Latest / The Index Podcast / Build a Solana App With One Prompt? Amit Ishairzay & Bilal Shahid of Poof Are Making It Happen
Transcript
- 0:00SPEAKER_02: I finally found a hardware wallet that isn't a pain in the ass.
- 0:04SPEAKER_02: Every hardware wallet I've tried, ledger, Tricer, whatever, they all have the same problem.
- 0:09SPEAKER_02: They're annoying enough that you stop using them.
- 0:11SPEAKER_02: The Soulflare shield is the size of a credit card.
- 0:14SPEAKER_02: No cables, no batteries.
- 0:15SPEAKER_02: You literally just tap it to sign transactions, and that's it.
- 0:18SPEAKER_02: It's built into Soulflare app, the same app that's been around since 2020, securing$14 billion for 4 million people.
- 0:25SPEAKER_02: And because it's Soulflare, you get real P ⁇ L tracking, DeFi positions, and all your Solana stuff in one place.
- 0:31SPEAKER_02: Cold storage that you'll actually use.
- 0:33SPEAKER_02: Link in the bio, go take a look.
- 0:59SPEAKER_02: They're building the infrastructure that lets you deploy a full Solana application from natural language prompts.
- 1:07SPEAKER_02: And really excited to talk to you guys about this because I have so many questions about how this actually works in practice, especially when you're dealing with like on-chain programs and stuff like that.
- 1:18SPEAKER_02: But first, just thanks for coming on the show.
- 1:20SPEAKER_02: And I would love to hear, maybe we'll start with you, Amit, like what gets you out of bed in the morning?
- 1:25SPEAKER_02: Why are you doing this?
- 1:26unknown: Yeah.
- 1:27SPEAKER_03: Thanks for having us on, Alex.
- 1:29SPEAKER_03: Happy to share about Poop.
- 1:31SPEAKER_03: And um what gets me up every morning is um I I've always been like a nerd by nature.
- 1:38SPEAKER_03: I care about technology and I get super interested in things.
- 1:41SPEAKER_03: And crypto specifically, and and now more so Solana, um always caught my interest as this global computer that um just makes so much sense from a both a developer on-chain kind of geeky perspective and also um for just the world and finance as a whole.
- 2:02SPEAKER_03: Um and ever since I discovered that, I've just been on this mission of kind of finding ways to spread that or making it a reality.
- 2:09SPEAKER_03: Um and so that gets me up at just knowing that the world isn't on these rails yet and knowing that they should be.
- 2:16SPEAKER_01: Um yeah.
- 2:18SPEAKER_01: What about you, Belal?
- 2:20SPEAKER_00: Yeah, honestly, similar to what Amit said, um, in general, I'm a technologist at heart.
- 2:25SPEAKER_00: I love working on the frontier of technology and just kind of pushing it to its limits.
- 2:31SPEAKER_00: And I work in AI and crypto, so I'm kind of doing it from both uh domains.
- 2:37SPEAKER_00: And yeah, it's just thrilling to see kind of what we can build that is making things capable that literally weren't capable the day before, right?
- 2:45SPEAKER_00: And like every day just kind of building on top of each other and and really making things that just can't be done elsewhere.
- 2:53SPEAKER_02: That's awesome.
- 2:54SPEAKER_02: And what like walk me through Poof, how did you get started?
- 2:58SPEAKER_02: What was the inspiration?
- 2:59SPEAKER_02: Like, give me the Genesis story.
- 3:02SPEAKER_02: Sure.
- 3:03SPEAKER_03: Um, so Bilal and I, we um we go way back like since high school, but uh we both kind of got nerd sniped by NFTs back in 2021.
- 3:13SPEAKER_03: Um, and that was with NBA Topshot.
- 3:16SPEAKER_03: And uh after learning about that and getting deeper into it, we actually started an agency to help other people build on-chain.
- 3:24SPEAKER_03: Um, so bringing our development expertise into crypto.
- 3:28SPEAKER_03: Um and that's how we got started, and that kind of took us to Poof eventually because as we started building for more and more clients, we found so much infrastructure pain was the same for almost every client over and over again.
- 3:43SPEAKER_03: And eventually this pain that we were facing building for everyone um became it made it clear that crypto as a whole was missing a unified platform for the infrastructure.
- 3:55SPEAKER_03: Um and poof was kind of originated off of that thought.
