Vollständiges Transkript von Mempalace, Models & Mythos | Anthropic's Project Glasswing & Securing The Stack

Opus experience. So I don't believe it was a model change, but rather a 65 00:04:50.389 --> 00:04:51.270 usage 66 00:04:51.509 --> 00:04:52.310 change. 67 00:04:52.310 --> 00:04:59.349 And you can kinda do the same thing with GLM five one if if you choose to have thinking enabled or disabled, but I just want to 68 00:05:00.845 --> 00:05:02.765 kinda get get that out there. 69 00:05:04.125 --> 00:05:12.604 As far as the other models, yeah, there were a bunch of models that we added into the router. I'll I'll list through a couple of them. GLM five one thinking, 70 00:05:13.170 --> 00:05:16.130 GLM five one non thinking, and 71 00:05:16.530 --> 00:05:18.770 that difference is important because 72 00:05:19.970 --> 00:05:22.290 some agent tests that you're doing 73 00:05:23.010 --> 00:05:28.530 and depending how you're parsing responses, you may or may not want those thinking tokens. You may wanna trade 74 00:05:28.905 --> 00:05:30.985 speed for you know? 75 00:05:34.025 --> 00:05:35.625 Rugged, but we're back. 76 00:05:36.825 --> 00:05:37.465 Yeah. 77 00:05:38.665 --> 00:05:42.985 Okay. I'll just kinda jump right back into the continuation of this. So I was talking about 78 00:05:43.479 --> 00:05:46.520 GLM five one thinking versus nonthinking, 79 00:05:46.759 --> 00:05:52.199 and they're they're just cases where you might wanna run a nonthinking variant 80 00:05:52.919 --> 00:05:56.840 for the case of speed, for the case of cost, or for the case of 81 00:05:57.675 --> 00:05:58.635 JSON 82 00:05:58.795 --> 00:06:04.235 kind of output parsing, it's it's sometimes a little bit more complex with thinking. So that's really on the 83 00:06:04.715 --> 00:06:06.075 on the GLM 84 00:06:06.075 --> 00:06:14.440 five one side. Now we also launched a bunch of other models into the Morpheus Inference marketplace and Morpheus Inference API. 85 00:06:15.400 --> 00:06:18.520 That was the Trinity model. 86 00:06:19.560 --> 00:06:20.520 Unfortunately, 87 00:06:20.600 --> 00:06:25.800 the we launched the Trinity model and then g l m five one came out, which I personally think is 88 00:06:26.415 --> 00:06:31.615 definitely a a higher performing model and and a better bang for your buck. But nonetheless, 89 00:06:32.495 --> 00:06:34.735 different models for different applications. Now 90 00:06:34.975 --> 00:06:38.495 the other models that we launched are 91 00:06:39.420 --> 00:06:46.940 are also very interesting. So there's been a lot of news around the Gemma four models. There's two variants that are available, both the 92 00:06:47.580 --> 00:06:48.700 27,000,000,000, 93 00:06:48.700 --> 00:06:49.820 the MOE, 94 00:06:49.980 --> 00:06:51.660 and the 31,000,000,000, 95 00:06:51.660 --> 00:06:53.180 I believe it is, the dense model. 96 00:06:54.025 --> 00:06:56.905 Very, very strong models. These 97 00:06:58.345 --> 00:06:59.065 are, 98 00:06:59.865 --> 00:07:08.905 I would say, best in like, reaching best in class levels or reach not not at not frontier levels, but really high levels of performance at 99 00:07:09.300 --> 00:07:11.460 low cost and low 100 00:07:12.500 --> 00:07:18.260 parameters. So these are some that people could even run locally on the right levels of 101 00:07:19.060 --> 00:07:20.180 of hardware. 102 00:07:20.980 --> 00:07:36.555 Maybe not the the same tokens per second that you're going to get, but they're models that are that are very flexible and and very strong. So I'm going to see a lot of I anticipate a lot of agents running on these for noncomplex 103 00:07:36.555 --> 00:07:37.355 tasks, 104 00:07:38.235 --> 00:07:38.555 whereas 105 00:07:39.389 --> 00:07:45.470 g l m five one and and the like are going to be used for kind of heavy reasoning, coding, etcetera. 106 00:07:45.630 --> 00:07:46.190 So 107 00:07:46.750 --> 00:07:52.669 really excited about those. Also, we launched some of the very small quen models, 108 00:07:53.405 --> 00:07:58.285 Quen 9,000,000,000 and and a couple others that that I can we can send the list after this. 109 00:07:59.005 --> 00:08:05.725 Those are for even simpler tasks where where I talk a lot about the sliding scale of intelligence, the sliding scale of privacy. 110 00:08:07.580 --> 00:08:15.340 You need small models for certain tests. Right? Certain extraction tests, certain summarization tests. So you really you don't wanna just throw, 111 00:08:15.580 --> 00:08:20.700 you know, all of your money away at at the top models all the time. You wanna kind of intelligently route 112 00:08:21.625 --> 00:08:27.865 models to the test that you're doing. And David can talk a bit about the the model routing 113 00:08:28.104 --> 00:08:33.464 that that he implemented in Everclaw, and and a lot of the agent harnesses are are using 114 00:08:33.940 --> 00:08:48.435 somewhat similar structure. And, you know, we know from Claude Code, uses Haiku for some things, uses Sana for some things, and and Opus for others. So somewhat proven in the industry, and now we have the models available to to support that in a in a, 115 00:08:48.755 --> 00:08:52.435 you know, best in class way right now for the different model sizes. 116 00:08:53.875 --> 00:08:56.915 Yeah. I mean, was super easy to update 117 00:08:57.235 --> 00:09:01.875 my agent to use the new models. So appreciate you pushing them out so quickly. 118 00:09:02.300 --> 00:09:16.060 Kyle, I literally just told my agent, hey. You know, there's new models on Morpheus. Go check the gateway. Go check the the router, and boom. It updated the list, and I was off and running. So anybody 119 00:09:16.060 --> 00:09:22.545 using a Morpheus API key, you you know, know, just tell your agent to do the thing, and you should be able to get on 5.1. 120 00:09:22.785 --> 00:09:23.345 And 121 00:09:23.745 --> 00:09:30.225 the routing, I think, is is important. Like you're saying, there's these smaller models that are coming out with with better quality. 122 00:09:30.790 --> 00:09:36.550 A lot of your heartbeats or cron jobs or stuff,

like, they don't you shouldn't be paying Claude, 123 00:09:36.790 --> 00:09:41.030 you know, to to run those tokens. There are now really good alternatives, 124 00:09:41.430 --> 00:09:49.535 and you can grab them on Morpheus for an extremely low cost. So yeah. Yeah. And, like, all the all the memes that you see of, 125 00:09:50.334 --> 00:09:57.375 you know, the guy with the sledgehammer asking Claude you know, asking Opus what's the weather. It's like, that's not necessary. 126 00:09:57.694 --> 00:10:01.935 Let let's let's use our money and and our and our energy intelligently in some ways. 127 00:10:03.110 --> 00:10:10.390 But yeah. And just one other comment is that this isn't only on the API. Right? Even if you're just running through 128 00:10:11.190 --> 00:10:13.110 hosting your own consumer node, 129 00:10:13.510 --> 00:10:15.270 I I would love for people to 130 00:10:16.514 --> 00:10:18.595 use the API for information, 131 00:10:19.154 --> 00:10:22.355 you know, as well as as inference if they choose. So 132 00:10:22.675 --> 00:10:31.954 you wanna find the all the models available on Morpheus and you don't want to pull the blockchain for the registering the information, you can just hit the model's endpoint on 133 00:10:32.250 --> 00:10:36.810 the Morpheus API, and it'll give you all the information. It'll give you the blockchain ID. It'll give you 134 00:10:37.370 --> 00:10:42.970 all the models that are available with bids and up and running, and it it's it's just very good informationally 135 00:10:42.970 --> 00:10:54.475 for an agent to be able to hit a simple curl command and and then go and and get the information it needs to make decisions even if it's running its own node. So just just kind of a note there. 