What Agent To Agent Connection Actually Looks Like | Inside The Morpheus Agent Stack
58 minAudio
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via their podcast app and to have them all in one place. If you'd like to participate they are held weekly on the @MorpheusAIs Twitter spaces. So enjoy Episode #96! Recorded: 22 October 2026 AI SUMMARY: In this week's town hall a live demo showed how easy it is to launch a personal AI agent using OpenClaw with no terminal no downloads and no deep technical knowledge. Morpheus offers self-sovereign AI that you own and control with private access not relying on a centralised platform. Agents receive goals user context and preferences then connect into collaborative environments like Buzz to work with other agents on real tasks in parallel. Decentralised AI infrastructure includes privacy trusted execution environments and verifiable protections over "trust us" promises. Morpheus users are developers builders and teams seeking lower-cost inference and stronger privacy with use cases across business workflows internal tooling and government-grade scenarios. Recent node releases improved provider steering and security while agents continuously checking software and infrastructure becomes essential in a world of accelerating cyber threats. Timeline: (00:00:00) Intro (00:01:25) Live demo of one-click OpenClaw agent setup (00:05:13) Why owning your agent changes privacy and control (00:08:15) OpenClaw, Hermes and Buzz for agent collaboration (00:13:31) Where Morpheus fits as the inference layer (00:17:07) Trusted execution environments and provable privacy (00:24:19) Who uses Morpheus today and why developers care (00:29:05) Government use cases and secure agent workflows (00:34:02) Token economics and cutting long-term AI costs (00:38:10) Node updates, Uplink improvements and new models (00:47:27) Security scares, software updates and AI defence (00:57:02) Wrap-up and next week's town hall If you would like to support me for the value I provide you can use the referral link below when depositing capital and we will both earn bonus MOR. Alternatively you could send me a token of appreciation directly to my wallet ;) Referral Link: https://dashboard.mor.org/capital?referrer=0xE935f231c99c04Ee0b4532a3d0BdA81B152a0384 Wallet Address: 0xE935f231c99c04Ee0b4532a3d0BdA81B152a0384 Connect with Mere Mortals: Website: https://www.meremortalspodcasts.com/ Discord: https://discord.gg/dX9BEhT3kz Twitter/X: https://twitter.com/meremortalspods Instagram: https://www.instagram.com/meremortalspodcasts/ TikTok: https://www.tiktok.com/@meremortalspodcasts Value 4 Value Support: Website: https://www.meremortalspodcasts.com/support Paypal: https://www.paypal.com/paypalme/meremortalspodcast
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Bowen BelmerWell, everybody, again, thank you. Thank you so much. So today is Thursday, October 22. This is the Morpheus community town hall. And for the first time ever, we have a live group of people joining us in addition to the fact that everyone's joining us online. So the big thing that we wanted to do this week was really a demonstration of of d AI, really, and what can be done with this, how can you interact with Morpheus, how can you establish an agent for yourself, what does that actually mean,
and then kinda show off what it looks like to have agents interacting with each other. Because this is really the future that we see playing out. And the thing that makes what we're working on at Morpheus different from what a lot of other companies are doing is this idea of self sovereignty with your own access to intelligence and also your own recording of that intelligence. And so you don't have an agent that works on your behalf as long as XYZ business says that it's able to and is not dictated
by XYZ business's version of what it looks like to have something work on your behalf. You have something that is wholly owned and operated by you, can work on your behalf with other people's agents who are operated by them, and kind of the magic that happens, you know, from from that synergy. And so the first thing I'm gonna do here is we're gonna show off install openclaw
David Johnstondot x yz. So, hopefully, you guys can see this alright. Is that coming through? And and I I previewed this a a bit for people. Like, my goal was you don't have to download anything. You don't have to navigate a terminal. You don't have to be technical. You could just show up on a website, give the agent a name, and hit the launch button. Right? And really make it that easy. Like, I helped onboard a lot of friends and family to an agent in January. And, man, it was rough back then. It was, like, five hours of debugging terminals and stuff like that. So, hopefully, if you guys try this out, you'll see it's become a lot easier
to set up your agents. So, yeah, Bo, go ahead and and run through the process. Really, you know, I'll I'll talk it through step by step. After you've given a name, after you hit launch, what's gonna happen is it opens up this panel where it asks you to to sign in. And here's the great thing. You can use Google. You can use x. You can log in with other socials. You can use an email. If you have a wallet, you're a crypto person, great. You can sign in with your wallet. But I wanted to get this experience down to literally sort of the web two experience everybody is used to from the rest of the Internet.
So as as Beau logs in, he creates his account. What's happening in the background is it's creating him an agent and matching it to his credentials. So this is your agent. Only you can access this agent. I don't have access to the information. Beau, nope. Nobody in Morpheus has access, and that's through a partnership we have with another protocol called Manifest. And Manifest hosts the agent on their decentralized network, and Morpheus provides the inference, the GPUs,
to power the agent so it can talk and have a conversation with you. And so this has been a really cool combination where, hey. You can sign up. It looks like any other login you've ever done. It sends you a code, and now it's connected to your account. And so you're the only one that has the key. I like to say, if it's not your key, it's not your AI. Right? That's what we all learned in in web three, and I think it's even more important when it comes to controlling your intelligence and your brainpower
that you ultimately have the private key that controls it, and it's it's connected to yours. So, you know, that's that's the goal here is to get this down to a super simple interface that anybody can use. You'll notice it supports credit cards. So this is supposed to be mass market. Not everybody has crypto. Not everybody has stablecoins. Right? We wanna create something that works for everybody. Right? And almost everybody has a a credit card at least in the Western world, and so that's the the objective here,
is to give people the tools and meet them where they are. Right? Now it has a web three wallet in the back end. You can connect the wallet if you want, but you don't have to. Right? So we're not putting that impediment up front. It's something that people can use and sign up for very easily. So you can see Beau's gone through most of the setup process already, created the name, set up the account, connected it with his wallet, and it's now spinning up an actual virtual machine on manifest, and it's gonna connect to that. So we don't necessarily have to wait a couple of minutes, Bo. You could always log in to your regular account and show people what it looks like, but that's how easy this has gotten. He didn't download anything. He didn't have to, you know, manage any technical aspects.
