Town Hall Now With Live Video! | Morpheus Turns 3 Years Old
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 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 #89! Recorded: 3 September 2026 AI SUMMARY: In this week's town hall the show went fully live on video while examining Morpheus network stats and their ecosystem implications. Strong growth in native and total throughput was highlighted with August setting a fresh high and usage nearing 50 billion AI tokens this year. Morpheus' three-year anniversary was celebrated reflecting on evolution from white paper to a live network with hundreds of models dozens of providers and real products people use daily. New visibility tools were explored including dashboards for model activity provider health and marketplace competition plus discussion on why 2026 marks the shift from "come and build it" to "come and use it". The conversation covered product experience private AI social coordination with agents and experiments like BuddyBots and OpenClaw aiming to make personal agents easier to use. Topics also included Morpheus' growing traction across Asia model naming challenges the rise of ultra-cheap high-performance open models like GLM 5.3 Flash and why cost per completed task matters more than benchmark theatre. The session wrapped with thoughts on the OpenAI/New York Times copyright ruling and its implications for open models access to knowledge and the future of intelligence. https://www.youtube.com/watch?v=dWh_82HjsHM Timeline: (00:00:00) Intro (00:00:30) First live town hall and Morpheus network growth (00:03:40) Morpheus turns three and the vision becomes real (00:06:24) New dashboards reveal models, providers and network health (00:11:11) From building Morpheus to getting everyone to use it (00:18:42) BuddyBots and making personal agents easy to launch (00:23:46) Asia expansion, events and model naming challenges (00:33:03) GLM 5.3 Flash and why cost per task matters (00:48:04) OpenAI copyright win and the future of accessible AI 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 BelmerFor those tuning in, whether it is live right now, if you're watching this live right now, you are witnessing history because we have stepped into the twenty first century and are not only doing this call via audio but also video live. Obviously, David and I do a lot of podcasting together. Others on the show have been on podcasts as well. Never done it truly live like this, though, which is funny. You know? Of course, we don't edit, too much
in the recorded stuff, but it still adds a level of of you know, makes it more real. But, anyway, for those joining this or watching the recording, thank you for being here. You're in the very early days. So we are gonna be using this call, which should be streaming on YouTube and also X for the purpose of our town hall every week to talk over everything Morpheus, everything built on Morpheus, and then also, of course, news related to DAI and and crypto and other things relevant to the community, and sometimes just fun stuff that we wanna talk about. The good news with video is we can also share our screens, which is huge.
So we could actually show you what we're talking about kinda as we're talking about it now going forward. So to kick off this week, as always, let's take a look at the network throughput. Let me see here. Let's do
David JohnstonAs, as you pull it up, Bo, go to the go to the native section. We just about got to the doubling of the previous all time high. So, yeah, if you go to the top and select native and then go, there on the right instead of, BOLT or gateway, Epic native. And, yep, there you go. So pretty small time high, 2.2, 4,200,000,000, for August, which is really awesome to see. It's a good,
solid organic growth of the usage.
Bowen BelmerYeah. So exactly like David just said. I mean, so we really quick, total throughput, we are now nearing 50,000,000,000 tokens. Again, those are AI tokens, not crypto tokens, which, of course, can sometimes be confusing. But we are nearing 50,000,000,000 tokens of throughput just this year, which is insane when we consider the fact that a you know, even twelve months ago, there was essentially no throughput. I mean, there were some, but, like, not, you know, not like there is today.
But now if we look at the growth, we are mostly up into the right. If we look at native only, we are also mostly up into the right. And as a reminder to those watching, native only suggests that this is inference coming directly through the decentralized compute router, not through the API gateway. Although the API gateway is fantastic, and I personally use it. I know everyone on this call uses it. The native throughput is also very important because it shows growth, like, ecosystem wide, which is net positive, just generally speaking, for the project. But, yeah, in August, I mean, we doubled up the previous all time high month.
We're on pace right now to have a similar or better month already in September. This is after the trimming of some of the, what's a good word for it? I was gonna say spoofed throughput, but spoofed isn't correct. People using Free free credits. People were using a lot. Yes. Free free credit throughput. This is actually paid throughput, meaning all people have staked more tokens to access this compute. So generally speaking, just just huge for the project.
