Mere Morpheus

Mere Morpheus artwork

Official AI pDOOM is at 10%? | Morpheus Crosses 10 Billion P2P Tokens

56 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 #90! Recorded: 10 September 2026 AI SUMMARY: In this week's town hall a huge moment for the Morpheus network arrived as it closed in on 50 billion tokens processed year-to-date and crossed 10 billion peer-to-peer tokens live. Real usage is growing with rapid new model releases from players like DeepSeek and GLM and cheaper stronger flash models are bringing practical everyday AI workflows closer. Personal multi-agent setups are already useful in daily life from automated research to custom morning briefings making intelligence more accessible and more personal. The gap between fear-driven AI headlines and meaningful solutions remains wide while Morpheus keeps focus on open transparent and decentralised access and proof of identity and verifiable systems grow more important as AI-generated scams and synthetic media improve. A simpler OpenClaw launch through installopenclaw.xyz allows start in under a minute with email or Google login plus Morpheus-powered inference. The latest node release brings provider health-check improvements better visibility into active bids and community contributions making the network more reliable and easier to navigate. https://www.youtube.com/watch?v=26y0_hBRdNo Timeline: (00:00:00) Intro (00:00:30) Town hall opens with network growth milestones (00:02:57) New model releases and the Flash race heats up (00:06:06) AI safety fears versus open and practical solutions (00:10:02) OpenAI maths claims, data use and trust issues (00:16:30) A decade old case for blockchain and strong AI (00:23:02) Fruit fly brains, simulated neurons and strange futures (00:29:18) Install OpenClaw demo with easy agent setup (00:33:32) Agent custody, pricing and mainstream onboarding (00:40:50) Node 7.11.0 update and provider health checks (00:49:49) Verification, attestations and why trust is not enough (00:52:32) Doom predictions, cosmic timelines and next week teaser 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 BelmerWe should be live for the Morpheus community town fall on 09/10/2026. Let me refresh our page here. We're still alright. Woah. Alright. We're live. Cool. Sorry to mute myself on the other channel. So we are live on YouTube on X and really just those two platforms. That's pretty much where DAI lives at this point. But, yeah. Hey, guys. How are you? How's everybody?

David JohnstonLife's good, man. Good to be here. Doing great. Another Yeah. Very good.

Bowen BelmerGood. Glad to hear it. Lots to talk about this week. As people can tell, we are making small incremental tech improvements to this every week, and it'll be a real live, great livestream pretty soon before know it. So give us give us some grace on that. Or or in other words, we appreciate the grace that we are given. But as always, a ton of great stuff to talk about this week. The best place to start, and and, you know, we start here every week, but is what is going on with Morpheus from a network perspective.

We are very, very closely closing in on 50,000,000,000 tokens processed sent, year to date, so since January 1, which is obviously something to celebrate. I mean, that's just a ton of tokens. And and what's crazy is, like, you know, the actual, use cases and people using it are increasing, you know, by the day, and so we're just gonna continue to see that number accelerate. If we look once again, particularly at

native access to, to Morpheus, We are on pace right now to have well, we're we're on a similar pace to where we were in August. So it's I guess there is still potential that we'll eclipse it and can be, you know, up into the right again month over month. But I won't be shocked if that doesn't happen, but it's still like, the jump from August to July especially and June, you know, prior to that was so large. And I just realized, and again, guys, we're still

working on the tech here. I can share my screen. So

David JohnstonI hope everyone And you'll notice in the top left bow, we have just officially crossed 10,000,000,000 tokens peer to peer. And that happened, like, literally in the last few minutes. So, yeah, looking at looking at the numbers, I think we'll be just past August at the current rate. It'll be interesting to see if it accelerates. One thing I'd love to see is more people offering GLM 3.5 flash or 5.3 flash

because that seems like the up and coming model that everybody's switching over to. I see a few providers, but if you go to active, I'd be curious how many providers have picked up. But I still see a lot of the usage on GLM 5.2 and DeepSea flash.

Kyle (Bow Tied Blue Fin)Well, maybe maybe not for long, David. And a new model just came out today. Ah, which one's that? The updated DeepSeek v four Flash. I don't know if they're calling it 4.1 or if they're calling it four with a different date on it, but that one seems to have come in above their last pro version update. So it seems like a considerable improvement needs to be tested outside of their own published benchmarks,

of course, but they're ranking it higher than their than the prior version, and the prior version is very good.

David JohnstonYeah. It's very much Morpheus timing. Every time we get on this call, another model seems to drop as we jump on. I think it is called 4.1. I hadn't got a chance to look at the benchmarks the last few minutes, but that's exciting to hear. It's like every new flash version exceeds the previous, you know, pro version Yeah. Is is pretty amazing. So, yeah, I'll have to try out the latest DeepSeek. But it's good. It's almost like our rivalry is DeepSeek, GLM. DeepSeek, GLM,

you know, they keep wanting up each other. So just keep doing that every week for the next year, and, you know, we'll all be in a pretty good place for us access to intelligence.

Kyle (Bow Tied Blue Fin)Yeah. And then when comes in every now and again. They

David Johnstonmake an appearance. They're like, Alibaba's still here. Like, oh, yeah. That that that's right. And and Mini Max is doing their thing. H three on the video side is awesome. It's good that we finally got an open weight model that, like, get text accurate and all that for the videos. So but, yeah, that started to jump on you, but, yeah. No. 10,000,000,000.

