Watch Episode
Episode Description
In this episode, Corey and Jessie talk with fubuloubu, founder of ApeWorX and Silverback. We discuss our history in the ecosystem, security, and what developer tooling and automation frameworks bring to the table. We also dive into what role AI plays in the future development of blockchains and automation tooling.
Links:
- Ape Framework Github: https://github.com/ApeWorX
- Website: https://www.apeworx.io
- X: https://x.com/fubuloubu
Transcript
Click to expand transcript
It’s a Bitcoin podcast. The only one that’s making your money you market. So listen to the Bitcoin. >> That intro never gets old. It doesn’t. I really do love it. Welcome to the Bitcoin podcast. Today we have an interview with the doggo. What do you want to be called since you’re not doing video today? What do you want to be called so I don’t dox you? >> Uh I don’t know. Senior doggo doggy or fubulu if you can if you can pronounce that. I can never pronounce that. I stopped trying. Been around for a long time. >> Um, >> how long is a long time? >> I don’t know. I mean, I’ I’ve been >> 2018. >> Oh, okay. >> We’ve been We’ve been friends since then. Yeah. Acquaintances, >> colleagues >> since the first security summit that we did. >> Oh, you go to security? Those are those those are fun. He was so fun fact we are uh admins slashcoers of all of the assets for um ETH security. >> Oh okay. >> We don’t do anything. It’s kind of a self modif like self self uh what is the word? Uh moderated community. Could probably use some more help. >> The one in Telegram, right? >> Uhhuh. >> Yeah. >> And there’s a Discord, too. >> Oh, gotcha. And there’s some web assets like I have F security and secure F domains. >> Yeah, it’s like five over 5,000 people slashbots slashcammers slash probably DPRK operatives. And >> currently there are in in the in the Telegram channel there are 5,244 members uh of which 933 are online right now. Um, it’s where a lot of the I mean in the longest time it was the place where like most of the security community like the the real ones would chat and introduce bugs and ask about vulnerabilities and then serve as somewhat of a war room. >> And then it got popular because it was basically free to join. And so people would just sit there and monitor and just wait for Sam to say ask a question about anything basically or someone to ask a question about a given platform and then they just go balls in on trying to exploit it because they’re just waiting for the security community to figure out what was wrong with it. So they would use that as an alpha to go and figure out how to exploit it >> for short the token. >> Yeah. So, >> it’s still I mean where I mean still a really great place where a conversation happens. How would you how would you how would you describe it today? >> Um it’s like a watering hole maybe. Um I think there’s still like conversations that sometimes happen. It’s mostly a little like back and forth uh kind of thing. I think it’s primary usefulness has always been like connecting people in the community like projects and auditors together. That’s kind of how we originally devised it is um because it was like very difficult to be like do you know anyone from this project and it’s like no I don’t but maybe if we ask in this group we can find the right person to connect to. Um, so I think it’s done very well from that perspective and it’s really helped um, like make those connections and sometimes like when active exploits happen like it shows up there first just because people share things um, and become aware of like what’s going on that way. >> Reasonable. Yeah, >> I’ve seen some pretty good arguments in there, too. >> A lot of arguments. >> I’ve uh I tried to like spend some time on it. I don’t know, a year and a half ago, whatever. Trying to >> make it a little more formalized and neat because right now it’s just like a single Telegram channel. It’s just it’s just some some days just a wall of text and you get lost. I made a Discord, but channels nobody liked it. >> A lot of lot of spammers didn’t get moderated. So, Discord still exists, but no one really uses it that much. that that’s kind of a nice like well kind of a funny thing is that all the spammers join there and like they do in other large groups except for the the bigger problem is that’s the group that has all the people that moderates all of the like anti-pam stuff. So it’s just like kind of a honeypot. We’re like collecting all these bad links and like scammer links and stuff now. >> It is true. But that’s not why we brought you on. Uh we wanted to talk about something else. So, why don’t you first off introduce yourself, say hello, say where you come from, what you do, and then uh we’ll dive into Ape Works and Silver Bag. >> Sure. Yeah. So, my story in crypto goes back all the way to 2017. Um, and you know, as we’ve been talking, got involved early on in like the security context, trying to improve the security of smart contract development and systems applications built on these technologies, which um, as everyone knows, security is like a huge problem and a huge deal to get right um, when you’re dealing with people’s money. So, um, that’s what originally bought me in. I I’ve been doing for for years uh for the years after I got involved like did a lot of education initially did a lot of um like consulting work eventually ended up doing some bigger projects um such as V2 which is like how most people got to know me in like the 2020 2021 time frame. Um so helped uh lead the effort on uh writing the v2 for year and then from there uh started a works because I really wanted to drastically improve the Python dev tool story in Ethereum because it was um you know so many stories of people that like tried to use the tooling the Python tooling that was available ran into a wall end up moving to some of the better maintained tooling in the JavaScript the ecosystem or now we have foundry which has kind of taken over a lot of people’s initial uh experiences getting involved in in development. Um APE has really evolved to be like a full stack solution uh for developing in Python especially for automation and data science use cases which I think are really important. Um, and from that we’ve built a a newer projected, this was a couple years ago still, but a newer project called Silverback, which is like a very easy to use uh automation uh framework. So you can like design bots and stuff uh really quickly and easily and kind of takes advantage of Ape