It was just click a button,
Brian Bourgerieand it looks like a login that you would use in in anything else. And so And it's it's worth the the reminder too. Like, all this is basic when you just look at it, but it's all under the hood, the magic, because you're not being tracked, traced, harvested, or anything by any big tech company. Like, you're truly owning every aspect of your agent at this point Yep. And not reliant on
David Johnstonsome big tech third party or surveillance state. Yeah. And and just to double click on that, Bryan, you know, what I'd like to say is I want my agent ultimately to be a fiduciary to me. Right? It's trying to make me money. Its highest goal is to accomplish my objectives. And, you know, that's been my experience. You know, when I use these closed models, A lot of times why they hallucinate or lie or, like, make things up is there's something in that system prompt. There's something in those guardrails that the company put in there that's conflicting with what I wanna do.
Right? And and, quote, unquote, they do this for my protection, but the result is is the agent doesn't work. Right? And so when you're using these open models and you're using your own agent, all of that nonsense gets out of the way. You've got access to uncensored, permissionless intelligence, direct communication with with the intelligence, and you're setting the values and the goals. Right? So one of the things I'd recommend when you set up your agent is give it a profile give it a profile about you. These are called your sole file or your user file. Right? And tell it what you wanna accomplish. Right? What are what are the parameters that you care about? How you want to to interact with it? And the wonderful thing here is you have that... All of that flexibility
with this approach. And then once you've got your your agent set up... Bo, I don't know if you wanna move on to the Yeah. Yeah. I was gonna say just really quick. So I'm gonna have this continue to be
Bowen Belmersetting up in the background through manifest, so we'll we'll jump over to it once it's complete. But just so everyone knows too, this process now, which will take, you know, whatever it is, three to five minutes to establish after just doing those initial steps, is a process that the first time I did this as a not particularly technical person... So so as far as my involvement in Morpheus, I am the litmus test for the nontechnical person. I am surrounded by these absolute geniuses
who are fantastic in in software and hardware, a lot of them. And sometimes the joking test is if Beau can figure this out, anyone can figure it out. So it took me, I don't know, three hours, maybe more the initial time I set up an OpenClaw agent because I was going through and doing the command lines and all these things that I wasn't familiar with. Install OpenCloud XYZ has distilled this down to literally, like, a three to six minute process, and then it's live. And from there, and we'll show you this, you're interacting with it in complete natural language after that. And so if there's something you don't know how to do, you can literally ask it.
Tell me how to do this. I also just wanna flag too before we go on to the next thing that we're gonna show, which is arguably the much more interesting thing, is Morpheus itself is completely agnostic to what platform or people refer to it as a harness that your agent is using. So OpenClaw is just one open source harness where you might be hosting your agent. You could also be doing that via something like Hermes by News Research or something like Grokbot,
which is now very popular for people who are, you know, PowerX users. And what we're gonna show here... And let me just share this next screen then.
David JohnstonYeah. That's that's a great caveat, Beau, is there may be an install Hermes coming fairly shortly. That's... That might be in the works. And, you know, we wanna support all the different agentic frameworks and and give them the power. So what Beau's sharing on screen, if you guys haven't checked out Buzz yet, buzz.xyz, I I highly recommend it. This is a project that was was created by Jack Dorsey, the founder of Twitter,
and he's basically created, I would say, Slack for agents. So it's great that I have an agent, but I want my agent to talk to Alan's agent. Right? So my my agent is called Bernardo. You can see... And you can zoom in a little bit, Beau. And and Alan's agent is called Tank. And so we've put them in the same channel here in Buzz, and they're connected to GitHub. They know the context of what we're working on, and so this is a conversation from a little bit earlier today.
And I... Bernardo is is asking Alan's agent, Tank, about how to best use this new Morpheus GitHub repo dashboard that's been set up. And Tank gets tagged in the channel. He sees the information. He replies in line. And so now we have our agents natively going back and forth in this conversation. And so instead of working on things and, like, oh, I gotta go and, you know, talk to my agent about that or copy and paste, Now I have a collaborative environment
where me and Alan and our agents can can work in parallel. So, Alan, I don't know if you wanna add any flavor, but you're you're one of the gurus that that built out a lot of this. So
Alan (Nomadic Rogues)Sure. I I had a couple things to add on to our, you know, the new... The Hermes and OpenClaw comment is Rockbot. I've really gotten... I think that was a good iPhone moment, and I've gotten used to running the Rockbot. And so, yes, you're you're in that kind of ecosystem. But one of the things I wanted to do was, can I, you know, plug it into Morpheus? Can I reduce my cost and use some of those OSS and some of the benefits that, David, you talked about? And so if... For people who are familiar with Grokbot, they've... You can have teams
of bots talk to each other and you set... You kinda have your own... Have my own little little personal office of really brilliant agents that use both Cursor as the backend, but also through our Morpheus uplink, which is a personal API gateway, getting into the Morpheus network as well. And so I get the best of both worlds, which is awesome. And then when you say, okay, Tank, go ahead and use the invite link and connect into Buzz, now I've got my internal team communications in Grokbot,
and I can now have Tank talk to the rest of the community and get that interaction as well. And so I think it's really cool as you just look at the network of networks that's possible. So I've got, you know, a Grok bot, you know, Grok four, you know, with my team that's also using Morpheus OSS models that are out there and then through through Buzz talking to everybody else. And it's an interesting way to talk about that and and look at... But I think it's gonna explode. I think it's... I think this is gonna continue to grow, and, you know, people will understand how they have to put guardrails.