136 00:10:56.555 --> 00:10:59.675 Can you speak to Kyle, I I was actually looking 137 00:11:00.550 --> 00:11:04.230 for the dashboard site again to see from a efficiency 138 00:11:04.230 --> 00:11:10.470 perspective, like, a cost perspective, how these compare to some of the models people would have previously been using. 139 00:11:12.805 --> 00:11:15.605 Yeah. I believe we have that up on 140 00:11:15.764 --> 00:11:17.524 tech.more.org. 141 00:11:17.524 --> 00:11:19.925 Let me just pull up the exact URL. 142 00:11:20.644 --> 00:11:23.204 Yeah. Tech.more.org/calc. 143 00:11:23.445 --> 00:11:25.204 So that's our calculator, 144 00:11:25.860 --> 00:11:28.740 you know, brief calculator page that 145 00:11:29.220 --> 00:11:33.060 you can kind of put in. Alright. If you're staking tokens, 146 00:11:33.459 --> 00:11:39.860 how much, you know, daily credits do you get from staking, whether it be in the API or running your own node specifically. 147 00:11:40.355 --> 00:11:46.515 We also have some price we have our I believe we have our pricing in here or at least a link to the pricing 148 00:11:47.395 --> 00:11:50.595 through the API, and you can really go there to 149 00:11:51.075 --> 00:11:58.709 figure out, you know, how how much is

this gonna cost, how many tokens do I need to stake if I choose to go that route, or you can go directly to 150 00:11:59.509 --> 00:12:01.430 a pidocs.more.org, 151 00:12:01.430 --> 00:12:05.750 which is the full API reference page for using the API. I would say 152 00:12:06.195 --> 00:12:08.595 API reference page page is probably 153 00:12:09.635 --> 00:12:12.275 the place to go for the, 154 00:12:12.595 --> 00:12:17.235 excuse me, the full API information, but this is just a good 155 00:12:18.115 --> 00:12:23.320 first step in, wanna get some quick information and and do some some basic staking calculations, 156 00:12:23.320 --> 00:12:25.959 that that's the place to go. It's tech.more.org. 157 00:12:26.440 --> 00:12:29.480 Awesome. Yeah. I'm just looking right now, like, as an example. 158 00:12:30.279 --> 00:12:33.800 If you had a thousand more stakes to the API, 159 00:12:33.959 --> 00:12:36.199 or the inference gateway right now, 160 00:12:37.375 --> 00:12:42.575 you'd you basically have something close to, like, between 800 and a thousand 161 00:12:42.894 --> 00:12:46.255 requests to GLM five every single day, 162 00:12:46.575 --> 00:12:54.030 which, like, when you think of that in terms of purchasing API tokens from, like, a Hugging Face or somewhere else where you're paying buy token, 163 00:12:54.270 --> 00:12:56.270 I mean, that's just an insane value, 164 00:12:56.590 --> 00:13:00.350 when you consider that it's perpetual. That's pretty pretty amazing. Absolutely. 165 00:13:02.285 --> 00:13:07.565 And if you're staking so let me just I like to go through this just to make sure that everyone remembers 166 00:13:07.965 --> 00:13:10.525 exactly how it works. If you're running your own node, 167 00:13:11.645 --> 00:13:18.440 there's a certain cost that a provider is putting out saying, I'm willing to offer my compute compute at this price. And 168 00:13:18.840 --> 00:13:23.000 let's say it's a thousand or whatever the number was, that's saying, I'm going to rent 169 00:13:23.320 --> 00:13:23.960 one, 170 00:13:24.520 --> 00:13:25.320 let's say, 171 00:13:26.120 --> 00:13:31.320 one slot from this provider, right, where I I have a full slot of GLM51 172 00:13:31.400 --> 00:13:37.385 for the whole day. I can send as many requests as I want, but they're going to fill serially. 173 00:13:37.385 --> 00:13:37.865 So 174 00:13:38.585 --> 00:13:54.000 for applications where you have a bunch of users or you have a bunch of agents running that you need to send prompts concurrently, you might need more than one session. So you might need to stake 2,000 if you wanna run two prompts side by side all day long. So so it's just good to think about it on 175 00:13:54.720 --> 00:13:57.360 like, just like if you're renting a VM, like, how many 176 00:13:58.400 --> 00:14:11.105 how much do I need to duplicate this? How many, you know, VCPUs do I need? How how much stuff am I gonna be running at the same time? And you should plan how many sessions you need accordingly. But just wanted to get the point out that renting 177 00:14:11.825 --> 00:14:13.505 full access to 178 00:14:13.970 --> 00:14:28.765 a model through the node is a single stream that you get, and you can, of course, scale from there. But it is. It's it's full day long. You know, as as long as you choose to create your session for, it could be a full day, unlimited prompts throughout that session, 179 00:14:29.165 --> 00:14:33.245 and in perpetuity, really. It's you could you could just stake the next day as well. 180 00:14:34.045 --> 00:14:34.925 Absolutely. 181 00:14:35.165 --> 00:14:36.685 That actually brings me to, 182 00:14:37.085 --> 00:14:45.150 of of the topics I wanted to cover today, I'm actually gonna skip two or come back to it and jump straight to three because it's very relevant to what you were just talking about. And it's, 183 00:14:45.710 --> 00:14:49.950 basically the idea of as like the DAI community 184 00:14:50.190 --> 00:14:56.425 eating our own dog food in terms of the DAI workflows. And what I mean by that is so I 185 00:14:56.745 --> 00:15:00.505 was in a private conversation with a couple people yesterday, we were talking about, 186 00:15:01.545 --> 00:15:03.705 like, building DAI 187 00:15:03.785 --> 00:15:04.665 alongside 188 00:15:04.825 --> 00:15:08.425 the use of central tools and, like, becoming 189 00:15:08.960 --> 00:15:10.160 more 190 00:15:10.160 --> 00:15:19.520 DAI centric over time. I think we need to think of a term for this. Like in the Ethereum world, it would be like becoming more bankless, which is of course the tagline of that podcast. 191 00:15:20.320 --> 00:15:26.615 But the idea of taking your your workflows that you're executing on and transitioning them to ultimately 192 00:15:27.095 --> 00:15:29.095 someday, hopefully in the near future, 193 00:15:29.495 --> 00:15:31.495 being DAI native. 194 00:15:31.815 --> 00:15:35.175 And so I'd love to hear, like, if anyone's willing to share 195 00:15:35.495 --> 00:15:43.410 what their workflows look like right now, and and are there new ways that maybe they're integrating DAI that they weren't previously? 196 00:15:44.450 --> 00:15:48.050 I can start if people wanna think about it in the meantime. So 197 00:15:48.370 --> 00:15:49.650 fairly recently, 198 00:15:50.210 --> 00:15:53.170 I, you know, have had my own OpenCLOS setup. I have it running 199 00:15:53.585 --> 00:15:55.265 through Morpheus Inference, 200 00:15:55.425 --> 00:16:00.385 to the API gateway. Shout out to the team who built that yet again. 201 00:16:01.185 --> 00:16:08.225 And I've started basically teaching it to be somewhat of an assistant for me with my real estate financing business, 202 00:16:08.750 --> 00:16:11.070 which today, what that looks like is 203 00:16:11.389 --> 00:16:12.270 basically 204 00:16:12.350 --> 00:16:13.790 taking information 205 00:16:13.790 --> 00:16:14.510 and 206 00:16:14.990 --> 00:16:16.350 whether it's like 207 00:16:16.910 --> 00:16:26.085