Bowen BelmerHere we are on the peer to peer. Yeah. No. No problem at all. It's it's it's amazing. And I think, you know, speaking about, like, the models jumping and then ultimately becoming a flash version, so then it's hitting, from an economics perspective, like, it's becoming super cheap as well. We're I I swear, we're only, like, one or maybe two iterations away for my use cases not being able to tell the difference. And so then just at least not, like, a real meaningful difference and then just having a workflow, you know, on a daily basis that is effectively

free, quote, unquote, because of my more stick. I mean, it's obviously not really free, but, you know, perpetual is a better way of of looking at it. And in fact, I mean, I just set up a new multi agent workflow using Hermes, and maybe maybe I'll start off some time, all on Morpheus inference where I have an agent doing research every single morning by 07:30AM. It's then sending the research to a secondary agent who's organizing it into a newspaper specifically for me and then sending that to a third agent who's making it pretty. And so I have, like, my own personalized

newspaper now every single morning that my agents are just making for me. Because I like reading in that format, but I wanted something, you know, I'm sick of hearing just doom and gloom mainstream, you know, everything all of the time, basically. I basically was like, tell me good news unless if it's really important about these topics. But,

David Johnstonanyway, yeah, stuff like that. It's I mean, we're so close to just Well, if you wanna jump into the doom and gloom, I think the big thing this week was the Anthropic employee who came out. He's like, AI is dangerous. Like, I know we've said this before, but it really is. And I don't know. There's a lot of caveats. One the guy was there six weeks. So it's like it's not like a founding member of the team or, like, somebody who's been there for years. Dude went straight from OpenAI to Anthropic, and then was like, nope. And then just ducked out. So it's hard to know because they've played this, like, fear mongering marketing thing so many times.

Though a few other employees that are still there or, like, came out, oh, we have the same concerns. The thing that always drives me crazy, though, with these posts is he offers no solution. Yeah. Yeah. Yeah. And and it's like you've been doom and glooming for four years since '22, and you have no particular, like, we ought to do x other than, like, they should be in control, and they don't have a plan, but them being in control would be better than somebody else. And it's like, dude, that's not a real proposal.

And and that's the thing that, you know, keeps me so focused and passionate about Morpheus is it feels like we have one of the few communities who have a clear vision, like an articulated goal, a set of principles that, like, let's get this power in the hands of as many people as possible, transparently, uncensored, and, like, coming out and, like, I have concerns, but no solution. It's like, okay. I don't know what to do with that. And it's you know, maybe it's boy that cried wolf,

you know, and and there certainly are, you know, dangers, but I don't see any scenario in which we get a better outcome by, what, centralizing all the power into, like, one hyper scaled person that, you know, totally won't make mistakes or isn't aligned to everybody. So, like, I don't know. I saw all of those. Like, again, here we go again.

Bowen BelmerSo Yeah. I'm over it personally. You're muted, Alan. Alan, I'm over it personally. I don't believe them at this point. I think they're totally full of it. And it's one of those things where okay. I look at this similar to how I would look at, like, the presidential elections. They those people will stand on stage and call each other literal Hitler. And then the day after the election, they'll have coffee together in the White House lawn and be, like, yucking it up like they're just buddies. You know? Like like, if they actually believed in their core, the things that they were saying, their actions would look completely different than they do. And and that's how I feel about this. Like, if these people

actually believed what they were saying, then they have a like like, if this person actually believes that about Anthropic, he has a moral obligation to, like, actually do something about it. But they but they won't because, like, they don't actually believe that, you know, and it's it's just It's

Alan (Nomadic Rogues)more about the clicks. Yeah. Was you you read into it. You clicked into it, which which was the goal. The goal wasn't to solve the problem. It was to it was to click count. And so it's resist resist the need to click in and say, maybe he's got a solution. Melody. No, they don't. Although It's a cynical way to look at life, but it seems to be more and more the case is, you know, amazing headline, no substance. Click. That's all we needed.

Anyway

David JohnstonYeah. And and most of it's not even written by a human anymore. Right? So, like, the whole, you know, what what's the theory referred to? The empty Internet or the dead Internet? Dead Internet. Yes. The Internet theory. It's like yeah. I think we're we're pretty much there. And so I use signal and other sources like my proof of human for like who you know, where's the real signal. Though on this subject is, like, what would advances look like? There was an interesting breakthrough with OpenAI, or at least it looked like that at first when they're like, we solved this math problem that is one of the millennial goals, and, like, it's this 100 year old thing.

And and the two interesting things came out the next day. One, mathematicians were like, that's not how any of this works. And, like, the way you did this was basically just brute force with 10,000 agents exploring all possible solutions until they came to some proof, which I don't necessarily have a big objection to. I get the mathematicians don't like the brute force approach. But if you brute force the cure to cancer, I would be perfectly okay

that you guys brute force the the cure to cancer. So so I'll give them the brute forcing is okay. But the interesting thing is the next day, apparently, a lot of the breakthroughs that they claimed were actually some of the mathematicians who used their system, who then they were training and using the data to fix the thing. And then the mathematicians was like, wait a minute. Is this is this our work? And they opened their eyes like, maybe,

Bowen Belmerbut Was your work, bro. Gotcha.

David JohnstonYeah. Like, you know, we're not gonna claim the prize. You guys can claim the prize. And then, like, wait a minute. This is really, really messed up. And and the people are like, you know, maybe we can ruin your career. And, like, the whole transcript of, like, them going back and forth about the announcement came out. And it's like, wow. Yeah. These aren't good people that are doing this. Right? They're they're very maniacal or, you know, sort of very

mercantilistic, you know, transactional. They're like, oh, but we need an announcement before our IPO. And, like, what can we steal? It's like, jeez. Like, this this whole thing went from, like, oh, wow. That's a cool breakthrough to, oh, wow. They did it by stealing a bunch of other people.

Bowen BelmerWhich that's how it's all gonna happen, though. Yeah. I mean, none of it it it like, none of it's proprietary going. It's even I have a friend who works at a very well known AI company that I will not dox because I care about this person. But when when this person first found out, you know, about Morpheus and kinda that I was, like, you know, working on this with this community and, like, hey. What's this thing all about? Why are you so passionate about it? Whatever. This person was under the assumption that, well, we have these contractual agreements with these inference providers, so they they you know, it's fine. Like, they're not taking the information.

