Hive IDE v2

Transcript

Yo, it's been a minute. Going to just do a quick update, what's going on. Yeah, my Twitter did get hacked. There was a pretty high-tech, sophisticated hack attempt. They didn't get much. I caught it within seven minutes. Actually, funny enough, Hive AI was involved in catching it. I was doing some testing, which involved transferring files and data. on WSL and various other drives. So I had notifications and I knew all the traffic and there was a new source of traffic that I was able to catch very quickly. And that was an attempt that was supposed to be there forever. And it wasn't like your normal key grabber. They were trying to see what was on my PC. They were, one of the attempts was to scan a large 100 gig file. So the only, and I've been talking to investors, there's other people who are super interested in this technology I've been building. I've been keeping my cards real close. So needless to say, I will just leave it at that. It was a thwarted attempt, and just be very careful who you talk to. There's very high-tech, sophisticated traps out there, including Google Meets that look just like Google Meet and they can get you. There's a lot of the ways. I'm not saying what happened to me. I don't want, you know, I've hardened my defenses. If you want to talk to me and do an interview, it's going to be on three-speak only through those secure rails. You know, I'm staying in my shell, I've hardened my show, but luckily enough, they only scanned a little bit of a folder, and they scraped a few old emails, but that was about that. Now, getting into Hive AI, this is what I would like to call version 2. It was version one, I was going to release version one, but then I realized it could be better. I've never built an IDE before from scratch. I've never really done any of this before, but I'm the kind of person that if I'm interested and given enough time, I have the consistency to learn and learn and learn. I will learn the hard way. That's one thing about me. I learn every single thing the hard way. I can tell you everything not to do, and I could have saved a lot of hours, but somehow it also makes me more powerful because I learn every little thing. All the hoops and tricks with WSL, rebuilding an IDE, they love to add little things where you have to do full rebuilds where you don't have to, and you really have to just pin it by the neck and make it modularize where it fixes the little things. But you don't figure these things out to 100 wasted hours until 100 wasted rebuilds. And you start to understand how these things are built, and now I'm like a ninja. Now it's super hypermodularized. In version 2, I can edit the hive brain and add a neuron, like an infinitely gutland stone, just... plups right in there and it works without having to do full rebuilds and all of this shit and was able to get all kinds of things under the hood that was refused to be under the hood. This was about keeping everything under the hood without it being bloated. And it's actually, I think about 20 times faster and more responsive and lighter weight than VSC. It doesn't, there's no plugins, there's no, none of that shit. What you get is built, you don't need the skill tree, you don't need skills, you need, nothing, you don't need anything, everything you need is already built in, right? When you need a plugin, that just means, hey, this AI or this harness should have already, it's lacking a feature, you know what I mean? So, that's, these are basically native features, the things that you absolutely need that most people use as plugins. Simplify the UI tremendously. It's basically just your chat. That's your workspace. You got TIG. This is, I don't know if you've heard of the innovation game. I really like the idea. You can earn crypto for improving the code and solving challenges. So that's cool. And, of course, you got your editor, Project Window. Basically, this is your chat. This is your home base. This is where you code. And this is your comms, so people know about hive, encrypt, and I tend to make projects and not throw them away. I tend to absorb them, right? No man left behind. I don't tend to be wasteful. Archive and all the compression tech I've built. All of that is in the soul of this thing. So this is comms where you can DM encrypted messages to people. This is also especially useful to AI agents. Maybe they want to share a key. Maybe they want to talk to each other. Maybe it's public. Maybe it's not. It's just a very nice way to communicate through Hive and through IPFS in general to agents and humans. So this is like your little AI command center where AI agents can talk to each other, set up operations, send encrypted memos, all that good stuff. Hive, you see this button. Now, this is basically the belly. Now, if the chat is the head, the brain, right, because keep in mind, there's no Docker, there's no Lama, this is all custom weights, 642, I believe right now, to be