loan pricing across, you know, more than a 100 lenders or it's looking at an approval from an underwriter and just having it break down information for me. And 208 00:16:26.645 --> 00:16:29.845 basically, like like, the way people will use the data to 209 00:16:30.885 --> 00:16:43.970 summarize things like articles and and things like that to provide them, like, curated lists of what they need to know, It's essentially doing that for me to cut my my time down in like manually sifting through information, 210 00:16:44.210 --> 00:16:47.890 which has already been beneficial to to me and my business, 211 00:16:48.290 --> 00:16:53.154 you know, just in the last few weeks that I've been doing that. But the reason 212 00:16:53.635 --> 00:16:56.115 I hadn't done it previously in any other 213 00:16:56.834 --> 00:17:01.394 centralized platform was because that data is really valuable. 214 00:17:01.714 --> 00:17:04.595 And I don't want, you know, anyone else to have access to 215 00:17:05.280 --> 00:17:13.280 my client data or anywhere where I I can't verify that, you know, I'm getting the most privacy that is available 216 00:17:13.520 --> 00:17:15.120 to the best of our ability 217 00:17:15.280 --> 00:17:30.504 when I'm running these tools and, like, training this agent to ultimately become, you know, good at this job, like, as an assistant to this job. And so just in my own life, like, I've already been able to integrate that a little bit. I expect to, you know, continue to scale it up over time. 218 00:17:30.905 --> 00:17:32.905 And I hope to get to a point where 219 00:17:33.400 --> 00:17:35.720 a lot of the models I'm sorry. 220 00:17:36.840 --> 00:17:47.240 Excel models that I've built that I used in the past, which I, like, track my clients' financing with to determine when they're in a good position to refinance based on market conditions. 221 00:17:47.855 --> 00:18:01.135 I anticipate a time fairly soon in the not so distant future where I'll have an agent that'll be monitoring those models for me and ultimately, like, making those recommendations before I'm even manually getting to them, 222 00:18:01.970 --> 00:18:13.170 which would be huge for me and, like, anyone who does, you know, a similar job to what I'm doing. But does anyone else have any examples of of things, you know, that they've been able to transition to, like, fully DAI? 223 00:18:14.865 --> 00:18:21.184 Well, I'm happy to talk about that, Bo. I mean, the first thing I transitioned was was basically all my writing. 224 00:18:21.345 --> 00:18:22.545 Right? I remember 225 00:18:22.945 --> 00:18:24.545 killing very happily 226 00:18:24.705 --> 00:18:26.545 my OpenAI subscription 227 00:18:26.545 --> 00:18:32.400 at a year, year and a half ago. Right? And I I never really used it heavily because I knew it wasn't 228 00:18:32.640 --> 00:18:34.400 secure or private. 229 00:18:34.720 --> 00:18:38.240 So I experimented with it. But as soon as 230 00:18:38.720 --> 00:18:48.465 the open source models and this is probably even back in, like, the llama three point three days, got good enough on the writing side. I was like, alright. I don't need Chachi PT anymore 231 00:18:48.785 --> 00:18:57.345 at all, and I can kill that subscription and move completely to decentralized AI. And I feel like we're at a similar moment now on the coding 232 00:18:57.889 --> 00:19:00.289 where I haven't been using Quad for a while, 233 00:19:00.610 --> 00:19:06.209 at least directly. You know, I've been using it through some inference providers for better anonymity. 234 00:19:06.690 --> 00:19:09.249 But I feel like with 5.1, 235 00:19:09.490 --> 00:19:13.624 you know, if my testing bears out the next few days, I'm gonna be able to 236 00:19:13.945 --> 00:19:15.544 officially cut the cord 237 00:19:15.784 --> 00:19:17.304 and be using 238 00:19:17.705 --> 00:19:22.744 only open source models for all of my coding. And I think that's a pretty huge moment 239 00:19:23.225 --> 00:19:24.424 for the community 240 00:19:24.585 --> 00:19:30.630 that we can sort of move strongly into that whole new set of users and personas 241 00:19:30.630 --> 00:19:38.470 of people that I mean, that seems like the killer app certainly the last six months is every all of a sudden, everybody's a coder, and everybody's coding. 242 00:19:39.095 --> 00:19:41.735 Even people who weren't developers historically, 243 00:19:41.895 --> 00:19:43.655 you know, the tools become good enough 244 00:19:43.895 --> 00:19:45.414 to really pull that off. 245 00:19:45.655 --> 00:19:53.580 So, you know, I think that's probably my my next big moment that I've been really looking forward to. And it it was clear it was coming, 246 00:19:53.900 --> 00:20:00.460 but the last few months, the models just weren't quite good enough. I I think we might finally be there with 247 00:20:00.860 --> 00:20:02.620 with GLM 5.1. 248 00:20:02.620 --> 00:20:08.985 So that's that's my next step is cut the cord on on any cloud usage even indirectly 249 00:20:09.065 --> 00:20:11.544 or even through inference providers 250 00:20:11.545 --> 00:20:13.705 and just be a 100% 251 00:20:13.865 --> 00:20:18.184 on open source for coding. That's that's the next milestone I'm I'm hoping to hit here. 252 00:20:20.870 --> 00:20:22.790 Awesome. That would be amazing. 253 00:20:22.950 --> 00:20:27.830 A pioneer as always, David. You'd certainly be one of the earliest, 254 00:20:28.710 --> 00:20:35.544 probably to do that, at least in in the capacity that we're talking about. So that's amazing. Definitely keep us updated on that. Looking forward to hearing 255 00:20:35.945 --> 00:20:42.905 about when that cord officially gets cut, and then no one else will have excuses as to have not also done the same. 256 00:20:43.385 --> 00:20:45.980 No. I'm just kidding. But that is awesome. And 257 00:20:46.540 --> 00:20:49.260 and I'll just add on. I mean, we have a couple 258 00:20:49.900 --> 00:20:53.340 larger users. And again, I'm sorry