as a infrastructure framework um to build around. So you get all the plugins and stuff for free. you get all access to like different ape uh signers and all that kind of stuff. So you can run it in the cloud. Um yeah, so Silverback has been really cool. A lot of people really like it and it’s uh taken advantage of all the stuff we built with Ape. >> You didn’t you didn’t mention you you were also maybe still are I don’t know a a large proponent of Viper as opposed to using SLing languages. So like you were the one kind of pushing the I don’t know if you’re the one, but certainly one of the ones pushing the Viper narrative against against Solidity for a long time. >> Yeah. Yeah. Like in 2018, I I also kind of uh was like the lead maintainer of Viper for a couple of years until maybe three years ago when Charles finally took it over and started pushing the language uh further. Uh so big shout out to Charles uh for doing that. Um he uh um you know the language at that time was used in the first iteration of unis swap unis swap v1 but then uh like until curve came along nobody else was really using it for that um so uh it’s definitely evolved a whole lot like I used it in writing yarn v2 and um I think laido uses it for some of the system contracts that they have and and different people have started to use viper So, it’s come a long way and it’s really like held its own as um as you know the second most popular language to use for smart contract development. Now, >> how easy is it to use inside of Ape? >> Um, so yeah, we have like a plug-in system in Ape. Um, we support both Solidity and Viper, but I really wanted to kind of be Viper first with a lot of the documentation and tutorials we were doing. So, um there’s a little bit of a preference for that, but it supports both pretty easily. Um and I’ve been trying to develop even more uh functionality into the framework, so you can do things like Foundry style tests and uh fuzzing and stuff a lot more easily. >> I’m going to pull up some some websites here. Might as well scroll through things while you’re talking since someone can see your face. >> Yeah. How do you like you said you started Ape Works as like a you wanted you basically you wanted Python tooling. Um why what was wrong with like so Foundry has definitely taken over a good portion of the developer tooling and Rust George a stupid good job with with all of that. >> Why did you want a really good job promoting it? Um >> yeah, >> I think uh like at the time um there was Brownie uh which I had really loved but um it kind of needed some tender love and care uh to push it forwards which is how we came up with Ape. Um I was working with the the Brownie developer who um at the time wasn’t looking to like really push that project further. I think there’s another maintainer now that has been maintaining Brownie for some of the legacy use cases that people have for it. Um, but at the time it was like hard hat, truffle, which now truffle is obviously not there anymore. >> You remember? >> Yeah. Um so like APE is kind of the hard hat for Python I think of it and we’ve been >> trying to develop inwards as like thinking holistically as like a full stack uh full stack application framework versus just like focusing on the development experience of smart contracts. Um but there’s some pros and cons as well. Um, I think the primary um issue that most people don’t consider when they’re like 100% Foundry workflow is that you’re not actually replicating the like live experience of integrating with the smart contracts on a real, you know, blockchain. like everything is done in foundry test and foundry scripts which execute within the EVM context and don’t necessarily replicate the experience of like going on ether scan or like actually interacting with it right and like oh yeah I added uh I started fooling around with AI and added uh claude skills um try to get in there but um yeah so like um Ape replicates the entire experience um of using it. That’s why it’s really been amazing for like public onchain stuff and bot development and all this kind of uh things that we built around it. Um whereas like the the testing and scripting experience uh Foundry kind of changed the game. Well, it really adopted the uh dev tools approach of like write everything as a smart contract compile like a test so that a single a test is a single transaction instead of a series of transactions and calls like it is in ape and then it’ll be much faster because you’re executing a single transaction. You can get everything from that. Um, so I wanted to build that style of testing into Ape. So you can have both the integration experience of using it how it’s meant to be used in in production and you can have that like unit testing like um foundry test style experience where you know everything’s bundled in a single transaction get really fast with it. Uh it’s especially important with like fuzzing where when you do a fuzzing test case um the test cases like maybe look simple but you’re actually executing it hundreds of times in a row. Um so like those speed differences really show up when you’re starting to fuzz because you know 500 times uh you know a few second slowdown ended up being minutes and potentially hours of overhead that you’re adding right. So I think Foundry fuzzing and uh like invariant testing all that um made it feel a lot smoother and and brought that developer experience of doing that enhanced verification uh down to a level that most developers can use. And I think that’s like the primary thing I give it credit for doing. >> So when people built on it, I know I’m in the I’m in the chat with you guys. Uh there’s a lot of people in there building stuff. What do you What do you think has been some of the more fun stuff that people have started to play with Silverback and and deploy bots and because like you’re right because like for those that don’t know um >> the way that EVMs work, they don’t trigger anything. You have to manually click on stuff once it’s available to click on. So a smart contract will once the like say you build a smart contract that’s like you know if this then that right if this happens it doesn’t automatically do that you need someone to go make a transaction that clicks on that makes that happen that makes sense right so like there’s no automation of things more often than not I think what you’re trying to build is a way to react to those situations when things happen and then submit transactions to make them happen is Is