And then does a little bit of digging into it and was like, oh, wait. Actually, there is two massive ways that they can they can use it. And the first is cloning the data because now it's a new dataset. And so it's not your dataset. It is our dataset that looks eerily similar like your dataset. And the second is, and in some cases, like, if there is potential for, like, security or any type of, like, public safety consideration or something like this where they can make the claim, oh, well, we had to retain this, you know, for, like, basically the greater good, like, for public safety.

And and and and who's the arbiter of that? You know? Like, there is no actual ins and ins and outs of, like, what that framework looks like yet. And so, yeah, it's I I'm I'm bummed but not surprised that this is how that went. Yeah. Yeah. But but that's

David Johnstontelling about base assumptions. People that are coming out of the AI community are coming from academia or they're coming from a big corporate world. And all they've ever had are contractual trust me, bro. We signed on the dotted line and swore we wouldn't do something that's, you know, not in that contract. But crypto people are used to an adversarial environment where bad actors will do anything and everything they can in the system, and the only thing that's real is a verifiable

cryptographic proof. Right? The whole fifteen years of crypto can be summed up as verify, don't trust. Yeah. Right? And so all these people coming in now to these adversarial open systems, they're still on trust, and they haven't gotten to verify yet. But if they wanna survive this change, they gotta get to verify and not trust. Right? And and we can get there, but that's the cool thing we have to offer to the world is, hey. Here's all the cryptographic primitives

we built for identity and validation and whatever. The last fifteen years, they're free. Come and use them. Right? And so but it's taken a while because crypto has become the word so associated with, like, meme coins and speculation Yeah. Opposed to cryptography, which is the math behind the proofs, which is the real foundation of Web three, crypto, and blockchain. So we'll we'll get them there, but, like, that's the education people in that ecosystem have to go through is is how we all learned

Bowen Belmerthe last fifteen years in crypto. We're so close to the breakout event of, like, the average normie realizing why it matters. And I I just had an example. I was just meeting with a group of people the other day, and they were talking about spam text messages that they're getting and, like, spam phone calls. And I was saying, well, just wait till, you know, your family members' voices are cloned, and now that spam phone call because one of them that somebody got was a text that was like, hey, this is, you know, their kid's name. My phone died, so I'm texting you here to get ahold of you, and it wasn't their kid, which is obviously scary. Well, what happens when,

you know, your kid has a video of them talking on the Internet, and now all of a sudden their voice is cloned, and now you're getting a voicemail from someone who you'd swear is your kid because of the the similarities in the the voice and mannerisms, that's So a huge something like that and and then, of course, people, when I was meeting with them, were asking the question, well, like, how do you how do you stop that? I'm like, proof of identity. By the way, we have the technology and have had the technology for quite a while. It just hasn't been, like, largely adopted. But I'm sure, like, all of that all of those types of things will also be in some way, shape, or form on chain.

Because it's the only way to verify that it's actually them. Right? I mean, like, it's it's coming, without a doubt. Yep. Speaking of all of this, I do wanna draw people's attention to and give a shout out here to you, David. This article that you wrote in 2016. So Soothsayer David wrote this article that I'm showing here, which you should check out on Medium, How Blockchain Check Tech Will Help Us Survive the coming of strong AI.

This is a decade ago, you guys. And, I mean, I I definitely encourage I encourage everyone to check it out. But, David, can you just give, like, a really quick overview on, like, what it is that you wrote here and kinda why at that point in time this was something that you were thinking about? First of all, it's hard to believe it's been ten years.

David JohnstonBut I remember like, I'm not oblivious to risk. Right? And I'm not oblivious to being concerned about where these trends are headed. And so in 2016, I remember walking into a restaurant, and if if you show that that image on the article, like, saw this sort of weird cartoon of, like, the chef, and the chicken is, like, self aware, and it's running away, and he catches it, and he kills it. And it's, like, pretty, you know, explicit in the sense that they had, you know, anthropomorphized

this this chicken, and he slaughters it and and eats it. Right? And I and I was I had been thinking a lot about AI. I'd been building, you know, projects in AI all the way back in 2011, 2012. And I was like, we really gotta avoid being the chicken. Like, how how do you not be calories for the more intelligent entity? And so I thought about it very deeply, and what I came to is sort of the same thesis that I have today, and I think it's played out pretty well, which is TLDR.

If good AI creates more outcomes, more money, more, you know, economic activity for people, people will reward it with power and compute and all these things. Right? So if there are bad AIs and good AIs, I think the good AI will win as long as it's creating more value in the world. You know, it it's it's very much if you've ever read Tolkien. Right? You know, evil cannot create. It can only corrupt.

Right? That whole idea, by definition, all the theft in the world, all the credit card fraud and and crime is a tiny fraction of global GDP. Right? By definition, like, the the thieves can't steal something somebody didn't else create. Mhmm. Right? And usually, in a healthy society, the builders are better resourced than the thieves. Right? They have more money and and power and ability to protect their wealth than the thieves do have to steal it. Right. Any place that's not true is basically a completely broken dysfunctional society where, there are no property rights and it's just banditry or whatever. Right? But mostly,

in a functional society, people that create value are gonna defend it, and they have more resources to defend it. Right? And so if you're concerned about bad AI, you really want good actors to have the best intelligence and the most access to these agents and able to protect what they already protect, their house, their family, their property rights, their wealth, you know, their food, their economic outcome. Right? And so I feel pretty good about that. Right? And that's why I was so concerned

when the proposals were made to ban open weight AI back in 2023 because that's one of the edge cases where it goes wrong. It's when, you know, all of the power ends up concentrated in a potential bad actor, whether that's a government or a large company, and everybody else is prevented from having access to the same intelligence to protect themselves. Right? If I were to use the historical analogy, you know, in in the industrial revolution and the age of guns, germs, and steel,

you really wanted to be on the guns, germs, and steel side of history and not the one that didn't have access to that. And so it wasn't that guns weren't dangerous. It's that you wanted to decentralize their power to everybody so there wasn't as huge asymmetry where those with took advantage of those without. Right? And so I feel like we're in a pretty good place. Looking back on this article from ten years ago, we've got open weight models. We've got broad access to them. We're using, you know, through Morpheus crypto right now

to offer access in a way that's is permissionless. So I I feel like we're in a pretty good place. But, yeah, I have I have been thinking about this for a while, but that's my answer is I think it's an economics answer. It's not a perfect technical solution. They're gonna be hacks. They're gonna be stuff like that. But what do you do with a 160 IQ bad actor? I want a 200 IQ defender.