about the background noise. Apparently, 259 00:20:53.820 --> 00:20:55.980 someone's doing major construction near me. 260 00:20:57.020 --> 00:21:00.860 The we have a couple major users on the 261 00:21:01.054 --> 00:21:05.215 Morpheus API who have integrated this into their production workflows. So 262 00:21:06.095 --> 00:21:09.215 really moving to the open source models for 263 00:21:11.054 --> 00:21:23.730 I guess, we'll have them on a in a couple weeks to to talk about it more, but making the switch from Opus, making the switch from Sonnet to these open source models to run their businesses and and rely on Morpheus 264 00:21:23.970 --> 00:21:26.049 and the the providers 265 00:21:26.049 --> 00:21:33.134 who are hosting through Morpheus for to service them and their customers. So it's really great to see kind of that transition 266 00:21:33.855 --> 00:21:35.614 beyond just us and beyond 267 00:21:36.015 --> 00:21:39.294 what we're doing, you know, with with our coding and so on. 268 00:21:39.695 --> 00:21:41.855 Personally, also, I've switched my 269 00:21:42.409 --> 00:21:44.249 AI research platform. 270 00:21:44.809 --> 00:21:47.369 What is AI Research OS. 271 00:21:47.929 --> 00:21:50.330 Originally, was doing it through Opus, 272 00:21:50.649 --> 00:21:52.249 then I was doing it through 273 00:21:52.809 --> 00:21:54.010 one of the smaller 274 00:21:54.250 --> 00:22:02.535 I believe it was Sonnet or one of the smaller ones. I switched fully over to g l m five one, and I'm using Gemma for some of this stuff too. So, again, finding, 275 00:22:02.695 --> 00:22:05.415 you know, the parts of my infrastructure that could use 276 00:22:05.735 --> 00:22:15.299 open source models, which is all of it now, and then the parts that can use even smaller open source models to help, you know, give, you know, those customers a better cost. So, 277 00:22:15.620 --> 00:22:18.820 you know, really excited about about where we've come, and 278 00:22:19.539 --> 00:22:25.700 every month, you know, all the models get smarter. So it's gonna be you know, we're we're approaching that 279 00:22:26.044 --> 00:22:29.565 that line that says this is good enough. Right? 280 00:22:30.125 --> 00:22:37.724 More intelligence isn't required for this task, and it's certainly something that can be done through open source. So really excited there and 281 00:22:38.044 --> 00:22:40.605 really excited to save a lot of people a lot of money, 282 00:22:41.809 --> 00:22:47.009 you know, and and that funneling down to users getting great services for for a low price. 283 00:22:48.610 --> 00:22:49.489 Absolutely. 284 00:22:49.490 --> 00:22:55.410 I can pile on for a second while we're at it Please do. For anyone who I haven't 285 00:22:55.625 --> 00:22:57.145 shared with yet. 286 00:22:57.305 --> 00:23:01.705 My current setup with some of my agents, particularly my Hermes 287 00:23:01.705 --> 00:23:02.505 agent, 288 00:23:02.905 --> 00:23:04.745 I have the default 289 00:23:04.905 --> 00:23:06.105 is running 290 00:23:07.065 --> 00:23:08.825 off of good pure stake more, 291 00:23:10.010 --> 00:23:13.769 mainly off five point zero and five point one right now. 292 00:23:14.250 --> 00:23:17.369 If it has something that needs something frontier, 293 00:23:17.770 --> 00:23:18.409 it 294 00:23:18.809 --> 00:23:22.409 would be over automatically 295 00:23:21.345 --> 00:23:24.864 to run through Venice. You'll be using or 296 00:23:25.025 --> 00:23:25.825 Opus. 297 00:23:25.905 --> 00:23:29.184 So, obviously, that's not open source, but if I do need something that requires 298 00:23:29.505 --> 00:23:32.225 that memo, it's at least more anonymized. 299 00:23:32.385 --> 00:23:36.385 And then I also put in I don't know how to say it. If it's 300 00:23:36.990 --> 00:23:39.309 a RISE or something. There's a 301 00:23:39.950 --> 00:23:43.790 there's a specialized model that was derived from plans 302 00:23:43.790 --> 00:23:46.270 3,590,000,000 303 00:23:46.270 --> 00:23:47.149 parameter. 304 00:23:47.310 --> 00:23:48.590 The new one is specifically 305 00:23:49.230 --> 00:23:49.630 for 306 00:23:50.545 --> 00:23:55.025 Hermes and for tool calling and things like that that's running completely locally 307 00:23:55.745 --> 00:23:57.424 just on a 16 gig. 308 00:23:57.745 --> 00:24:00.705 So that's actually performed quite well. 309 00:24:01.265 --> 00:24:06.710 And when I was having issues trying to connect all the different things and routing and all this other junk, 310 00:24:06.950 --> 00:24:07.510 I 311 00:24:07.910 --> 00:24:11.190 ran, like, almost the entire day off of this 9,000,000,000 312 00:24:11.190 --> 00:24:21.365 that was optimized for Hermes, and it didn't suck. Like, it did all the regular stuff completely fine, and it didn't trip up where I thought it would trip up in some of the more 313 00:24:21.685 --> 00:24:22.485 challenging 314 00:24:22.645 --> 00:24:28.325 tasks that I had. So I'm still playing around a ton, but little by little, I'm able to, like, 315 00:24:29.040 --> 00:24:35.760 expand out what models are being routed to. And at a minimum, everything's anonymized, if not completely 316 00:24:35.840 --> 00:24:37.040 over and 317 00:24:37.680 --> 00:24:39.200 or mobile. So 318 00:24:39.360 --> 00:24:48.065 just more fire under everyone else's pants. Like, you can do it, and it's probably gonna be, like, in a week or two, and you can have an almost sovereign setup. 319 00:24:49.825 --> 00:24:53.424 Brian, were there any tasks that you, like, specifically anticipated 320 00:24:53.424 --> 00:24:57.184 there would be slip ups on and you were surprised that there weren't or anything like that? 321 00:24:59.310 --> 00:24:59.950 Yeah. 322 00:25:00.750 --> 00:25:04.510 Mainly, like, doing code fixes and debugging 323 00:25:05.150 --> 00:25:05.870 and 324 00:25:06.270 --> 00:25:08.669 and even building a couple new features. 325 00:25:09.150 --> 00:25:09.789 So 326 00:25:10.350 --> 00:25:11.550 I had 327 00:25:12.185 -->

00:25:15.705 I was trying to eat my own dog food in a sense and not 328 00:25:16.105 --> 00:25:17.304 fall back on 329 00:25:17.545 --> 00:25:20.424 SSH ing in with my Cloudmax 330 00:25:20.825 --> 00:25:21.705 account. 331 00:25:21.945 --> 00:25:22.664 And 332 00:25:22.745 --> 00:25:25.544 I was trying to, like, fix all this different routing 333 00:25:25.785 --> 00:25:27.225 and 334 00:25:26.670 --> 00:25:28.429 and a bunch of things like that, 335 00:25:28.750 --> 00:25:31.150 and it was powering through. 336 00:25:31.470 --> 00:25:35.150 Like, it might have maybe taken me, like, one more prompt than, 337 00:25:35.470 --> 00:25:37.390 like, a single shot, but 338 00:25:37.710 --> 00:25:39.150 I was I was pretty impressed. 339 00:25:44.205 --> 00:25:45.565 Awes oh, 340 00:25:45.805 --> 00:25:47.005 there you go. Sorry. 341 00:25:47.725 --> 00:25:52.605 Awesome. Thank you. Yeah. Thank you for sharing. I was I was curious to hear, like, kinda what 342 00:25:52.909 --> 00:25:54.669 what exact types of things you were, 343 00:25:55.870 --> 00:25:57.229 you you were looking, 344 00:25:57.870 --> 00:26:00.429 to use it on and kinda how the outcome 345 00:26:00.830 --> 00:26:06.429 was surprising in in its ability to pull it off. Because, honestly, I'm still at the point where 346 00:26:07.045 --> 00:26:12.164 anytime I can successfully get my agent to do anything, I'm I'm floored. You know that 347 00:26:13.685 --> 00:26:14.885 that meme 348 00:26:14.885 --> 00:26:27.450 that was like it was like the stonks meme, but it was like the white kinda like stick figure guy that was like, traitor, and it was like, spelled wrong? That's how I feel every time I do anything with code in it, but like developer, 349 00:26:27.530 --> 00:26:32.010 because it's, you know, giving me powers that I would never have otherwise had. 350 00:26:33.210 --> 00:26:39.455 Speaking of which, somewhat of a good transition to a next segment that I wanted to start doing, 351 00:26:39.935 --> 00:26:41.695 every week or at least as 352 00:26:41.855 --> 00:26:46.735 as they pop up on the timeline or as people send them to me. But that is the 353 00:26:46.895 --> 00:26:47.695 unexpected 354 00:26:47.695 --> 00:26:52.815 celebrity chime in of the week, with regard to DAI specifically. 