NEMOCLAW... NVIDIA is going ALL IN on OpenClaw

All right. So, you remember that whole open claw thing? Is anybody still using that thing? Oh, yeah. Everybody. And now to pour even more fuel onto the fire. Jensen Huang of Nvidia says the following to a packed arena at the GTC, the Nvidia's conference. He says, "Every single company in the world today has to have an openclaw strategy. " All right, so that's kind of a bold claim. What does he mean? First of all, he called OpenClaw the most popular open-source project in the history of humanity. Here's his framing. He's saying Mac, Windows, Linux, those are the operating systems for PCs, for personal computers. Open Claw is the operating system for personal AI. By the way, if you've been watching this channel, that should not come as a surprise. We've been talking about this idea for quite a bit. AI will be its own operating system. You're not going to be using 50 million apps. websites. You're going to have some sort of a chat interface, whether it's a voice or you type it in, and your agent's going to go out there and do stuff for you. Here's Jensen Huang's thesis for what's coming next. He's saying every single SAS company, right? So that's a software as a service will in the future become a gas company as he put it. If you're wondering what a gas stands for, it stands for agents as a service. So just in case you didn't catch that, SAS is SAS, software as a service. He's saying in the future they're going to be agents as a service. The acronym of which would be a AAA, but we're not going to pronounce it that way. We're going to say a gas like agents as a service. Just in case you see that spelled somewhere and then decide to use it in a polite conversation, I'm just warning you it's not pronounced the same way it's spelled. But moving right along, the big point of Jensen Hang's speech is that the Agentic Revolution isn't coming. It's here and Nvidia will be building the infrastructure layer for it. So, really quick 10-second primer on what Open CI is for those of you that missed it. I'm sure most of you have heard about it. The problem is there's a lot of these news and applications and AI things that come out that make a lot of noise when they first appear, but then some time goes by and you kind of stop hearing about it. For me, a simple way to tell how impactful an AI application is like how often do I use it? Do I keep coming back to it or not so much? Here's the thing. Open Claw for me from day one got just kind of deeply embedded in my life. I use it I mean almost every single day. There might be a few days here and there where I don't touch it, but they're rare. I use it for health related stuff, for finance related stuff, for business related stuff, for building various code and websites, software applications. There's some things I use it for that you don't even want to know about. And in the last few weeks, I've actually been setting up friends and family with their own little OpenClaw accounts. Or not even accounts, I just install it on some old piece of hardware they have lying around. Usually, you can squeeze quite a few agents on one piece of hardware, like if you have a mini PC or an old laptop that's been lying around collecting dust. I just do a Linux install on it and just throw a couple agents on there. And I've been kind of handing it out to a few members of, you know, the friends and family group, so to speak. And I got to say, people get hooked. People get borderline violent if that thing ever goes down. They want to keep using it. They want to have access to it. Now I of course explain the issues with it, the security issues, the various things that could go wrong and in fact that has always been kind of like the biggest problem with these openclaw agents. This was the thing that really prevented widescale adoption. They can very easily screw things up and leak your information online and do all sorts of stuff and then apologize for it later. Like you're absolutely right. I should not have leaked all your personal stuff onto the dark web. You're right. you're correct in being frustrated. This is very frustrating and yes, I should not have done that, right? As you're there on the floor like having a meltdown like what did you do? But here's where invidia comes in. So Nvidia comes in and says, "Hey, we're going to solve these security issues and we're going to make open claw be secure and safe enough to be useful for enterprise uses. " All right. So what exactly are they doing? So first and foremost, they're introducing Nemo Claw. So Nemoclaw is not a competitor to OpenClaw. It's sort of an enterprise wrapper around it. So you have OpenClaw at the center, Nemoclaw kind of surrounding so to speak to use for enterprise use cases. Okay. So Neimoclaw is Nvidia's stack that makes OpenClaw enterprise ready. So it's a oneline command file just like OpenClaw is now and it adds three things that OpenClaw does not have natively. It has privacy controls. So it's policy based data routing. It has various security guard rails. So it takes the AI agent that kind of sandboxes it. So it's not allowed to do certain things outside that sandbox. And of course they have their own opensource models. So now you're able to run OpenClaw with Nvidia's open source models. You're able to run them locally. So Nvidia has a whole host of models called Neotron and they are getting very good. Nvidia has in fact quite a bit of things as open source. They are a pretty big contributor to the AI open-source ecosystem. All right. So you can think of openclaw as the engine and you kind

