Daily AI Update: Agent OS, Fusion VS Fable 5 + Goldiebench...

So I wanted to kick this off by doing something new and what I'm planning to do is basically do a new daily update, breaking down everything that I'm working on, what I'm focused on right now, what's working, what's not working, what I've learned about AI, just kind of have a nice place to document and create a different style of content that I think with AI and AI avatar video content and everything else, like it's a good way to connect with people and, you know, yourselves who are watching right now. So I've actually kind of seen, for example, Nick Sariv, he was talking about doing this recently, I thought that's a fantastic idea. I've seen people like Jackie Chow as well, obviously being in the SEO space, his videos work really well and that's something I'm going to do, so if you have questions and you're watching these videos, drop them inside the comments every day and I promise you, I will answer every single one of your questions, so that just means that I can help you as much as I can, you get a lot of value out of watching them, you learn something new and also, if we're all learning and growing together, we're all going to win. So let's kick it off, I mean, first thing that I'm working on recently is really growing the AI profit board and this is my school community, we actually started from zero in February last year and now we're at the point where we are top five all time in our category, which is pretty nice, you can see here, making 173k MR, which is pretty nice, I mean, if we look over the last week, we are growing the most in the top 10 of all time, in the tech category, which is pretty shocking.

Honestly, when I grow a school community, I never expected to get anywhere near this, but it's exciting to see that and I think we've got a fantastic community, I mean, at this point, if you watch my videos, you're probably aware of it anyway, but at this point, we've got over 186 pages of testimonials, not 186 testimonials, 186 pages of testimonials, which absolutely blows me away, it's like, you know, we're actually growing something that's useful to people, people find value from it, people enjoy it and I think one of the things that people have loved recently is the agent operating system that we've built, just because it's fun, it's nice to use, it's pretty chill, it does some really cool stuff and also, recently, it's been so much fun to work with, like, let me show you an example, so if we go to my recent live streams on my main channel, you can see here that we've been doing, like, daily live streams for maybe three to four hours each time and quite often during the video, I'll just put on some music, you know, some AI generated music, which I've actually created with the agent OS system over here and then I'll put that on, usually a bit of drum and bass, that's my favorite to work to and then for like three or four hours, I'll answer questions, speak to the community, build out the agent OS, listen to what they want, build it in as well and it's kind of like building a full SaaS tool, so essentially what this is, is that we're building out the agent operating system inside Claude and then every day, I just add new features to it, so for example, if we have a look inside the agent OS, today, what we actually added is some new features with Hermes, where you can set up blank slate, which is pretty cool, that update just came out a few hours ago and then, you know, there's some really amazing stuff, like, for example, loop engineering that we've plugged in here, NotebookLM got a recent research update, so we added that in too and also Fusion, which is supposed to achieve like Fable 5 level intelligence, we plugged that in and we're actually doing some benchmark tests, actually, as I'm talking to you right now, we're running some new benchmark tests with Fusion and it's pretty mind-blowing, if you've never heard of Fusion, it's a way to have like five models work together on API and then you can essentially get the judge, who's also inside that API, if that makes sense, to judge all the answers from all the AR models that created an answer and then from there, it fuses the best answers together and so what that eventually means is that you have the best answer possible from those five models and on Draco scores, apparently, and I am testing this and it is pretty impressive so far, but apparently it can beat or achieve Fable 5 level intelligence, which is pretty mind-blowing in itself, which leads me to my next point, so we are generating something called GoldieBench and bear in mind, it's going to be less me, I mean, I'll keep you updated every day, but in the next video, it'll be like me answering your comments, me answering your questions, etc, if people leave questions, maybe they won't and then I'll just be talking to myself again, but if you have questions, I'll be answering those in the next one, but for this one, I'll just show you what I've been working on lately, so you can see over here that we have this leaderboard and it's called GoldieBench and basically from all my tests, you know, you saw that I do four-hour videos where I'm just like live streaming and showing you what I'm working on and everything else, so when I'm doing this, I get Claude to track everything that we built and then we put it into a benchmark table like you can see right here, so we can see visually which models perform the best, what we've created and whether it's actually good or not, so for example here you can see that we've been building out a Nordic Crypt game inside our leaderboard, so if we have a look, for example, we've got Kimi, Opus and GLM who have all had a crack at this and this is like a pretty cool way to just benchmark, okay, based on what we're building, which one creates the best stuff, you know, which one creates the most useful, so if we have a look at this one as well, this is another version of the same Crypt game, totally different, and then here's the final version, which is pretty awesome, but there's no enemies in there, as far as I'm aware I can't see any, but the game itself looks great, so you get the point, like we just run these tests, we build cool stuff, then we