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 #73! Recorded: 14 May 2026 AI SUMMARY: Ryan Condron launched mordiem.com, letting Venice DIEM holders stake unused inference to Morpheus (boosting compute and earning yield) plus options to stake MOR for user inference or to the MorDiem builder pool at 20% APY. This strengthens Morpheus as a permissionless inference aggregator. Hermes has overtaken OpenClaw as the top open‑source agent framework due to better memory, self‑improving skills and easier migration. Security discussion covers recent npm supply chain attacks and dependency audits. The US Clarity Act clearing committee may impact decentralized markets. The episode closes with best practices for agent permissions, connecting Hermes with the Morpheus skill and popular models like GLM 5.1/Gemma on the network. Timeline: (00:00:00) Intro (00:01:58) Introducing mordiem: putting unused DIEM to work (00:04:03) Why this matters: incentives across Morpheus, Venice, and users (00:06:17) Morpheus as an inference aggregator, DeFi analogies (00:10:17) Hermes agent overtakes OpenClaw in usage (00:18:54) Migrating from OpenClaw to Hermes: tooling and tips (00:21:11) Quirks and credit drain issues with OpenClaw (00:24:01) No single winner: evolving agent frameworks and skills (00:29:55) Halftime note: Brainpower podcast with Emad, AI research highlights (00:33:08) Supply chain attacks explained: TANStack, Mistral, UAI path (00:36:46) Practical security guidance and agent‑assisted checks (00:42:35) US Clarity Act progress and implications for crypto (00:44:51) Community segment: sharing inference and value capture (00:48:03) Community chat: events, daily agent use, and open models (00:51:06) Getting started with agents: Hermes + Morpheus skill (00:54:04) Operationalising agents: permissions, SOPs, and safeguards (00:57:59) Builders corner: Discord sharing and top models on Morpheus (00:59:41) Wrap up and next steps 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 BelmerAlright, everybody. It is another week of the Morpheus town hall. This week, handful of things to talk about. The kind of headlining story out of the Morpheus ecosystem though is from the brain of Ryan Condren, we are gifted more dm.com. Ryan is actually he's in I believe he's in flight right now, so we were unable to get him to jump on this week. But we'll definitely have him on in a subsequent week to, you know, hear from hear from the horse's mouth, if you will,
some of the things that he's working on with this. But definitely wanted to touch on it because it's very timely. You know, over the last ten days or so, there has been a massive run up, in terms of more token price, VVV token price, and more focus shifting into DAI. And what's awesome about this is as the meta, is shifting to DAI, this is, at least from my perspective, the first time
that DAI is getting a lot of, attention from the crypto markets, but not through the scope of, like, meme agents. In 2024, or I guess it was spring twenty twenty five, there was that moment of decentralized AI mania, quote unquote, in the markets. But it was really I I I mean, I'm always one to kinda poke some fun. But but at the time, I won't pick on any specific protocols or anything like that, but at the time, most of what was being called DAI was really just,
you know, telegram bots with tokens attached to them. Now, we're at this moment where proper decentralized AI tools, things like Morpheus, things like Venice, things like Dolphin, some of these others are getting recognition both on the timeline and also obviously through through investors and token price. And through this, Ryan shipped more dm.com, that's mordiem.com,
which is a new way for people who are DM holders from from Venice's project to actually if if they're if they have DM and it's unused inference, meaning that they hold the asset but they're not actually consuming the a dollar a day of inference per DM, they can stake it to more dm.com, which is ultimately offering it up as compute to the Morpheus network ecosystem and earn more proceeds for doing so. So they're kind of pseudo acting as a compute provider in doing that, which increases the amount of total available inference on Morpheus.
The other two things that you can do with this website are buy inference. So if you and by buy, I mean stake for inference. So if you want to stake your existing more tokens similar to the API gateway, you have the ability to to do that, stake your MOR to use inference as a user, and then also the ability to stake to the MORDM builder pool, which is the subnet, and you'll earn 20% APY more on more for doing so. But the real like flagship thing that this is doing that's, you know, most different from everything else is giving you a way
to put your unused DM to work, providing more compute to the Morpheus network ecosystem, but also providing an economic opportunity for you as the DM holder. Kyle, Thomas, do you guys have any notes on this? I mean, this is, of course, pretty big. It's another connection point between Venice and the Morpheus ecosystem. And and I think there will be a lot of participation in this probably. I mean, there's already 400 it looks like 471
DM stakes to this, and it's only been live for a couple days.
Kyle (Bow Tied Blue Fin)Yes. So I I could speak to it a little bit. Definitely, I think when when Ryan is on here, it would be he can explain in much in much more detail. But, yeah, overall, it's it's really an amazing opportunity. Right? It it allows people to, how do I frame this, utilize some of that excess capacity that they have, monetize it while also creating much broader access within the Morpheus network. So it's kind of a win win everywhere. Right? It's a win for
for DM holders, right, to to get commentated. It's a win for Morpheus token holders that the the system's going to be be more broadly utilized and have much larger scale. It's a win for consumers within the Morpheus marketplace that now have access to, again, deeper scale and larger amounts of capacity. And then it's also a win for Venice, right, where they're getting more utilization of their system. So it's really a great application of of what the Morpheus infrastructure is kind of built around. Right?