I'm gonna stop sharing this really quick. Do Oh, I was gonna say
David Johnstonhappy three year anniversary. Yeah. It's the birthday of Morpheus, September 2 every year. You know, coincidentally, I noticed also the birthday of Keanu Reeves, and apparently Atlas shrugged. So there's a bunch of random things that happened on September 2, but the community has sort of been celebrating. And I was thinking, like, how far has this project come in the last three years? Right? You've now got dozens of providers providing hundreds
of models that have served hundreds of thousands of sessions that are now outputting billions of tokens. Like, considering it was a white paper and that it was, like, a beta implementation to, like, where it is today, that it's it's just nice to take for a moment and and pause and be like, wow. It really has come a huge way in the last three years. So I just wanna say I really appreciate not over everybody on this phone call, but everybody in the community
that has, like, kept focused on the vision and seen it get to where it is today and the open source models catching up. And everything that's happened has has really been wild. So appreciate you, Bo and Kyle and Renee and Alan and everybody that's put in the work to make that a reality.
Bowen BelmerYeah. I mean, absolutely. It it is amazing. It's it's one of those things where now, like, all the things that were just being spoken about even you know, I got I got involved, like, a little over two years ago now, so there was already a ton of things happening before I even, you know, discovered the project. But even from those days to now, there's so many things that were theoretical and being spoken about, and we would meet and talk to other members of the community and say, hey, isn't it gonna be cool when we can do isn't it gonna be cool when we have 10 models hosted? Isn't it gonna be cool when we have redundancy in all the models? Isn't it gonna be cool? You know, the API,
all of these things that are fully functional today. And it's it's what separates, in my opinion, Morpheus and the community from a lot of other things out there. I use it every single day. You know, even even DeFi. Like, there's things that I use in DeFi and I would consider myself, and I think most would consider me to be like a power user in some regard. But I don't use it every day. Like, I'm not every single day doing something. I'm quite literally I bet it'll get to the point where I and others will be using Morpheus every hour of the day, you know, pretty soon.
Yep. And and I there's probably I'm sure even people on this call who are well ahead of me in that and and will be, you know, continue to be. Well, and and just to that point, I I love that we now have the video capability of showing off the screen. So tech.morr.org
David Johnstonis great, but you should pull up active.morn.org. Yes. I do. This is my my new favorite website because it's the other set of statistics that really matter, which is like there's 221 routable models. Like, that's insane. Right? And you can see right away which ones are the most popular, like DeepSeek v four Flash, you know, GLM 5.2 web. And you can see, like, the real health. Like, there are 16 providers now offering DeepSea v four Flash. Right? You basically got almost perfect coverage because you have so many backup providers.
So credit to Alan for creating this. I think this one is is awesome. If you also go to the network detail tab there at the top, it shows you provider by provider, like their reputation if you will. Like if you scroll down, it shows you whether, you know, they're online, they're offline, they're healthy, how many sessions they're doing. You can actually get a pretty good view of like, okay, I've I can see how good this provider is versus that provider, or you can sort by what they're charging, you know, per model, and all that kind of stuff. So, Alan, appreciate you you,
you know, kind of bringing this stuff out because, man, it's so much easier to understand once I can see it.
Alan (Nomadic Rogues)Yeah. Absolutely. This is I'm I'm glad everyone's digging it. The the data is there. Right? The data is all on chain. It's how do you expose it and how do you share it and then how do you consume it. So, you know, this helps us understand how the, you know, it helps us check is the provider rating system accurate, you know, are you getting a good marketplace of those 16 providers that provide your deep sea v four flash, what's the best one? What's the most robust? And
that's what the network was built for and so as we get that that mass, that center of mass that you talked about 24 providers, you know, all of the bids and all the models is great. That's gonna continue to add data to say, alright, have we built a robust system? Have we built a robust network that's gonna survive? And true marketplaces, right? So, all those deep seed before, are you actually having a good competitive marketplace where stability and reliability
and speed are rewarded on the network. And I think that's showing a lot of that support. Yeah.