- 4:00SPEAKER_03: What were some of the pain points?
- 4:02SPEAKER_03: Um yeah, so you would think it's just smart contracts.
- 4:05SPEAKER_03: A lot of people might assume that at first, but there's so many other layers of pain points, such as um the indexing side is always extremely complex.
- 4:14SPEAKER_03: Um actually storing that, um, making sure your access control on those things are accurate, um, your deployment processes, your front-end wiring to all of those other components.
- 4:25SPEAKER_03: Um each of these it it like takes a specialist to get it right.
- 4:28SPEAKER_03: And there's so many um things that you have to get over, so many deep decisions, and so many opinions you have to form.
- 4:35SPEAKER_03: And so facing those and kind of working with clients with on those um repeatedly over and over, and learning about like what makes sense per project is kind of what led us to believe that we have a platform issue because when you're working in Web 2, you don't have as many decisions to make and it's usually quite a bit more straightforward to build.
- 4:55SPEAKER_02: What are some of the pieces of infrastructure that you guys built under the hood that kind of make Poof work?
- 5:02SPEAKER_01: Yeah, Pilot, do you want to take that?
- 5:04SPEAKER_00: Yeah, so we kind of designed a protocol from the ground up that makes it much more easier for people to build applications on top of the blockchain.
- 5:12SPEAKER_00: Um so it is tailor-made to build dApps on Solana, and it is more akin to how you would build an application in Web 2, uh like non non-crypto applications.
- 5:24SPEAKER_00: You can kind of think of it like a Firebase or Superbase or Convex and all these things that are just like blowing up in the Web2 space.
- 5:32SPEAKER_00: So Web3 and crypto in particular always kind of lack something like this.
- 5:37SPEAKER_00: So we built something like this for the ground up, and this is kind of what came from our just experience running the DevShop.
- 5:42SPEAKER_00: And then with the whole like LLM boom, what we realized was AIs are trained on this type of like language more so than writing Anchor and Vanilla Solana smart contract.
- 5:55SPEAKER_00: So they could uh write end-to-end uh smart contracts with a higher degree of accuracy.
- 6:01SPEAKER_00: So we tested this and it was true, and that's kind of what Poof is the bedrock's built on top of.
- 6:09SPEAKER_02: Understood.
- 6:10SPEAKER_02: And then you guys develop Poof, like just in uh very succinctly, just tell everybody what Poof is.
- 6:16SPEAKER_02: Because I don't think we've actually even uh brought that up, but that was the solution to the problem here.
- 6:20SPEAKER_02: So what is what is it?
- 6:22SPEAKER_03: Yeah, uh good point.
- 6:24SPEAKER_03: Um so Poof, it it's quite a simple product.
- 6:27SPEAKER_03: Um it allows you to visit the website, type in a big text box there, what your idea is of an app that you want to create on Felana.
- 6:35SPEAKER_03: Um and after you do that, it will help define what we call a poof policy for you.
- 6:41SPEAKER_03: This is when our natural language AI side comes in.
- 6:44SPEAKER_03: And your policy is what Bilal was describing, a very simple format that defines your app's permissions, your on-chain behavior, and the rules that dictate how your app works as a whole.
- 6:56SPEAKER_03: Um then afterwards, we'll generate a UI and a backend that fit around those rules that we generated already.
- 7:03SPEAKER_03: Um and the goal with this and with Poof as a whole is to make it so that you as a builder can think in terms of your product and the logic of your product itself and not in the deep technical details of how things should work behind the scenes.
- 7:17SPEAKER_02: Yeah, it's super cool.
- 7:18SPEAKER_02: I mean, I I I've definitely wanted to see something like this in the past where I could just like with prompts make a uh an a Solana app.
- 7:25SPEAKER_02: Um, and I it's something I've been learning about, you know, because I use Claude Code like every day and build lots of different things with it.
- 7:33SPEAKER_02: And uh I think for domain-specific things like this, it needs special configurations and skills and knowledge.
- 7:42SPEAKER_02: And you guys, it sounds like I've kind of packaged that into Poof.
- 7:45SPEAKER_02: Do you use Claude Code like on the back, like Claude on the back end for writing the software or like what what what LMs are using?
- 7:55SPEAKER_03: Yeah, um we our whole team uses Cloud Code and also uses Codecs.
- 8:00SPEAKER_03: Um we're always tinkering and trying everything out, really, to see what's working best for us.