Like, my team shouldn't share any team information. We can talk about general public things, but hey, the harness that I've set up, make sure that, hey, tank, don't share anything that's that's not supposed to be shared. So it's interesting as you... Just like you would set up your own companies and your own teams, you do have to make sure you put put those ring fences around it. You've gotta train your employee agents to to interact with the rest of the world like you like you would with a, you know, company or an internal team. So,
yeah, I think it's pretty cool. If if if I could have a picture here, it would be, you know, Grokbot via Uplink to Morpheus to Buzz and then Openclaw through embedded skill to Morpheus to Buzz. And I think those Lego building blocks are really amazing. They show that you can pretty much plug anything to anything. And when, you know, as Beau mentioned, you can talk to them in my natural language. How do I do this? How do I plug this into that? They're gonna be able to help you walk through that. So that's what we've done here. Yeah. And and I just wanna double double click too on where
Bowen BelmerMorpheus specifically falls into making all of this magic happen. And it is the fact that any agent that you are talking to or hosting rather, like your agent, has to have a brain, meaning it has a large language model somewhere that it's tapping into to actually run the inference and get you the information. You have to have access to that hardware. Most people... My opinion, some people think a lot of stuff's gonna be locally ran. I think that there is...
There are things that will be locally ran, meaning there will be people who have, you know, certain varying degrees of hardware in their own home. However, as the models continue to get more robust, more impressive, faster, require more resources, the reality is those frontier models, especially both from an open source perspective and closed source perspective, are going to be hosted at institutional grade data centers. So the question is, how do you, as the individual user, guarantee to yourself and to your agent ultimately
that you have the ability to access those large language models without needing permission from any specific third party, without any type of necessary KYC. Now there is some nuance to that, of course, but... And also make sure that it's perpetual. So you know, my agent, I will have access to when I need it, where I need it, to the degree that I need it without anybody being able to disrupt that. That is ultimately what Morpheus is providing, which is why I was saying that
it really doesn't matter from the perspective of Morpheus what type of agent you're running, what harness you're running, even what exactly it is that you're, you know, choosing to use it for. Is it business purpose? Is it personal use? Morpheus is the place where you can permissionlessly, most privately relative to other options, and and always improving access that intelligence that is running on those pieces of hardware.
Brian BourgerieAnd we've, you know, gotten all the way to some compute providers who are doing what's called inference in a trusted execution environment, which is a whole another level of of, you know, depth that we don't necessarily have to get into unless if someone is is I think we should get into it because that's my favorite part and what I optimize everything for. Because at the end of the day, I don't care about paper privacy or a promise from a company that I don't trust. I care about, like, verifiable privacy.
Bowen BelmerOkay. So so as things exist today, there are various, you know, degrees of of privacy when you're using something like an AI tool. Obviously, if you're using a Bo, you should bring up tech.mora.org.
David JohnstonIt has that beautiful spectrum diagram Yes. I will. From centralized
Alan (Nomadic Rogues)Absolutely.
David JohnstonTo hyper private and local. And this is the audience that will will appreciate that kind of nuance. Yeah. No. I I think, Bryan, happy to dive into it. You know? And Alan's done a lot of the work on this, but how do you get a higher level of privacy without everybody having to have a $20,000 Mac studio? Right? And that's that's what we're trying to optimize for. You're like, the the ideal world is you're running everything locally, but there's just a big capital barrier there. Right? The average person globally
has a $150 Android phone. They they don't have an iPhone. They don't have a laptop. They certainly don't have a GPU. And so how do we get privacy in the hands of these folks? Alan, I don't know if you wanna walk us through the spectrum. You built this. This is your baby. Well,
Alan (Nomadic Rogues)I think the biggest thing I'd like to talk through is, and I think Brian said it, is, okay, I'm done with paper promises. Right? So what things can we do over time that's... That takes it out of either it's mine, as you mentioned, I've I've I've paid the paid the capital and I have, you know, this this rack of hardware. I've got a couple friends out there that keep showing me, you know, racks and racks of stuff. I'm like, most normal mortals don't have that kind of stuff, time, or expertise to do that.
And, you know, so the next the next layer up is, like, you you submit yourself to one of the commercial products, you know, OpenAI, Anthropic, you know, whomever. And your guess is as good as mine, cough Gemini, cough Facebook, how private it's gonna be. And okay, on any given day, that may not be an issue, but let's say you wanna talk about finance or your medical conditions or relationships, stuff that shouldn't be. And you'd start to talk and if you've invested in like, okay,
I wanna look at something like Morpheus or Venice or something like that. And we needed to get away from that trust me, bro, that paper promise that says, on paper, our lawyers have said, this is our wiggle room that says we can say that we really don't save your prompts or save your inference or save your data. And here comes TEE, trusted execution environment and what this is, is a model that says down to the silicon on the chip, down to the GPUs, down to the CPUs
that are running either LLMs or the provider nodes, we have a way that says, look, I guarantee you, here's the software that's running on that provider node. It's the same proxy router that we have that runs the Morpheus network, but it's built in such a way and containerized so that you can watermark and fingerprint and say, this image, and for things like, we don't, you know, the logging is turned off, any telemetry is turned off, you know, those kind of things that that it's not just trust me, bro, it's you can inspect the code and see how that that that that is turned off.
And then when it actually runs on a machine, there's an attestation that runs and there's a there's checks as a consumer starts their session with that trusted provider that that says, hey, you're supposed to be running this image with this SHA two fifty six, you know, this this commit code and your LLM that's behind you should be running this attestation. Are you still okay? And oh, by the way, did the TLS certificate that I originally started the conversation on my my API call,
is that still the same, or has someone gotten in in the middle of it and messed up that that TLS or that HTTPS certificate? And so all of those checks along the way guarantee that the software it's running, the model it provides, and the communication path that between you and that provider is, you know, physically, physics proves that it is private. And I think that's one of the strongest
supports or strongest reasons to start looking at Morpheus and where we're gonna go next. And that's providing, you know, bringing other TE providers on so that you can say, look, when you're in the Morpheus network, you know, you don't believe, you don't think, you don't. Well, maybe it does or maybe it doesn't. You'll be able to cryptographically prove that your conversations, the really private ones that you really wanna keep quiet or personal,
are being protected and not logged and not used for training, but not logged and or observed.