share what we build and then we compare and we see which one does the best thing, so here's another example, comparing each one and from what we've seen so far, like Fable 5 is, sorry, Fusion is creating some of the best stuff so far, which is pretty impressive in itself, so we're just building out cool stuff like this, testing them out, seeing which one performs the best, you can see Fusion's responses here on the left hand side, you can see what it's built, how it performs, etc, and then if we go to each individual model, this is, I can't even log into this website, so I can't direct what the content is, I can sort of tell Claude what to include, but it will basically create most of the content on this website, and then we'll talk about, okay, what is Fusion, how we use it inside our agent operating system, what we built with it, all the demos that we created with it as well, and then how it compares against every other model, so what are the scores, so you can see here, for example, it leads 7 to 1 versus GLM, 6 to 1 versus Grok, 4 to 1 versus Opus 4.8, and basically as a leaderboard, Claude will judge the outputs and see which one is the best, and so far I think Fusion is the best, which is pretty impressive in itself, so that's basically what we're building, how we're building it, how we're creating all this cool stuff, as you can see, and then we have these comparison pages as well, so that's been a super fun project, something else that we've been doing as well is like, every time I create a guide, or every time I create a new feature inside the agent OS, I will build out a guide, like you can see here, and this basically breaks down the whole process with nice diagrams, it's actually a skill I've got inside Claude, that just creates nice stuff, and yeah, what else I've been doing, so every time we get a question inside the Air Profit Boarding Community, I answer them, day to day, in a video tutorial on the live stream, and I think that's great as well, if you're trying to build a community, or you're trying to connect with your audience, fantastic way, I mean, number one, it's super easy to create content like that, and number two, it helps people, and if you're helping people, that always leads to good things, from what I've seen so far, something else that shocked me actually this week is like, we can get Claude, or any of these models, with a combination of hyperframes, Haygen, 11 labs, and Grok to generate the B-roll, and it puts it all together like you can see inside this video, so there's some really cool stuff with that, actually something else that's pretty interesting as well, and this kind of shocks me day to day, is if I create a video, for example, like Notebook LM, I cannot out-compete my avatar, my AI, and there's a lot of people who watch my AI avatar videos, and they're like, it's total trash, we hate it, you should never create videos like this, and there's a lot of people who watch my human videos, and they're like, I'm so glad you created a human video, but when you look at it, considering the amount of energy, time, skillset, and you know, let's be honest, it costs money to make these videos, when you look at it, this is a human video, and this is an avatar video, and the avatar video is outperforming the human video by 4 to 1, and I've seen this cycle repeat over and over again, every single day, and it shocks me, so it's very interesting, because you know, some people say, well, views don't matter, but actually, if you want to grow a channel, you kind of have to depend on views, and you might say, well, eventually, that burns out, but we've been using avatar videos for over a year, and our channel has grown dramatically as a response to that, far more than we were growing as a human channel, so in every way that I see it, avatar videos are outperforming human videos on every level, you know, like in terms of the growth of our community, in terms of the amount of people who watch, in terms of the amount of people who come back and keep watching, and in terms of just the overall growth of the channel, it's actually shocking, so something to bear in mind there as well, do I like it? You might be watching this and thinking, I wish Julian only made human videos like this, but I wish too, but it's like, if we did that, we would miss out on growing the business massively, and we would miss out on a lot of growth potential, just because of that, so it's actually shocking, I mean, you can see this comment right here, like, A, says, I'm impressed by your avatar, it's so realistic, and it is, it's super realistic, it's getting to the point now, I think only the voice needs to be improved, and I can't wait for a good update to come out on voice notes, whether it's just my voice is super monotone or not, I'm not sure, but basically, right now, when I listen to the voice, I'm like, ah, could be better, could be better, but this is the worst it will ever be, give it 3 months, give it 6 months, give it 12 months, who knows what's gonna happen next, but, what I will tell you is, this channel, you're watching right now, will be human only, only human content, so if you like this stuff, subscribe, keep watching, that's it from me, leave your comments on this video, or your questions, and I will cover them on the next one, happy to help with any questions you have, happy to go super deep on anything that you have as well, as long as it's relevant, and I will see you on the next one, in the meantime, if you want to get the agent operating system that I showed you, if you want to connect with me personally, if you want to connect with an amazing QP, if you want to connect with a community of 3,700 business owners, like us, and get me to help you with anything that you need, because you can DM me, or you can ask inside the community, either way I'm gonna help you, you can get all of this inside the airpowerful boardroom, so the agent OS is inside this section, as you can see, you can get the latest version right there, and then inside the calendar, you can jump on weekly coaching calls, inside the map, you can meet people in your local area, who are building with AI, just like you, and I hope to see you on the next one, cheers for watching, bye bye.