horrible.

So this is actually a great point. So both of you kind of hinted at it. But I'd like to now get into what Lilypad exactly is and it it is definitely solving this problem. But like how does it solve this problem? Yeah, would love to get into that.

Yeah, cool. So I mean, at how I like to think about it almost from a a layman's or how I explain it to my mom point of view is I kind of call us the Shopify of AI to a certain degree. So we have a three-sided marketplace where one of them is our hardware network, so this is the compute providers that provide their GPUs to our network, and they are obviously rewarded by being paid for running jobs on our network. We have users who are looking to say use one of those protein folding models or use a specific model that's been created by a scientist and then the third part of it, which I think is like the Shopify storefront, is our module marketplace. So people that have created these AI models or specific models for, you know, potentially for you know these marine analysis or for rare disease analysis, they kind of have this option to monetize that model while still keeping it available to the public. So I think that's kind of a special thing that we do here. And that's basically it, right? So three-sided marketplace you as a user you can, you know, go on and choose a model, run whatever you like, get the results back. As a hardware provider, you're paid to provide hardware to the network, and as a, you know, AI data scientist or someone that likes to create models or even if you just want to wrap a Hugging Face model for Lilypad, you have an opportunity to kind of monetize that model and you know, have a have a kind of living out of it and have this third option that's outside of being paid by someone proprietary and having to like have the rights go to them or open sourcing your model and and getting nothing for it. So I think that's what's really powerful and did I miss anything there Stan? I think that's kind of Lilypad.

Well, I think that that three-sided marketplace, currently the triangle is not fully sided, one of the sides is broken. People building modules are just doing it as a labor of love love and it ain't necessary. But but I do have a fun thing I would mention and Narb just for the audience's context, we are talking on a very special day. Ally is actually here in LA with me and and we're about to meet up and hang. And Ally, I believe Amelie is joining us? So I am so excited to introduce you to Amelie. Narb, Amelie is I think considered one of the most influential and high-level open source developers on Hugging Face working in biological machine learning.

And she is to me such an important example of what Lilypad is going to do. This is a person who is in love with molecules. When she talks about the molecules that she's designing, you can just hear the passion and the fascination come through in her voice and you can kind of tell within a couple minutes of talking to her she probably thinks way more about molecules than how she's handling her bills or like where her parking space is, right? And and so I became aware of the fact that probably literally 10 well-funded startup companies had launched in the past six months based on the work that this open source developer was publishing and she didn't have any support. She was just kind of piecing together GPU access in her local network to do, you know, this work that was guiding a revolution. And and then here's the idea, let's pay her right? Let's like make her comfortable, let' her, you know, do that work that's in demand. And and yeah, we have this enormous resource of untapped compute. And and by putting that to work, you know, we can kind of create a an economy that sort of supports those kind of really important developers.

Yeah, that's amazing and quite the the game-changing tech. Especially since it sounds like this side of the world really needs that compute power and have been pouring their blood, sweat, and tears into the work they've been doing. But can't get to that next level because the the hardware may be too expensive for them or it just doesn't make economic sense. So from that standpoint, the hardware, like who who can be the provider of this hardware, what does it take to join the Lilypad network?

Yeah, currently we're we're fully decentralized. So if you have a GPU you can join our network and provide compute power to run jobs on our network. And I think that's an important thing that maybe I I wasn't clear about as well, we're a job-based network so rather than renting out your GPU for a certain amount of time, you know, like most other kind of Vast.ai or or similar platforms like that, we do job-based networking, so we're coordinating jobs on the network rather than coordinating time-based, you know, rentals. And that's kind of powerful because if you that means we can we can kind of collaborate with other DePIN networks out there that maybe aren't being fully utilized for example. So, you know, we could potentially collaborate with Akashi and have them run Lilypad nodes while they're waiting for a time-based, you know, rental for example. So I think that's kind of a powerful thing. But yeah, just jump on the docs and you'll you'll find out how you can contribute compute to the network and yeah, grow the network with us.

It's such a important part of the model too because it creates an incentive for developers to build modules that don't just perform well but are reliable. I think every machine learning engineer has that experience of wincing when you log into a Hugging Face and seeing that it hasn't been updated in three or four months. You know, another part of the Lilypad project that's been so interesting is is this continual feeling of trying to aim past where the target's moving. Because I think we've all seen the space changing so quickly. In terms of what hardware is in demand, in terms of, you know, what kind of workflows we're using. So yeah, kind of like staying on on top of that making sure hardware runs reliably very non-trivial jobs, so almost becomes critical that we find a way to support the developers doing that work.

