ISOLATE AGENTS

“We have a couple of people at NVIDIA. We've been working with security to bring agents really close to compute. So we now have stuff where you can tell Dynamo, like, go run some experience with Dynamo on x cluster and just try it right now. We've actually been able to one shot problems. We used to have this problem where, with Dynamo, you have to find the right configurations. We've just had an agent just completely one shot that. It goes. It gets the compute. It runs a couple experiments. It's like, this is the best. Go run this. And then we just give that to people, and it's faster than anything that they have.”

“Coding agents have been so much more effective than general purpose agents. And I think a large part of that is it just has access to the terminal, and that means it has access to everything that you've installed into your terminal. It can write code, and it can compile the code. And if there are errors, it can fix it. It can run your suite of tests because that's all just in your terminal. Computing began with a terminal with a shell, but we said that it's not empathetic to humans, so we built these nice user interfaces. And then now we have LLMs navigating our user interfaces, and ironically, we're not empathetic to the machine anymore.”

“Don’t install it on your main computer. Put it on a Mac Mini or an old laptop to keep it isolated from your primary workspace.”

“Nothing we can really do about that apart from, you know, clean up our act as it relates to some of these things we've just kind of brushed under the rug, like, having one of one multisigs for, you know, securing nine figures. I think, like, 97% of layer zero DBN setups were one of one or two of two. So, like, that's a problem. That should be addressed. Yeah. Can I can I just add one small thing? You know, back in the day at Chorus one, we looked at a lot of these interop networks, and we used to optimize as both investors and users sort of the most secure bridge.”

“I think someone resurfaced a tweet from Carl Samani this week about him saying that, you know, security doesn't matter. Like, all all that matters is speed. I think for a while, the last few years, the industry has optimized for scalability and speed and whereas in the early days of blockchain, we really optimized for security. And I think that pendulum is swinging back. Now we know that we can have fast chains. I think we now need to work out how we have secure chains, including smart contracts, and that, of course, is under the same umbrella as DeFi.”

“I would just say that if you look at the composition of deposits and lending activity on these borrow lend protocols, people are interested in RWAs much more so than crypto native assets. I think it's pretty clear that the driver of the let the next cycle or at least a big part of the story is gonna be different types of actual real world assets and people using it, like, RWA looping, this broader, trend. Shout out three f, small but proud bag holder there. But, yeah, I think that that's gonna be a big part of the story.”

“Setting up an agent is like onboarding a real assistant—you have to give them a soul, an identity, and a set of memories to actually be effective.”

“I have a twenty four seven agent running. I hooked up to run pod. It doesn't shut down instances, and I've tried prompting it. I've given the instructions, shut down when you're done. It's like, I need to keep it warm. I'll need it soon. It's horrible on time estimates too because it's like, yeah. I'll need it in forty five minutes. Forty five minutes of human time is actually three minutes of agent time, so it's like, I'm booting it up. I'm waiting. I'll just leave it on all night.”

“How can you get in front of these agents? And, like, potentially, there's a lot of value to be captured there in terms of owning that distribution channel. And so they're not really like money Lagos anymore. It's just more like a back end as opposed to, like, composable primitives that developers use in their own smart contracts. So I think the Lego sort of conversation is changing more to, like, a direct to smart contract back end kind of conversation. Like, do you need to compose anymore? Yeah. Maybe maybe you do.”

“There's the other concept that is explored a lot at NVIDIA, which is this idea of a $0 business. Market creation is a big thing at NVIDIA. Jensen says we are completely happy investing in $0 markets. We don't care if this creates revenue. It's important for us to know about this market. We think it will be important in the future. It can be $0 for a while. An org doesn't have to ruthlessly find revenue very quickly to justify their existence.”

“North Korea is actually playing a really big part of this story. So Lazarus Group obtained about $2,000,000,000, in hack proceeds last year from DeFi and crypto. That is a non trivial amount, and the exploits here are getting far, far more sophisticated. So a part of this story is we could be doing better in terms of multisigs. Like, at least on the Kelp DAO side of things, there was a layer zero DBN, which was operated by them, and it was one of one, which was a central point of failure.”