Bowen BelmerYeah. I just still can't believe that this was stuff on your radar a decade ago

David Johnstonand anyone's radar a decade ago. Like, how many nerds out there are there like you? What do you what do think? Just like a solid couple 100 maybe? It's it's probably a short list. But to be fair, AI's been sort of thought about and talked about and researched for forty or fifty years. Yeah. We just didn't know when it was gonna get to that tipping point. Right? And Transformers, right, the the seminal article by Ilia and the others at Google didn't come out till a year later in 2017. And that was sort of the moment like, oh, okay. You know, attention is all we need. If we put enough compute and data with this formula, we're gonna get better and better results. And for most people, it wasn't till '22

when ChatGPT had that moment of breakout. They're like, oh, all the things people talked about AI doing now can actually do some of them.

Bowen BelmerYeah. That was for me, admittedly, the first time that this became, like, remotely real in my brain. Prior to that, I just looked at it as, you know, you might have software that has a lot of canned answers, so you can ask it a lot of things, but ultimately, like, it will have a limit and, like, it's not actually, like, creating anything new. And then, you know, GPT was the thing that was like, okay. Wow. This is like. But even by comparison, like, look at GPT, what was it, two in 2022 compared to everything now? It is, like,

laughably different. I love that Will Smith eating spaghetti, you know, meme thing that they use as a test. This actually segues really nice into one of the craziest things. This was, like, something that just happened the other day and is now kinda happening in real time and was another one of those moments for me where it was like, oh, this just got real in, like, a very different way than it was previously. And it ties into what we were talking about right before we hit record where I told you, hold it because I wanna hear your guys' thoughts on this on air. But

the the guy who mapped the fruit fly brain, like, all, looking at it right now, it's all a 166,700 neurons inside Minecraft so that they could have simil simulated neuroactivity driving the brain the fly's movement. And now, I don't know if you guys have seen this, but people have been taking that because he open sourced it and, like, giving it a $100 and attaching it to a Bitcoin wallet. And, like, doing all these things to basically say, like, hey. How much brainpower do we need to do these, like, menial tasks? I saw

a video of somebody attaching it to, like, a driving simulator. And I don't know how real any of this is, but it was the first thing that I was like, okay. Wow. This is, like, now another level of intelligence, you know, that I didn't even consider really before.

David JohnstonYeah. Have you guys seen this? Oh, yeah. I saw it beating Beat Saber. Saw the Beat Saber. Yeah. Crazy. That was that was entertaining. It reminds me of if you ever read The Singularity is Near, which was Kurzweil's 2005 book, This is basically sort of he puts everything on a nice curve, and it's like, okay. This is how much you compute you need for a fly brain and a a mammal brain and a monkey brain and a human brain, and here's the years according to Moore's Law that we get to those different milestones.

And I think it was 2025, end of twenty twenty five, we got to the first you know, with an enormous amount of compute for the first time, you had enough of it to get to, like, a human level of cognition across, let's say, a large language model given this many tokens equals this much human verbal output or something like that. So we've got sort of to that level, and now running a fly brain or whatever, it's gonna get weird, man. It's gonna get weird. Really weird because, like, where's the line where, like, okay. Now you're running enough neurons to be a human brain. Well, that's the part, okay, that's the part that I'm confused by is, like, what exactly it's measuring. So, for example,

Bowen Belmermy understanding of this is they're using the same number of neurons that a fruit fly brain has, but the way that it's arranged is not the same way that it's arranged for a fruit fly. So in other words, it's saying if we took all of the computational, like, theoretical possibility of this many neurons, this is the outcome. Like, oh, it can play Beat Saber or it can do whatever. And then, yeah, to your point, like, we'll get we'll get to the point where it's like, okay. Well, now we've matched the number of neurons of a human brain, and so we can effectively, like, have

an isolated incidence, or or instance of of a human, you know, and and there's a billion sci fis about this. Right? Like uploading your consciousness and measuring your neurons and doing that. I mean, it's feeling more and more just weirdly, like, maybe this can exist. I don't know. Maybe it's not just theoretical. I don't know. Yeah. It brings up a lot of questions. A lot of questions I don't think we really have answers for yet. So This was a jump for me, though, where I saw this and was like, okay. We're now doing something that is, like, truly different and goofy and

David Johnstonis not is goofy today, but the goofiness of today is, like, very serious, you know, development of tomorrow a lot of the time. It's it's a lot of orders of magnitude, but it's gonna be interesting when we get to, like, the cat level. And they're like, oh, I'm gonna scan my my pet's brain and upload it, and now it plays in a metaverse. And, like, long after it passed away, like, you know I mean, there there there's a lot of difference between a fruit fly and a and a cat that's several orders of magnitude. But, yeah, like,

Bowen Belmerit's gonna it's gonna get weird. It's gonna get weird and interesting. Well, and this is what what I had mentioned before when you guys were talking about, like, the human battery scenario. Right? Like, we're like, we're talking about the potential for p doom, and do we just effectively become, like, computational batteries for

David Johnstonsuper intelligence in in the way that the original and, David, you were saying this, that the original matrix was that's what it was supposed to be. They they dumbed it down to batteries. Right? Yeah. The the studio had them dumb it down because they didn't think people would understand that brains had lots of neurons and computational power, and the machines were using that to power the matrix and all this stuff. And they're like, no. Use the battery analogy. We're using them for power. And but it was really confused. It's like one of the few confusing things about the movie. It's like they're the heat of the human body is being used for powering this? Like, okay.