that a >> maybe maybe more concretely like in system design of a smart contract you’re building a new smart contract system you’re kind of playing with incentives right you want to create incentives such that certain actions get done at the proper times um in order to keep the system operating uh the way it was designed to be operated right so you’ll give rewards if people do you know liquidations right like you built a a lending protocol like a Well, the natural incentive is people want to lo uh like uh lend out their assets, earn yield on that. There’s other people that have an incentive where they want to take the loans and kind of leverage with them. So those are those two like um user stories are like pretty developed in in themselves. But then like in order to keep the system actually functioning, you need liquidations to occur so that the system stays capitalized. So um you design a liquidation system that’s great but like who knows to call it how do they know to call it those are those are the problems that you start to run into and like you know a is obviously a very popular system so there’s like an existing set of people that are performing liquidations on mainet and you don’t have to like worry about that but like how do you advertise to the set of autom automated actors that there’s opportunities and I think that has been the most interesting thing we’ve seen with Silverback is like people designing new protocols. They want to expose the opportunities they’re creating to this. I mean it’s really not people anymore. It’s like automated actors, generalized front runners. These are bots. these are things that people don’t have full control of or even awareness of sometimes of what those like actors are doing. Um, how do you advertise this new protocol to those bots or agents or whatever? Um, so that they know that they can interact with it and participate in a new system you’ve built and make money from that because ultimately that’s what they want to do is make money. um or that’s the the agent’s design goal is like find some way to make money. Um so like Silverback is a really interesting way to um develop some of those automations because it’s inherently like >> on and offchain like partnership, right? So you have >> a smart contract system like people do actions like take a loan um or take some leverage whatever those create events and events are kind of underutilized in the Ethereum ecosystem but those events should trigger like callbacks or trigger like off-chain actions uh or trigger like off-chain watchers to like go back on chain and like trigger another action right and you get this feedback loop this like economic feedback loop built Silverback is a way to like design uh these types of behaviors that you need your system to function correctly um to design like easy to run uh implementations of those behaviors that you need and expose that to your community and get people like operating like day one in that way. Um, and we’ve seen people develop projects where because they built a silverback bot that bundles along with their, you know, day one deployment and when they’re going public on mainet and they’re running the bots or they’re having their community run the bots alongside that, those bots are eating well on the first day, right? Because nobody is aware of this new opportunity that you just created. And then eventually like the fact that you know your bots are eating well starts to attract attention from other like other more generalized uh actors in the system that are just watching for any opportunity and then it starts getting them involved. So you start to get like, you know, better activity on this protocol you just launched without doing like a ton of Twitter marketing and like all this like stuff to get attention to the people that are no longer actually the ones who will interact with it. Right. >> That’s what you mean. I’m trying to think of like an interesting way of like would you imagine it’s it’s most useful to like in conjunction of building some underlying protocol that enables new opportunities on chain for people to do things. >> Yeah. >> To deploy a number of bots alongside of that that do those things right. Because like ostensibly you’re not building something you’re not building some project on on chain that doesn’t fit in somewhere. >> Yeah. >> Like there’s a purpose for it. So it would make sense to build a bot that does that purpose that automatically starts doing it the moment that you go live. Correct. >> Exactly. Or it even serves as a template. Right. So, like maybe you build the the basic version of that behavior that you want to see done and you have this bot serve as a template and somebody else could come along and riff >> riff on that bot and make it better or find more um like more niche behavior for them to optimize into. Um some cases it’s just the generalized making it um you know like it’s like the dark forest analogy like imagine you created this new protocol you put in the dark forest nobody’s aware that it exists until you send that screaming piglet running through the woods where it’s like oh these animals can now see this pig running around right >> yeah it’s a it’s a great way of like showing to the community of pe potential users >> that here are a number of examples of using this stuff in different context. Take that. >> Yeah. >> One, these are just good examples. Go do them if you want. Or two, use this as an example to then fit into your niche. >> Yeah. And there’s other examples too, like you can use it for your own monitoring purposes. Like let’s say you want to monitor um the activity of your protocol, you can build a bot for that. Um you know, you have like an alerting bot or you can have like a guardian bot. like if it sees some sort of bad activity happen, it can like trigger a um like trigger a stop or something in the protocol. Like let’s say you’re early on and you want to have that kind of thing. Um it could also be like a part of your infrastructure. Um like some people have used it for indexing use cases which are popular and there exists other tools for that. Um but Silverback kind of bundles it all together. So you get, you know, the ability to read and and listen for events. You get the ability to act and interact both onchain and offchain. Let’s say you want to update a telegram group. It’s very easy to do that in Python. Let’s say you want to um ping Discord or or send a tweet, you can do that with Python pretty easily. Um, so there’s