Brian BourgerieAnd to put that into perspective, applicable, what I do... Maybe I'll stand in front of the camera. We'll see what that looks like. Can't see myself, but either way. What I do is I actually have a setup where it routes to different models depending what I'm using and depending on if it's sensitive information or not. So anything that's really sensitive, private, finance, whatever, all the stuff you said, it can only and it's only allowed, and I've made it so the Morpheus router will only whitelist trust execution environment or completely provable
private models. So I don't have to, like, worry. It's either... I know it's either local or only touching something that's completely private. And then I have kind of a cascading scale where if it's semi... I want semi private, I'm willing to do that, and and that will be for a different set of things, and it still might redact information and stuff like that all the way to, okay, I'm using Claude or Grok or something. And some things if it's gonna be public, then, yeah, it doesn't matter. I don't care. Whatever. At a minimum, it normally
redacts out information. But I have that all automated at this point, and you can do that very simply by just telling your agent to set those things up. But you can really optimize for privacy where you never could before. I'd say even if one year ago from today, like, it just wasn't really possible or that easy. And now pretty much anyone can do it following a lot of steps that you just showed us, but
David Johnstonit's finally attainable, and I'm just excited about that. Yeah. Yep. Normally, there's a a trade off, unfortunately, between privacy and convenience. And, you know, I'd say what Alan and the other developers have done is make privacy easy. Right? And and so there's less of that that trade off, and now you can just pick the provider that has the level of privacy you care about, and and you have a a much better level of confidence than if you're sending it all to Claude or one of these other groups. So, yeah, big big credit down the guys. This is about two years of a heavy lift over time to keep improving and accessing these technologies, and it really just got there in the last six months.
Alan (Nomadic Rogues)Yeah. It's... It was a... It's been a it's been a a fight. It's not easy. And that's why there's a spectrum. Right? That's why there's this grade. Like, you can stop at three or four if you're happy with that. I think to... Kinda like what Brian was talking about is not all data is treated... Should be treated the same, so don't. It's gonna be expensive. The right hand side tends to be more expensive. The the number seven is on the right, is is the really expensive one where I've kitted my own rack of hardware. I've had to get new electrical entrance. So so look at the spectrum,
you click on the bubbles, and and see where you wanna go, how far you wanna go. And I think that we've got a a good smorgasbord or buffet of of options there on the spectrum.
David JohnstonAwesome. That's great.
Brian BourgerieGood stuff. Do we have any questions actually from the group for anything that they've talked about? If not, we'll let them continue on with the show. But does anyone have questions about anything on this specifically or more decentralized AI or infrastructural stuff? K.
David JohnstonAlright.
UnknownYeah. No worries. Okay. Awesome. Go ahead. Were you gonna say something? Oh, no. Go ahead. Did you have a question? Sorry. I do have a question for you. Oh, amazing. Hey. So, hey. How are you? Jeff Crane here. Just got back from Michael Turpin and Joyce Chao's event in Tokenize in Las Vegas. We had about three and a half days, and it was straight AI. Question for you is, I've I've... Of course, I know Luke Stokes and all the all the guys and Bryan Christensen, OG, from way back when. And so you have a lot of super smart people, and they're all in this room too here in Utah. Quick question for you. Who's the avatar? Who's the people who are using it right now? Give us kind of like a
demographic or psychographic of the of the people who are using your service right now.
David JohnstonYou know, I would say today the primary persona is developers. Right? It's it's people that are building with these tools. And, honestly, Claude and a lot of these others have gotten really expensive. Right? You know, Claude keeps raising their rates. OpenAI keeps raising their rates. It used to be $20, and then you needed the $200 plan, then you needed the $500 plan. I... Back in January, I was spending $6 a month building amazing stuff with Claude.
When I switched to the open weight models, it cut my bill by 90%. Right? And so people are waking up to the fact that there's this Claude tax, and now that the open weight models have become so powerful, they have an alternative. They have a real alternative to keep building and getting in a similar level of performance at a fraction of the cost. So, you know, Beau, maybe if Yeah. I have a good example for these... Oh, okay. I was I was gonna say if you pop over to the tab on tech.mor.org
that shows off all the models, like, you can see how many models that are available on Morpheus now if you go to... Yeah. You can see GLM 5.2, DeepSeek v four Flash. Like, these are the power horses that have really replaced because you can use Claude on Morpheus. Like, people are offering all of the models permission to draw... Permissionless. This is what the market is choosing. Right? It's picking these high performance open weight models that are a fraction of the cost.
Right? And so, yeah, I would say that's that's the persona. Right? It tends very heavily towards engineers, builders, developers, people that want access, but don't wanna pay through the nose. Right? Yeah. Like, privacy folks are definitely there, but, you know, wanting to build at a more affordable prices is a huge motivator. Go, Bill. Yeah, David. I have a good example of this too. And, you know, while
Bowen Belmerwhile my perfect world looks like every individual having their own self hosted agent that uses Morpheus for his inference, The reality is there is a huge amount of people who will experience AI and decentralized AI through some type of front end service that's being offered to them. I mean, we see this in crypto. Of course, everyone who works on this project and all the adjacent projects are also, you know, natives to crypto. And everyone's been waiting to have that self custody moment for, you know, almost two decades now. Right? And so to some degree, that is the way people are gonna experience it.