Like, they gotta wave their hands like, and a form of fission or fusion or whatever. And it's like but clearly, this studio kinda shoved that in there to make it easy for for people to to run the analogy. But, yeah, the brain one makes a lot more sense. Like, the human brain has an enormous amount of computational power, even compared to an entire data center full of h one hundreds. Right? The human brain, we we we've discovered over time is much more complex and much more powerful and works in ways we don't really understand.

Maybe quantum's involved. Who knows? Oh. But like, Yeah. The plastic the plasticity alone

Alan (Nomadic Rogues)to be able to move and rewire compute computers are finite, right? They have this pathway or that pathway. The the brain is amazingly plastic to go and rewire and repath and reinforce. It's there's yeah, you're right. There's some things in there that were not all transparent just yet. Yep.

David JohnstonExactly. But coming back to like the takeaway, you know, I think the the goal ought to be, okay, how fast can we get an agent into the hands of as many humans as possible and preserve agency and privacy and transparency with with these systems. So, Bo, I don't know if this is a a good segment to to cover it, but I know we wanted to take a look at the current state of install OpenCall. And Oh, sure. And I'm a bit tempting the demo gods here because I literally just pushed him out to to

the code base. And it's like, you know, yeah. Okay. Let's let's test in in live. But, know, if it doesn't work, you know, we'll we'll we'll we'll do it again and and record a little session. But I was working earlier today. Let's let's see if it works now. I but we need somebody Bo, I don't know if you have an account, an email account that you haven't set up and install OpenCloud with. So if you have a Gmail account that you haven't set up yet, it should work. And you'll just go to installopenclaw.xyz.

And my hope here is and I've got a couple 100 beta testers now, and thank you everybody who's joined and is using it. It's awesome to see more people flowing in every day. You go, you give your agent a name, you push a launch button. That's it. Click on Google, log in with whatever you want, But I'm trying to get this, like, under sixty seconds so that anybody who wants an agent doesn't have to download anything, doesn't have to touch terminal, all that stuff. So

let's see if it works. There's there's some new new instances in the buffer. So, you know, let's let's give it a name and Cool. Let's see. We want this to be I hope you like the title. Get Open Claw. This time on easy mode. Like, after after helping a lot of friends, like, spend hours trying to do it, it's like, no, no, no. Gotta have an easy mode. So doctor Samajay junior. Okay.

Bowen BelmerI don't know if that just it just came to mind. Sure.

David JohnstonThat's it's all good. Whatever whatever you wanna call your agent. So hit the launch button. Let's let's see what happens.

Bowen BelmerGive me just one second. This is this is where we get boat accidentally docks all his credentials. I know. I also just realized that I'm gonna have to actually look at this name. So we're gonna change it to Sama Sucks. Okay.

David JohnstonMuch better. Why you want your agent called that? I don't know.

Bowen BelmerAlright. Launch.

David JohnstonAlright. So now it should let you log in with Google or whatever account you want. You could use email. You could use Google off. It has other social options. You can connect to Wallet if you want, but I tried to set this up so it works with web two social people are sort of used to. And because you picked email, it'll ask you for a code. So you just check your email and put in that code, and it should connect your agent to your credentials. So this was really important for me. I wanted it to be an agent that's really owned by you.

And because this is your account, only you can access this agent. And there you go. All set. So there's a big green button. This is launch agent now. Let's see if that does a thing. And if it does a thing, it will connect you. Remember, this is hosted on manifest network into a new tab, and it'll show your agent, and then you'll be able to talk to it. And it'll look like an install

of OpenClaw, presuming

Bowen Belmerthe, you know, refresh Oh, can't see it. Share this. Sorry. Oh, yeah. Alright. No. You got it. Yeah. Sorry. I thought it was gonna switch over. Yeah. Boom. And we're live already. New agent. Yay. Sam Assass. So so yeah. So, you know, that's it. Like, now you've got an open claw.

David JohnstonIt's connected. You can see it's defaulting, I think, to DeepSeek as the model. Won't tempt fate too much more. Like, that's that's it. Name it, launch, log in, boom. Now you've got an agent, and and it should just do a thing.

Bowen BelmerSo yeah. For some reason, I'm not connected. I don't know what happened, but that's okay. But but, yeah, I mean, there's obviously, there's something happening. But, yeah, that that's crazy. How, David so let's say I was using this, you know, to set up I mean, you could set up a job like what I was talking about earlier, like, make yourself a newspaper, make yourself some daily research, like, have it do a cron job for you. Let's say that I did that. Right? And now I have this agent, but I wanted to, at some point, export it and, like, you know, quote, unquote, take custody. Like, what does that process look like?