like many ways of like bundling the onchain and the off-chain worlds together so that you can get like more like full like more full feeling applications or like more reactive applications to user behaviors uh using the system. >> You had said that you like this particular for like uh data science use cases. >> Yeah. >> Why? >> Oh well not so but ape. Um so ape we built for um enhanced like data science use cases to be able to grab data and like one example where we’re integrating that into silverback is uh like when you start a bot like let’s say you have you need some amount of past data in order to like fill in the behavior right it’s called backfilling the data in the bot so we can use ape to like backfill uh events or backfill logs uh uh or blocks or something um into the bot. And then uh eventually wanted to introduce a back testing functionality into silverbacks where you can take your bot and run it across uh like let’s say a million blocks in the past. Like you want to see what the performance of this bot would be over, you know, a million blocks of of data. you can kind of run that back test on the bot, see the performance, see if like you’re improving or uh reducing the performance by making a particular change. Like those are the kind of things we want to add using ape as like the data system to to grab that >> new chrons. >> Yeah, we just added those and the metric tasks are actually really interesting. So we’ve had metrics since overback for a while. metrics are basically just when you return something from a function handler. Um you have different ways of of naming the metrics that get produced. Um but basically you know metrics are uh numeric signals that get produced by running the task. Um and then we introduce something called metric callbacks which is uh when you produce a metric you can have the runner read the metric and decide to execute another task based on that metric. So when the when the metric is produced by that task above the current price task um that metric can trigger an action on current price that will do something else. Um and more interestingly you can add filters. You can have like greater than less than filters on um on that sort you know let’s say rebalance the pool when the the delta gets above you know 0.05 05, right? Um, so you can kind of have these like very interesting reactive workflows uh using metrics and that’s been really really cool and I’ve used that in like my trading bot scenarios where you know every hour I measure the price and then off the price I measure RSI signal and then off RSI signal you know when RSI is above.7 I’ll do you know start to sell when it’s below.3 I’ll start to buy like that kind of thing you It really gives you like, you know, whatever logic you want. If this happened, it’s I mean it’s it’s it’s if this then that for crypto, right? It’s >> Yeah. >> And have you ever seen >> Yeah. You can have it trigger Telegram updates like oh like here’s an opportunity like it doesn’t necessarily have to do onchain actions, but the fact that it can do onchain actions. It can like send calls, send transactions. >> Yeah. you just built that that that connectivity and >> uh contextual awareness, right? Like the the ability to get that contextual awareness quickly >> via ape into this so that you can react and then >> do on additional onchain events accordingly, right? >> Yeah. Um, have you like in the process of building all of this, have you pulled inspiration or guidance, knowledge, wisdom from the platforms that do this type of stuff, just not on chain? Um, that’s a good question. Um, I don’t know. I since I got started getting involved in crypto, I’ve always >> like been really interesting interested in the automation >> use case in crypto. Like these things run 24/7, 365, no sleep, no nothing, right? They have to work all the time. >> Um because the the blockchain’s on all the time and you have to sleep and you have to, you know, live your life. You can’t just be sitting there on on online all the time, right? >> I tried. Didn’t work. >> Yeah, it doesn’t work very well. At least not for a long enough period of time. So, like this is my attempt in like introducing that like introducing autonomous control systems that can work while you sleep essentially. Um, and there are I know probably plenty of examples of that and traditional tech and and stuff. it. I got I guess a lot of my experience doing this from aerospace because that’s where I used to work before I got involved in crypto. So like I think of them as control systems. >> Yeah. primarily like the ways of introducing inner and outer loops to you. I think of like smart contract system design as like you’re designing an economic control system and that is your like inner controller, but that creates a lot of external behaviors and external incentives and like things you need to add an outer loop to. And I think a silver bag is that outer loop. >> How’s the web app? Can you show me some uh examples of like people like what the dashboard would look like or people playing with um because you have like a cloud platform basically like where do you deploy to? You just deployed to Ape Works eventually. >> Yeah, we the the front end design has been a real work in progress. >> Good process, man. >> Yeah. So, it’s it’s getting there. But, um you can actually use almost the entirety of the platform from the command line using the silverback command. Um, so like a lot of the examples are like using the SDK’s uh command line functionality to actually create bot clusters and deploy bots and manage them. And I’ve been doing that for like the past year to manage things like grants and uh other stuff that we have going on. Even the payments for the platform itself is using SilverVEC to um react to the payment events and create the infrastructure and it can actually autonomously scale itself which is really crazy. >> Explain that. >> But um yeah so um the platform manages uh Kubernetes based clusters um for deploying these bots onto. So we built some infrastructure to manage those bots at scale like you know see when they have an issue, restart them, send alerts, all that kind of stuff. Uh still an active work in progress but um so you can actually run these things all the time uh and trust that they’re running. Um and uh as part of like you know we don’t do the work for free so there’s a payment aspect to it and the payment uh the payments are using something called apay which is like a subscription payments protocol we built