David JohnstonThat's really awesome. So thanks for all the visuals. But yeah. Beau, I think every time it's like, where's the token usage? How many providers? Like, what what's the number of models? Like, these are the things that are dynamic and, you know, sort of up into the right and interesting, right, is is just watching all these numbers. Because, you know, I remember the early days of Bitcoin, like, people would just go and refresh the websites that showed me, like, number of transactions and number of wallets and blockchain.info.
Before it was blockchain.com, it was blockchain.info. Like, people would just they wanna see what's happening on the network. So, yeah. So I'm really excited about that. But share those those links with people. That's what makes it real. When I go talk to people now, I show them tech.morn.org. They're like, oh, wow. Look at that chart. Look at the usage. Or, oh, here's all the models. Active.more.org. Then all of a sudden, it resonates in a way where, like, talking about it theoretically
Bowen Belmerdoesn't hit the same way as just showing them people are using it. I actually I had a moment just like that this weekend, David. I was I was in the city at a at a friend's birthday party who he works in the crypto space, and somebody that he works with apparently used to work at blockchain.com, and they don't anymore. And the guy walks up and was like, you look really familiar. And I'm like, oh, okay. We start talking. And he was like, oh, you you and that guy David did the Morpheus podcast with us a couple like, whatever. Was like a little over a year ago, and I was like, yeah. And he was like, well, how's that going? Like, is that you know, where are you guys at? I pull up tech.morrow.org,
active.morrow.org, and I'm like, you tell me. And he was he was blown away by how far the project was because it I I just it's still such an undiscovered gem. It's the signal to noise in Morpheus is, like, off the charts, which is a great place to be, and and, you know, the stride will happen. Like, the the thing whatever the catalyzing thing is, it's going to come because all of the plumbing is already there. You know? And so, yeah, this is, like, really fun to show off, though, and it was one of those things where, like, yeah, it happened in real time, totally unexpected, and it was like, wow, this is this is fantastic.
To that end, on on the Morpheus side of things, yeah, actually, Kyle, I was about to turn it over to you. Do we have any updates on on the protocol side, API side? Anything that you wanna share?
Kyle (Bow Tied Blue Fin)Well, let let me just comment on the the last thing first is that I I think that's why we're here. Right? That that's why we're doing these spaces and doing, you know, doing the podcast is, you know, the last you we talked about the birthday. The last three years or the last, let's say, two and a half years was kind of building this. Right? And now that we have the usage, now it's it it really has turned from we need this in the world to we have this in the world and and seeing the usage and,
you know, having the data that you can, you know, trust that, hey. These transactions are happening. These tokens are moving. We've kind of gotten to the point now where it's like, let's use it. Like, let you know, it's let's go realize the value of the system that we've created. And, I think that as more people have that conversation of, hey. You you promised years ago that that this would, you know, come to fruition at some point, and now it's here. I think all of those people who were interested then
will be more interested now. Right? Because it's an idea, but now now I can come and use it. So I think, you know, something that we all want to do is get this out to the masses. Right? Get everyone in this arena who wants this access, get them using it, whether it's through the API, whether it's through some of the node directly, whether it's through some of the other, like, clients that people are are looking to use. We don't care. Right? The we are,
you know, in in infrastructure play here of the underlying backbone of how you get access to decentralized, you know, and private AI, and we want people to build on top of it. We want this to be that core that people create businesses on top of or, you know, run all of their businesses inference through. So, now is kind of that time to scale, and we invite the people who, you know, fundamentally agree with this to share it, right, to get it out, to help get this message, you know, further and and not you know, people did an amazing job of that in the beginning when when we were talking about the the vision and the mantra of what we want. I think now we need to
make that call to all our supporters to to help get that message out again, but this time as a, you know, start using it rather than than help us think and help us help us plan.
David JohnstonYep. No. I think that's the fundamental difference in 2026. It's instead of come and build it, it's come and use it. Mhmm. Right? And it's a hell of a lot easier to use it than it was. Build it. Right? The building That's for damn sure. You know, detailed, bespoke, you need a lot of coding background, whatever. Using it, copy and paste the key. Go. Right? Here's where you get an agent. Use it. Have fun. Right? And I'm having a blast, man. You know? I I get to I get to use it with my kids. Like, they wanna make a cartoon,
I just tell, you know, my agent and it uses Morpheus and out pops some cool content. Right? Like, that we've been talking about that. Now I do it every day.
