NEW OpenClaw Update is MASSIVE! (Sub-agent Flow)

All right, folks. So, Open Cloud just dumped a new release on this. It's just uh >> massive, >> massive, massive new release and we want to talk about the new features that are all in this version 225. So, we made a presentation and we tested it out prior to making this video. Um, and one of the key um improvements I would say there's two key improvements. >> One is heartbeat DM delivery. >> Yep. uh this actually applies to you because you actually been constantly asking it to reply to you and make sure everything's okay and checking up on you. But the other uh very big update is the sub aent delivery feature. And um I actually we can go into this in a little bit of detail because sub aent delivery uh is overhauled, but there's still some tweaks that we found were particular usely useful in this update that um we'll just relay our experience with it. So, if you guys want like watching AI content, we do real world AI content, not like, oh, let's make $10,000 SEO, whatever, but real deliverables uh on here. So, make sure to stay tuned, subscribe to this channel, uh, whatever. Let's get started. So, first and foremost, okay, open clock key updates. Uh, 225. Yes, we just talked about the key features. The key one that I would really want to say is, you know, other than gateway hardening and stuff and security, that's always important. uh I would say is the sub agent delivery feature. >> So um sub aents are really interesting because they actually spawn sub agents automatically a lot of times. >> Not for mine. >> Not for yours. >> Yeah. Not for mine. Miniax doesn't. Not for me. Not unless I explicitly tell it to. >> You have to explicitly tell it to and then once you tell it to spin up sub agents, it will do it more often because it actually um figured it out. So let me give you an example here. So you know if we made this presentation, right? So I said, "Hey, you know, uh, look what's in the newest open claw." Um, and he did the research. So this is kind of like a single agent. >> Um, actually wait, no, actually wait, actually incorrect. He actually spun up a sub agent to handle the task for me. So this is like um with Stark, the the AI that I've the open claw that I've already trained, he already did the sub agent to get uh what is new um here kind of what's new. Oh, he's back. He's like digested this. And by the way guys, if you don't know how to update your open claw, you just tell it. Um I said open uh I said after the open claw update, can you digest what um and what people are saying on X? I'm interested in sub agent improvement. What actually changed? And also update yourself. Just a little line like this also update yourself, [laughter] you know, just just by the way. >> Yeah, don't assume anything. Just update yourself. Please please >> please just update yourself. Right. So he he came back, he updated uh here's a sub agent research. This is what we're going to talk about in detail. and he spun sub agents to do a quite a few tasks for the research right then more importantly um we were building this presentation yet again real world use case right we want to build a presentation for you guys and we have a previous presentation there already >> so I said hey why don't you update that so he automatically went and used sub agents to do so I said can you uh can you update our presentation to include the recent stuff and he also suspun the sub agent automatically to do so yet again if your uh if your bot's not doing sub agents It should. >> Yeah. Right after they update, immediately test sub agents. >> Yeah. Test sub agents. Uh why sub agents are really useful too u rather than getting your main agent to do this um is because a sub agent doesn't need all the context >> because Stark's been working me with me on a day-to-day basis. >> He he's he his his brain is full of other things like crypto updates or um you know how to build a car, how to wash a car. you only want why you want to spin up the sub agent is because the sub aent has a very clean context, right? It's just all it has to do is like go do some research, go update the presentation. So that's why the sub agents are really powerful. >> Now you can already see 2.5 is a little bit um better because >> um there were times when the sub agents actually timed out. So this is actually even during making of the presentation >> uh during the update because the tasks were too complicated, it actually timed out. So you can actually see the new uh uh open claw at work here where it now tells you when um it's uh it times out. So before this it's like it tries to do the sub agent, it vanishes and you use you never see it again. I' I've experienced this before where like the agents the sub agent's gone. I'm like Stark, >> you know, you told me to wait five minutes, >> right? Yeah. >> And then where's the stuff? >> Yeah. >> Right. And I'm still waiting. I'm still I'm still waiting. I'm like yo yo yo yo yo yo yo. and then you know uh he doesn't tell me what's going on. So now they changed the pro procedure the whole mechanism so it tells you oh the sub agents done with their work. So I think that's that's just kind of a very big visibility update on their end. >> In terms of intelligence nothing really improved yet again because we're still using the same AI models. Nothing's really changed. It's just the process the procedure is a little bit more vocal about what is good and bad. M >> the last thing I want to say here um just for on this slide is that it also tracks the crown jobs. So you were saying on your bot uh it actually sometimes doesn't even run the um cron jobs. >> It doesn't run the crown jobs and >> or maybe it failed. You don't know. >> Yeah. And if it does run the crown job, it's at a wrong time. >> Right. So Right. Right. >> Uh um >> that might be your time settings on your PC. That might on your VPS. >> But yeah. >> Oh