- 8:04SPEAKER_03: Um our team isn't just engineers, it feels like it's a full uh team of AIs as well.
- 8:10SPEAKER_03: Um we constantly have AIs running on loops and parallel with us.
- 8:14SPEAKER_03: And uh most of our jobs as engineers have become reviewing code and ensuring that we're working on the right things.
- 8:21SPEAKER_03: So, yeah, definitely using cloud code a lot.
- 8:24SPEAKER_02: And and like what's what are some of the things that are the hardest unsolved problems for like turning just normal English into a secure on-chain program or app that uses an on-chain program?
- 8:39SPEAKER_02: Do you guys like deploy?
- 8:40SPEAKER_02: I mean, can I write a product that deploys a contract on Solana, or does it have to use off-the-shelf stuff and it's really just an app layer?
- 8:49SPEAKER_00: No, so with Poof, you can actually deploy custom smart contracts.
- 8:53SPEAKER_00: And we see this already.
- 8:54SPEAKER_00: We have like users building actual like prediction markets with smart contracts, and these markets are being resolved and and whatnot.
- 9:00SPEAKER_00: So uh and to answer your question about what's the most difficult part, I would say security is probably the most difficult part, or if not one of the more difficult parts.
- 9:10SPEAKER_00: And we designed Poof in with security in mind from the ground up and kind of what Amit was talking about earlier, that like policy system, what this does is it makes it much more easier for like an AI system to kind of come and review your application's smart contract.
- 9:26SPEAKER_00: And um it's a language in AI is much more like it's much better at reviewing than if it was to review like raw Rust code and it can just focus on your application's business logic.
- 9:37SPEAKER_00: So it reduces like the attack vectors and things that are like common that you see in smart contract programming, like re-entrancy and text and things like that.
- 9:44SPEAKER_00: It doesn't need to worry about those per se, and it focuses on your application's uh business logic.
- 9:49SPEAKER_00: So that's how we have seen users be able to deploy things like prediction markets and token escrows with different types of managements and and and who can do what and take what type of actions.
- 9:59SPEAKER_02: What's the what's the craziest thing you guys have seen people deploy?
- 10:03SPEAKER_02: I mean, the funny thing about like Poof to me anyway is that it's basically a crypto app launch pad.
- 10:10SPEAKER_02: And like launch pads have a unique history in crypto, uh, you know, where they get pretty crazy, like pump, for example, like one of the first things on that scrolling animation on your site is like a prompt of like, I want to launch a pump pump fund clone.
- 10:25SPEAKER_02: Like, how many people have launched Pump Fund Cl clones, right?
- 10:27SPEAKER_02: And like it used to take like 30k and a couple engineers to like build one.
- 10:32SPEAKER_02: Um and then now it's like one prompt, and I'm sure it comes out and it just works.
- 10:36SPEAKER_02: But like what's that been like for you guys?
- 10:38SPEAKER_02: Are the DGens coming and just like shipping like crazy, uh crazy things?
- 10:43SPEAKER_03: Yeah, it it feels less and less DGEN these days.
- 10:46SPEAKER_03: I feel like Solana's gotten more serious.
- 10:48SPEAKER_03: Um and uh but there are some crazy apps, of course, and we have had launch pads shipped through us.
- 10:54SPEAKER_03: We've had a lot of experiments in that realm.
- 10:57SPEAKER_03: Um one of the weirdest apps I've seen is um someone made a D-pin app for incentivizing people to pick up dog poop in their community.
- 11:06SPEAKER_03: Um, really funny one, and I don't think they actually shipped it and got users on it, but um it was cool to see that creativity being able to be flexed and used on-chain in that way.
- 11:17SPEAKER_02: Yeah, they were like so pissed off about it, and they, you know, they're like, I'm just gonna figure this out, you know.
- 11:23SPEAKER_02: I'm gonna try and use crypto token economics to solve the dog poop problem.
- 11:28SPEAKER_02: Um I hope someone out there listening solves a bigger problem than that.
- 11:32SPEAKER_02: But that's pretty funny.
- 11:34SPEAKER_02: Um uh what else has happened that you didn't expect?
- 11:37SPEAKER_02: Like, what were some of the surprises?
- 11:39SPEAKER_02: Because that's the cool thing about this, right?
- 11:40SPEAKER_02: Like it's it's not just a token launch pad or something, right?