Yep, yeah, 100% 100%. Ally, were you going to chime in with something else here?

I don't think so. Maybe, but I've forgotten it.

All good. I wanted to double click into the concept of a module a bit. So for people who aren't familiar with the nomenclature of Lilypad, could you give a high-level explanation of what a module is?

Yeah, so basically a module is is just a dockerized it's just a container with a Lilypad spec over the top of it. So I mean you could tell everyone Narb, you are one of our main protocol engineers out there now. But yeah, so a module is basically just a Docker job with a Lilypad spec and the Lilypad spec enables us to match you up with the right kind of compute and the right kind of client for for running that job basically, yep.

I mean of course I couldn't have told the story but I'd rather hear from both of you.

Oh, and and I hope you won't mind if I I give a slightly not not even a different answer but an answer that's really reson or a thought that's really resonated with me on this question. And actually like hearkens back to those early days of multi-internode GPU computing engineering. Because Narb let me tell you, it was not easy. This was the pre-Docker days. So we were literally for each node type having to write scripts and test scripts and you know, do pretty messy deployments. And you know, typically like if we got on a thousand-node deployment, like 70 to 80% of nodes turning green on deployment that was a good day, right? And then we would spend a week or two SSHing into 200 machines and trying to figure out what was going on. And I I felt at that time a pain, a psychic pain. And and I feel like that pain that I and many other developers were feeling kind of led to Docker and and the container paradigm. And I I feel a similar pain. I think decentralized AI is challenging because the hardware is complex, because the models are complex, they require complex objects that need to be available on runtime with performance constraints on interaction between the cores hosting the model and the data artifacts. And that's before you even get into these things needing to talk each other to to each other. And the fact that agent-driven systems have very specific and challenging sort of data and interoperability concerns. And you know then you also build in these kind of social and recommendation elements where like a module isn't just its own software runtime, it is the the marketplace and the data about that model's usage that's going to define the usage of a a future user. So what I'm really getting at is I believe, and I I hope it will be the Lilypad module or a standard that we are able to contribute to, but in a year or two we will have a new Docker container. And it will be the kind of standard atomic unit of decentralized AI compute. And I can't wait because like I said Narb, I'm I'm feeling the pain.

And yeah, that that would be quite the quite the feat. That would be amazing if we can do the same thing that Docker did for the software world, but for the AI agent world here. And I guess, kind of around the same point, we see a lot of other decentralized compute networks in the field now, I mean it's to be expected with the rise of AI agents and whatnot. But like how would you both kind of describe how Lilypad differentiates itself from those competitors?

Yeah, I I'll take that one a little bit. I think I kind of suggested that the differentiation is what we've just been talking about, one, this standard atomic unit, I love that name for modules to a certain extent, creating kind of this standard for running them. Also we're a job-based system. So I would say that a lot of who you may initially think of as a competitor, so some of the DePIN networks, are really a collaborator for us. I think we can we can work with most of those networks as long as there's some sort of standard to, you know, provide them with more revenue as well, and you know vice versa they're running jobs on us and and helping us build out our DePIN network. So that's one way, I think we definitely differentiate and yeah, I mean the standard for kind of modules is is the second way, so a job-based ecosystem. I heard this word the other day on a podcast that I quite like which is instead of being competitors, instead of being competition it's coopetition, and I think I really like that word now. I think a lot of that is is what's happening across the decentralized AI space in Web3 as well. It's not necessarily competition but it's each coming at different parts of the problem in a different way. And I think we'll see a lot of that coming together over the next year. There's lots of lots of bright minds working on different parts of this problem and we need lots of different minds to be able to essentially make a decentralized cloud, that's what we're doing here. I mean it took Amazon 20 years, and making a decentralized cloud needs needs a lot of bright minds working on a lot of different parts of the project which we're seeing in in Web3 at the moment as well. So I think it's really powerful and I'm looking forward to the next few years to see what happens. I really honestly believe, and I've been in Web3 for a while now, but I honestly believe this is kind of the breakthrough use case, these agents and and AI. This is this is what blockchain's kind of been waiting for, something like this and we're mature enough to do it as well. We wouldn't have been a year or two ago.

Definitely. And it's quite interesting to see this narrative playing out now where you see AI and Web3 much more intertwined whereas a year or maybe not even a year, but a couple years ago people were like oh no, it's like there's nothing to do with each other. I know there was a few projects that came on the show and I started to probe into like what that potential could look like and it's just amazing for me to see how...

Explosion. Cambrian explosion.