Let people provide into the marketplace in whatever way that that they choose to and work as more of this decentralized permissionless consolidator of of inference and and help people monetize some of that, excuse me, excess capacity. Right? This is one great avenue. Right? We we we see people with AI credits. Right? We see other larger opportunities of, you know, when we work up the scale, it's like individuals with with,
with capacity. People with API excess credits or capacity. People with data centers who are who are running, who aren't getting full utilization. So it's kind of we're checking we're trying to check each box within that, and this more DM product fits perfectly kind of right in the middle where it's way better and way stronger than than a simple consumer scale running through Venice. And then we hope over time we get to the the larger scale data center level access as well providing local. So it's it's really great. I think it shows,
you know, the strength of the system that's built, and we encourage people to build similar things like this on top of it for for other platforms as well. And, you know, hopefully, they they have a good opportunity to monetize through through their products, through through the builder's protocol, which which, more DM is using. And, I think I think it's it's a great opportunity overall.
Bowen BelmerYeah. I so this is a really good example of how Morpheus as a project like, this really highlights how Morpheus as a project is an aggregator of decentralized AI tooling or, like, in this case, decentralized AI inference. And I I've used this example before on other spaces, probably even our our this specific space of how if you compare what's happening in d a I right now to what's happened in DeFi in the past,
you have in DeFi, you have decentralized exchanges that ultimately have their own liquidity pools. So for example, you have Uniswap, you have Compound, you have some of these other places where on Uniswap as an example, Uniswap has its own liquidity pool of tokens that people are providing token pairs into and they're earning yield for doing so. But then you have this next layer of something like one inch and and cow swap and other aggregators where
those aggregators are allowing you to tap into those liquidity pools across different exchanges. Morpheus, and this is more DM is a really good example of this, is providing that same type of additional layer to the inference market as something like a one inch would be providing to the capital markets. And so what I mean by that is now that people can use their Diem and ultimately stake it to provide compute to the Morpheus ecosystem, that means that
every more staked from a user's perspective is effectively getting you access to Diem. As other people come online, it'll be the same thing. So I I said this on Monday that all all DM is not more, but all more is DM from this perspective. I'm not saying they're one to one in terms of in terms of value, but what I'm saying is staking more as a user now inherently gets you access to Diem, or at least the same compute that you would otherwise have if you were staking Diem. And as more and more inference sources come online and things like this become more robust,
Morpheus can become a central place where you can access all of them permissionlessly without needing to go to any one individual inference provider. And so this is kinda just like the next step in that evolution, but it's an important one because it it highlights, you know, that that's like the direction that things can and will ultimately head in. And hopefully, eventually, Morpheus will be like the singular or one of the singular places to stop for all decentralized inference because you'll just be able to access all of it in one place as opposed to having to go to, again, like any one individual provider.
But, yeah, it's really cool to see. We will definitely have Ryan on on stage with us at some point soon to talk about it in-depth. But for now, if you are a DM holder or a more token holder and you want to engage with more dm.com, definitely go check it out. Especially if you're a DM holder and you're looking for a a way to have that DM token be productive for you beyond providing you the dollar a day of inference that, you know, as an investor, you may or may not be actually chewing through on any given day. Of course, if you're a developer, that's probably a little bit different. But
but, yeah, I wanted to definitely kick off with that this week. The second thing that I wanted to talk about this week, and it is changing gears here a little bit, more focusing now on the agentic side of things, is that over the last four days, Hermes agent has now become the number one most used global open source agent platform, overtaking OpenClaw and Claude code, which is pretty amazing. So the figures that I'm seeing
in terms of token throughput let me just make sure I have this correctly. It looks like Hermes has done is it 224,000,000,000 tokens? Which is wild and beyond, you know, what OpenClaw's done at about a 186. Thomas, I know that you are a huge proponent of Hermes and have made the switch from Open Claw. Why is that? Has has this are you surprised, I guess, to see them jump Open Claw? And and why do you did you make the jump yourself?
Thomas BorrelNo. I'm not I'm not really surprised because, you know, when you look at the the evolution of of the tools, every time, you know, there's a new tool, it builds it's built on the shoulders of the ones before and improves some of the key limitations that they had or add some capabilities or things like that. Right? And I think it's particularly at the stage we're at of Vagintic AI, it's a big surprise to see
Hermes surpassing the consumption of OpenCLAR. And it's likely gonna happen, like there's likely gonna be another framework coming up in a quarter, two quarters, three quarters that's gonna surpass what Hermes is doing today. So I think it's just the nature of those tools evolving, the profile of the users adopting those tools changing
as well. And for me, the switch from OpenCLaw to Hermes was driven predominantly by memory and some of the challenges I was having with memory using OpenCLaw. But I can see like, some people I've spoken to have made the switch because OpenCLO still required a certain like, a high degree of technical understanding
to set up and maintain on an ongoing basis, Hermes has managed to lower that barrier. So, yeah, I mean, I started using Clawbot, Moldbot, Open Claw, and over the last sort of month or so transitioned over to Hermes. So went like I went from 100% Open Claw, 0% Hermes, to 90% Open Claw, 10% Hermes, and then there was a switch.
And what triggered that switch was interaction that I had with both platforms. OpenCLR was a bit of a frustrating one. I had one of those agents that builds company profile when I do research on particular companies, and I want that output to be very specific. Like I have this form basically that I want the agent to fill. And a few times I had
OpenClaw come back with the output, and the output was completely different from the template. And so the the last time I asked OpenCLR, like, why are you not using the template? How can we work better together so you always use the template? And and the answer was, it's not a memory problem because it is in my memory MD that I have to use this template. I just didn't consult the memory.