- 11:44SPEAKER_02: Like you're building applications.
- 11:45SPEAKER_02: So, like, what's the most unexpected thing aside from dog poop apps, d pins?
- 11:52SPEAKER_00: Um, sorry, belonging.
- 11:55SPEAKER_00: Uh we have like different types of users.
- 11:57SPEAKER_00: We have like one cohort of users that like build hundreds of products and they're just like obsessed with building as many products as they can and launching them you know on web apps and launching them on the speaker.
- 12:08SPEAKER_00: And then we have like another cohort of users that just obsess over a single product and it's just like just shit the thing, you know?
- 12:15SPEAKER_00: And and they'll like they're like perfectionists, I would say, and just just keep iterating and keep iterating until it's like, you know, uh I mean you can always iterate on a product, right?
- 12:25SPEAKER_00: So um I think those two user behaviors have kind of been interesting to see the contrast between the two.
- 12:32SPEAKER_01: What about you?
- 12:34SPEAKER_03: I mean yeah, um expanding on what Bilal was saying, it's it's crazy how many people are making like very personalized apps um where they add almost everything you can think of in their own lives into their one app for themselves.
- 12:49SPEAKER_03: Um so it's both their trading terminal and also telling them when to go to sleep at night, um, and also telling them like what the weather's like the next day.
- 13:00SPEAKER_03: Um so it's really interesting seeing the personalization coming into it and how deep people go into that and how long they're willing to build out um their own personal kind of experience for themselves.
- 13:11SPEAKER_02: Well, it's it's interesting to hear you say that because one of the one of the things I've been seeing is just like adaptive software.
- 13:16SPEAKER_02: You know, a agents create adaptive software.
- 13:18SPEAKER_02: Um and I'm curious how you guys have been have you been experimenting with a agents um as a product because like they can do all because like Poof in its current instantiation isn't an agent.
- 13:30SPEAKER_02: It's probably it's using the agents to code, but it's not an agent.
- 13:34SPEAKER_02: And but it could be, you know, it could be that adaptive.
- 13:38SPEAKER_02: I mean, it's sort of like that between zone.
- 13:41SPEAKER_02: Um I'm curious what your thoughts are on that.
- 13:45SPEAKER_03: Yeah, um, it's it's like hard to ignore agents in their future.
- 13:50SPEAKER_03: Um we've been pretty focused on what the future of Poof looks like when agents are more in control and when maybe a user isn't always there to prompt it.
- 13:59SPEAKER_03: Um and we we have some things uh around accelerate time we'll be kind of showcasing around that realm as well.
- 14:07SPEAKER_03: Um that where Poof is very usable by agents and for agents, and um having an agent autonomously use Poof on its own is something that's a good thing.
- 14:18SPEAKER_02: Yeah, like VLI or something like that.
- 14:20SPEAKER_02: Yeah.
- 14:21SPEAKER_02: Um I would look like hook it up uh to some of the other tools I use and get that agent connected with Corbett's as well would be really cool to give it like some other APIs um that it can pay for.
- 14:33SPEAKER_02: Because it is a nice cross-section of users who have USDC RDM wallet and they like know how to use this stuff.
- 14:38SPEAKER_02: Um and so you don't have to worry about on-ramps and whatever.
- 14:41SPEAKER_02: You know, you can just give the agent money.
- 14:42SPEAKER_02: There's and this is where things have changed like so much in the last six months, you know, um, in the agent space.
- 14:49SPEAKER_02: It's really just kind of accelerated a lot.
- 14:53SPEAKER_02: Um what's the business model right now?
- 14:55SPEAKER_02: Like, how do you guys make money off of like as a business?
- 14:58SPEAKER_02: How are you sustained?
- 15:00SPEAKER_03: Yeah, so there's um a few sources of revenue for us.
- 15:03SPEAKER_03: Um, one is just AI credits.
- 15:05SPEAKER_03: Um, to build on Poof, you do have to pay for credits.
- 15:08SPEAKER_03: Um, and that helps us um pay for the AIs that are actually building your app.
- 15:13SPEAKER_03: Um and another one is we have um some small on-chain fees for when your app reaches production and is serving customers, um, when they're taking actions that actually store and um you can think of it similar to like Metaplex's fees on-chain.
- 15:27SPEAKER_03: Um we take a small, tiny um cut from uh some transactions that might go through us.