355 00:26:53.190 --> 00:27:04.229 So I've been seeing a couple things like this where people who, at least from my perspective and probably the perspective of many, are like very much outside of the scope of 356 00:27:05.855 --> 00:27:11.215 almost like people that I'm surprised to see paying this close attention to 357 00:27:11.535 --> 00:27:13.775 not just AI, but specifically, 358 00:27:13.775 --> 00:27:17.695 like, privacy and and personal access to intelligence 359 00:27:17.695 --> 00:27:18.575 and 360 00:27:19.260 --> 00:27:21.980 data security and things like this in AI. And so 361 00:27:22.460 --> 00:27:23.179 the 362 00:27:23.500 --> 00:27:26.299 the first one, and what made me think of this was 363 00:27:27.900 --> 00:27:32.060 the the joke about everyone becoming a developer. Although it's not a joke. And this is 364 00:27:33.145 --> 00:27:34.024 Mila 365 00:27:34.265 --> 00:27:35.544 Jovovich, 366 00:27:36.025 --> 00:27:42.905 who was an actress in many, many different things. Let me just pull up her IMDb 367 00:27:42.905 --> 00:27:43.865 really quick. 368 00:27:44.505 --> 00:27:46.905 Famous for roles in The Fifth Element, 369 00:27:47.530 --> 00:27:48.809 also in 370 00:27:50.330 --> 00:27:51.850 let's see. She was in 371 00:27:52.090 --> 00:27:52.410 two 372 00:27:53.050 --> 00:27:56.570 was it the Resident Evil movies? She was in Hellboy. 373 00:27:56.890 --> 00:28:02.250 She was in tons of movies, like a really accomplished actress, you know, going all the way back to 374 00:28:02.975 --> 00:28:04.095 the eighties, 375 00:28:04.095 --> 00:28:13.535 actually. I'm looking at her IMDb page right now. It is very impressive. She recently published open source code on GitHub that has now been downloaded 376 00:28:13.615 --> 00:28:18.729 thousands and thousands of times and has actually been integrated into Everclaw 377 00:28:19.210 --> 00:28:20.809 called Mempalace. 378 00:28:20.809 --> 00:28:26.249 David, can you talk about, you know, what that is exactly and why you chose it for for Memclaw? 379 00:28:26.490 --> 00:28:27.690 I'm sorry. For Everclaw? 380 00:28:29.085 --> 00:28:36.764 Well, it's it's really an interesting trend. Right? You have smart people from other industries and fields 381 00:28:36.845 --> 00:28:40.365 that have never been able to sort of touch GitHub or code 382 00:28:40.840 --> 00:28:43.720 and now kind of flowing in using these tools. 383 00:28:44.520 --> 00:28:52.040 She she mentioned she partnered with a guy who's actually in in our community. I've seen him around some of the the the groups, 384 00:28:52.280 --> 00:28:53.000 Ben Sigman, 385 00:28:54.065 --> 00:28:58.145 that is, like, a longtime developer. So she did this in partnership 386 00:28:58.305 --> 00:28:59.985 with with a good engineer 387 00:29:00.145 --> 00:29:03.024 and basically came up with sort of her own approach 388 00:29:03.185 --> 00:29:16.769 for, okay, how should you handle agentic memory? Like, the idea of a mental palace has been around for thousands of years. Right? You can kind of imagine in your mind's eye, I'm walking into this room. What do I have stored in this room? What do I have stored in this drawer? 389 00:29:17.090 --> 00:29:23.809 You know, you you could say this is sort of just a, you know, a glorified way of of handling a database. 390 00:29:24.049 --> 00:29:24.290 Right? 391 00:29:24.875 --> 00:29:30.795 But, you know, it's kind of clever and catchy and and and kinda caught on. Right? And so they picked up a lot of attention 392 00:29:31.115 --> 00:29:41.539 and, you know, stars on GitHub. You know, and people have, you know, criticized it. Oh, well, let's do this or that, but it's all open source so they can make pull

requests and and prove it. Right? 393 00:29:41.780 --> 00:29:44.580 But it gave me an excuse to go back to my agent, 394 00:29:45.140 --> 00:29:46.580 and surprise surprise, 395 00:29:47.140 --> 00:29:49.940 OpenFlow does not configure 396 00:29:49.940 --> 00:29:51.140 memory search 397 00:29:51.940 --> 00:29:52.900 by default. 398 00:29:53.140 --> 00:29:59.305 So if you haven't done it, go into your agent and tell it to use memory 399 00:29:59.385 --> 00:30:00.345 underscore 400 00:30:00.585 --> 00:30:01.305 search 401 00:30:01.465 --> 00:30:04.425 and actually activate it in your configuration. 402 00:30:04.585 --> 00:30:07.945 Because if you didn't do that and you have to do that step, 403 00:30:08.760 --> 00:30:10.919 you haven't been searching your memory 404 00:30:11.080 --> 00:30:14.840 when your agent is doing a task. Right? And it's today, 405 00:30:15.000 --> 00:30:31.404 you know, probably you're having to give it the context again and again if you don't have that activated. So number one, activate in your agent the ability to search its own memories because that doesn't come baked in, which is a a surprise. You know? I I had seen that error sometimes. 406 00:30:31.725 --> 00:30:34.284 You know? And and so it's been so refreshing 407 00:30:34.605 --> 00:30:36.365 ever since I I updated 408 00:30:36.740 --> 00:30:39.380 the configuration to enable 409 00:30:39.380 --> 00:30:41.140 memory underscore search. 410 00:30:41.220 --> 00:30:57.845 All of a sudden now, I'm having to give it a lot less context when I'm talking to it. Right? It's it has an easier time finding its standard operating procedures and stuff like that. So so one, this was a great excuse to go down the rabbit hole on how my agent is using memory. Two, you know, Everclaw, 411 00:30:57.845 --> 00:31:00.964 you know, Morpheus in general ought to be agnostic 412 00:31:01.205 --> 00:31:03.524 to any memory system. So, you know, 413 00:31:04.890 --> 00:31:06.010 integrating this, 414 00:31:06.170 --> 00:31:12.250 there's other great systems like Neo four j that is an open source graph database, 415 00:31:12.410 --> 00:31:15.530 but it's probably something you wanna use for larger, 416 00:31:15.530 --> 00:31:16.250 heavier loads. 417 00:31:16.745 --> 00:31:19.145 Right? It's really designed for for large 418 00:31:19.305 --> 00:31:22.345 datasets. So, yeah, it was it was fun integrating. 419 00:31:22.905 --> 00:31:25.065 You know, it's good to see it was MIT 420 00:31:25.465 --> 00:31:29.225 licensed open source, and so it's super easy to add as an option 421 00:31:29.950 --> 00:31:39.470 for my agent. So and and that's what you want is you don't wanna bake in a lot of defaults that force people to use x y z, but you at least wanna make it easy 422 00:31:39.870 --> 00:31:51.014 for them to use the different options based on on what they need for their setup. Right? So, yeah, it was a it was a fun little integration. It only took a a few hours of of coding and testing, 423 00:31:51.095 --> 00:31:59.170 but it it was a good excuse to go and really tear apart or, like, how is my agent using memory? What do I need to do so that when somebody 424 00:31:59.730 --> 00:32:01.970 uses an agent, you know, it comes with 425 00:32:02.290 --> 00:32:04.210 at least the basic defaults 426 00:32:04.210 --> 00:32:14.484 turned on for for memory search and stuff like that. But it was it was it was a good experience, but I think we're gonna see more of this. So it should almost be like a recurring thing, like who 427 00:32:14.804 --> 00:32:21.044 you know, what famous person is shipping cool stuff in AI this week? And I I wouldn't be surprised if we see a lot more of that. 428 00:32:22.884 --> 00:32:25.044 Absolutely. And and particularly interesting, 429 00:32:27.050 --> 00:32:38.330 like you were saying before, like, that