So that was obviously quite quite disappointing.
Bowen BelmerI have a what might be a dumb question, Thomas, but and this is just I mean, I'm a retail user myself. Is what what about the framework of the actual agent is different in that regard? Like like, I couldn't OpenCollege just have a specific memory skill set up?
Thomas BorrelDoes.
Bowen BelmerIt has what? No. No. I know. But I guess what I'm trying to understand is what about the memory skills that exist within OpenClaw, like, can't match what Hermes is doing? Like, are there specific skills on Hermes that I just can't access on Openclaw?
Thomas BorrelSo Hermes has this self improving loop that is quite that is quite powerful, and it's it's really built into the system. I try to do something like that with OpenCLOS, so I had a look at all the logs, look at all the errors, like every day at two in the morning, three in the morning, look at all the logs, look at all the errors, identify issues that you see in those logs, and then propose a way to fix them.
So there was already sort of this self improvement loop that was like, I had already added that self improvement loop to OpenCLO. And every morning I'd get the message saying, hey, these are the issues that I found. This is how I think we can resolve it. Why do you think? If you agree, I'll implement it. So I'd review this, agree, and then it implemented. Would With Hermes, I don't even need to do that. It happens completely behind the scenes. And there's
a lot of discussion around what are the implications, what's the impact on token consumption. Hermes has a high degree of token consumption, that OpenCLO, things like that. The way I look at this is if I can get a one shot answer, and that one shot answer consumes more token than one answer from OpenCLaw that I'm gonna have to go back and ask for more,
then it's worth it. Because at the end at the end of the day, to get the output that I wanted, I would probably have used the same or even more tokens than Hermes, and what it cuts is the frustration and my time interacting with the system to get the output that I want. Even the memory management, right? I spent quite a bit of time digging into memory management for OpenCLO and implementing different strategies from
like, using a vector database, implementing QMD, looking at MEN panels, different different approaches to memory. I could never really get it right. With Hermes, the experience that I had was the complete opposite, where I was doing some research on a particular topic. And then maybe a week later I went back to that topic and wanted to drill a little bit more on it. And when I went to Hermes and say, so
I want to do research on this topic and I have some specific questions, you know, these are the questions, Hermes immediately responded with, yeah, we looked at that last week. Let me look at where we left it last week and then we can go from there. So that right there told me that the way memory is managed across both systems is fundamentally different.
And I was getting a much much better output from Hermes than I was from OpenCLO.
Bowen BelmerGot it. Okay. I have another kind of dumb question potentially. If I am currently a user of OpenClaw and I want to switch over to Hermes or give it a try, what are what are the things that I need to do to, like, capture as much of the existing context and tooling as I can to kinda, like, package it up and bring it over? So obviously, like, I could go in and just download Hermes onto my machine, but is it is it just the sole MD file? Is it more than that that I would need to transfer over?
Thomas BorrelSo as part of the the install wizard, if you will, of Hermes, it asks you for a pointer to your OpenClaw install, and it goes and gets all the files that it needs.
Bowen BelmerOh, so do it just do it itself? Yeah. It just does it itself. Oh, that's interesting. Can I have it wipe itself too? Like, could I literally download the the Hermes wizard, say, hey, go grab everything you need and then wipe Openclaw?
Thomas BorrelI don't know. I I I didn't try that. And I the the permissions that I have on Hermes would not allow it to do that because giving an agentic framework the ability to wipe the machine that it's running on isn't really a best practice, right? Because if it can't delete OpenCLaw, what else can it delete?
Yeah. Yep. And removing it, removing OpenCLOF from your system, you go to like, you know, tilde slash, which is your your home directory, and then in there, there's a dot OpenCLOF folder. So you just delete the dot OpenCLOF folder, and that's it. That makes Okay.
Bowen BelmerYeah. I'm not necessarily saying people should do this. It was more so I was just curious if you did wanna make the switch. I mean, I I know for myself, even just getting my OpenClaw set up the way that I wanted it to on a hardened machine and all of those things was like it was like a challenge for me. Like, I had to learn, you know, more about code than I ever have in the past, somebody that does not have a software background. And so any point of friction to change it up is like it's kinda like, oh, do I really wanna do it? Even even though it might be the better tool.
And ideally, you know, if there is some type of like one shot solution, that's always the thing that I'll I'll tend towards just because, you know, not having the technical background of many others. It's it would just take longer, so I'm trying to say. There is a migrate tool, by the way. Yeah. There's a migrate tool.
David JohnstonFrom Hermes.