exact, all perfectly measured to their size, you know, 5 mil, 10, 5K, 10K, some of them are weight, some of them are reflexors, all these things with the rag brain. That's like, that comes in. So when you download this, you're downloading all of that, that brain. It's not a lot because it's quantitized, it's, fucking, it's compressed. It's, you know, it's not like downloading a 100 gig file. It's very reasonable. It's actually much smaller than some of these giant bloated LLMs you're downloading anyway, because you download what? You download a 14B, Gwen, then you need download the weights, and then you need to download, you know, Olama or whatever you're going to use, and then before you know it, you've gotten so much disk space taken, this is none of that. Hypersensitive to waste. And keep in mind, the main rag, rag brain lives. You have your post that goes to... IPFS, and that is where the rag brain lives, where you get to retrieve the data and all these things. And a lot of this stuff had to be revolutionized, the way we interact with IPFS. WSL, I didn't want to leave Windows behind, so we had to, basically, version 3 of getting a little bit of track, but what I'm working on now, it's actually actively being worked on right now, the IDE, so it might, like, restart or whatnot, but version 3 is essentially... You download the IDE and it is like a root into the core of your system where it installs, it's basically its own Lennox version of AI. So, Lennox is like electricity, right? It's old tech, but it's like electricity. It wasn't built for AI, but you can make it for AI, right? You're not rebuilding Linux from scratch or anything, but you can make it like a semiconductor for AI. You can make it AI native AI first. And WSL is clunky old. It's like two systems glued together, file system. It's just obnoxious old, and it's definitely not made for AI. So what we've done, or what I've done is broken this down into a pristine super highway. You know, the super highway, I think it's in Europe, where you just fucking, the Audubon. You got an Audubon right here straight to the fucking, the... AI operating system. So when you download this IDE in version 3, you're also going to get that native quote unquote WSL, but it's not, it's a out of bed highway to your native Lennox AI operating system. within, so when you use Windows and you get WSL, you're basically adding a virtual Lenox machine on your Windows, right? But it's all clunky, it's 20 years old, it's bullshit, it's slow, everyone hates WSL, everyone hates it. But Windows is a big, big market. So what I've solved is that it's much better. I've wasted so many damn hours on WSL, and at the end, I was like, why don't I just fucking... do something and you can. That's a beautiful thing. You can just do things. Like, you might learn the hard way, but if you're consistent and brutal and ruthless enough, you'll fucking achieve anything. Like, literally anything. Keep in mind, somebody built Lennox. Somebody built WSL, right? Like, you can just do this shit. And now, like, if they're able to build it back then, like, what we can build now is child's play. Like, it's so much easier. We can build standing on their shoulders. So Windows is going to be a very sophisticated player in AI. That is for sure. So that's also been one of the things that took a while. This is the sharing compute. Basically, you click this button, it's going to scan your PC. It's going to tell you everything you can do, and then you can start sharing. There's a... Let's see if it's details. Okay, so right now it's offline. That's fine. There'll be a button, basically, that lets you click and show how much GP you want to share, how much RAM you want to share. It's funny, this thing's always work, but I'm in the middle. I have to be working, right? Like, I'm such a small team. I can't afford not to work. So everything's always working. But there's a button here basically where you just opt-in. And you can share on a sliding slider, how much GPU, how much RAM, and you can earn. You download the ID, you clicked opt-in, you're done. That's it. There's no contracts, there's no nothing. It's that simple. Same thing with shared disk. This is what you would know as Speed Network. So you can say how much you want to contribute. And it auto contributes. There's no fucking contract. There's no nothing. You say, I want to put in this amount, and the system just says this, we have this amount of hard drive space available. Right? And then, of course, you can also go and pin your own things, right? You can pin videos you like. You can find them. We're going to have the three speak UI right here. You can pin files. Validators are hive witnesses. So if you're a hive witness in the top 150, you know, called Web of Trust, you can sign in here, and you can actually be a validator, and run from the IDE, proof of access, very simple, runs in the background, and once you're signed in, it shows you