Kyle (Bow Tied Blue Fin)Absolutely.
Alan (Nomadic Rogues)And I think I'd love to I'd love to put a plug in for what we've been doing behind the scenes with our Buzz communities. Right? You wanna talk about drop in, right, is when, you know, we're setting up our, you know, internal Buzz and some community Buzz environments, And you have an option to create a a global agent. And instead of going and picking one of the, you know, Frontier paid for, I hooked it up with Morpheus. So our Buzz communities are backed by our own dog food. This is a drink your own champagne time, right? Is, you know, the the interactive
agents that you get. We're going we're putting through our own system, so we're using it on that production as we collaborate internally or as we collaborate with the larger builder community. It's driven by what we have built here. And I think that's pretty cool is setting that up and and having to integrate work work like a charm. So it's not Build It, but we're we're certainly using it, leveraging it.
David JohnstonYeah. 100%. I and I I wanna call it right now. Like, using Buzz the last few weeks has been wild to see agents and people in the same channels going back and forth. Just like the speed you can get things done. I think the social layer is gonna be the breakout moment when this goes mainstream. It's not people talking to an agent by themselves in a corner all isolated. No. It's gonna be you and your friends building whatever you want with your agents
right there. Right? And instantly sharing that context and and creating things. So, yeah, that's it it's it's pretty cool. I think Buzz is like the early implementation. We'll see if, like, that goes mainstream, but, like, somebody's gonna build some easy to use version of that where the agents and the people are all coordinating together.
Bowen BelmerSo, yeah, can't can't wait. I get mildly stressed anytime you call something in the future, David, because you're right a lot more than you're wrong. And so it forces me to, like, lean into the chair a little bit and listen a little closer. And it's it's the same feeling I get when when the market is absolutely crashing and it's just like a a red wedding type of day, I'm never stressed out because, like, the conviction is high. So I look at it like, oh, this is a great opportunity, like, add to the bag if you have capital. I get more stressed when things move up very quickly because it always feels like, oh, no, the boat's gone, like, the ship's leaving, you know, that was it. And
that's that feeling of, like, weird twins of twinge of stress when things are going well, that's how I feel when you make predictions. Because I'm like, oh, I better pay attention because he's probably right. Well well, it is a common it's a common pattern. Right? Anything
David Johnstonhighly technical people figure out for themselves is only six to twelve months away from somebody building a simpler version of that for millions or billions of people. Like, look what happened. Like, to give credit where credit's due, like, it's the Morpheus community who was experimenting with personal AI even before OpenClaw. Mhmm. Right? But OpenClaw wasn't easy to use. Like, I spent a lot of time trying to set up friends and others. It took, like, four hours, like, you know, going through a terminal and stuff like that. So Noose was a simpler version of that. Right? And they had a moment and and saw an explosion of growth. And then you could say,
you know, Grock Bot has been an even simpler and easier version of that. And it's had a huge explosion, right, in usage. So much so that I I saw somebody posting like, the Grock Bots are are out of inference. Yeah. Elon's got a lot of inference, which means it's been a very popular product that people want. And so just look at the same thing. Like, we were talking about coordination of agents before Buzz existed. Mhmm. And now Buzz is the engineer version where you have a Noster ID and and you could do these things, and then boom, there's gonna be some simplification
of that UI. And maybe it'll come out of the Morpheus ecosystem, right? I've been experimenting with a project built on Morpheus called BuddyBots.
Bowen BelmerIs there a front end for this yet, David, that I can show? Yeah. You can you can go to buddybots.org.