- 15:33SPEAKER_03: Um and yeah, those are the two ways that we monetize right now.
- 15:37SPEAKER_02: That makes total sense.
- 15:39SPEAKER_02: Yeah, pretty straightforward.
- 15:41SPEAKER_02: And have you guys raised any outside capital or are you just bootstrapping it?
- 15:45SPEAKER_02: Like what's the where what's the stage there?
- 15:48SPEAKER_03: We have uh we um we went through Alliance and we also have raised some other money, but we haven't announced that much yet.
- 15:54SPEAKER_03: So right now, uh we haven't really talked much about what we've raised yet.
- 15:59SPEAKER_02: And and like I I mean, how like I'm curious that you know the last couple weeks, Claude kind of struggled to put it gently.
- 16:10SPEAKER_02: Um how did that impact you guys?
- 16:13SPEAKER_03: Uh it's it's been tough, honestly.
- 16:15SPEAKER_03: Um a lot of our users they don't keep up with AI.
- 16:18SPEAKER_03: They don't even see that Claude and Opus is having these issues and like us downstream.
- 16:24SPEAKER_03: Yeah, exactly.
- 16:25SPEAKER_03: Yeah.
- 16:25SPEAKER_03: And and it feels really bad because like we are like it feels like we're doing them a disservice when really Opus is just performing way worse than it was.
- 16:34SPEAKER_03: Um so yeah, the last two weeks have been pretty tough in that respect.
- 16:38SPEAKER_03: Um what's cool is it did force us to find some really creative ways to make the experience better for users.
- 16:45SPEAKER_03: Um we made more AI models available.
- 16:47SPEAKER_03: We started auto routing between different models depending on the complexity of your requests.
- 16:52SPEAKER_03: And also that also helped us reduce the cost for everyone, where if we're routing to Quen instead of Claude for certain requests because it doesn't need that level of reasoning, then we get to lower your prices by like five times or more.
- 17:07SPEAKER_03: Um so yeah, it it's nice in a creative perspective, but it has been quite difficult to deal with um to keep everyone happy.
- 17:14SPEAKER_02: Yeah, we always have been like laughing internally on our end when like everybody wakes up at 9 a.m.
- 17:19SPEAKER_02: and like Claude just isn't is five is throwing 500 hairs and you're just like, oh, everyone just got to work, you know?
- 17:25SPEAKER_02: And it's giving me um, I was thinking the other day, it's giving me like 2021, 2022 Solana infrastructure vibes, you know, where shit was just breaking.
- 17:35SPEAKER_02: Um and I, you know, I I was the the founder, one of the co-founders of Olaplex back then, which I don't know if you knew that company, but we were like one of the original NFT marketplaces storefronts for artists, and we dealt with like the users didn't know that it wasn't us.
- 17:50SPEAKER_02: You know, it was really like an underlying issue with Solana um back then.
- 17:54SPEAKER_02: And obviously, like that's long gone now.
- 17:56SPEAKER_02: You know, the infrastructure's way better on Solana.
- 17:59SPEAKER_02: Um, but more it's funny how history just repeats itself um in any kind of transform transformative technology that requires a lot of infrastructure um and tools.
- 18:10SPEAKER_02: And it's what's been interesting, it's been interesting to watch the communications coming out of Anthropic and you know, feeling like a little bit gaslit.
- 18:18SPEAKER_02: I think they've not handled it that well, honestly.
- 18:22SPEAKER_02: I don't think it's gonna matter at all because their models are the best.
- 18:26SPEAKER_02: So, you know, like even like I'm still opening.
- 18:30SPEAKER_03: Yeah, you're still I'm still on cloud code every day, even though we're facing it.
- 18:35SPEAKER_00: They're also active on X too, which makes it much more like in your face.
- 18:38SPEAKER_00: Uh like it's and whatnot.
- 18:40SPEAKER_00: And yeah, no, I totally agree.
- 18:42SPEAKER_00: I think if you checked my like browser history, like in the top 10, status.enthroping.com is on that list.
- 18:48SPEAKER_02: Like it should not be on the list, but it is Yeah, we're all like, I guess we'll hop on you know a huddle right now and drink coffee together.
- 18:55SPEAKER_02: So it's just like what you're not doing anything until it's not 500.
- 19:00SPEAKER_02: Um So do you I'm curious, like, where are these apps living?