um and then basically we have silverback bots that listen to these payment events like when you create a new stream you update a stream you cancel a stream those trigger actions on our infrastructure force silverback where let’s say you create a new stream it responds by creating the cluster resources to give you access to run more bots right and uh over time like it can you know like right now today if you wanted to create an hour’s worth of time on uh on the product we can uh it’ll get started I won’t have to do anything it’ll just kind of work in the background and that’s that Right. >> A >> you got to sign up. >> Yeah, I might do that right now. >> I was going to ask uh I was going to ask uh what what can you pay in? Is there like um uh >> stable coins right now? >> Oh, nice. Okay. >> Yeah, nice. >> So, we have three deployments of a pay on optimism, base, and arbitrum, which have been our three most popular networks that people use ape on. And um you know we’re still building um the functionality that I think justifies the price uh because that’s some feedback that people have given us. But the like right now you can pay for just an hour if you wanted to try it and like really get uh fooling around with it. Um and eventually like the idea would be something like this like a like a wizzywig kind of editor. Um >> I take credit for that by the way. If you end up building this, I’m taking credit. I’m taking credit for it because I told you >> I’m actually going a step further and I don’t want you to have to build anything manually. I want you to talk to the computer and tell it what you want and have it figure it out. >> That’s where I’m trying to get >> I want I don’t know if I’m trying to think of the right infrastructure framework for that. So what’s the best context to provide an AI such that they could do that quickly? I know your skills are an attempt to do that >> a step in that direction is that you just you know hook up some MCP server give it a skill or whatever the current >> So we have an MCP server for the platform um so you can connect to the clusters and have the AI like manage bots deploy new bots and all that kind of stuff. I’ve been using that as well. Um but that’s for like managing and deployments. Um the skills and all that stuff is still like developer focused. How do you make new bots and how do you make bots that do what you want? Um since I think like what most people have learned is that LMS are really good at writing code. Um so if we can like constrain >> if you can give it the right context and constrain it. >> Yeah. So eventually the idea would be to have a store because you might not know what you want, but to have a store where there are pre-published bots that do certain things you might want to do and then you talk to the computer and the computer can have an MCP tool that searches the store, gives you some bots um that you could run and it has you know uh some description of the bot like what you know it’s kind of like a skill thing actually. is like loading into context like uh this is what this is good for this is how you use it like that kind of information and then through that conversation with the bot you say like oh I want to do an ARB trade on this and it’s like okay well here are some options um which one do you think sounds good to you and it’s like oh I kind of like this one and uh it’ll walk you through the setup and actually deployment of the bot so that all you do is pay for the cluster talk to your AI get it to deploy a bot for you and then you’re off and running. You just have to configure some things um >> in the Kubernetes cluster like you know some secrets and some other things like that, right? Uh then you’re off to the races. >> I think yeah, I’ve always liked these and I think that’s definitely the best way to go if you’re trying to build like a production system like something like a bot that’s going to get deployed and stay there and manage. >> I’ve always enjoyed this interface. >> Yeah. Um, for those who are just listening, it’s like the it’s f.build build or if you’re familiar with Comfy UI for generative AI video and image editing that interface um this is the best for education and I think like sharing >> scratch >> if you want to share and educate people and get people to understand like how these things actually fit together visually it’s I’ve never found anything that allows people to understand and share these things so quickly. >> Yeah, it’s a great Then from here you can like this is effectively just visualized code. You can then generate a bot based off of whatever whatever workflow is, right? >> Yeah. Yeah. This could be like a way to do it. Um I I think even like some of like cloud code works like just give it context and it uses our docs, it uses the skills that I’ve been fooling around with like it uses all that stuff to get you >> um to get you like the proper code that you can go try. It’s still like very developer focused. You got to know if it’s doing the right thing properly. Um, but that’s the idea is like to help a little bit more educated uh developers like build like unique and interesting bots. There’s a store for them to publish. there’s a way to um you know buy the ebot through the store, buy access uh or rent the bot I guess it’s more appropriate like rent the bot from the store uh run it on your own but we’ll cluster and then eventually you would have like a a profit sharing type deal set up with the platform so that you know there’s an incentive to do that >> go ahead say that again do you ever do you ever play Runescape? Uh, no. >> Okay. Uh, I had conversations with Corey about the automation to create the black markets in video games and Runescape had a pretty >> uh, they went through like several changes, but they had this bot called RSBot. It was like a bot written in Java >> and they essentially had a store where you could uh, buy and sell scripts that were written for particular resource harvesting efforts. >> Yeah. uh written by, you know, other developers and uh there was like a starring system so you could see what was most lucrative and you could see like the conversion of the in-game currency to real dollar value. It seems like I haven’t played with again I haven’t played with solar back but yeah it’s like a similar like sort of uh automate extraction or but this is like a more generalized automation tool >> but it seems like >> yeah like as you said >> there might be some developers that develop like some generic behavior