David JohnstonSo the nice thing is and if you go back to the other page, you'll see there's a dashboard button, And this is where you can kinda control your agent. The thing I'm adding in right now, it doesn't work on this current version. This is what I'm working on today, is the download agent button. Oh, got it. And literally, it works in the code like you would have to type in, you know, export my agent and it would do it. But I'm trying to get it to, like, a one click. I wanna put it on my machine. And so that's what I'm working on right now. This is also a way to get back to your agent. If you, like, close that tab at the top, it's you know, open your agent, and it'll reopen

your agent. You can see your usage on here. You can see the models. You know, you see it loads the gateway, and then it populates the agent and and opens it up again. To go back to the dashboard yeah. Open call. Oh, there it is. Alright. So you go down there. You can upgrade plans like, oh, I ran out of inference. Cool. Upgrade to pro, upgrade to power, you know, whatever scale you need. You don't wanna subscribe to anything? That's fine. Here's credits. Just buy whatever amount of access or inference you need. Honestly, DeepSeek v four is so cheap. It lasts for a long time. You can connect wallet like you see there,

the whole thing. Right? So this sort of just gives you instead of having to poke around your agent in, like, the really complex and confusing settings inside of OpenCloud, this just gives you a way to control all these things with one click. So, yeah, I'm I'm pretty excited to where it's gotten to. I'm seeing five, ten. I had 18 people sign up randomly the other day, which is a good record considering we haven't done much announcements or anything. But I wanna get this down to, like, okay, here's a quick video.

Click, click, click. Now you've got your agent, and people are off and running. But it's it's a free tier, anybody can set it up, use it how they want. Yeah. I see it's defaulting to DeepCP for Flash. And so, you know, I I think this is the key. Like, let's get Morpheus. And again, this is defaulting to Morpheus. Let's get all the major harnesses. Hey, success. Yeah. There you go. It was on my side. It wasn't your side. Yeah.

Bowen BelmerI mean, this is, like, so incredible, David. This is the exact type of thing community so that you can literally show your friends and family how to set up an agent and set it up for them with an email, a two FA code, and a button click, and they're live. And and they can use it with the free inference. Obviously, like, there's only so much, like, free inference to kinda get them going, show them this, that it works. Upload credits, stake more, you know, connect to wallet, whatever makes the most sense for you, and boom, you you have an agent. And, like, the alternative to this, especially for those nontechnical, by the way, is spending, like, five hours

learning how to set this up on your own. Ask me how I know. And it is did I learn a lot? Yes. I, you know, I I now understand command lines in a way that I didn't used to understand. So that was cool, but the reality is most people aren't gonna wanna do that. You know? Like, it's it's not I didn't even wanna do it. No. I'm just kidding. But, like, yeah, most people are not gonna wanna do that, and this is just like an instant onboard.

David JohnstonYeah. Awesome. And and you'll notice too when you go to the subscription or the credits on the dashboard, it uses Stripe. So if you wanna pay with a credit card, you don't have crypto, it's an option, but it's not a requirement. Right? So hopefully, this has removed all the barriers. Right? You you can pay with a card, you can log in with Google, It works on on mobile. And think about it, it's now cross device. If you logged in with Gmail, you can do that on your phone or your computer or your laptop or wherever you want, and you can take your agent with you, and you can download and back up the context onto whatever additional machine you want. So

I think we're we're this is this is huge. Right? Now we have an easy install for OpenClaw. I wanna do this for Hermes. I wanna do this for PY. I wanna do this for the other popular harnesses. And just give people an easy way because even using the new version of OpenClaw, man, the place you always get stuck is like, okay. Now I need a key. So I have to subscribe to something else, and I have to go sign up to Claude and figure out how to, you know, get their model connected in here. And that's usually where people get lost. Yeah. So my take on what happened with OpenClaw

is you had all these people excited about a personal AI. They rushed in. Some people got through the door, but it was really hard to use, and you sort of exhausted the number of people that could get through that process. And it hasn't gotten a lot easier. A little bit, but not a lot. Right? Yeah. In Herbie's, it was much easier. That sort of took over. Right? Grock Grock Bot's even easier than that, but you're sort of getting away from from really owning it. Right? It's more of a third party.

And so the key is let's let's make it easy. And that's my whole mission right now is I think how we get a lot of people onto Morpheus. And and huge credit to the Manifest guys for supporting me in the whole development process and offering a real alternative to something like AWS. Yeah. Because this would be pretty lame if it was on Morpheus, but it was also just entirely centrally hosted. Yeah. Right? As opposed to on a web three alternative. So it's cool. The timing's great. I don't think you could have built this a year ago. You probably couldn't have built it six months ago. You needed the open weight models to be good enough

and cheap enough, and for Manifest to to host all the, you know, containers in a decentralized way, and Morpheus to be ready with all the pieces. So, yeah, this is a really cool milestone.

Bowen BelmerYeah. Absolutely. I I think, like, the the magic part about it too from a user's perspective is I remember, like, when I was when OpenCost first came out and I was first learning how to set it up, I just could not get wait to get to the point where it's like, can just talk to it in natural language and it can tell me what it can do. This just bypasses everything before that. Like, someone can get into it immediately and then say, hey. What can you do? How like, teach me how the like, teach me how to teach you, basically. Teach me what you can do and how I can use you. And this just gets you from zero to one instantly,

which is it's awesome. I love it. This is fantastic. Looking forward to the Hermes version. I did switch over my personal agents all over to Hermes a little while back, and it's been a good experience. So, yeah, looking looking forward to that. It's it's awesome. Let me just look here what else we had on the list this week. Actually, Kyle, we haven't heard from you yet. What's going on over there?

Alan (Nomadic Rogues)Maybe he's not in the spot. He can talk. That's okay. I'm not there. Just jump in if you are at any point. Yeah. We can we can talk about the the node release this week. I've got some provider messages if you want to Please. Yeah. Talk some of that. Yeah. So for all of the of you fans out there for the Morpheus Lumiren node, we did update to 7.11.0 a couple days ago. This was a pretty cool update for a couple reasons. The biggest being that it had three big community submissions in it.