- 19:06SPEAKER_02: Are they like the front ends?
- 19:07SPEAKER_02: Are you deploying them to like AWS or something like that?
- 19:10SPEAKER_02: Like what's the infrastructure?
- 19:11SPEAKER_02: What's under the hood?
- 19:13SPEAKER_03: Yeah, so um we use Cloudflare for a majority of our apps that um are built.
- 19:18SPEAKER_03: So um whenever we're building a front end, we're hosting it on Cloudflare's R2, um, which is like their S3.
- 19:25SPEAKER_03: Um and your back end, we host it on Cloudflare workers.
- 19:29SPEAKER_03: Um for your actual like um on-chain indexing and things of that nature, we handle that for you.
- 19:36SPEAKER_03: And so we use various providers to help us out as well and to index your data and then put it into a database that you can actually use and kind of reason about.
- 19:46SPEAKER_03: Um and so that's our own on-premise kind of machines.
- 19:50SPEAKER_01: Um and yeah, that's the maturity.
- 19:53SPEAKER_02: That database is why are you doing the the the the the data?
- 19:57SPEAKER_02: So like I deploy an app, my people start using it, transacting on it, that's all getting indexed.
- 20:03SPEAKER_02: Like the app has probably an ID on chain, so you're like associating that ID with the app, and then um Triton or Helios or somebody is indexing that stuff for you, and you're sending that data to a database that's like just more easy for me to access via your UI or something like that, and that database is on-prem somewhere.
- 20:23SPEAKER_03: Yeah, it via our UI or via um us directly and uh or via our APIs.
- 20:30SPEAKER_03: And yeah, it it just takes everything that's defined for your app through our policy system and indexes it automatically.
- 20:38SPEAKER_03: Um and that can mean on-chain data that is still on-chain and stored on-chain, or it can be um on-chain data that like kind of just passed through on-chain and like a transfer that happened in the past.
- 20:50SPEAKER_03: Um, or off-chain data as well.
- 20:52SPEAKER_03: Like we store all of that for you.
- 20:53SPEAKER_02: What uh give me some examples of that.
- 20:55SPEAKER_02: Like, what's the policy stuff?
- 20:57SPEAKER_02: Like, what am I looking at if I log in and I'm like, Yeah, for sure.
- 21:00SPEAKER_03: Yeah.
- 21:01SPEAKER_03: Yeah, so um let's take like a super basic staking application, um or let's say like an escrow app where you have to deposit to some escrow in order to chat in some chat box.
- 21:16SPEAKER_03: Um, what you'll see is Hoof will generate a policy for you um where one we call them collection in your policy will be uh your escrow.
- 21:28SPEAKER_03: Um and then another collection in your policy will be the chat messages for your actual chat.
- 21:33SPEAKER_03: Um the escrow would be marked as on-chain because that needs to be on-chain to actually store um money.
- 21:40SPEAKER_03: And it will have some rules attached that say if someone wants to use this escrow, they have to deposit money here.
- 21:46SPEAKER_03: And this is how much money they have to deposit, like point one soul.
- 21:50SPEAKER_03: Um and then the chat messages can reference that on chain um escrow through its own rules.
- 21:59SPEAKER_03: So it can Have a little rule there that says if this user is deposited in that escrow, which is on-chain, um, then allow them to actually create a new um database entry for a chat message here inside the chat side.
- 22:13SPEAKER_03: Um so it's pretty much a JSON structure that defines both your on-chain and off-chain database structure.
- 22:19SPEAKER_03: And uh the on-chain side is of course the more interesting part there.
- 22:23SPEAKER_03: Um where that same structure that we use for like it looks like a database schema, except that it has extra kind of configurations that let you do on-chain actions.
- 22:33SPEAKER_02: Why are you storing that on-prem?
- 22:34SPEAKER_02: Why not in the cloud?
- 22:36SPEAKER_03: Well, sorry, when I said on-prem, I meant like that we have our own servers.
- 22:40SPEAKER_03: I mean, it's we depending on the app, there are different like databases in the back end for it.
- 22:47SPEAKER_03: Uh we've actually been experimenting with that.
- 22:49SPEAKER_03: That was the wrong wording.
- 22:51SPEAKER_02: That's okay, yeah.
- 22:51SPEAKER_02: It just that's why I was confused because I was like, yeah, you have you have like a rack somewhere, you know, and it's not me to say on-prem.