yeah. Yes. You're right. Because it's on a different time zone, right? >> Okay. >> Yeah. So um but the key thing is sometimes you don't know why it failed. Oh yeah. >> Right. So which is why they changed it. So um it's a a lot better and there's a better cleanup system for it too. >> I'll give you another live example of um this work. So it rebuilt the presentation and because we had a previous presentation there I said just stack on top of it. Yeah. >> Right. And this actually turns out to be a really large task. So he u he actually said that their um the sub agent left the file trunicated. It's cut off midway. Let me do this myself. So, you can actually see that I would say the sub aent feature is still not fully fully baked yet. Um, I would still use sub agents because it has less context. But you can already see that um Stark was kind of double-checking on the work, right? Um, and making sure the sub agent completes his tasks. >> Um, Stark also tells me now that oh, what I'm asking is a very big task. So, I said I had a previous presentation about OpenCloud updates. You guys probably watched the video, but um I kind of wanted the new information to come first and I wanted to change the order of the slides and it was actually a really hard task for Stark to complete. So now Stark is a lot more vocal about oh yeah it's not it's not um there it's not finished and then the sub um there's sub agents delivering different tasks. Then I realized, oh, because this is so hard, I started to tell Stark, you know what? Uh, is it a good idea to kind of split these tasks up, you know, uh, should we also set up sub agents for art, >> right? So, I'm kind of expanding how I'm using my sub aents, too, just kind of inspired by this uh, newest update. So, kind of breaking down this large task of building this presentation for me, >> uh, spin up different sub agents who do different things. So one is for the research you know what's with open claw right you got the structure you got all this then you have um the the sub agent that can write the presentation but also I was trying to add a sub agent to uh fix any problems there are so sometimes you know some slides um you know might just get cut off I constantly see here like even on this slide it was cut off yeah right so my future sub agent will scan to make sure it's not cut off um before what's happening so that's kind of the biggest update uh in terms of real world practice of these sub agents are just much more clear of what it's doing. So it's much more functional. >> But I what I would like to see is maybe names for the sub agent because how do I know if my orchestrator agent and sub agent are in fact different agents? How do I know that Stark himself is, >> you know, gaslighting again and saying, "Oh, I deployed a sub agent." But it turns out [snorts] he's the one doing the work. >> I did I did that earlier. I named my sub agents and I realized, you know what, I don't care for names, but I think maybe each person's different, but I think you should try that like to see which one's activated >> um to see what it's going. I mean, technically speaking, these sub agents are all running Opus 4.6 or whatever model that you want on the back end. So technically, they're not different people. We just structure it like that. But you are good in the sense that we should see like more optics and I think you can tell Stark that like yo give some names to your sub agents and then we'll we'll give them a little bit of personalities to there. I think that could work, right? >> So So that's kind of what's new. Um to catch up as well, we didn't do a video on um oh just an Android feature that I didn't use Android. So yet again, sorry guys, I um there's a new Android update, but I don't use it. So I'll be honest, I didn't really um use it. So that was kind of the other big feature on here and then okay and then some uh some documentation updates. So those are kind of the key ones. Uh the previous ones was the kilo provider. Um that was actually quite interesting if you want to use different like save money and use different >> um providers and the vision I haven't tested out yet. This is actually a very big feature that I want to do a video on where uh moonshot kimmy vision can understand video content and I think that will be pretty cool to um do a video on. But those are kind of the key drops that um that were released previously. Um and then yeah, I think there's not much not too much previously. A lot of a lot of security updates. What I would also say that's interesting um after browsing through their updates is that they have a lot of people working on this. This is why open claw is being developed very quickly. It's you know contary you know we we thought it would be closed claw after Peter Steinberg got um like well he he got hired by open AI and we know AI open AI is very closed source right but the project itself seems to be at this current point lots of people contributing to it so that seems to be really cool that they have a strong community really building on every day which is yet again why we kind of want to play with it every day and you know test out the new features. There's also a lot of updates for just communication side. So if you guys use like Telegram, Discord, the visibility should be improved as well. >> But yeah, that's around it for the updates. Any other comments, suggestions? >> Uh, none so far, but I would like to get that sub agent >> running >> and give me some feedback as well. Like we did a pre uh pre preview test, but you know, there's going to be more to do. So with that guys, thank you guys so much for watching. We al we all read all your comments down below. There's a lot of new suggestions coming in, so keep them coming. Uh, we are really here to provide you solutions, real solutions, and not just hype. So, with that, guys, thank you guys so much for watching. See you next one.