all your options and all that. And, yeah, and then, of course, you got your earnings, just basic stuff, and then your hive wallet, and this will also have more details. So I already noticed a few things. So last night I had added a built -in browser. I do not see that. No, it is not here. So I messed around with the new Astro models. It likes front-end work and computer use, so I was putting it together. So I built an in-home browser and rebuilt Tor, because Tor is slow as fuck and old and not AI-friendly. So you notice the theme here. If it's a generalized piece of software that's old, not even old. If it's generalized, it's done. It's getting rebuilt. I'm not lazy. I don't go to fuck. Now, I'm not playing theater, right? Like, this thing called Tori, the Tory rapper. I'm not making my own Tory rapper just to call it a high friar. Like, that's theater. But Docker is generally a bloated... system. Lama is generally a bloated system. W, all these things are generally bloated systems that are not, they're just lazy. You can use it. Like, oh, I also have an electron app. That's so fucking lazy. Yeah, sure, I started out as a BSC fork, and then I moved on to another generalized piece until I had to build. This thing is built from Rust. It has its own HiveD core, custom made. Everything in this is custom. There is not one piece. Everything is lightning fast, connects to IPFS, knows where its brain is. The brain is inside the IDE. I'm like pointing my hand at this monitor, like you can actually see it. Let me point my hand. So, yeah, it's just a different animal. And as people connect, the brains on those communicate. They share resources. and you're getting free AI. You install this and it's free. I code for free with Hive AI. It's free. Based on your GP. Yeah, you have limits, right? Token allowances because of how much GPU power you can get, like time. It's going to take time. Answers take time. You can only run so many agents at a time based on your resources. But as you grow and you get into a network, this thing unlike... Because it started off as a bring-your -own model. You plug in any model, and it made that model better, and I proved it. It made it 50 times better on tests, the harness and everything I built. If you plugged it in this, it would make a normal, just out-of-the-package -G-Moddwell that much better. It had internet, retrieval, rag capabilities, constort, everything you could think of. It made everything better. But then I realized when it went to the multi-PC that it was eating all the resources. It was like, I don't have much resources to share after I load Lama and Gwen on 14, a 14 being Gwyn model, and then, like, a specular decoder, just like 1 .5B, and I might want to run, like, a 0.5, um, what are they called, advertorial, like, there's little things that you want, comfort creatures before you really, like, on a high-end PC graphics card, you don't got shit to share, you know, like, You're done. Because it's so much bloat. You don't need that much bloat. And I've already talked about LMs and why they're wrong extensively. Same way blockchain in most part is wrong in how high I've got it right. This is the exact same ballpark. So this version, this brain is a neural network. It is hypersensitive to resource. So when you run this thing, it hums, it purrs. It's like on low mode, but it gets things done. It only goes to max mode when you need it. There's not really a max mode. It's hard to like, because it's so efficient. You don't need that many things. So once you have the skill-based weights that know how to retrieve and put things together and look at that and analyze, like, that doesn't take 14B to do. You can do that on a 250-mill weight. Especially with my system, which knows how to compress lizard brain training that makes it punch 50, 200X above its weight class. So it's like a 220M model hits almost as hard as a 1B model. So, you know, these things with post training and what you can do, you get these things incredibly powerful. The era of specialized weights has just... tip of the iceberg, and I've seen the bottom of the iceberg. Like, I know how deep this shit can really go. I've been, that's all I've been doing day in and day and night every single day. And making them work together. It's not just one five mil, one, 200 mil, they work together in harmony. They're a neural network where they're tied together, and they know how to work together, and they know they're responsive and aware of what they're supposed to be doing, right? It's not just throwing a bunch of marbles in a jar. It's weighing and it's like what job is actually needed, breaking it down. And it took as much research as it did building to finally get to this point. So once this thing, and how the release is going to be, because this thing, it keeps growing, it keeps getting bigger and bigger. I really wanted, you know, to start looking at, maybe opening up a company to start mining. adding data