of think of Nemoclaw as the security cage around it so to speak plus a local brain. So a local model that can run specific tasks in the future maybe any task but for the time being you probably want the latest and greatest models running kind of like the high lever tasks but there's tons of stuff that can be run on AI models that aren't quite as advanced that you can run locally and that's exactly what Neoclaw plus OpenClaw that's exactly what they do. Okay, so I know there's like a million different names to learn here, right? So, so far we have Nemo Claw. Neatron models are those Nvidia open source local models and there's now also open shell. So, this is that new open- source runtime that can host these agents. It sandboxes the agents preventing what it can do outside, what kind of a data can send back and forth. So, this open shell enforces as Jensen said it, it enforces the companies, the enterprises their policy. So what the agents can access, what has to stay local, what can go to the cloud and the privacy router figures out if it stays and runs locally on the Neotron sort of local model or it can go up to the cloud depending on what data you're working with. If you're a big enterprise, there's tons of stuff that you can't be sending somewhere else to some third party to a cloud. It has to stay there on a local machine. And of course, there's some stuff that could you could get data, you can run other models that could do research online, get data back, etc. that can be run in the cloud. So part of this open shell is that data router. It figures out what has to stay here, what can be sent elsewhere and it routes the data to those locations as needed. Again, based on the organization on the org level policies that are set and of course this works, you know, the cloud stuff works with any agent, not just Nvidia's, but you can use OpenAI, Anthropic, Google, XAI, whatever. So this is kind of a big deal because this is what open claw has been mostly the biggest complaint that people have. The biggest criticism is its security. A few weeks back there was this kind of I don't want to say it was hilarious but it was a little bit hilarious. There was a meta AI alignment researcher and she was experimenting with open clot and seeing if it could sort her emails for her and while it was running in a sandbox environment it was performing great right it was following all the instructions. It was doing everything as it was supposed to. So this mail employee decided to test it out on like you know in the real world on her email inbox. Immediately open cloud proceeded to delete like half of the inbox just delete email after email. The employee did post it on X, but by the way, kudos to her. That was, you know, it was very helpful for people because I know it's kind of like one of those things where I mean, you post something like that, there's some people on X that are going to dunk on you and say, "Oh, it's your fault and it was foolish, blah, blah. " But it's helpful for everybody to kind of see some of the worst case scenarios and understand what can happen and how to prevent it. In that particular case, if I understand correctly, what happened was the context window reached the limit. So, got reset. So, you can kind of imagine, let's say you're doing some work for a while and then your short-term memory just gets completely reset. You wake up like, "Whoa, what was I doing? " Uh, let's see. Looks like I was deleting some emails. Let's continue. Delete, delete, delete. I think what it was supposed to do is to like mark them and then ask before delete them saying, "Are you sure you want all these deleted? " But that got lost in translation somewhere. So, you can imagine that these large enterprises, they would love to have these AI agents do some of the work, but obviously they can't deploy them without guard. So Nemoclaw and Nvidia, they're coming in and saying, "Okay, here's the layer, the guard rails that will protect agents like Open Claw that will allow enterprise level companies to actually ship the stuff and to deploy it. " So Nvidia is definitely positioning itself to kind of be the Switzerland of AI. They want to be able to work with any model, any foundation, and they're adding the security layer that everyone needs, right? So they're sort of like the glue that connects all of these pieces together. And the whole data privacy router thing is pretty clever. A lot of companies can't use thirdparty cloud AI to run their data through it. They have to if they're using AI models, they have to be local models. So if you're able to build some sort of a router that takes the company's org level policies that they've set and is able to intelligently route different data like sensitive data to the local models, everything else to the cloud models and then intelligently control what goes where so it doesn't get coingled etc. I mean that does allow these enterprises to securely use these AI agents. It's a big win for those enterprises and obviously Nvidia. They don't really care which local models used. cloud models used. At the end of the day, they just want to make sure that the tokens keep going in and out. The GPUs continue to go burr. So very exciting. Let me know what you think about this thing. Are you planning to use this? And do you think enterprises will start using something like open claw now that Nvidia is stepping in and kind of handling the security? Let me know what you think. If you made this far, thank you so much for watching. My name is Wes Roth. I'll see you in the next