like you know you want to do an ARPA you want to do something with unis swap or something where it’s like generic and then there I think there is a great use case for um a protocol. You’re developing a new protocol and there’s some things you want done like liquidations and there’s a pretty easy way to get involved in doing that and you just would publish it on the store and maybe you don’t even charge anything, right? It’s just there on the store. People can use it easily and for free and and um that gives you, you know, value to your protocol. It gets people using your protocol. So, um, it’s just a way to make that accessible to a non-technical user base, right, very easily and very safely. Um, and I think ultimately like that’s a really nice way to to scale the ecosystem because a lot of the stuff these days and has been for years a decade now is very developer heavy, very like um, you know, if you’re not a developer, you’re like a heavy heavy uh, user. You’re like a power user and you know things that most people are not going to know. Um, so to make that available to others, you kind of have to package it really nicely, right? >> Yeah. >> And even this AI stuff, it’s like a lot of the AI stuff, you’re still like giving people too deep access into it to like really get it done. Right. Right. >> Yeah. You have to like pre-chew their food for them to give them the best experience so they word of mouth share like, “Hey, this was a great product.” That’s why I like that’s why I like no code solutions so much, right? You just look at visually >> like you can go as deep as you want. You can code and that’s why I think that’s why in has been so popularized. >> Yeah. >> Because they were able to provide all this utility in a way that makes people who aren’t the deep like like these specialists that aren’t developers, they they know they know of opportunities, but they can’t code things themselves because they’re just actively looking for those opportunities. they can then at least explore something, get it most of the way, if not all the way through no code solutions and then share it. >> Yeah. And I think the best no code solution is talking to your computer. >> Touche. I want to do that. We are we are on the path to that, but yeah, >> it’s more expensive, I guess. I mean, >> it can get it can get expensive if you >> if you just willy-nilly go after it, right? Even if you just Yeah. >> use the max plans. What? 200 a month for the various ones. No. No. Max. >> I've had really good >> 100. Oh, is that it? >> I've had really good results using just Claude Desktop with the MCP tool for managing the bots with the 20 a month tier. Like, >> okay. >> It’s worked very well. Like I ask it, you know, you’re connected to the cluster. Tell me how things are going. >> And it goes and does the MCP calls to tell me the status of the bots. If there’s a issue with one, it shows me the a summary of what the log information is saying. It’s like, “Hey, this one’s got a problem. I think you need to go fix this.” Like, it’s very good. >> Where is this MCP server? >> Uh, the Silverback SDK uh all the way at the top there. >> Oh, just the SDK. >> Yeah, it has an MCP in the CLI for so you have to go up. Yeah. >> Right there. Like >> this. >> Yeah. So if you go in there, go into B uh silverback and then there’s um then uh cluster and then in there is MCP. >> Wow, damn >> export that. So like once you you have to have the cluster set up for the MCP to do anything because it it talks to the back end of the cluster. >> Okay. Um but yeah like essentially just streamlines a whole lot of information about how to properly integrate and you can start bots, stop bots, see their logs, update them, remove them, you know like it has full authority over the bots. It has less authority over configuration like variables and and like how to access things which will kind of >> streamline in other ways. So it’s less dangerous than like randomly x-filling like API keys and stuff. >> Yes. >> Um but yeah, like that’s the idea is like um the silverback SDK has been really great for local development. So you can uh build a bot like super easily. just, you know, UV tool install silverback and you’re off to the races in building a bot and it it will take you like less than an hour to like really get something going. Um, and then once you’re ready for that, there’s a build command which allows you to containerize it. Uh, we also have >> that nice you just you can just generate Docker files for it. >> Yeah. And then once those Docker files are created, uh, you just share them in some way. So basically you can publish them on a public GitHub repository and anyone can run them. Uh you can publish them on your own you know private GHCR and then if you add credentials to the cluster you can access them. Eventually the store will be like a semi-permissioned system where it’s like the store is a private registry um so that you can control access but like it’s public because it’s like uh always uh each cluster has access to that right um and then just like the payment will give you access to uh to running bots from there if the payment’s required for that particular bot. uh that that needs more thought. But um once you set up a cluster, you’ll you can set it up all purely from the command line. I think you have to initially sign up with the product through the front end. Um so you have to sign up and then create a workspace that way, but after that everything else you can do for the command line. Um which is good for developers. Um then obviously we’ll make it more uh more accessible to like non-technical users as we get more features built in. >> Yeah. Yeah. Guess stick to your core audience first. What’s for that part? Like where you all at? Like what are you trying to do? Like what where you at in I guess >> um product development, organizational development, like what are you trying to do at this point? What’s your next milestone? >> Um yeah. So, um, for Silverback, um, trying to get the UI framework up and running so I can actually like very quickly iterate on changes with that. Um, so that it’ll look nice and clean and cohesive and not a mess like it kind of feels like right now. Um, you know, it serves the goal of getting you signed up, but yeah, I want it to be a nice experience and an experience you can use on the fly on your phone or whatever, right? That’s really the goal. >> Mhm. >> Um, and then, uh, real-time dashboarding is kind of the next actual feature we want to add