Bowen BelmerOkay. Cool. I'll have to give that a try and walk through it, see see if I can't figure it out. Because, yeah, now I'm tempted to change. I have had a couple interesting, like, quirks with my OpenClaw. Like, there was a a point in time where it was pinging the Morpheus API gateway continually, but but not it wouldn't it kept claiming that it wasn't, but I could see my daily credits going down, and then I had to do a bunch of digging to figure out why it was only for to be like, oh, yeah. Actually, I have been doing that. I'm like, okay. Please stop. Well, that's interesting. That's interesting you you say that because
Thomas Borrellast week so I I wiped OpenClock completely from my machine this last week, and but I had basically stopped using it, but it was still running on sort of a segregated machine. But last week, when I was away, I I got a list that I was basically running out of running out of credit. So the the amount of of inference that I was getting access to through through Morpheus was basically running low, which made no sense to me because, you know, I was traveling. I wasn't using
Hermes at all. So why was it running out? And looking at the logs, it was all OpenClaw consuming all that inference even though I was not interacting with it
Bowen Belmerat all. Yeah. That's what it was like. Was it was giving input without requesting like, without output and just pinging it for some reason,
Thomas Borrelbut it was chewing through inference by doing that. Yeah. Yeah. So I I had exactly that. So I ended up having to remote into the the Mac mini that I have to run to run Hermes and and OpenClaw and shut down the gateway so it would not have access to anything. And then all of a sudden, my consumption went back to what what I expected. And that was the that was the nail in the coffin there for for OpenClaw because I didn't mind having it running in the background. And, at times I would sort of look at the output was and what the performance was of
both for the same type of request. But no, I was not going to let Open Clauses chew through all the inference credit that I have just because it feels like it. I actually needed to do some real work, so nope. You're not gonna steal it from Hermes.
Bowen BelmerYeah. No. That is funny. That's it was exactly that same issue. I think I was able to fix it, but I'm glad to hear that you were not having that issue with Hermes. It's yeah. I really I guess the only thing that surprises me about Hermes popping up and nothing should surprise me really in terms of things moving quickly and also people's, like I don't know if brand loyalty is the right word, but kind of, like, brand loyalty is sticking around with any one place. But because OpenClaw was so prolific and and basically
became, at least initially, like the retail standard for running your own agent, I really thought that was maybe gonna be like the Kleenex moment for this for, like, personal agents. Like like, that was gonna be everything gets referred to as an Open Claw or or Clawed Bot or, you know, they went through a couple branding changes. But it happened much more quickly than I thought it was going to that someone else stepped in. And now it makes me feel like there won't be a time where there's gonna be, like, one exact
agent framework, and it'll be just a ubiquitous thing that everyone carries their their context and and souls and things with them to from platform to platform to platform until they ultimately find the one that they wanna stick with for a while.
Thomas BorrelYeah. I think I think it's gonna take a while before we get to this. And and you if you look at, you know, the differences between both both platforms, it also speaks to a different user. Right? You can like, on OpenCLAR, you can you can curate the instructions. You can you can curate the skills. M d, and that's the that's how you improve them. You have to do it yourself,
which is great, right? It's going to speak to people that are ready to put their hands into the heart of the system and tinker with it to see measurable improvements. With Hermes, the agent writes its own skills. And as you ask sort of the same questions, so I'll take the example of the company profiles. So the first time I asked for a company profile,
indicated the format that I wanted, etcetera, Hermes created a skill. And every time I ask for a company profile, it reuses that skill. But if I ask for clarification on something, it goes back and updates the skill. So next time I ask for a company profile, I don't run into the same issue I did the last time I used it. So that's part of this self improving loop that you have. And so if you're a less technical user,
if you're intimidated by opening the skill. Md and making changes to the skill. Md, which is something you would have to do with OpenCLaw. It is not something that you have to do with Hermes. Hermes will do that itself without you even asking.
Bowen BelmerMy favorite type of work.
David JohnstonNo. That
Bowen Belmeris awesome, though. Yeah. I Kyle, do you have anything on this change up in in frameworks or popularity, I guess, otherwise, between these two?
Kyle (Bow Tied Blue Fin)Not not necessarily. I think that each of the frameworks are are good in different ways. I I think the memory management, like Thomas has been talking about in Hermes, is is much better. I think that there's some better aspects of OpenCLOF from the way that it handles the heartbeat and way that it handles the reasoning. But, overall, I think it's
I don't think that we're in the final place right now of what a good agent harness is. I think that we're gonna keep seeing nuance of these. We're gonna keep seeing forks of these that that continue to improve on it. So I'm just excited that that everyone's
Bowen Belmerusing them honestly, and and we're kind of collecting this global knowledge on them for for what's next. Yeah. I was gonna say the best part about this is, on this ranking sheet I'm looking at, ClaudeCode is fourth.
Kyle (Bow Tied Blue Fin)Yeah. Mean, Cloak Cloud Code is is great as well. Right? This is all based on
Bowen Belmersome of the infrastructure that Cloud Code created with Oh, I just mean, though, that the open source options are more popular.
Kyle (Bow Tied Blue Fin)That's Oh, oh, yeah. No. I wasn't talking necessarily about closed open source, but gen but just more generally on the the structure of the harness. Right? Like, Cloud Code I guess you could kind of call Cloud Code open source now after the leak. But regardless, the I think Anthropic kind of figured out the structure of the skills files, and then and then OpenClaw kind of iterated on that and and created the full platform, and then Hermes iterated on that. And now we kind of know what this basic infrastructure looks like in in how to handle agents and and what is an agent. Right? It's kind of just like a skill file.
Yeah. And an agent, you know, a a handful of files together, but I I love all of it. Right? Every time a new one of these forks come out that kind of optimize for a specific niche, I think it helps add towards towards what's next.
Bowen BelmerYeah. Totally agreed. Changing gears here a little bit, I'm gonna do some little halftime self promo here. We put out the latest episode of the Brainpower podcast yesterday, and it was with Ahmad, who is the founder of Intelligent Internet, and was also the founder of oh my gosh, stability AI. Holy smokes. Sorry, people. Stability AI.