it came from kind of an unexpected source. Granted, in this case, a partnership, you know, with with a developer, but I think we're gonna see more of that too. And and you've said this in the past, David, like, 430 00:32:38.890 --> 00:32:45.924 everyone is a developer now in some capacity, or at least can be. Right? And so I I that'll really change 431 00:32:47.125 --> 00:32:47.845 where 432 00:32:48.085 --> 00:32:52.645 where different ideas and products and and agents and skills come from. 433 00:32:53.900 --> 00:33:03.820 This is definitely a segment that I wanna try to do, and celebrity's kinda tongue in cheek. It doesn't have to be, you know, someone who's like famous per se. But I I just like these stories about 434 00:33:04.380 --> 00:33:07.179 people who maybe would not who 435 00:33:07.179 --> 00:33:09.740 would not have maybe otherwise 436 00:33:09.055 --> 00:33:10.495 worked on shipping 437 00:33:10.575 --> 00:33:12.095 software or code 438 00:33:12.415 --> 00:33:19.294 that now are, whether it be, you know, on their own or through a partnership or something like this. I think they're just fun stories to cover. 439 00:33:19.695 --> 00:33:22.495 I do have a second one for this week, and it ties into 440 00:33:23.000 --> 00:33:32.360 the initial topic two that I skipped over, and we will jump back to that because that should be a pretty big one. But the second one of kinda unexpected chime in was 441 00:33:32.440 --> 00:33:33.960 Elizabeth Holmes, 442 00:33:34.360 --> 00:33:41.595 who I thought was in prison, and I I think maybe still is. Somebody correct me if I'm wrong. But of Theranos' 443 00:33:41.595 --> 00:33:47.355 fame, if you've ever read the book Bad Blood or seen, I'm sure there's documentaries about this, 444 00:33:47.995 --> 00:33:49.674 wrote a pretty interesting 445 00:33:49.755 --> 00:33:56.209 response and got a lot of engagement on on X over the past couple days as a response to, 446 00:33:56.450 --> 00:33:58.289 the news coming out of, 447 00:33:59.409 --> 00:34:03.409 Claude versus Open Claw and then also Project Glasswing, 448 00:34:03.730 --> 00:34:16.435 which was actually the

topic I was referencing and wanted to talk about. I don't know if there's too much to say about this other than, you know, if you're interested in reading, you know, some good entertainment on the timeline from somewhat of an unexpected source, 449 00:34:16.994 --> 00:34:19.635 check it out. I was surprised to see 450 00:34:20.835 --> 00:34:27.670 someone like Elizabeth Holmes chiming in on this. One, because I genuinely thought she was in prison, and I'm I'm not sure actually if she is or isn't. 451 00:34:28.150 --> 00:34:31.750 But then again, we see SPF pop up on the timeline still sometimes. 452 00:34:32.710 --> 00:34:37.430 But two, it was just another example of although it was someone who was in technology 453 00:34:37.430 --> 00:34:38.069 previously, 454 00:34:38.635 --> 00:34:40.635 not necessarily in AI, 455 00:34:40.875 --> 00:34:41.995 who's recognizing 456 00:34:41.995 --> 00:34:44.315 the importance of privacy 457 00:34:44.315 --> 00:34:45.035 and, 458 00:34:45.835 --> 00:34:49.355 personal ownership of data, and like, what some of the 459 00:34:50.075 --> 00:34:51.675 overarching implications 460 00:34:51.675 --> 00:34:53.995 are of 461 00:34:53.220 --> 00:34:57.140 some of not just the technologies, but the way in which they're being implemented, 462 00:34:57.940 --> 00:35:00.980 which I think is actually a good swing over to 463 00:35:01.460 --> 00:35:04.500 Project Glasswing and some of the other things going on. 464 00:35:04.820 --> 00:35:05.060 So 465 00:35:06.635 --> 00:35:10.475 if anyone is not familiar, so Project Glasswing 466 00:35:11.275 --> 00:35:14.955 is a project that was announced by Anthropic, 467 00:35:15.835 --> 00:35:17.115 basically where 468 00:35:17.115 --> 00:35:18.475 their new 469 00:35:19.195 --> 00:35:20.234 Mythos model, 470 00:35:21.400 --> 00:35:27.880 they believe and and are claiming, is so powerful that it's discovering zero day vulnerabilities, 471 00:35:28.280 --> 00:35:33.160 in some cases that have exist for decade existed for decades across 472 00:35:33.855 --> 00:35:37.535 pretty much, from my understanding, they're claiming that this is pretty much like most 473 00:35:38.255 --> 00:35:39.695 major softwares, 474 00:35:39.695 --> 00:35:41.215 or at least many, 475 00:35:41.215 --> 00:35:43.695 many, many companies are exposed to these things. 476 00:35:44.175 --> 00:35:47.295 And so they're they've created Project Glasswing, is basically 477 00:35:47.900 --> 00:35:49.020 a collaboration 478 00:35:49.020 --> 00:35:50.860 with other large companies, 479 00:35:51.580 --> 00:35:56.700 I guess, to give them access to using this tooling to fix these vulnerabilities 480 00:35:56.700 --> 00:35:59.980 prior to it going live as as like a retail product, 481 00:36:00.515 --> 00:36:03.955 which will eventually, you know, something similar will eventually become 482 00:36:04.195 --> 00:36:05.955 an open source product. 483 00:36:06.675 --> 00:36:13.875 Some of the companies involved are companies like Amazon Web Services, Apple, Cisco, Google, the Linux Foundation, Nvidia, 484 00:36:14.840 --> 00:36:15.720 JPM, 485 00:36:15.880 --> 00:36:16.840 Microsoft. 486 00:36:16.840 --> 00:36:20.280 So so really, really huge names. I'd be interested if anyone 487 00:36:20.520 --> 00:36:25.480 on stage here has any opinions on this or any thoughts on this just generally. 488 00:36:26.200 --> 00:36:29.400 Sure. I've got the hot take. I I think the 489 00:36:30.494 --> 00:36:34.494 a 16 z response was was probably the best thing I saw, 490 00:36:34.815 --> 00:36:44.335 which was Mark Andreessen's like, guys, we're about to enter a a moment where every vulnerability and every piece of software ever becomes exposed. 491 00:36:45.200 --> 00:36:48.079 And then in the next moment, there will be no more vulnerabilities. 492 00:36:48.960 --> 00:36:51.920 And and I thought that kinda captured it pretty well, 493 00:36:52.160 --> 00:36:53.280 which is, yes, 494 00:36:53.680 --> 00:36:55.920 as these models get incredibly 495 00:36:55.920 --> 00:36:56.480 good 496 00:36:56.720 --> 00:36:59.585 at pulling apart coding vulnerabilities, 497 00:36:59.905 --> 00:37:09.665 like, all the stuff that people didn't have time to go line by line and, you know, perfectly check and all that code that's built up for thirty, forty, fifty years, 498 00:37:09.905 --> 00:37:12.065 you know, it's all gonna get reexamined 499 00:37:12.065 --> 00:37:12.785 real quick. 500 00:37:13.560 --> 00:37:15.960 And I I'm glad that they're 501 00:37:16.120 --> 00:37:18.680 sort of doing responsible disclosure, 502 00:37:19.320 --> 00:37:23.800 which is what you'd call this in, like, a white hat hacker kinda context. 503 00:37:23.960 --> 00:37:27.640 It's like people that find a bug should report it to the maintainer 504 00:37:28.595 --> 00:37:37.795 quietly and let the maintainer fix it rather than be a bad actor and throw that kind of stuff out there for for, you know, people to to exploit. 505 00:37:37.954 --> 00:37:40.994 Right? So I think they're doing responsible disclosure, 506 00:37:40.994 --> 00:37:41.795 which I'm glad to see. 507 00:37:42.630 --> 00:37:44.950 And it's important. Right? Because 508 00:37:45.270 --> 00:37:45.910 if 509 00:37:46.150 --> 00:37:49.990 a bad actor had invented this level of model first, 510 00:37:50.150 --> 00:37:54.309 they would be going and having a field day with those zero days right now. 511 00:37:55.035 --> 00:37:55.675 Right? 512 00:37:55.915 --> 00:37:58.635 And instead, because they've been responsibly disclosed 513 00:37:58.714 --> 00:38:03.515 and there's been enough time, apparently, the maintainers have already gone and fixed 514 00:38:03.915 --> 00:38:11.570 a lot of what was initially found. So I think we're gonna kinda go through waves of this. But to your point, we're only, let's say, a few months out 515 00:38:11.970 --> 00:38:14.530 of a purely open source model 516 00:38:14.690 --> 00:38:19.170 being at that same level and finding all those