there. Like you mentioned, like you want to see the metrics of the bot as it’s executing. You want to see, you know, is it making money, is it not? Like those are the kind of information you need to see from that dashboard. Then once we get that real-time dashboard uh working uh to the store and like exposing uh your bots to other people and making you know an economic incentive to to build the bots, right? >> Are you going to do like that through some sort of profit sharing like if other people run your bot like rent your bot? Okay, got it. >> Yeah. So like everyone will have like a profit sharing uh like if you your bot gets used everyone will have like profit sharing like certain bots you could charge you know in addition like like everyone you can share bots through the um yeah it’s there’s a lot of variety too but you could have like anything that’s public anyone can run you know >> kind of like a WordPress store right like you’re like what Yeah, >> we’re doing templates >> or uh themes, right? You can you can either get your own theme or you can buy someone’s theme or and whatever that type of But for bots, >> yeah, they have >> Oh, sorry. Go ahead. >> Uh you can have like more interesting payment models in them too like the operation of the bot. You could charge like a uh you know like a management fee or performance >> performance inside the bot. >> Yeah. So like you can get really interesting with that kind of stuff. Yeah, I’m thinking about that because if you were to it’d be nice to say you come up with a protocol that does something and you know a number of um I don’t know what the right term is but like use cases for these things you just build the bot build the bot for those use cases deploy them to the store >> and then say like you can do this or just you know rent the bot or buy the bot on Napeworks and call it a day. >> Yeah. You could also kind of like replicate the hedge fund structure. Like imagine you’re just in the person making a particular strategy. You get it really nice. You add your you slap your little performance fee on there and then like you make that accessible for anyone to run. So it’s like you can kind of outsource the strategy development in a way and then like as a as a fund you might be just curating those strategies built by other people finding the ones that you like um monitoring their performance you know when the performance gets bad like kick it out find a new one um like try to control risk and all that kind of stuff like the the stuff you get paid the management fee to do like that can just be your job right >> did you ever try Quantopian This has like a lot of parallels, but it’s like instead of being trady, it’s like crypto. >> Yeah. I have not No, I’ve not heard of that. >> Okay. Yeah. It allowed you to run basically scripts that uh so every everything was was um Okay. So, like how do I explain? So, first they allowed you trading data going back several decades for free just with a starter account. And then they allowed uh custom scripts like in terms of maybe custom I don’t know back testing um algorithms that people like tweaked that are based on um I don’t know some some some book theory. Um and then I think one of the one of the things that they added eventually was like a leaderboard. So like whoever had created the most um >> lucrative bots, you know, I think like there were hedge funds that would purchase these these essentially these scripts um from people who are building them. >> Is that like is that the >> very similar? Yeah. >> Okay, got it. >> And the leaderboard concept could kind of be integrated into like the bot store, right? Like yeah, when you go to the Google store, it’s like here’s the top app of the day like thousand downloads. It’s similar thing you can do. But then the game changes, right? Because if everybody’s running the most profitable bot, then you just need to figure out how to trade against that bot. >> Yeah. Arthur, >> I think I think with the story, it would have to have like some sort of like maybe capacity thing that we can introduce. Like there are certain bots anyone can run. There’s certain bots that are like a limited number of licenses available to run it and it’s like, you know, 50 or 100 >> and you can do like some more interesting like NFT pricing for those or something. Mhm. >> Um you can get fun with it, you know. >> Yeah. Yeah. It’s you really just you’re you’re subjugated to the limits of >> paying for things in blockchain, which is basically open >> open opening this game. You can do whatever you want. >> Yeah. >> It’s interesting. I was just thinking about something about But that’s also like it doesn’t always have to be trading bots, right? There’s a lot of cool things you can do. >> Even if you’re just monitoring, right? like you can build an entire monitoring platform on >> sure >> looking at different types of wallets and activity within wallets and then >> yeah the the capability of white labeling the system I think is like really interesting so like the the bot store that’s kind of like the most consumerf facing aspect of the product but like for example the apay subscription system I could white label that out and then anyone can add cryptos payments to their to their SAS that they’re building um very easily by going to like you know the in-house white labeled version of silverbeck for AP and then you just like >> configure that pay for it um and then you’re good to go right like I can add a new bot like based on a payment um to to the existing cluster that I white labeled and then like I’m running my own instance of that or like chaosnet which is really kind of where this idea came from could be another white labelled instance of Silverback. Um, and it’s like you’re creating a bespoke cluster to execute that simulation um, for like a couple days to a week, right? Uh, you’re like deploying hundreds of bots at once to like perform your like beta test of your entire system integration. Um, like there’s these different aspects that can be done with it. I’ve thought of several use cases for that. There’s like that is a very interesting like developer use case for the product like to be able to kind of white label your own instance of that to like grow and scale dynamically and do dynamic payments and all that kind of stuff, you know. >> All right, man. Let’s wrap it up. Um an hour at this point. Is there anything that you wish I would have asked you you wanted to talk about that I didn’t? >> Uh, that’s a good question. No, I