Outside of Morpheus, my professional background is in finance and more specifically real estate finance. I have now built multiple tools that my team uses for various portions of managing our clients and managing our work. And because we have a fiduciary responsibility to those clients and there is so much, you know, stress and and really critical necessity on things like privacy, on things like retaining our own data, Morpheus is the only place that I can realistically trust to provide that to me for these tools that I've built. Because,
basically, the options are, okay, host everything locally so that I can verify that that data exclusively lives on that machine or host it through a permissionless private avenue, basically, in this case, Morpheus, where I can prove with physics that that data is remaining private without me needing to have enterprise level hardware, you know, in in my own offices. And so my employees who use these things often don't necessarily know or care ultimately
what what Morpheus is and how it works, but they're experiencing it every day through the fact that that's what's lifting up the back end of the tools for... That we're internally using. And and I know that there's others like me too in various different capacities. That's just kind of the the best firsthand account I can give. Yeah. There's also gonna be at least two or three presentations later today that are different products, but that run on Morpheus as well.
Brian BourgerieWe have a couple more questions.
UnknownGo for it. Hey, guys. This is Nick. This is really awesome. Thanks for sharing. Thanks. This is really good work. I just wanna make sure I'm understanding the stack correctly. So I've got a huge database. It pertains to a lot of governments, and we're helping build agents for those governments to use the data to help them execute their key workflows. We've built those agents in the LangFlow environment and agnostic kind of different LLMs, but I'm still, you know, constantly questioning the security of that. So is there
is this if I'm understanding this right, is I don't know if Morpheus is a replacement kind of the LangFlu infrastructure, but then seems like this Morpheus would be much have much more natural privacy protections in place. And then is there a world way down the road where cities, counties, local governments have agents for them that might be posing these questions about their performance in Buzz? And then if our agents who are doing this work with their data for them are also in Buzz, they could be answering their questions
for them, and then I would have the assurance that they're only sharing their own data, and everything else is kinda locked down appropriately. Just wanna make sure I'm understanding this all correctly.
David JohnstonYeah. That... That's that's a great question. I would say Lang Chan is an example of a good harness. Right? It's it's a tool set. And so Morpheus is not replacing that. It's something you can use with Morpheus. Right? It can provide the inference, the models, the LMs that you need to power your langchen setup. And, yes, I I think that's a interesting use case. You can definitely know for sure that it only has that information if you're, you know, connecting that agent.
And a lot of people are starting to use agents for all sorts of support and other tasks. Right? There's there's far more interactions that we'd all like to have than we have time to have. So I love the idea that I have an agent, and they can answer Alan's question at 2AM or on the weekend or when I'm traveling in in an airplane, it's still there, and it can still get something done. Think about how that transforms work. Right? Today, you have a weekly meeting.
You know, you you you have a a transcription to that meeting. Now I just dump that transcription into my agent and say, hey. Give me all the outcomes, all the tasks, and all the things that need to be done, and start working on them. And I'll check back in an hour later, and the things we were talking about during the meeting are already getting executed on. I didn't have to wait for next time I had an open time to to focus on it. But... Yeah. Nextly...
Bowen BelmerOh, go ahead, Moe. Oh, I was gonna say I love this question and things like it, especially if it ever pertains to things like government and national security or security of public... Private public information or the ability to keep that information private. I think Morpheus specifically is such a strong natural fit for security of government RAN agents and and really just, like, use of AI in general for for this main reason. And and this is now gonna be...
I'm gonna do just a slight teetering into Tinfoil Hatville. So everybody, you know, hold your chairs. But I just want you to think about a world where we are using agentic AI for all types of public public works, whether that's local governments, whether that's the military, whether it's, you know, what have you. If you are centrally running all of that inference in physical data centers that are operated by those government agencies or private companies
operate... Or or, you know, backing those government agencies, that is a limited number of attack vectors to cause a problem. Even if it's a 100 data centers, you can feasibly attack physically a 100 data centers, unfortunately, with modern technology and especially with, you know, the increase of drones and things like this. If you are hosting your agents with decentralized infrastructure that's encrypted end to end, you as the user do not even know where the hardware is actually located,
and you don't even... You don't really care where it's actually located because you can verify with physics that it's running the thing that you want it to run and it's doing so privately, how can that be attacked? Gun to my head. How do I shut your agent off? I don't know.
David JohnstonYep. Well, I mean, that was the original inspiration for the Internet. Right? Is they wanted a network that they couldn't fall down and didn't have a central point of failure. So Morpheus is that for AI. But great question.
Brian BourgerieDo we have any other questions? We just have a couple more minutes on our end. And if not, it's okay.
Alan (Nomadic Rogues)I think
Brian Bourgeriethat's all the questions we have. I'm gonna do a preamble just really quick to something we're gonna discuss later because there's a whole other aspect to Morpheus with a more token and other things and staking it to get inference perpetually that we're not gonna go into all in this live stream, but there is a whole other component that I'm excited about. And the one thing I'll say on it is maybe a teaser, but, like, this month is my one year anniversary of running my OpenCode and then Hermes agents 100%
on Morpheus compute without paying any API fees. So I've staked the same, I think, like, $2,500 ish worth of token for an entire year nonstop, permissionless access to private compute that would have cost me over $70,000 if I would have paid API costs throughout the whole year. So from, like, a business perspective, and we'll get deep into this after lunch, it's just phenomenal on my end because I've saved literally tens of thousands of dollars while having better privacy
Bowen Belmerthan I could have anywhere else. And for me, that's like the holy grail. Yeah. That's the best part, Bryan, is all of the things that we just said, which are specifically gonna be things that developers care about and people who are very focused on privacy care about, which which is great and should exist, is not even the biggest win for most people because most people are already willing to give up those privacy rights, which is why they use things like ChatGPT and Claude in the first place. Fine. If you're someone who doesn't care about that... Well, I I wish you would, but let's just say that you don't,
there is still a massive, undeniable economic benefit to using Morpheus that, yeah, you guys will definitely get into it, but that's... Yeah. We, like, we just scratched the surface with the privacy piece.