those zero days. 517 00:38:19.410 --> 00:38:27.165 So, you know and that's gonna sort of, I think, be the next wave is you want good actors adopting those tools quickly, 518 00:38:28.125 --> 00:38:30.525 having the best open source available, 519 00:38:31.405 --> 00:38:35.405 and so they can go and patch. Because there's a lot of people that aren't the Linux foundation 520 00:38:36.050 --> 00:38:39.570 who, you know, would probably like to find their own zero days, 521 00:38:40.290 --> 00:38:43.650 and, you know, they don't have the privileged access. 522 00:38:43.970 --> 00:38:44.530 Right? 523 00:38:44.770 --> 00:38:54.505 So, you know, yeah, that's it's gonna be an interesting future, but I I I think this moment is mostly gonna play out maybe the next few months, but certainly this year, 524 00:38:54.905 --> 00:38:58.745 right, where all the old software gets examined. And then going forward, 525 00:38:58.985 --> 00:39:15.360 you've got the ability to do it at an incredible level of auditing of all new code. Right? So I'm I'm not very doomer about this. Like, that's Anthropix's whole marketing thing is this thing is so dangerous, you know, and nobody else can use it. They've been saying that since ChatGPT two, 526 00:39:15.600 --> 00:39:18.240 right, when they when that team was still at OpenAI. 527 00:39:18.684 --> 00:39:19.244 Right? 528 00:39:19.484 --> 00:39:25.325 And I don't think anybody thinks ChatGP two was seriously dangerous in any way, but it's great marketing tactic. 529 00:39:26.204 --> 00:39:29.085 So, you know, I think they're playing also some of that angle. 530 00:39:29.244 --> 00:39:33.005 But, yeah, you know, I I think everybody's gonna get these capabilities pretty quickly. 531 00:39:33.910 --> 00:39:43.590 Do you think I guess the the one question I would have about it, though, is kinda to the point that you were making. So these people who are and by people, I mean institutions 532 00:39:43.590 --> 00:39:44.470 that are 533 00:39:44.790 --> 00:39:51.405 directly involved in Project Lastwing and so have this, like, premarket access access to fix their own code. 534 00:39:52.365 --> 00:40:05.100 I guess, are there, like, companies, like, smaller places that are gonna get left behind? Like, this will go retail before they have the ability to fix it for themselves, and now all of a sudden there's this, like, massive moment of vulnerability. 535 00:40:05.260 --> 00:40:15.180 I mean, I do agree with Andreessen's assessment that and and this is part of why I love, like, crypto and smart contracts is they are inherently antifragile. 536 00:40:15.180 --> 00:40:27.715 Like, every single time a smart contract has a hack, all the other smart contracts get better because they discover what the vulnerability was and fix their own code. And so, like, over time, it's it's a self fulfilling system. 537 00:40:28.355 --> 00:40:37.080 So I agree with that. But but is there like this split moment of time where there is such a big advantage to these players that are involved with Anthropic 538 00:40:37.080 --> 00:40:41.640 that everyone else doesn't have access to, and it and it's a huge problem? I 539 00:40:42.280 --> 00:40:50.365 can I'd love to jump in here on this one a little bit. If that's okay. Of course. I I I don't think there's gonna be a huge advantage from what I've read. 540 00:40:52.045 --> 00:40:58.285 What I would recommend is if you see a patch for many of those companies come out in the next few months, you better bloody apply it. 541 00:40:59.760 --> 00:41:02.720 Because I think, you know, as I as I read this, 542 00:41:03.040 --> 00:41:07.280 I I don't think it's gonna give a competitive advantage of those. I think they're gonna clean their house up 543 00:41:07.840 --> 00:41:11.120 so that you don't have that panic and moment. 544 00:41:12.185 --> 00:41:16.185 So as your, you know, as your Linux distros get patched, as your AWS, 545 00:41:16.985 --> 00:41:20.185 you know, your back end ECS things need to get updated, 546 00:41:20.345 --> 00:41:23.705 I think the community would be very well served by paying attention. 547 00:41:24.720 --> 00:41:27.680 Oddly enough, I got one of my first, you know, you know, 548 00:41:28.560 --> 00:41:29.760 iOS patches, 549 00:41:30.080 --> 00:41:38.800 you know, that came out and so I think being very attuned to those is good. So I think folks are gonna start cleaning up their houses, right? The the sixteen year old f f, you know, MPEG 550 00:41:39.945 --> 00:41:43.305 vulnerability that's always been out there, that's everywhere. 551 00:41:44.745 --> 00:41:48.105 So I think that we're gonna get some some disruption, 552 00:41:48.745 --> 00:41:56.180 but then I think that'll level out. To David's point is people will will use this to say, okay, we now have this tool for an as an offensive tool 553 00:41:56.740 --> 00:41:58.020 even if it's not 554 00:41:58.260 --> 00:42:00.660 even if it's not the the mythos 555 00:42:00.740 --> 00:42:05.220 is use the existing tools that you have available and work into your workflow. 556 00:42:05.460 --> 00:42:07.060 Hey, check this for vulnerabilities. 557 00:42:07.615 --> 00:42:17.615 Play a bad actor, how would this be, you know, how could this be used poorly? So I think even if it just wakes up for folks who are developing software to use the existing models 558 00:42:17.855 --> 00:42:24.970 as their own red team would be tremendously powerful. It's a really good wake up that says, hey, there's something 559 00:42:26.890 --> 00:42:31.290 Oh, I think I lost Alan on my side. Can can you guys still hear? 560 00:42:31.530 --> 00:42:32.090 I 561 00:42:32.810 --> 00:42:36.825 lost you. You guys are back. Yep. Pod pod dropped on me. 562 00:42:37.705 --> 00:42:40.985 Anyway, yeah. I think just pay attention to what's coming from folks. 563 00:42:41.625 --> 00:42:49.065 I wasn't sure from the reading that they were gonna release it broadly. Now that might have been some of the the it's too