“Kimi two comes out, right. And it's an interesting model. The creators of Kimi k two actually talked about it in a blog post. Attention scales to the number of heads. They made a very specific barter in their architecture. They basically said, hey. What if we give it more experts? So we're gonna use more memory capacity, but we keep the amount of activated experts the same. We increase the experts' sparsity, and we decrease the number of attention heads.”

“Agents can do three things. They can access your files. They can access the Internet, and then now they can write custom code and execute it. You should really only let an agent do two of those three things. If you can access your files and you can write custom code, you don't want Internet access because that's one issue. Vulnerability. Right? If you have access to Internet and your file system, you should know the full scope of what that agent's capable of doing. Otherwise, malware can get injected or something that can happen.”

“Historically, models would be hosted with a single inference engine, and that inference engine would ping pong between two phases. There's pre fill where you're reading the sequence, generating kv cache, and then using that kv cache to generate new tokens, which is called decode. Some brilliant researchers across multiple different papers essentially made the realization that if you separate these two phases, you actually gain some benefits. You don't have to worry about step synchronous scheduling, and you allow yourself to split the work into two different types of pools.”

“SOL is actually I think of all the lessons I've learned, that one's definitely my favorite. The speed of light moves at a certain speed. So if light's moving slower, then you know something's in the way. So before trying to layer reality back in of why can't this be delivered at some date, let's just understand the physics. What is the theoretical limit to how fast this can go? And then start to tell me why. Because otherwise, people will start telling you why something can't be done.”

“Aave as evidenced by the launch of v four, is what is the model of lending here? And you have, like, the Cominos and the Morphos of the world who maybe started with a more, isolated markets or, modular approach to lending as opposed to, the more, you know, structured product of of a core. And there are trade offs to both. Right? But I think that's also something that people are thinking about here too. Yeah. Definitely. And, Mike, I know you spent a bit of time in this space, but I also think it's worth pointing out, you know, Aave, you mentioned the asset itself is was the thing that created risk.”

“The unlock is really multiple agents, not just one. Having specialized agents that know their specific job is what makes them useful.”

“Brev was just it's a developer tool that makes it really easy to get a GPU. It was actually Evan Conrad, SF Compute, who was just like, you guys are two dudes in the room. Why are you pretending that you're not? And so then we were like, okay. Let's make the logo of Shaka. We brought surfboards to our booth to GTC, and the energy was great. My wife was, at the time, fiancee, helping me put these vinyl stickers on and she goes, you son of a, if you pull this off. And so, pretty much after the acquisition, I stitched that with the acquisition. I sent it to our family group chat.”

“There's a summarization of that trend that I like to say to my team. This is the year system as model. Where, instead of having a single model be a thing, you have a system of models and components that are working together to emulate the black box model. So when you make an API call to something that's like a multi agent in the background, it still looks like an API call to a model. Under the hood, it's like a billion different models.”

“The risk was not that Aave had a, you know, a bad multisig or Oracle configuration or anything like that, but, another type of risk, which is the type of collateral that's been onboarded into one mixed pool, it caused these knock on effects. And you're you're essentially looking at a bunch of people right now. This is sort of without precedent. You know, I I I don't think in any of these, maybe, analogs that you could point to and TradFi, you've seen something like this, which is one, everything is a 100% transparent, and you have some folks who are locked a little bit or unable to, you know, unable to to move stuff on the wrapped east side of things.”

“WAP is, like, emerging competitor to Stripe. 2,000,000 businesses and, like, 20,000,000 creators hold assets on WAP. They had a partnership with Aave and Plasma and Tether, right, to just if you're holding money on their platform, hey, you wanna earn some yield? They're gonna deposit that into AVE on plasma in the background. And so so it's defi mullet cases, which I think could have gotten big, big chunks of depositors just because, you know, it's all abstracted away. Now that's gonna be kinda hard. Right? Because those risks and, you know, if I was just talking about an ETH depositor that didn't know about kelp kelp DAO.”