David JohnstonRight? And what's underneath is Morpheus providing the inference manifest for the agents and the the CPUs. It's basically a light client for install OpenClaw, which is already like one click give me an agent. My my goal with BuddyBots is I wanna be able to get my friend an agent in the chat. Just tell my agent to make them an agent, and instantly it spins them one up into the chat connected to their phone number or connected to their WhatsApp ID. So this is super super beta and early, and I still got some work to do on it, but like, it's getting pretty close where this could be like that mass
market moment is when it's simple and easy and one click and everybody wants to do it. So I put out a simple Chrome extension, and it's really just sits in your your browser and and links you back to install OpenClaw. Because I think my thesis is like maybe one person in the group will be technical enough to wanna, you know, click it and set it up, you know, the manual way. And then they can just do it for all their friends and family instantly through the chat. So yeah. I I did a bunch of updates. By the way, thank you to everybody who signed up for the beta. I'm happy to say since we talked about this last week or two, we've hit the first 200 beta users
of install OpenClaw, which which is great. That's, you know, giving us a ton of feedback on how it works. But, you know, this is what you want is I name my bot, I hit the launch button, boom. I've I've got an agent in under thirty seconds, and I logged in with Google or social or whatever, however I wanted to log in, right, and there was no download, there was no terminal, there was no complexity. So, yeah, pretty pretty excited about this. And this, to your point, is like, now I get to do this on Morpheus. I couldn't have done this a year ago.
Probably couldn't have even done it six months ago, but when we hit June and the open weight models get like GLM 5.2, this is using DeepSeek flash DeepSeek v four flash in the back end. It's a great model. It's highly capable. It's like quad quality, but a tiny fraction of the price. So anyway, this is free. If people wanna sign up, this is all open source. It's all on GitHub. Try it out. Give your give your agent a name, hit the launch button, and boom. You should be set up and good to go and
hopefully solves a lot of the pain point. Because I saw OpenClaw two point o just came out. I used one of my my my claws and immediately it died. And because it broke all the configuration stuff. And it's like, yeah. This I love you guys, but this still isn't easy. Right? So we do need these abstraction layers that make it easier.
Bowen BelmerYeah. Well, that's awesome. And I appreciate the efforts very much, David. I love having somewhere to point people to to, yeah, go from that zero to one moment for them with them having their first agent. And it's just, you know, being a seamless experience. So you don't have to do, like, a big learning curve, which I think is just gonna continue, to your point. And and also, it'll be, in my opinion, part of the catalyst for why people realize the importance of the private aspect. Because now it's like, oh, I can spin this up very quickly. I can use it. Oh, wait, it has access to all my personal information. Or maybe I want it to have access to personal information. And actually, I think that this is
maybe how more people will think about it, is I think most people will realize, I shouldn't upload this personal document to, you know, x y z central business, but, oh, I have a private version of this. Okay. I do actually feel comfortable doing that because I know and I can verify that it's staying, you know, local to my device or local to my agent. And, yeah, it it'll just, you know it's it's all the things I say often, but, again, like, it's this this process and, like, this succeeding, it's it's boiling
boiling the frog or boiling the lobster. It's not a a microwave solution. It is a slow build, but that's how you know? There's a great Ted Lasso quote about this in the most recent episode for those who are fans. He says something about winning somebody over, and he says it's like cooking a great steak. You can't just crank it with heat. You have to slowly rise the heat over time, and then you end up with the, you know, the perfect the perfect what's the word I'm looking for? Steak, I suppose. But I forgot what word he used.
Cool. Well, that was all I had on the Morpheus side of things. If you guys have anything else you wanna make sure we cover, let me know. Otherwise, we do have a handful of new stories to dive into. Yeah. Yeah. Well, one, I wanna bring Renee in.