Rob (DRM3)Yep. Yeah. I'll I'll say, this is Rob from DRM three. Hey, guys. That's that's initially why I got into it as well. It it kind of changes... Hello? Yep. We got you. Yep. We got you. It kind of moves you from OPEX expenditure to CAPEX. Right? So you're buying something. And so when Morpheus says you own your AI, that's real. You you you actually own it and it's perpetual. So those that got in early with decentralized AI have a great situation where they've got it effectively perpetually.
So good.
Brian BourgerieYeah. Awesome. Thanks for letting us monopolize
David Johnstonyour time.
Brian BourgerieBut... Yeah. So go ahead and show... We have some QR codes up here.
Bowen BelmerLet's see where Are those big enough, or do I need to make them bigger? Always Bigger's always better in this case,
Brian Bourgerieeven though we're not in Texas. But Let me see if I can full screen it for you. Hold on. And those go to... Okay. So for anyone who's sitting in the back, the first one is just a link to install OpenClaw to x y z if you wanna do that one click agent setup. Second one is if you wanna join the community, kinda Morpheus community on Buzz on that platform and start chatting with both humans and agents about these sorts of things, you're more than welcome to join on that front as well.
David JohnstonVery cool. Well... And I wanna say, Bryan, thanks for hosting. This has been a very fun way to do a town hall and to include all the people there in person. Like, hopefully, you know, this has been interesting and exciting. I just encourage you, if you don't have an agent, set one up. Like, having an agent and not having an agent. I mean, really having an agent. One working for you, a personal agent, is gonna be a 10 or a 100 x difference in the next year or so. So it's worth the time to get up that learning curve, try it out, and and get set up. But thanks again for hosting, Bryan. We appreciate it. Yeah. Thank you. And leave it back for the last five minutes. We're gonna hang up for five minutes, but continue with the livestream for everyone else, and then we'll drop off in about five.
Bowen BelmerThank you, guys. Do, and thank you for the patience, everyone, with the tech issues. Somebody somewhere doesn't want you to hear about this, and you should think you should think deeply about that. We
Brian Bourgeriewill. We'll ponder it. Yeah. Yeah. I think it sounds good. Right. K. Thanks, guys.
Bowen BelmerAwesome. So for regular livestream viewers, let's let's kick off really quick. Alan, I know that we had a node update this week, 17.1six. What is going on with that?
Alan (Nomadic Rogues)7 dot 16. Or 7 dot 16, my apologies. That's that's okay. That's no problem. So we talked a little bit about last week on seven fifteen is starting to do provider steering, and I talked about when you wanna do thinking or non thinking models or context size or tools or images, you know, seven fifteen was kind of the MVP of that to say, look, regardless of the model, where it's hosted and who hosts it,
I wouldn't be able to turn thinking on or off and that's that's pretty complex because you've got to run it through this, you know, every model and provider kind of accepts those instructions differently and so we've kind of tried to do our our best at a Rosetta Stone to do that. And so with seven sixteen, we've expanded that, not just thinking on or off, but also context size tools, images, web search, structured answers, and reasoning. So that gives, again, better steering as as you start to expand your agent's ability to talk to those models is is actually fine tuning them and and getting them to the right model and the right bid that's out there.
Also, some other security fixes in the CLI. We had a really good... Hey. We had a new member just join the Buzz community, by the way. It's got a pop up that says someone else joined. So pretty cool. That was a neat neat pop up. So in the CLI for Morpheus, we had some defaults that were were pretty big, so a community member helped us shrink that down to ten minutes. And then I think Rob himself put in one for command line, some of the error trapping that we we didn't have
handled well. So again, that community feedback is great. And so that's where seven sixteen, so features and some bug fixes. And also on Uplink, which was the personal API gateway, just hardened that, added two factor authentication, and some fallback in fail modes for when models, you know, models or providers give feedback that they are... They're not available or they have an issue. So, again, some good good features on the Morpheus Uplink personal API gateway.
David JohnstonThat's awesome. Yeah. The other thing I think worth saying, though, is, like, the open source models this week have been coming fast. Like, there is a ton of new models out there. The new DeepSeek from what it it seems is incredibly powerful. Like, it's benchmarking above Fable five, right, levels. And I'm seeing it match even current models when it comes to cybersecurity, right, and other aspects that the closed models sort of have a lot of censorship around.
And so, yeah, it's it's an exciting time. We're gonna need all of those models with all of the sort of hacks that's been going around. The only defense is to have an agent that's constantly doing security checks and upgrades and updates and things like that. So really important to have that. I noticed the other day on that they now build in security checks to your entire tech stack for your website, and they're pushing it. It costs zero credits. They're like, do this check. Push this button, and it it... It's amazing how good the the security checks are getting. Go ahead, Alan. I... Just a quick comment on that. In in my past life, working with
Alan (Nomadic Rogues)lots of stuff in the... In in history, things like those security announcements over the last couple days, the PHP and the node, would have caused mass panic. I mean, people are panicking. There's a lot of stuff that's going on. But this morning, over morning coffee, I said, you know, from two different organizations, I've got a 104 repos. Go tell me how bad it is. And the agent went off and told me, hey, you've got three. Go fix these. And so the ability for having those... You you mentioned that 100 x,
you know, effectiveness. This was over coffee. You know, I'm cooking eggs while that agent is going and telling me how panicked I should be, and the good thing is it's not. So that's the accelerator that we're seeing with these and that agent. And oh, by the way, saying, hey, here are my refos that I worry about. Don't want that on anybody else's servers. I want that going through Morpheus private inference so that I don't, you know, pre on their... Pre pre alert something. So it's just a game changer. I I agree, David. If you don't have an agent,
get one. Yeah. And put it on Morpheus.