It was a fantastic interview. The guy is a genius. I mean, like, without a doubt, had some really, really interesting insights and takes. And if you haven't checked that out yet, I definitely recommend that you do. And also check out Intelligent Internet. It's I I dot inc. They are doing some really deep heavy deep research with AI. One of the recent topics that they covered in papers that they published was, and he and he explains this better than I'm going to right now in the podcast. But, basically, they did I I I believe what he called the 1911
test, which is they took information that only would have been available from 1911, the year 1911 and earlier, and tried to see if a large language model could prove relativity just with that information. So basically simulating Einstein in a way. And spoiler alert, the answer was yes. But also, interestingly, they did what was called the the single or one postulate and were able to prove it with just one calculation,
basically taking special relativity from two postulates down to one, which was very, very interesting. He explains this again so much better than I ever could, or will pretend to be able to. But, yeah, it really was interesting. We also cover a handful of other things ranging from the fact that he is he he sets our p doom likelihood at fifty percent, which is kind of kind of wild. And then we also talk about like humanoid robots and some other things. He was one of the only people I've ever talked to who actually currently has a humanoid robot. I think he even has a couple of them.
So we talked about that and kind of what that's likely to look like in the future, which is also why I'm so glad for things like Morpheus and decentralized AI because if and when all of us have these types of robots in our life, or doing things, whether it's literally in our home, or in our workplace, or even out in public, having versions that we can have sovereign intelligence tied to. So it is acting on your best interest, not the best interest of
x y z entity that, you know, allows you to use it. I think that that's, you know, would be best case for humanity and certainly for all of us as individuals. The only other topic I had here, if you guys do wanna touch on this really quickly, and Thomas, this is definitely one that I'll need a little bit of help with, certainly from a technical perspective, but was the TANSTAK Mistral and UAI path supply chain attack that happened over the last couple of days?
Can you tell us, Thomas, what is a supply chain attack and and why should we care about this? Like, what is going on here?
Thomas BorrelYeah. So a supply chain attack, when you look at open source software, open source software is built on typically built on other open source software through what's referred to as dependencies. So when I launch an application, it pulls those dependencies from the creators of those dependencies that are open source developers. And so a supply chain attack is when you manage to compromise one of those dependencies to
effectively compromise the software that you're trying to run. Okay? So in the case of the attack that we've seen on Tansack and Mistral and so on, This is the mini Shai Vilute attack, which is not a it's not a new one. The the first time it was seen was in September '25, so September. It affected about a 180 NPM packages.
So NPM being the node package manager, the package manager for Node. Js. There's over a million packages in node. So you think that a 180 isn't that much. But the strength of Shaihulud is that it is a fully replicating war. So what it does is it steals credentials from a CICD pipeline, so continuous integration, continuous development. And then once it compromises a software, it tries to compromise
other software that call upon it. So the second wave was in November, and this time it affected 25,000 repositories. The third wave was the trust wallet hack. That one drained about $8,000,000 of crypto. That was in December. And so what we see now with Tansdag and Mistral is sort of the latest embodiment of
this worm called the Shaihulud that effectively compromises software so it basically stages a malicious payload and then injects it into published NPM tarballs, so basically applications. And then once it's done that, it hijacks the project's own CICD
pipeline to publish the compromised version of this software. And so now everyone that uses this compromised software is at risk of being compromised through the self replication of the worm. So I don't know if I I helped clarify or if I made it
Kyle (Bow Tied Blue Fin)blurrier or more complex than I should have. Well, it it made sense to me, Thomas. It was a great expert.
Bowen BelmerAlright. I'm a yeah. There are certain topics that are or aspects that are definitely above my head, but I trust Kyle's opinion of this. If this is like a it's like a sliding scale of comprehension in terms of, like, software knowledge. And I'm I'm the bottom of the scale. So just because I understand Yeah.
Kyle (Bow Tied Blue Fin)Let me TLDR. TLDR, when you run programs, they have requirements for them to run, which automatically get installed. Instead of hackers now targeting you, they're targeting the programs that you didn't realize you were installed.
Bowen BelmerAnd they're finding their way out here. Got it. And, also, in in a lot of cases, it's agents who actually install these things without you realizing it? Not necessarily,
Kyle (Bow Tied Blue Fin)honestly. Like, yeah. Certainly, yes. Right? But there's even some, like there's some major packages. Right? Like, OpenAI based infrastructure, like, in their company, like, uses many commonly used open source packages like everyone does. And the latest one, I think, found their way into their system. Like, OpenAI sent out an alert, I think, today or yesterday about it. So it's not just, like, agent specific stuff. These are code packages that are used in
millions and millions of of pieces of software, you know, across the world. So it's the recommendation for everyone, right, is to go and look at your code base, check your requirements files and your log files, see what versions of software being are are are being installed as part of your NPM and check if any of these malicious pack not malicious packages, but the packages that were impacted maliciously
are installed. And then, you know, you could search online to take the required steps, like uninstalling the package and cleaning some stuff out and making sure you don't hit these kill switches that they have. Like, it's it's it's kind of the first step is figure out if you're impacted. And then if you are, then it's gonna take a little bit more research to to find the steps on on what do I need to do. But people have written many, many pages on on what to do if you were impacted, so I would leverage those.