Exactly. Exactly, exactly. Exactly. I well I think, you know, this is an inflection point in history too. It was a combination of things coming together. Before we wouldn't have been able to do distributed compute very well on one side and then on the other side AI wasn't really kind of mature enough or mainstream enough now, it definitely has captured the mainstream interest but it wasn't mature enough either. So we've kind of seen this power play of both things coming together, this distributed compute, also blockchain enabling primitives like payment methods. I mean an agent can't be banked, so how else are you going to give it money? And then you know payment not just payment methods either but like that transparency and provenance guarantees that you can get with blockchain. These are things that AI kind of desperately needs to solve some of its problems. So I think these these are why we're seeing so much collab or cross-collaboration of the two arenas and I'm super excited about that.

Yeah, I'm sure lots of folks are. I see we're more or less coming to time. Just want to finish off with a couple more questions. Firstly, is there anything that you can or would want to share about the Lilypad roadmap for the upcoming year or beyond?

Yeah, so we're really heads down at the moment, heading towards our product launch. So we're hoping to launch early Q2 this year, and that means TGE and a production-ready kind of mainnet. We're not a full chain so when I say mainnet, it's production-ready code on a mainnet chain. And we're fully EVM compatible so like a lot of the AI projects and agent projects you're seeing, we're not necessarily tied to to a single chain here, which I think is also amazingly exciting. But yeah, so our whole team is fully focused on getting towards that, our engineering team's working hard on on defining the protocol and I know you are as well, Narb, one of our MVPs even. Getting to that mainnet launch. But if you want to participate, we have plenty of kind of we're running a a rewards program for builders, we're running rewards for building modules on our system, so bringing over even Hugging Face models to our system and doing that extra 2% of code that's needed to to transfer it over to Lilypad. It's a little non-trivial at the moment but that's Web3 for you. We're working on the experience as we go, only got so much time and engineers. But we'd love you to participate and the other thing we're doing is really promoting Lilypad for agents because one of the big value props here is you can run kind of parallel compute jobs on Lilypad. So your agent could run a thousand LLM jobs at once on uncensored LLMs or unbiased or censorship-free, I should say, LLMs if they're on there on Lilypad and have that return and that's something you can't just do on your local machine. So we're working towards having an API out actually I think due this week that you'll be able to kind of call Lilypad and run those kind of agents. So we'll have our next phase of our builders program will include porting over agents to Lilypad. So we're super excited to see what people might want to build with some of those agents because that's definitely exploding and I can see why, like it's just so practical across so many verticals. So we'd love to see what people build on our system with those kind of tools.

You know, oh I'm so sorry Ally. Yeah, yeah, yeah. Oh well so I I was just going to say, I love me some big fancy computers. Like when I was a kid I liked working on cars and now it's big clusters. But you know the right computer is the one that the user needs and the one that's set up for the user, right? So I've always felt that as a a data scientist it's like you really do have to kind of meet the user where they live, and that's usually on their their MacBook Pro. And the new MacBook Pros are really inference beasts and so I've had so much fun showing a lot of my friends in different areas of science how, you know, they can just run some basic agents on their their local laptop, but there is this common pattern where you'll have a a gummed up node in your agent chain of thought. One where you're maybe simplifying the problem or restructuring the problem on mass over your whole data set. So you need to kind of chunk the problem and have a bunch of smaller models process it piece by piece and and that can kind of take what would otherwise be a 20-minute agent runtime and make it an eight or nine-hour agent runtime. So having what we might describe for these like local agent users as like a pressure valve. So when they get to one of these nodes that's just like so overwhelmed with demand they can just connect that up to Lilypad and sort of, you know, get a system that might be based on their local but has just the right amount of cloud resources to call in. Yeah I think this will be just incredible enabling for scientists looking to be part of the agent revolution.

Amazing, and well said by both of you. And Ally, you answered both of my questions as well so great stuff, great stuff. Yeah, thank you so much again to you both for taking the time out of your busy days to come on the show and talk shop. Really really appreciate that. And all the resources that you need to get started with Lilypad are provided in the YouTube description below. And if you watched us live today, I am dropping the link for you to be able to grab your attendance badge on BitBadges. So you'll have an hour from now to grab that to prove that you watched the show live. So again, thank you. Fancy right? Yeah, exactly. There's a leaderboard and everything. Yeah, wow, that's really cool. I love that. Shout out to BitBadges. Go DD, exactly. And yeah, with that, just want to wish everybody a very happy Friday, a happy weekend wherever you may be and we'll catch you back here for another great episode of DevNTell. Thanks Stan, always lovely chatting with you too. Thanks Ally, likewise. Cheers. Bye.