- 22:58SPEAKER_02: No, it's okay.
- 22:58SPEAKER_02: It's okay.
- 22:59SPEAKER_02: Um so what are you guys most excited about for the next like couple months?
- 23:08SPEAKER_03: Um I'm super excited about bringing um agentic poof to the world.
- 23:14SPEAKER_03: Um, and so that is uh, as you alluded to, a CLI and some more features there.
- 23:20SPEAKER_03: Um where right now Poof is very good for iterating on your product and kind of building from scratch.
- 23:27SPEAKER_03: And I'm really excited for when we allow an agent to do that on its own and get to production.
- 23:32SPEAKER_03: And we've been working pretty hard on that now.
- 23:34SPEAKER_03: And it's pretty incredible that um you can build something that's complex and on-chain in this way.
- 23:41SPEAKER_01: Um and so I'm I'm really excited for that future.
- 23:47SPEAKER_00: Definitely excited about that as well.
- 23:48SPEAKER_00: But I'm just I'm also just excited to see some of these applications that we see some of our users building to kind of come out there.
- 23:55SPEAKER_00: Some some users are building some pretty neat things.
- 23:58SPEAKER_00: Um, and I'm I guess I'm talking more about those perfectionist users that I alluded to earlier.
- 24:03SPEAKER_00: And I'm excited some of these apps to hit the market and just see what they do.
- 24:07SPEAKER_00: I remember when I first started building apps and I was putting out there, I remember like that thrill and that experience.
- 24:11SPEAKER_00: So I'm excited for these users to kind of go through that.
- 24:14SPEAKER_00: And I'm fairly confident that some of these apps will do well based off what what I'm reading and what I'm seeing.
- 24:20SPEAKER_00: So I'm excited for that success for them and and for these people just to be rewarded for for uh being early and and just knowing what people want and what to build.
- 24:32SPEAKER_00: Super cool.
- 24:32SPEAKER_02: And as we get closer to the top of the show, I always I always ask this uh question, but what have I not asked you that I should have asked?
- 24:46SPEAKER_00: I just look at questions around like maybe just AI in general.
- 24:49SPEAKER_00: Mythos, we can talk about mythos.
- 24:51SPEAKER_02: Well, I was thinking about mythos because I was kind of like, you know, that seems pretty important to what you're doing specifically because the security is such a big issue.
- 25:02SPEAKER_02: And, you know, mythos for people who haven't seen, they uh is from Anthropic, it's a new model.
- 25:08SPEAKER_02: They didn't release it because it was so good at finding vulnerabilities on the internet.
- 25:12SPEAKER_02: Like it found vulnerabilities in um, I forget like BSD or what is the thing that like software has been on the internet for like 27 years, you know?
- 25:20SPEAKER_02: Um and they were like, oh shit, it's so good at this, we like can't release it.
- 25:25SPEAKER_01: Um what's your take?
- 25:29SPEAKER_00: I think we just uh we have no choice but to believe them.
- 25:33SPEAKER_00: Like you just have no choice.
- 25:34SPEAKER_00: Logically it makes sense, right?
- 25:35SPEAKER_00: Like the models are getting better and they're just naturally gonna get better at finding these exploits.
- 25:40SPEAKER_00: And it's nice that they're doing this due diligence and working with these um companies out there to kind of uh help them fix their vulnerabilities, but that clearly doesn't scale.
- 25:50SPEAKER_00: So I'm curious how it's gonna play out, right?
- 25:52SPEAKER_00: Like in theory, MIPPIL should lead to more software production.
- 25:56SPEAKER_00: So you can't just like take it away and fix these vulnerabilities.
- 25:59SPEAKER_00: So I'm curious to just see how that all kind of uh pans out and and whatnot.
- 26:04SPEAKER_00: But I think yeah, we'll we'll we'll just see.
- 26:07SPEAKER_00: Yeah, I'm I'm tapped in a lot and I'm following it uh really closely.
- 26:11SPEAKER_00: And um cool.
- 26:13SPEAKER_00: I'm also very curious about the pricing of these this model and what that's gonna look like.
- 26:17SPEAKER_00: So very expensive.
- 26:19SPEAKER_03: Yeah, very expensive up there.
- 26:21SPEAKER_03: Um one thing that shocked me is Claude just released like a new desktop app.