Bowen BelmerI really hope at some point here someone builds some type of agent or tool that can just be like the the Spider Man of, of security that can basically watch it for you and then flag it and fix it.
Kyle (Bow Tied Blue Fin)That's probably I don't know if David's on here yet, but I'm sure David is already doing that in his EverQuote. I'm sorry. I'm forgetting the new name of it. Morpheus skill. Come on, man. I'm skill. Doing that. But it it's it's really as simple as telling whatever coding ID you're using, am I impacted by the security hacks that went out yesterday? Search the Internet to see what security hacks I'm talking about. And it'll search. It'll be like, oh, this MPM package
is big in the news. You're not using it. So, okay. I'm good.
David JohnstonThat that actually is something I'm doing with my agent is I have it run a daily cybersecurity check and then compare it against my own configuration. And then I sort of have a a broader catch all where there's something in Morpheus skill called bagman, which has a bunch of best practices around how to handle keys and wallets and things like that. But there's also a skill around looking for prompt injection,
which is called, I think, prompt check or and then there's also one around, you know, sort of scanning the the space. So, yeah, if if you're using the Morpheus skill, you know, there's a lot of that built in. But, yeah, more more suggestions always welcome.
Bowen BelmerAwesome. I'm gonna once I have that through Morpheus go, I'm just gonna picture David you looking over my shoulder, making sure everything I'm doing is non malicious going forward. It's a great visual.
David JohnstonThere you go.
Bowen BelmerWell, David, thanks for hopping on. By the way, do you have any we we did cover more DM, and we covered Morpheus just as, like, the inference, aggregation layer earlier. Do you have anything you wanna add to that or anything specific you wanna say on that?
David JohnstonWell, you know, it's it's interesting. I I was just in a meeting, and I was previewing that to folks, and they're like, wow. That makes a lot of sense. There's all this unused, you know, inference where people already have the credits. You know? That's how Morpheus is always cheaper. That's how you guys were able to do this. And I think it sort of clicked in in place for folks in a way that they they didn't understand the arbitrage that was going on before. So big credit to Ryan for pushing that out. I think it's a real breakthrough
to sort of simplify the way people understand, you know, that Morpheus can be this this marketplace reference and just put it in very practical terms. Like, if you're holding Diem, go to morediem.com, you know, and put them to work. Right? If you're not using all those credits, you might as well, you know, put them to work. So I think I think that's really great.
Bowen BelmerYeah. Absolutely. Totally agree. That was kind of a a similar message to what we were saying before is that it it highlights the way in which Morpheus as a decentralized permissionless sandbox, essentially, essentially, can be connected to all of these other things and then provide opportunities for people, in this case, DM holders who want their DM to be productive and ultimately want the compute to be productive, but aren't necessarily chewing through the inference themselves.
One other thing, guys, so this was big. This actually broke just now while we were talking, was and it's not directly related to AI, but it's indirectly related to us. It looks like the Clarity Act has passed the US Senate Banking Committee and now will be moving to the full Senate. So that is a huge win for decentralized assets. And I don't know if you guys have any specific stances on this. We, of course, don't have to go too down the rabbit hole of the Clarity Act, but looks like we're gonna be getting some legitimate legal framework here in The US for crypto assets.
David JohnstonThat's great. That's Just got a little Yeah. Well, it it it's a good signal that the bankers were all in up in arms at the last second trying to scuttle the the because I think the the biggest fight was something that was already settled in the Genius Act, which is that it was okay for Coinbase, let's say, to pass along the yield to users. And I think the banks sort of blatantly realized, wait a minute. Why would people use a bank account if I can get yield from my USDC?
Right? And and sort of, you know, try to relitigate all this in the Clarity Act, trying to prevent their own users, their own savers from, you know, having that yield that is is created, you know, from stablecoins because most of the stablecoins are holding them in treasuries. They're collecting four or 5% annually. And most of the stablecoin issuers like USDC and and Coinbase are passing that along. Right? Which you'd think the banks would, you know, wanna see, but, you know, it really doesn't fit their model. And so, yeah, it'll be interesting to read what actually gets into the final version,
you know, and and then I could say, you know, more definitively whether this is good, whether this is bad, and, you know, how it all plays out. But yeah. I mean, at least the way it looks from the outside, it seems like a big milestone for the industry.
Bowen BelmerYeah. Totally agreed. And and certainly long overdue in terms of just having this framework in place. So that's excellent. That was all I had on the official things we must cover this week. I know economic junk junkie has jumped up on stage. What's going on, man? How are you? Hey. Yeah. Hi. I
Irida TVI had nothing specific. You know, I'm just I I just wanted to say hi. I was just listening to your podcast with Iman. Thought it was amazing. Super fascinating topics you covered there. And, also, you know, when David just explained the so I don't know who just explained the whole more DM thing. It it just now clicked with me that we can do something
that, you know, OpenAI, as far as I know, can't do. You know, you have people paying for inference, but can they easily share that inference amongst one another? Not that I'm aware. So not another win. Yeah. Not only can they not, but
Bowen BelmerOpenAI likely wouldn't even allow them to.