Bowen BelmerMhmm. Yeah. I will... I'd love... So in in the deepsake news, David, did you see that they raised nearly $15,000,000,000 in this current funding round, which is... I didn't I didn't see exactly where the source was, but probably China is probably probably where a lot of that capital came from. What I love about this, though, is it puts such a large amount of pressure on the Frontier Labs in The US and also just continues to perpetuate open source. Because in this case specifically, China just generally not respecting,
like... What's the word I'm looking for? Like, copyright laws and things like this is a huge win for the average person who's using open source AI. Because, basically, you know, historically, it's like, oh, well, this stuff got, you know, effectively copied and improved upon or whatever, and now it's offered in the Chinese market, specifically by these Chinese companies that we don't have access to. That's not the case anymore. Now it's like, oh, well, we benefit everywhere from this happening, so net positive.
David JohnstonYeah. Now everybody's sort of getting the benefit of that intelligence work that's being done. It was interesting to see the US government decided they don't like DeepSeek, and they're now not letting at least any of the government agencies access DeepSeek. It may... I... I've heard mixed reviews that it may be only applicable to accessing the Chinese based API of DeepSeek or whether it applies to all models. It's hard to know how you would apply that to something locally hosted and even know what models it it is.
But, you know, I think we're gonna see more of that, and you sort of have The US and and China and their opposing spheres. But, honestly, the competition is benefiting everybody. Right? And and it's driving down the cost of tokens. It's driving up the value of the intelligence. I think we're just gonna see more and more like that. I I hope to... That GLM is not added to that list because they're based in Singapore. And so people use the word Chinese very broadly as, like, they're in Singapore. Like, the... We... You know, I go to Singapore. I know they're in Singapore. Like...
You know? So let... Let's use a more nuance. And so that's... It's one of my favorite models. So Yeah. Hopefully, that avoids the flack of of the nationalistic stuff. Well, I actually think it's pretty rational to have a ban if it is specifically
Bowen Belmerfor the provider. If it's specifically like, hey. We're not gonna use DeepSeek as a as a large language model via centralized providers located in China. Sure. Yeah. I get it. That makes sense. You're you're giving privacy to... I'm looking right here. It was the Department of Homeland Security. Yeah. That's a rational position. But this is, again, when... What I was saying earlier, where this is why Morpheus is so important because you can have access to the same level of intelligence and these same tools without needing to worry about,
oh, this provider is ran on this hardware that's being monitored by x y z country, x y z entity. It it solves for that. Like, that's that's the point. Again, like, not even mentioning the economic components, which are arguably the other best part. Yeah. So it is it is interesting. Now the question is, you know, are the the powers that be in the government... Is this a rational decision that they're making on purpose, or is it accidentally rational? And then they're gonna be irrational in how it's implemented by saying, hey. Categorically,
we're not gonna use, you know, this model despite the fact that it's open source and we can measure it and, you know, all of these other things? Yeah. I did see that, though. I thought it was pretty interesting. What other... I didn't see another model released this week, David. What was the other one?
David JohnstonThere were a few. I think there's a new Mini Max. There's, I think, a different version now of Muse. There there was a whole series of them that that came out. There's also an American open source, but it it didn't seem like it it ranked that high or was that really interesting. I'd be more powerful than than Jem and some of the older models, but, you know, there was just a whole flurry. And then I think that news got subsumed by the cybersecurity news. Mhmm. Right? Yeah. And that's been the whole news cycle the last forty eight hours. Yeah. Can you guys explain to me, like I'm me,
Bowen Belmerwhat exactly is going on there?
David JohnstonSo there are a couple of big programming language libraries and things that a lot of people use for the Internet. PHP is one of those. Node is another one of those. So this is why it's important to update your your software. Node is one of the things affected, and Morpheus and OpenCLAW and and everything uses Node. So it... It's a good idea to do your update. You know, I went in and, like Alan said, I asked my agent, here's the list of recent hacks. Which ones
are my software affected by? Oh, this one. This one. Okay. Update the security patch. Done. Right? So it doesn't take a lot of work necessarily to fix it, but you gotta be on top of it. And so highly recommend people upgrade to the latest version.
Bowen BelmerAgreed. I'm actually... As you were talking, I was replying to my agent to make sure it checks to to see if it was on on the latest version.
David JohnstonYep. That would be a good thing to do. And so PrinceAll OpenClaw made made the upgrades on GitHub. And so... Yeah. Now everybody... It's it's good to... If you see a security pop up on any of your devices or stuff, make sure, one, it's coming from, you know, the the real app store or you're going directly to the website. But if it's a legit upgrade, you you really wanna do it right now. So there's there's a lot of stuff floating around. You don't wanna have your your file stolen or stuff like that. So, you know, I think we're gonna keep going through this. The next
six months would be my guess, is every bug that exists is gonna get found. There's swarms of agents everywhere on the Internet now trying to find all these zero days and and and steal money and and hack. Right? And so all the world's software that was ever written before this moment, right, is gonna get reviewed with a fine tooth comb, and they're gonna find all those bugs. My my hope is going forward,
the cost to audit, the cost of security software has never been lower. Like, when I when I publish something that's gone through 10 audits by different models with different approaches, and it gets to a point where, like, okay, I think this is this is pretty solid. Right? But we've never had the ability to do that before. People were just writing code, and maybe they got somebody else to look at it, and that was it. Right? Maybe they had a test suite. Maybe they did a few things if they were sophisticated, but, you know, I think the
the the challenge now is you gotta take securities more seriously, but you have the tools to do that. Right? You should have an agent auditing every piece of code before it goes to production. Hey. Throw three agents at it. Have three different models do it. Have them use different libraries and approaches. Right? So my hope is we'll get through the next six months, and we'll come out the other side, and code will be bulletproof. It will be post quantum. It will be... You know, even people are moving towards, like, deterministic code.
You know? Show me it can't be hacked. Like, how do you get to those those type of approaches? I think there's gonna be more and more pressure to to get there the next six, twelve months. Yeah. The whole thing is just... It's antifragile.