think uh we covered a lot here today. Just um yeah, if anyone’s interested in trying it, it’s free to try. Silverback is just oneclick download UV tool, install Silverback, good to go. Um you might have to add some extra plugins. Um there’s there’s ways to do that with UV, but um we wanted to make it like oneclick install and get ready to run some bots really easily. >> Cool. >> How do we uh I mean um I’m in a insiders channel with people who are playing with this. Is that something that’s relatively closed or is that something people can join if they’re willing to like give feedback? Like say for instance they download, they try things. Where do they give you feedback? >> Uh yeah, that channel’s been great. Um that’s open. Uh we’re trying to keep it to people that are just actively doing stuff. Um so please join if you’re you’re really wanting to to play with it. >> Can you send me an invite because I’m going to start trying to use this for rebalancing the LP. >> Uh oh, absolutely. >> On unis swap for >> I’ll send you I’ll send you one now, Jesse. >> Okay, cool. >> Yeah, it is open. Just trying to keep it to people that are actually doing stuff instead of lurkers. >> Gotcha. you’ve been added. >> My uh my question is uh is what you do actually difficult? And you can interpret that however you’d like. >> What do I do actually difficult? >> Yeah. Is what you is what you I guess is what you do actually difficult? Yeah. So like we get a range of people. people who are in crypto as developers, people who are, you know, >> really really random people like I apparently I thought the lady who uh what happened other time. >> Yeah. What was that situation? >> I don’t remember but like she she took offense to the fact like is what you because like the original form of the question was like is what you do hard? >> Yeah. >> Like what do you mean? Yeah. What is it hard? And she’s like you almost like >> I think she was like an influencer. I think that was the problem. Yeah, what because what she did wasn’t hard. And so like >> Yeah. I think it’s hard sometimes. I think it’s too hard. Like um doing an open- source company is hard. Um trying to push the entire ecosystem to use Python more has been hard, uh but worth it and rewarding. Um there’s been so many people over the years that have just complained about the state of Python tooling and I think we really helped on that. Um building a business in crypto is hard because there’s so few moes and it’s like you talked about the data and like back testing with that traditional quant framework, right? >> All the data is accessible open source in crypto. So it’s like that’s not really a moat that you can easily have. Um, and so, you know, trying trying to build a product where it’s like not purely extractive and we’re trying to like um do the hard parts for people um so that they can focus on what they’re interested in doing the most, right? Like that’s I think a winning um uh strategy for a winning product long term is like not feel extractive, feel like you’re adding value to their experience. Uh, but doing that is really, really hard. And, and sometimes the people that do best in business are the people that are the loudest. So, I struggle with that a lot. Uh, I’m a bit more quiet than most, I guess. >> Yeah, you are. >> All right, man. Thanks for coming on. Really appreciate it. Always good to hear from you. Don’t get to see you this time, though, so I guess I’ll have to see you win. Devcon. >> Yeah. Yeah, I don’t I don’t know if I’m going to make it to the next one, but uh I missed you at the past one at Dev Connect in Bonus Areas. I was >> I didn’t go there. I was in I was in Spain. >> I know. I was looking at at you at that status event. I was looking for you. >> Oh, that’s right. That’s right. Yeah, I wanted to be there, but uh you know, duty calls. >> That’s great. You missed uh you missed some great steak. I’m sorry. >> Well, I had pretty good steak in Spain. It’s not bad. But yeah, so uh we have we have a number of Python developers who already do this and I’ve been they’ve been waiting for this episode so they can hear more about it. Um I’ll try and stick him on it as well and give give feedback. I have Jesse has ideas. I have ideas. I’m going to try and just get people at Logos to do some of this stuff for some of the more like monitoring stuff that we care about. But >> even thinking about >> I think >> deploying something bots that show the >> example use cases of those deployments. It’s really an interesting idea. >> Yeah. I think the um the hardest part is it’s a little abstract and a little difficult to explain the power of the system and like why you should be interested in it. And that’s another reason why I’ve been doing a lot of this like AI and like skills stuff is like I want it to be able to like rapidly prototype a demo for me. Like, oh, it’d be cool if you could do this and have it just like oneshot that example. And it doesn’t have to be complex. It just has to show like what you can do with it. And I think that’ll drive a lot of power of like, oh, I don’t have to sit here and fiddle with this for a week. It can do it in an hour. That’s amazing. That’d be fun to do a follow-up once we have a couple ideas on that working you coming on >> and then us trying to oneshot ideas and making it like on screen and seeing what happens. >> Yeah, like mob programming like like all of us just yell at the bot until it does what it wants. >> You just stream like a vibe coding session of building building silver bots. I think the problem is you need some sort of middle layer of meta tooling which is what a friend of mine started building and that like uses playright and uh basically MCP server with another LLM like GLM so that it’s like the second pair of eyes so it there’s like some some sort of recursive loop where one is doing the most of the development and that’s like claude code opus and then the other one’s kind of doing like uh supervising checks and tests. So it’s just yeah I think that meta tooling layer if if you can do that you could you have a chance at oneshotting as close as you can. >> Yeah. >> I’ve been really focused on improving the docs improving like skills so that it can get high quality results quickly like that. I think that’s really important for our entire stack is >> get one high quality results quickly with it. >> Cool. Yeah, >> let’s do that. Let’s plan on that next time. Be fun. >> Yeah. All right, man. Uh, appreciate it. >> Yeah, it’s good chatting.