- 26:26SPEAKER_03: And if you try it out, it's quite buggy.
- 26:28SPEAKER_03: And I'm like surprised if they have mythos in house.
- 26:31SPEAKER_03: Like, how did this end up like coming out this way?
- 26:34SPEAKER_03: Um Yeah, so uh I'm also very curious of like what that really looks like.
- 26:39SPEAKER_03: But regardless if Mythos is or isn't what it's being made out to be, we're not far from that future of zero-day vulnerabilities being capable to be found by AI.
- 26:50SPEAKER_03: Um, and so that is a really interesting future we're like finally reaching now.
- 26:54SPEAKER_02: Well, I think you know it brings it brings up a really important topic to me, which is open source software and the future of open source.
- 27:00SPEAKER_02: And we just saw Cal Cal uh AI, which is a software that we use for um scheduling, and they're open source.
- 27:09SPEAKER_02: It was one of the big reasons why I wanted to work with them.
- 27:12SPEAKER_02: Um, and they now are closed sourcing, they're being like core closed or whatever.
- 27:17SPEAKER_02: And a huge reason from the founder was that AI is making it easier than ever to hack a service.
- 27:25SPEAKER_02: And I think the other thing is like the opposite is true.
- 27:27SPEAKER_02: I mean, I don't really think anything has changed about the dynamic of open source versus closed source.
- 27:33SPEAKER_02: My opinion is that if you're open sourced, you can catch these things much faster.
- 27:37SPEAKER_02: Like, why don't they have Claude agents like 24-7 looking for vulnerabilities?
- 27:45SPEAKER_02: And we do it all the time at our company.
- 27:47SPEAKER_02: Like we find, you know, we run our agent teams and they are like patching things and uncovering like whatever various degrees of issues and bugs and whatnot.
- 27:58SPEAKER_02: And I think when Mythos comes out, it's even better, right, at doing that.
- 28:02SPEAKER_02: Um so something does need to change.
- 28:05SPEAKER_02: Like the pair there has been a really big paradigm shift in what it means to be an open source company.
- 28:10SPEAKER_03: Um uh the lag time between making some code and it actually going open source might need to be increased so that you have time for those agents to run.
- 28:21SPEAKER_02: Sure.
- 28:22SPEAKER_02: Um but also like um in a world where people can just use Poof to make anything, right?
- 28:28SPEAKER_02: Like what's the point?
- 28:30SPEAKER_02: You know, I mean it it does, it is like right now, there's just like what was it that App Store was getting so many submissions?
- 28:38SPEAKER_02: Um, GitHub, another great example, right?
- 28:40SPEAKER_02: Like the volume of pull requests and GitHub's being created is like through the frickin' roof.
- 28:47SPEAKER_02: Um and it does, again, going back to 2021, 22, Solana vibes of like people the change is getting stress test, right?
- 28:55SPEAKER_02: Like I think all these different infrastructure providers are getting just massively stress test.
- 29:01SPEAKER_02: Um and the reality is people are just shipping, right?
- 29:06SPEAKER_02: Like that, and that's what they're gonna do.
- 29:08SPEAKER_02: That's just what they're gonna do.
- 29:10SPEAKER_03: Um ship faster and way more people can ship.
- 29:13SPEAKER_03: It's just never ended really.
- 29:15SPEAKER_02: I mean, we saw it with OpenClaw, right?
- 29:16SPEAKER_02: Like the first version of OpenClaw that took off, like a million people instantly started using it, um, was really insecure.
- 29:24SPEAKER_02: It was incredibly insecure and risky, and no one cared.
- 29:27SPEAKER_02: Like, no, that they didn't know or didn't care because they were too caught up in the utility.
- 29:33SPEAKER_02: And you know, maybe that will catch up with us.
- 29:35SPEAKER_02: I don't know.
- 29:35SPEAKER_02: We'll see.
- 29:37SPEAKER_02: We'll see.
- 29:37SPEAKER_02: It might not.
- 29:38SPEAKER_03: Yeah, we'll see.
- 29:39SPEAKER_03: Let's hope it doesn't, I guess.
- 29:41SPEAKER_02: But um, well, guys, thanks so much for coming on the show.
- 29:43SPEAKER_02: I really appreciate your time.
- 29:44SPEAKER_02: It's been great having you.
- 29:46SPEAKER_03: Yeah, thanks for having us.