David JohnstonThey've Oh, yeah. Sorry, Renee. You know, I don't think people in the in the West realize how popular Morpheus is in Asia. Like, I went to two conferences they took me to. Like, you know, I'm on stage, huge crowds, like, there's robotic, you know, robots, you know, escorting me to stage. Like, it it's really known there and, you know, they've got university students and programs and meetups and all these people using like, you see that usage. A lot of that's coming
from Southeast Asia. And so I just wanna give credit to Renee and Eric and all the crew that they've got there that it's sort of you know, this use case is so powerful for people, you know, in Asia. If I don't know, a $100,000 max studio, but they wanna access this intelligent, like, this is what that is made for. Just the way crypto was popular in Asia. Right? It was a big deal in China and parts of Asia long before
it hit mainstream in The US. So just wanna call out Renee and and all the efforts that they've done and say, I'm gonna be at token $20.49
Rene Bernardcoming up Yeah. 3 or 4 weeks. If you guys are gonna be there, come come and talk to us. But, Rene, the floor is yours, man. Thank you for the very kind words, bro. Yes. Good. Good. Good. Nice that you recognize the efforts we are doing over here. Yes. We're trying to be somewhat the bridge between the continents and and connecting to you guys over there, which is obviously hard for the time zones. I realize it's early morning for you. It's bedtime for us here. Man, we need to figure that out how to better connect. But, okay, I I just wanna mention that
we had many events mostly focused around Singapore and Malaysia, which is in the center of Southeast Asia. We are, in Indonesia. We we had a big event, the the Indonesia blockchain week just happened now around two weeks ago. We were present there. I had a team there on the ground. We handed out flyers, and that I wanna say we did this one four par. We had printed lots of flyers with this $10 free credits for the API gateway in July, and we still handed them out because we did not get the memo from our friends from US The $10 free credits was abused and switched off. Oh, man. So now we need to reprint and and and change that approach. Too successful, Renee. You Yeah. But I don't know who it was who abused that system then later. I I understand what you're saying. This is an abusable thing when we give send a lot of free credits. I get it. Yes. But then we have printed it on flyers. Right? So now we are reprinting them. Never mind. We need to be in calls like this more often. So but but then
I have one one important these events we are going, we have I I can't even tell about universities. Now we have Cambodia coming up. We'll be in Phnom Penh with a team of four people from office Asia. We have university events lined up. We will have a small booth at the Cambodia Blockchain Summit. It's I'll be we do many things happening here now, but what I want to now say, one important thing. Everybody of us is using this API gateway, and thanks. Shout out to Kyle for helping to build this wonderful piece of software.
And and we are, as in the internal people now, we are mostly trying to shift to this native Morpheus proxy and shift away some of the load from the API gateway to native. And I came across that one issue with the canonical naming of models. I you guys know there was this DeepSeg v four update of the 07/31. Right? So I configured all our these these agents and Morpheus proxy to exactly that model because I wanted the latest version. I was not sure whether the v or Flash
is the latest version or not. Is that the April version or the July version? And I still didn't figure out how do we get that naming right to prove which model is which. So now, okay, we switched our Morpheus Asia load over to v four Flash, hoping it's the last version, but we don't know. How do we tackle that one, guys? What's the approach here? Well, that's why I highlighted the active.more.org
David Johnstonbecause it's got so much of that information you're looking for. But I'll hand it over to Alan.
Rene BernardThanks. Yes. Yes. Please help us.
Alan (Nomadic Rogues)Yeah. It's what you pointed out is one of the biggest frustrations we have in the network today Because it's permissionless, because it's open source, everyone is able to register their own model and model names. And so we've been working very hard. Of the buzz conversations that's that's going on in the background right now with with our agents is how to maintain permissionless, maintain anonymity and but register models
such that everyone can then bid on them. So you don't have all the different naming conventions, all the different naming variants, spaces, dashes, skewers, etc. So that you can know, alright, I go to Hugging Face as a reasonably canonical source, Can I find the same model on active.more.org and know that that's the latest or the greatest? And so that's where we're, I'd say drifting to what we've got right now is a proposal on the table for a better model registry
that closer aligns with Hugging Face because that helps us understand because you can go to Hugging Face and look at release date and weights and all that kind of stuff. And now with Nvidia's announcement that they will care and help care and feed for Hugging Face, I think that is amazing for the open source community and allows us to help standardize on that. So, absolutely hear you on the trouble with what model should I go after if I want the latest and greatest
consolidation of the model names so that there's one box of cornflakes for the providers to bid on. I had a great conversation yesterday with with one of our members on it's difficult for the providers. Which one should I bid on? Do I bid on them all? Do I have to have this this nasty models config so that I bid on every variant of GLM 5.3 I could find? No, that's that's terrible. That's a that's a it's not a I hesitate to say it's a terrible design. It's a design you find when you're at scale, realize, oh, this needs to be taken care of. So we do have a proposal up on
deck, we're working through it. It's difficult. How do you how do you create a model name that allows itself to hang out, wait for it to be griefed if there's an issue with it and then finally get locked in. But the net is, we hear you it's a problem for consumers and for providers and we're drifting toward canonical source of hugging face as well as the frontier providers, their naming conventions and hopefully we'll build that into the network over the next few months.