Thomas BorrelFrom their perspective,
Bowen Belmerunused inference is like, that's basically more revenue coming to them with less overhead. Yeah. It's free. Yeah. Yeah. It's literally free money. I mean, it's like the gym membership that no one uses. Right? Yeah. It's it's like a it's like gift card, you know. Walmart makes, like, I don't know how much millions every year just from unused gift cards. Yeah. I think I saw one time that Starbucks with their, like, Starbucks dollars in their app, was it like they were they would be like in the top 40 banks or something in The US if they were a bank? It was something crazy like that. Like the amount of money that's just sitting in these Starbucks
like dollar accounts is massive. But but, yeah. No. That's exactly right, is it allows you to actually get some productive benefit as a user for the inference that you're ultimately paying for. Now in this case, you'd be paying for it through the ownership of DM, but you're entitled to that dollar a day. I mean, that's that's the, you know, conflict with Venice,
Irida TVultimately. You could I have a buddy of mine who'd he just, you know, invests. He holds in Venice because I do. And, you know, I tell him, okay. I believe in this, but I'm not telling you what to do. But he does it, and he but he doesn't use the inference. So then I told him, here, you can just you know, you can stake your DM here, and, yeah, he'd happy to do that so you can earn some extra yield. So he's just interested in basically some speculation, but at the same time, somebody else can benefit
Bowen Belmerfrom that. Yeah. Exactly right. And from a user perspective, this is why more the asset is so powerful because it gets you access to any available inference across the whole network. And so, like, DM in this case would just be one compute provider that's allowing you to access the inference that you need to run your agent, but they could be any number of providers. And so it's Yeah. Yeah. It's it's quite amazing, and it's awesome to see it start to play out in real time.
Guys, we definitely should keep the conversation going here. I, unfortunately, do have to drop. I have a a call I absolutely have to be on in six minutes at two, but I definitely, like, keep the space open, and let's keep it rolling.
David JohnstonYeah. Awesome. Thanks, Bo. Good. And good to see you on stage. Economics, I know you've been in the Discord for a long time. Oh, yeah. Yeah. Your story and what brought you into Morpheus.
Irida TVYeah. Yeah. And I saw you in Miami recently. I I was so mad that I couldn't make it, but it was just I wasn't paying attention. I didn't know that that consensus thing was going on. I wish that. But next time, whatever, you know, whatever's going on around here, I definitely wanna show up and, you know, meet up in person to talk some more because I always love talking to to to you all.
David JohnstonThat's great. Yeah. No. I think consensus is in Miami the next two years in a row. So definitely be Nice. More opportunities for that. But it's always cool to see people in the community come out to all the events. You know? And I think the big thing that's changed in the last six months is now we're all using agents. Now we're all accessing the inference from Morpheus. Like, I use it every day. Right? Yeah. So I don't have to sort of explain in theoretical terms anymore. It's like, hey. Here's my agent. I'm powering it with Morpheus.
Irida TVYou know, it's routing to this privacy and layer on Venice. It's great. It's a great point that Bo made on this recent podcast episode, or I don't know who made the point, but he was somebody was saying that the challenge with cryptocurrency and decentralized money has been that there has been this there there's this super old entrenched centralized system that we're competing with. But with AI, that's not the case. The the decentralized and decentralized systems are growing together.
And I think that's such a huge plus for for us in decentralized AI right now.
David JohnstonSure. Every time a new open source model gets released, you know, decentralized AI becomes more compelling, more powerful, more useful. So, yeah, it's amazing to see, you know, Google released all their Gemma models. Great. You know, bunch of Asian groups release, you know, GLM five or 5.1 that makes it better on coding. And so sort of all the work that's happening on the AI side is just making it more and more exciting.
Irida TVI mean, it seems like, basically, the closed source AI model is completely dead.
David JohnstonYeah. Well, I mean, their their vault gardens become a lot less compelling as the open source model start to saturate the benchmarks and can do a lot of the same things.
Irida TVYep.
David JohnstonAnd so, yeah, it's gonna be interesting. But I think Morpheus community has has done the hard work the last three years building out all these primitives. Right? The inference marketplace, all the technology, and, yeah, now we now we get to put it to work. So
Irida TVAnd you think that so because I've been doing I have not gotten into agents yet. I'm I've been, you know, doing client work, consulting work, and all that stuff. I can't wait to get into that, and I ordered my Apple I think it's it's a mini m two or something like that, and it's just sitting there. And I can't wait to fire that up and connect to think it's Everclaw that that you recommend, everclaw.xyz. And then it
from what I understand, it should be pretty easy to to get it set up from there.
David JohnstonThat's great. Yeah. No. I'm glad that you're gonna get an agent. You know, I started using OpenClaw back in January, you know, and I wanted to connect it into Morpheus. And so I started rolling and open sourcing these different skills, and I've branded them now the Morpheus skill. Right? So initially, I called it EverClaw, but I think people thought it was an agent as opposed to a a skill set that you could use with any agent. Right? And Mhmm. And since then,
to their credit, Hermes has really taken off. I see that they're using now more usage for Hermes agent than there is for Openclaw, which is what I've seen among my own friends. You know, a lot of the people that started with OpenClaw ended up using Hermes because it had better memory and better defaults and sort of was configured more more powerfully. And so, you know, I really recommend when you set up your agent, I would go to Hermes. I would use that, and then just tell your Hermes agent to use
the Morpheus skill. I've recently actually rolled a Hermes skill specifically for Morpheus and put in a poll request. Hopefully, they'll accept that poll request. I've published all the code. So, yeah, you'll be able to connect your Hermes agent with Morpheus, hopefully, pretty easily. So you recommend that when I get started instead of doing that curl
Irida TVget dot EverCloud x y z that I I do something else with with the Hermes?