Both two from our our Asia folks, Renee and Zac Zac 90 five and then from the from closer to Shore Bob, one of the contributors and discuss people, he's been on the spaces before but really good support on the on the proxy router, found some issues, some improvements. And one of the key things that came through was I talked last week about our active.more.org,

our status site and one of the things that's happening, there was some there were some issues and I do shout out to Bob for pulling that out and we found some some pretty pretty gnarly errors and what they were is that we've started having the providers self report their models and their bids and some of the behaviors to make sure that they're healthy or can you really provide them because no one wants to go open a session and realize that that bid really isn't there. And one of the things that was happening is that if if the model as

it was registered didn't have any tags, the provider didn't know how to probe it and so just let it alone. And that left a big gap, lot of a lot of bids out there that were that were unknown when they were really real. So seven point one one fixes that, is that it does attempt some of those probes to make sure that if it's a healthy bid, it shows as a healthy bid on the status page. So, encourage for all providers who see this, watch this, please communicate to your friends or your non friends to please upgrade the, if you're a provider upgrade to 7.11.0

and that will help to make sure that if you're bidding on models like Cloud Opus five was untagged, it some of the bids didn't come through because that said that self report wasn't there. Also because of that activity, we updated on techmore.org a new explainer page called active.html. So tech.more.org/active and what this is, is talks a little bit more about that status page and the filters we use. You have the raw models and bid data that's out there on chain.

Well, what we did for active and for the for the API gateway was very similar to what the consumer node will do when you go to open a session, it knocks on the door says, are you healthy? Are you okay? Well, instead of having, you know, multiple consumers do that for all the providers and all 486 or so bids is we've got our own cron job that runs every five minutes that pulls those providers. Are you good? Are you okay? What's your latest health check? Are you fine? And then consolidates that data into that S3 and into that status page.

And what was happening is we did not have clear views of different layers of the network. Raw you can get on chain all the time. All means all active bids, all active things that are not expired, bids and models. Then there's the active which says, look, we remove things that are broken, unhealthy, unreachable, a T failed attestation, you know, those kind of things we remove because you don't want to, again, you don't want to knock on the door and have a failed session That cost gas, that cost time and at the bottom was the gateway

layer. That says, we even remove the ones that we skipped. It says, look, the gateway is only gonna use things that we know are healthy. We're not gonna guess, we're not gonna punt and so really we've got three sets of files. Each one has like all models dot JSON or all bids JSON or gateway at the at the very bottom that make sure that if you build an application, so the API gateway is an application and you want to say, hey, what's the list of models and bids that I have access to? You can choose as an application developer, an actual application builder to consume either any of those three layers

to do what you want. And you could, you know, so you could say look, want the all and then all filter in my application, I'll filter stuff out. So that was pretty good. So again, I thank Bob for bringing that to our attention and I think it's a much better and as providers update to 7.11we'll see that fix. And then what's coming up next is as we continue to enhance that provider probe is we'll pull out capabilities. What we don't see behind it is, are you a thinking model? Are you non thinking bid

on the same model? And so now we're trying to get that subtleties of bids and providers and capabilities that we can expose on those health checkpoints. So you can better steer your inference for context size, for web, can you search web, are you thinking, non thinking, what's the temperature? So there's more coming with those provider probes. It's really helped show that network status. So that's all I wanted to just upgrade to 7.11 and things will be better. And thank you to the community members who contributed. That was a good set,

good release this week. I'm happy that we had some good submissions.

David JohnstonThat's awesome. Yeah. And big credit to Renee and Mazin and Bob and everybody that contributed. I know there's a lot of hard work with that went into that. So you guys are awesome. Yeah. And that's what it's all about. Right? It's open source. Anybody can jump in there. And I think these are really important upgrades, Alan, because as people have better tools to, you know, like, curate the list, especially as we have more and more providers Mhmm. That's gonna be really helpful. It's, like, getting good results. Right. Yep.

Alan (Nomadic Rogues)So It also helps. I mean, Bo, if you're if you're on there, if it also helps the providers. Like, hey. Why is my bid not getting getting taken? Scroll in the search. Go to Cloud Opus five. If you would type that in.

Bowen BelmerRight here.

Alan (Nomadic Rogues)Yeah. And then open him up. So click on click on him and that'll open up. Oh, yep. There you go. So you can see the here are the providers that are actually providing that and you can see at the top level, I'm pointing at the screen, I hope you can see I'm pointing at the screen. At the top, next to the model IDs LLM and it's untagged, that means on our model registry that's untagged and that's why the two bottom providers in that list are skipped because they don't know how to guess. Now, found a couple in the wild, people have patched their own

proxy router provider to to come overcome this. So, we've rolled that into our core product, but people could say oh, that's why it's skipped, that's an actual provider signal that says, hey, we did not approve that, that should go away if those two providers upgrade to 7.1.1.0. So they would have active good bids for Cloud Opus five and other things can be admitted on that same line like, t attestation. I think if you look at

well, search for TEE. I think this should work. Yeah. The scroll down a little bit, please. Yeah. It's not I don't think it's gonna come up with the the free text search. Can I filter for it? Yep. Yep. Yeah. It should be. There you go. And just remove yeah. Open him up. He's actually good. Yep. So he's healthy. Remove the

the name the name field if you would. If you go back to your filter, make sure there's no names. Yeah. Use t.

Bowen BelmerFor some reason, it's showing just Oh, yeah. You're clear.

Alan (Nomadic Rogues)Yeah. There's well, because the other two, I think, are defunct, and they should fail. I wanna show failed t attestation. So all the way back up to the top, hit t, goes up farther and see the other two pill filters, blue, green, yellow, skipped. Click on unknown and offline to the right. They're off by default. Oh, there we go. Yeah. There we go. Those click on Gemma. Yeah. There we go. That's what I wanted to show. So that Silverren VM secret labs

is a t provider, but behind it, it can't verify the model. So, says look, I can't verify tea so I'm gonna peace out. So, this helps providers go, oh, why isn't my bid getting? You can actually do some some tests. So, if you click on Silver Ren real quick, that should there you go. So one is verifiable and there's something wrong behind the provider node that says t is unverified and most likely that's because, there's an endpoint that's wrong or endpoint that's missing. And So Alan, just to make sure, like,

Bowen Belmerjust layman's terming it down a little bit. So so basically, this is confirmation that they are hosting what they're claiming they're hosting. So they can't say, hey, we're hosting Gemma, but actually it's v four flash or something else, but they're registering it wrong on on the network.