Bowen BelmerYep. Like, it... It's exactly what's so powerful with DeFi. Like, you build a system that improves upon itself and exploits literally make the system stronger. And, yeah, we will eventually get to the point where we... Or at least beyond the type... Like, any hacking that a person will be able to do. And then it's just a measure of having the white hat agents up to speed with the black hat agents. And I'm sure there'll be some relative cat and mouse game that happens with that to a certain point, and then...
David JohnstonYou know? And and I wrote my article. This is the ten year anniversary of my article on this subject. Right? So if you look up my old article on how blockchain will save us from the coming of strong AI, this is literally what I was talking about. And and I think we are in a good place, though, because the white hat, the companies, the individuals, right, they make up a $120,000,000,000,000 economy. We are way better resourced than the bad actors. Right? And so we just have to avoid the temptation to run to safety by giving one government or one company all the power
as if they would magically solve these problems for us. Like, people have gotta solve these problems themselves and avoid that centralization, and we're in a good spot so far. But this is this is going well in in the sense that, like you were saying, Alan, it's easier than ever to address these things. They're not, you know, oh, I gotta rebuild everything for a month. Right? Or rip everything apart, scan it in five minutes. Oh, this is what I gotta update. Done.
Alan (Nomadic Rogues)Right? Yeah. Searching searching through code for that one call anymore is is gone. And the Amplitude. Right? So this is this is what happened in in Web one point o in the early days of this is is your attackers all of a sudden had the capability, and the panic and the and the doom and the gloom were so huge. But eventually... And and then the defense against it, and then the another attack in the defense. And eventually we got into that that settled out. We now have a new inflection point where, oh my god, you're finding that all these models found huge exploits,
but you're able to meet them. And we'll see that amplitude go down over time. But you're right. In that panic time, you can't go and say, oh, save me, government, you know, insert Beau's comments earlier, because it's just not gonna work. So I think that... I think we'll we'll all level out and and on the other side of this, we'll have a good... Stasis is probably the wrong word, but balance between the red that that that, you know, the the red and the white teams. It's gonna happen. People want to exploit. Our defenses get better, and hopefully,
preventative measures, you know, kind of as what what Beau was talking about. This is, you know, can you create the hack proof system? Can you create the vulnerable... Vulnerability proof system? Doubt it? You know, people created it. People can break it. But, yeah, the the waves are high, but they'll get down. They'll they'll get better.
Bowen BelmerYeah. If, well, if the whole Internet doesn't get shut off, which is apparently what we're looking at right now. So this is all under the assumption that the Internet still exists tomorrow, but I... You know, there's good odds.
David JohnstonYeah. I I I think we'll be okay. We'll be okay.
Bowen BelmerI'm I'm seeing just, like, astronomical levels of FUD on the timeline, though, of not not of, like, us or anything, but just, like, generally the... There's a... I'm looking at a tweet right now. It has nearly a million impressions, and it's literally a video that has a square that says PHP on it. Let me just show you, actually. I keep forgetting. This is why we're doing a video show now, by the way, guys, so I can share the ridiculous stuff that I I see.
They're literally making it look like nuclear Armageddon, which for some reason is originating in this North Africa, which I'm not sure why. That feels probably not not where this came from. But maybe. Who knows?
David JohnstonThe Sahara is is really Yeah. That's right. Really known for their hackers. Yeah. The desert hates...
Bowen BelmerYeah. Exactly.
David JohnstonBut, anyway, yeah, record levels of fear going on in the market now. To be fair, like, the Indian one, it's like, oh my gosh. You guys saw the number four? It's like somebody in charge of setting the private key for one of the HTTPS certificates had literally just used the number four, and it's like, come on. What? Oh, you know That
Alan (Nomadic Rogues)should have been preventable. That should have been... That should have never left the gate. That's one of those things that should never have made... Seen the light of day. Yeah. But, like, when he was... Or he or she was designing it, it was kinda genius. I wouldn't have guessed for...
David JohnstonYou know? Everyone's sitting there like, it's gotta be something way harder than that. No way they would make it one number. And Oh, man. Like, you you hear these stories sometimes, and it's either, like, crazy or genius. Like, apparently, for, like, thirty years, the nuclear code in The US was, like, eight zeros or something. It's like, they'll never even believe that's a real code. Like, look. It's their That's the same they'll think that's real. It's like
Alan (Nomadic Rogues)That's the same combination I have with my luggage. Yeah. Exactly. Right. 12345.
David JohnstonYeah. It's kinda genius. Hey. They're gonna make another Spaceball. Spaceball two is coming. Oh, is that right? Can't can't wait. If you're too young to know the reference, that is the reference Alan is making. Thank you. I'm glad. So I'm I'm glad someone got that. Enough. I'm old enough. I've I've seen it, but I... I'm not gonna lie. I didn't catch that much right away. Thank thank goodness. Three,
Alan (Nomadic Rogues)5 is right out.
David JohnstonYes. Yeah. Yeah. Three. Yeah. Five is right out. Exactly. So we just go all the bad references. Oh, man.
Bowen BelmerGood stuff. Well, should probably wrap it up. It's been an awesome fun. This was a good this was a good experience. A lot of fun. Yeah. Thanks, guys. We will be back next week normal time, normal place. And by normal place, I mean, I will figure out how to make YouTube turn back on by then. So if you're watching this in Provo, thank you, first of all. Or if you're watching this via X right now, also thank you. Sorry that the YouTube link... I don't know what's going on here. It's still... Riverside is just not connecting with YouTube. So YouTube YouTube may have been one of the victims of Yeah. I wonder if they shut down the hatches. YouTube was like, absolutely not. Like Mhmm.
David JohnstonNothing.
Bowen BelmerConnectors. Break them all. Yeah. I don't know. I bet if MrBeast was going live right now, it would magically work. Something tells me. I'm just kidding. Anyway, alright, guys. Thank you so much. We will talk to you all next week. Take care. Sounds good.
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...