David JohnstonWell, it works with both OpenClaw and Hermes. So depending on which agent you wanna use, you can import the Morpheus skill either way. And I'm gonna post Hermes specific install instructions.
Irida TVOkay. Cool. Cool. Yeah. Yeah. I'll I'll I'll look look out for that then.
David JohnstonYeah. But for for me, it's been a game changer. You know? I've been able to, you know, execute more coding the last four or five months than I have the last twenty years. I mean, it's been
Irida TVtotally not night and day. Right? I know you're you're saying on the podcast that all your your projects you've been tabling, you're able to tackle now. Yep.
David JohnstonI had quite the back backlog of ideas and concepts, and all of a sudden, I could just tell my agent what I wanna build, and it'll access that model on Morpheus and code it up. You know, now it it moves at the speed of which I can articulate to my agent what I wanna do next. So, yeah, it's it's pretty refreshing. Right? It's like having a coworker that is online twenty four seven, always executes what you want. Yeah.
Irida TVYeah. So I I was wondering, would you feel let's say because, for example, I have a client who I do consulting work for, and I work in their QuickBooks system. I do stupid little tasks occasionally for them where I have to just work through line by line. And would you say you're comfortable enough with your setup to unleash it on somebody's billing system to do menial tasks? I I'm kind of terrified of that because I don't know what if it makes a mistake, then that that would be a huge disaster
David Johnstonfor the company. So But So so the way I I think about it is I tend to put the agent in its own isolated environment. Treat it like an employee. Right? Give it its own email. Give it its own phone number. Give it its own GetUp account. If I were gonna give it access to one of those systems, I would treat it like an employee. Right? You can only read these files. You can only have access to these levels of permissions. Right? You don't wanna give it access to the admin, and it makes a bunch of mistakes. Mhmm. But I feel pretty comfortable at this point. Let's say you give it read access to those systems,
and then it's giving you all of its work products. Right? And you have the edit power. You have the admin power. You can then take that information and output that it gives you and use your expertise and wisdom to execute faster. Right? And and that's, I think, where I'm where I'm comfortable doing is treat it like an employee, isolate it, give it permissions like you would with an employee, and then it's in a pretty good position. Like, if you make a mistake, well, it's only in its workspace. Yeah. It's only, you know, in in the things that it read
and the output, oh, it needed to be tweaked. That's fine. Right? And so I'll have it always come back to me before it pushes anything to production. I built that into I actually open sourced this last week. If you look at my X profile, you can see the standard operating procedure that I published for developing code and deploying code, and it has all those steps and checks and audits built into the process. So that's the way I would think about it is give it give it the permissions you would a junior employee
so that it can execute, but, you know, not make mistakes into other workspaces.
Irida TVRight. Okay. So that's the tricky part, I I guess, because let's say I give it the access that I have to my client's QuickBooks account because it's not just read access. It's write access because it has to be transaction has to be classified and stuff like that.
David JohnstonBut, again, that should be a way that you can do that locally. You may be able to dig into the permissions and see if there's some more advanced permissioning. Mhmm. There's m t yeah. P Go ahead, Scott. I'm sorry. There's MCP tooling that you can use with QuickBooks. So the Yep. It there's built the QuickBooks that you can tap into.
Irida TVOkay.
David JohnstonYep. Okay. Yeah. A lot of these programs are starting to give you an agentic method of interacting. So it doesn't have to go through the GUI. Doesn't have to go through the software. For example, Google recently released all their CLI tools. So if you gave it access to Gmail, you can now that do that through the CLI, the command line interface, which is more native to the agents. Right? As opposed to having it take screenshots and kinda try to poke around the GUI, makes a lot more sense for the agent to be using the CLI or, like Scott's saying, the MCP connection.
Irida TVGotcha. Yeah. Yeah. I'll I'll try to get that set. I'll I'll start small, and I'll try to work my way to to that confidence level.
David JohnstonWell well and what I would suggest is, you know, more and more people are popping into the Discord and posting about what they're billet what they're building, and that's where sort of we get this flywheel of people see what each other are building. Yeah. Right? And what they're connecting to Morpheus and which models they're using and where they're hitting problems, where they're being successful. You So I just encourage you. There's a whole builder section in Morpheus Discord to to really put what you're you're doing to the point that, you know, of what information you feel comfortable being public. You know? I think that kind of those best practices are really, really useful.
Irida TVYeah. Yeah. Yeah. Great. Yeah. I mean, I I don't wanna monopolize a lot of the oh, sorry. Just Didn't wanna wanna place the space here. Yeah. I just wanted to say hi.
David JohnstonSure. No. This is great. Appreciate the questions, suggestions. I think we're just getting to the top of the hour so we can wrap it up. But, yeah, this has been really exciting week. It's great to see all the traction and the usage increasing. You know, I think one of the suggestions by the community, Thomas, was to, you know, look at and and just share with people every week, like, what are the most popular models people are using at Morpheus. I looked at the analysis before this. It looks like right now the the two top models are GLM
5.1, GLM five, and then the Gemma models have really picked up a lot of usage, and so that's another one to highlight. So, you know, touching on those kind of, you know, metrics, I think, is really interesting as people use the tech. Great. Well, why don't we go ahead and wrap it there? Appreciate everybody jumping on. Another great week, and hope everybody has a good one, and everybody keep being awesome.
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...