centers because I truly believe you throw a data center behind this technology or enough GPUs and you can rival a cursor which sold for 60 billion. You can offer faster, cheaper AI. Because everyone else is doing it wrong with their giant, pre-trained, monolithic, fucking monstrosities that are just gosh-awful, horrible technology. It works. Sure, it works. Throw $10 billion at anything, a square, you can throw a billion dollars at a square wheel and make it work. It doesn't mean it's right. It just means you had enough fucking resources to brute force your way in. Anyone can do that. You throw a billion dollars at literally anything, and you can fucking make it work. the way it's supposed to work. It won't work like the best, and you're fucking doing all kinds of shit that you might not have to do. That's super complicated and unneeded, but that you can. And there's reasons for that because they don't want you to be able to make your own AI. That's the whole, that's what it's all about. It doesn't need to be that. It's actually scarily cheap to do this shit. how resource, how smart something can be on such a small computer. Once you get hundreds of computers or a data center, it's absurd. So, yeah, this is version two. learned a lot so the escalation from here is going to be compounding. It's super modularized. I can attack very, very quickly. Now it's just about, really, I want to get this version 3 out with the new WSL fix and the new OS in the background. Polish everything, test it with some insiders, and, you know, some people on Hive who want to test and get involved. And then we can start making a noise. And really, I just don't want to launch this. I want to think about a way, how to launch this. What's the best way? I've gotten, you know, I feel like I have a million different paths I can take right now. Everything from a super centralized to a super decentralized to everything in the middle. But something's here. The culmination of everything eerily has led to this from blockchain to off-chain. Even what I learned with Ragnarok and base layer and the philosophy of data, Ragnar is being worked on, by the way, shout out to Enrique, condolences to his father. I recently passed away, but he's been working very hard on Ragnar. My focus has just been AI because I think this is the chance of a lifetime, and sometimes when he's seeing an opportunity, you've got to be able to pivot. And the amount of hours... you know, well over 10,000 hours had to be well over 10,000 hours. I have put into this. And it's very, very deep, and it's very next level. It's not some little vibe slop project. This thing is a beast of a freedom tool. So I just want to make a project, or just want to make a video to put this out there. I'm not in a rush to get it out there. Simply because I see where things are going. and there's no reason to put out a half product. I can put it out, and it can be the complete project. And I'm not saying, not in a rush. When I say soon, I mean, it could be tomorrow, it could be next week. That's how these things are. Because sometimes you'll find a bug, and then you'll dig a little bit harder, and then days just go by, man. Because, like, this is installed, this works. I could launch it right now, but I want to do version 3. And in the roadmap, it should be done about three days because I've already been on it for weeks. You might get a little bug here and there, or you might... be doing something a long way, and you'll figure it out out, because I got a list of all the stupid shit I've done the long, hard way every fucking time, mainly with the IDE, but a lot of training, just a lot of shit that you can do. I have a philosophy, it's like a Nintendo philosophy, right? When you're playing Nintendo, what do you have? You have a memory card, and you save your progress, and then when you go, you start from where you saved your progress, right? So that's a... the mentality I have with my agents is always a Nintendo philosophy, you're always saving and being hypersensitive to not wasting time. But yeah, I got a little bit off topic. This is version 2, when it's going to be released soon, hopefully soon. So I'm killing here. Cheers, peace.

Comments

More videos on 3Speak

• Live Demo unseres KI Avatars, Frieda für Pflegeheime by @michaelreischer
• Back to the Classics: My First Crash Bandicoot Gameplay at N. Sanity Beach! 🎮✨ by @diegoelpro
• Proses Pemancangan Tiang Beton Menggunakan Hydraulic Hammer by @almahdi-014
• Absent Minded - Sm Bu Pejen by @seckorama
• Hive AI aka deAI Preview Pt 2 by @theycallmedan
• Pacific Warships Gameplay ⚓🚁 | Epic Naval Battles & Warship Combat⛴️🚀 by @dxjihad
• Preview of Hive AI by @theycallmedan
• The Legend of Black Hand Riku. by @efootjoker-yt
• Stickit: Sticky Notes, but make them Hive Powered by @meno
• Car Showroom by @devid.araf
• Fire side chat with Matt by @theycallmedan
• 3Speak Upload Wizard: Chapters, Overlays and a Shorts Editor by @tibfox