Alan (Nomadic Rogues)Correct? Okay. Because well, that's mostly in the t add the the t. If you're running a t provider and a t model, it must attest all the way through the chain. And so this is what this tells me is that that provider node which is Silverren and Secret Labs is up and reachable. So the nodes good and one model behind it is fully valid. There's something else wrong between the provider node and the actual LLM behind it for Gemma and Quinn that it's not there and so

and basically for the gateway, like if we don't attest, we shouldn't provide it for the gateway. Yeah. Because then you then we're serving things through the API gateway that is not that are not truthful and that's why, again, why we had those different views like, look, you can show them all, you can give it a shot But for the gateway, it's

Bowen Belmerjust a waste of resources to try to bang on the door that we know something's not gonna work. Which just comes right back to the idea of verify, don't trust. Like, build build as many trustless systems as possible effectively. Physics.

Alan (Nomadic Rogues)Well, I'll go back to David's earlier comment. I had a I had a laugh. You come from big enterprises and big corporations. The reason you trust is because you've got an army of lawyers. So my new catchphrase is physics, not not lawyers. Physics, not lawyers. I like it. Yeah.

Bowen BelmerIt's true. No. A 100%.

David JohnstonIt's it's good to see people getting that though. I tweeted out a compliment to Peter, the founder of Open Claw. And I was like, honestly, OpenClaw has done more for alignment than every Frontier Labs ethics group ever has. And it was cool to see him like that that that tweet. Right? Because he's like, yeah. Like, these are the practical things you have to do to make sure it's the model you expect and the outcome you expect and the harness you expect. And, like, those are the real things you can do. Not this, like, wave your hands about, you know, somebody do something. Like, no. We have we have a solution. Let's do it. Right? So

100%.

Bowen BelmerYeah. Sorry. I was getting distracted because as you were talking, Josh Wolfe, who's the founder of Lux Capital, just tweeted that after further reviewing the work being done at the Frontier Labs, he has come to the conclusion with a 100% certainty that we're all going to die, and just unclear if it's five years or fifty years. So,

Alan (Nomadic Rogues)you know Well, I hate to say, you know, Ben Franklin in death and taxes. He's not wrong. It's just are you speculating on the cause?

Bowen BelmerGuys, I'm gonna make a bold claim. We are all gonna die sometime between the next tomorrow and three hundred years from now. So Come come on, Bo. Between now and the heat death of the universe. And the heat death of the universe. The slow, slow heat death of the universe. There's actually this amazing YouTube video. What's it called? There's this YouTube video that's like obnoxiously long for what it is. I think it's like a 30

David JohnstonThough to give an optimistic note, you got a few trillion years until the heat death of the universe. But, you know, technically speaking, black holes are the most computationally dense entities in the universe. If we could just figure out how to use those for compute then, you know Yeah. Have we simply considered harnessing black holes? Yeah. Then then maybe we can go past that. So it does heat death of the year doesn't have to be the end. Was just, you know, maybe. Maybe.

Bowen BelmerIt's true. There there's this YouTube video called time lapse of the future, a journey to the end of time. It's from seven years ago. I highly recommend everyone just, like, grab a beer sometime and throw it on. It's twenty nine minutes long. And it just goes through, like, the entire heat death of the universe. And for example, like, Earth dying happens in the first, like, fourteen seconds of this video, and there's another twenty nine minutes to go, basically, and it gets all the way down to, like and then things will just continue to collide into each other throughout space time over the course of, like, hundreds of billions of years until it gets broken down so microscopic that it's undetectable, and, like, it's it's great. Everyone should watch it.

But, yeah, we in that time frame, we'll probably it'll probably be curtains for everybody. I guess I guess we'll see, though. Yeah. I think you guys like, in terms of news and stuff, that's all I had. I did wanna make one note. I saw in the chat earlier someone had mentioned that we had mentioned the Will Smith, like, spaghetti eating test for how far AI video is coming, and they had said that they haven't seen anyone do a good skateboard video. Like, it can't get the physics down for skateboarding.

And so I think that's a good supplemental test. Like, I'd love to see if, if anyone knows who Chris Jocelyn and elf Tre Flipping El Toro, what that means. If someone can make that the new gold standard for AI video, like capturing an AI video of that event happening, that would be that would be amazing. Oh, David, you're muted.

David JohnstonReminds me of all the, like, gymnastics, you know, comma you know, comedy wherein, like, AI tries to end and the bodies all just twist up and you're like, oh, okay. Not not quite. But, you know, a little more physics, a little more phys a little more training data, it'll it'll get there. Yeah. Yeah. For sure.

Bowen BelmerBut, yeah, guys, that that was all I had this week. Appreciate you guys hopping on. I I guess we'll throw a little teaser for next week. We're we're in talks with, a pretty big, well known, successful project, to potentially come on and join us on on air next week and talk about what a future collaboration with the MOR token could look like with them. So I'll kinda kinda leave it at that, but it it'd be a pretty big deal if if it, you know, does happen and there's adoption.

Open the transcript file

Hosts and Guests

Hosts

Guests

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Video Is Coming! | Morpheus Is About To Get Far More Entertaining

    52 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...

    Credit the people on this episode

  2. Local Models, Foreign Affairs | The Case For Private AI Archives

    53 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...

    Credit the people on this episode

  3. Let's Talk Node NEO | Step By Step Progress

    51 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...

    Credit the people on this episode

  4. Hello From Hermes! | A New Way To Onboard To Morpheus

    43 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...

    Credit the people on this episode

  5. 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...

    Credit the people on this episode