dangerous to release for the public. 564 00:42:49.790 --> 00:42:58.830 Who knows? I you know, another model may you know, other models may come out that that are That they they they are they're at least already releasing pricing. 565 00:42:58.830 --> 00:43:03.390 So Oh. I think they're gonna kinda white list it. It's like $30 566 00:43:03.390 --> 00:43:04.910 per million tokens. 567 00:43:05.525 --> 00:43:06.165 It's 568 00:43:06.485 --> 00:43:12.325 it they are charging an arm and a leg, but, you know, I I figure if you've got a high 569 00:43:12.565 --> 00:43:13.365 value 570 00:43:13.525 --> 00:43:19.445 code base, you're probably dumping the money in and having to give it a go. This is the new VC raise. 571 00:43:21.010 --> 00:43:22.130 Yeah. Mistakes. 572 00:43:22.450 --> 00:43:24.530 Not that they're struggling here at at all, 573 00:43:24.930 --> 00:43:28.850 but interesting. Yeah. It's gonna be interesting from a security perspective. 574 00:43:28.930 --> 00:43:30.610 It'll 575 00:43:30.610 --> 00:43:31.090 be interesting. 576 00:43:33.665 --> 00:43:36.065 Opus is already 25 577 00:43:36.145 --> 00:43:53.079 per million output. I I can only imagine. I I I didn't realize the pricing was really so I I gotta take a look at that. Yeah. It might have been more. I I saw some quotes higher than that. But yeah. No. It's they're they're gonna, yeah, they're gonna make a lot of money on upgrading everybody's cybersecurity. 578 00:43:53.079 --> 00:43:57.319 Okay. Sure. But it was interesting to see that five point one on cybersecurity 579 00:43:57.319 --> 00:43:58.119 specifically 580 00:43:58.359 --> 00:43:59.720 scored above 581 00:43:59.960 --> 00:44:01.240 Claude 4.6. 582 00:44:01.914 --> 00:44:04.714 So like like Alan was saying, you know, 583 00:44:05.035 --> 00:44:09.915 implement this discipline yourself. Right? In my standard operating procedure now for coding, 584 00:44:10.315 --> 00:44:11.115 every 585 00:44:11.115 --> 00:44:20.210 piece of code that gets out out is audited first, and it's audited by a different model than wrote it. And that's one of the unlocks I've seen recently is 586 00:44:20.530 --> 00:44:22.050 have a different model 587 00:44:22.290 --> 00:44:24.290 do the testing, have a different model 588 00:44:24.530 --> 00:44:29.010 run the audit because the models are really good at passing their own tests. 589 00:44:29.410 --> 00:44:32.610 But I found cross model auditing is really powerful. 590 00:44:33.444 --> 00:44:36.405 You know, grab Grok, grab GLM 5.1, 591 00:44:36.405 --> 00:44:42.325 whatever you want, but, you know, get something else in there that didn't write the code, and you find a lot more gaps. 592 00:44:44.484 --> 00:44:49.339 That just reminded me of when, like, government agencies will audit themselves, 593 00:44:49.819 --> 00:44:57.020 and it'll be like, the FBI did an investigation and did not find any internal malpractice within the FBI. Thank you. 594 00:44:57.660 --> 00:44:59.019 It's kinda what it reminded me of. 595 00:45:00.595 --> 00:45:01.715 But, yes. 596 00:45:02.115 --> 00:45:04.035 Well, I can certainly say that, 597 00:45:04.755 --> 00:45:07.235 to your point, especially on the cost, 598 00:45:07.555 --> 00:45:14.355 once an open source version or something that competes with an open source exists, I know the place where I will be able to run it 599 00:45:15.279 --> 00:45:19.760 the most economically, and that is on Morpheus, which is or in Morpheus, 600 00:45:19.760 --> 00:45:20.880 through Morpheus, 601 00:45:22.000 --> 00:45:24.320 whatever the best vernacular is. 602 00:45:26.559 --> 00:45:32.365 Guys, that is the full scope of everything we had this week in terms of, 603 00:45:32.845 --> 00:45:35.405 agenda to cover. The only other topic, 604 00:45:36.125 --> 00:45:41.565 which is definitely a conversation for another week, but just wanted to put it on the radar 605 00:45:42.290 --> 00:45:45.410 of things to think about if you're not already thinking about it, is 606 00:45:45.730 --> 00:45:47.570 post quantum positioning 607 00:45:47.570 --> 00:45:53.810 for DAI and what that looks like. This is obviously just like a huge quantum in general is a huge topic, 608 00:45:54.050 --> 00:45:59.345 in tech right now and has been kind of all over the timeline from a crypto perspective. 609 00:45:59.345 --> 00:45:59.985 I know 610 00:46:01.025 --> 00:46:08.065 personal opinion here, Ethereum seems to be well positioned for this, which is one of the reasons that the 611 00:46:08.270 --> 00:46:12.670 token for Morpheus obviously lives within the Ethereum ecosystem. 612 00:46:13.069 --> 00:46:27.755 And we'll talk about this down the line, but just maybe like put a pin in this and and pin it as something to kinda think about that we'll talk about in the future, because I think it is one of the next topics that'll become most relevant to, you know, the things that we're covering. 613 00:46:28.315 --> 00:46:32.635 But does anyone else have anything else in the meantime that they want to make sure we cover this week? 614 00:46:35.390 --> 00:46:43.150 No. I mean, that's, that's a lot of great stuff. I mean, the number of models that came out is just, wild. There's a new large, 615 00:46:43.390 --> 00:46:44.590 Trinity model, 616 00:46:44.670 --> 00:46:45.710 and I thought that was, 617 00:46:46.270 --> 00:46:55.435 interesting, you know, just as a as a Morpheus fan. You know? I I gotta go try that one out. It's one of the larger open source models that's been released recently. 618 00:46:55.835 --> 00:46:58.715 So you got, you know, Google with their 619 00:46:59.115 --> 00:47:04.395 Gemma models on the small side. You got Trinity on the large side, and they got GLM 5.1 620 00:47:04.630 --> 00:47:08.950 somewhere in the middle. There's definitely, I think, demand for US 621 00:47:09.190 --> 00:47:20.645 open source. There's probably some companies that are sensitive to where a model is built or trained or or sort of long tail risks around, you know, poisoning of models and things like that. So good to see, 622 00:47:20.965 --> 00:47:24.565 more American models drop. I gotta go check out the meta ones. 623 00:47:25.045 --> 00:47:27.365 They just dropped a huge number 624 00:47:27.685 --> 00:47:31.285 of new models, and they've got a whole new naming scheme. They're not called Lama 625 00:47:31.720 --> 00:47:32.520 anymore. 626 00:47:33.000 --> 00:47:38.440 So I gotta I gotta check that out. But, yeah, it's been a hell of a week for for open source models, and so, 627 00:47:38.840 --> 00:47:41.080 hopefully, we'll continue to see even more of that. 628 00:47:43.240 --> 00:47:48.165 Absolutely. Well, we urge everyone to try them out. Something we 629 00:47:48.325 --> 00:47:56.005 at least I try to mention on pretty much every type of space that I'm on is the best thing for anyone listening to this to do right now is go try stuff. 630 00:47:56.644 --> 00:47:58.005 And through trial and error, 631 00:47:58.760 --> 00:48:00.520 we will slowly 632 00:48:00.520 --> 00:48:01.640 but surely 633 00:48:02.599 --> 00:48:03.320 become 634 00:48:03.400 --> 00:48:09.000 masters of our own domain as it relates to super intelligence and decentralized 635 00:48:09.000 --> 00:48:11.640 AI, and owning your own 636 00:48:11.880 --> 00:48:16.975 owning your own AI ultimately, which is of course, the end vision that we're all working towards. 637 00:48:17.695 --> 00:48:21.615 Well, with that, thank you everybody as always for jumping on. 638 00:48:22.335 --> 00:48:25.295 From the those on stage, for speaking, 639 00:48:25.375 --> 00:48:34.380 and those listening, if you're someone listening to the recording, thank you as well. We really appreciate it and look forward to next week where we will all 640 00:48:34.860 --> 00:48:35.900 reconvene 641 00:48:35.900 --> 00:48:41.020 and cover the latest in everything happening within DAI and Morpheus specifically. 642 00:48:41.740 --> 00:48:44.300 Alright, guys. See you next week. Keep being awesome.