So we hear you absolutely. Thank you. It's great. Okay. Good. Good. And I know that you guys are working really fast. So this probably in a week or two, we'll already have that resolved as I know you guys. Thank you. That that one that one, no. That this one's this one's a sticky wicket because we also have we certainly all always also have to have backwards compatibility. We can't we can't can't hand grenade the the existing network. So live live live wire activity on a contract change is is not fun, but I appreciate the I I thank you for the confidence
for sure.
Bowen BelmerToday's I don't know. What I heard is we'll be Afternoon. It'll be shipped by next call. Yeah. Next call. Yeah. Yes, boy. Yes. Yes. Okay.
Rene BernardNo. Thank you. I I I didn't really prepare well for this this call right now with the proper list of updates. Like, next call, we will give more updates, and let's hope we can we can schedule so that we have overlap for our awake times.
Bowen BelmerThank you, guys. Sounds great. Thanks for jumping on, Renee. My dream is that Morpheus just absolutely explodes in Asia first and just gets massive over there. So then, you know, we'll have this huge community, on the other side of the world. I think that's so cool. I mean, we already do, but, like, in terms of just, like, general use case, I I think that that's
Rene Bernardboth likely and also would be awesome. So we appreciate it. But then, Bo, Bo, let's do a roadshow. You come over. You've never been here yet. You've not been in my house yet. I would to. That yeah. No worries. It will Pack your bag. We will make that happen. Exactly. One day. And then we do a roadshow together, and then we make your dream happen. Yeah. We definitely will. Although, I don't wanna get
Bowen Belmerpart of an international incident like David was. Kidding.
David JohnstonWe solved it in the end. Yeah. It'll it all worked out. It was it was a lot of fun. But but no. I mean, the the famous Bowen,
Bowen Belmerlike, you know, you you just kill it over there. It's like a rock star going to Japan. Yeah. That'd be great. The the best outcome ever would be, you know, brainpower in Morpheus is massive in Asia, and then I can go home and, you know, no one ever knows me. But but but over there, there's a huge community. Oh, that'd be great. Well, no. Actually, though, I will I will at some point get out there, Renee. I'd love to do that. Let's let's dive into some of the news here because I know we're a little bit short on time, but we can zip through a lot of this. So one of the biggest pieces of news over this last week was
GLM 5.3 Flash, and just Yeah. How incredibly performative it is. And and or I shouldn't say performative. That makes it sound like it doesn't it's fake. How incredible it performs, but also how it costs, like, negative dollars to use, basically, which is just insane. I mean, I was looking at some of the the stats, and and this brought up a good conversation that we had yesterday, David, about, like, we as a community and as people who are paying attention to this stuff need to start reframing
the, like, how we think about cost. Like, it should just be in terms of cost to complete tasks as opposed to, like, straight up token throughput be cut because ultimately, like, you're gonna be using it for things. And centralized providers and players can kind of, you know, strategically present numbers how they want, but it's not necessarily true when you put it to task. I mean, David, do you know you have comments on this.
David JohnstonSo, yeah. As as my friend would say, you don't always need a Lamborghini or to go to the grocery store. Right? Like, the idea that you're using Fable 5.1 to do most tasks is just absurd. Right? And and actually, this came this really came to me the last few days when I saw the release of Fable 5.1. And it's the first model I've kind of been like, okay.
And does it do anything the last model didn't do? No. But it's five times as expensive. And and to your point, we should stop playing their game of benchmarks. They're just gonna infinitely invent benchmarks that it's five percent better, 10% better. And it's like, but do I care? What's what's the cost per task? Yeah. It starts to mean nothing. Yeah. Yeah. Right. And it doesn't mean anything to me because I already have a great model that can do it for
10¢ for that task, a dollar for that task versus $50 or a $100. And so there's this great it's a Pareto distribution. And if you go to my x profile, you'll see I've posted this curve a bunch of times, but I think it's the best visual out there today. Like, for getting the task done, does it cost you $10 or $5 or $1 or 10¢? And then there's this curve. Right? And, yeah, you get slightly better accuracy. That's not heck. It is. There it is. So, like, yeah, okay, I get slightly better accuracy
if I go further and further up that curve, but look at the number of dollars at the bottom. Right? The the cost just explodes. And so my thesis is the middle of that curve is like the ideal spot. And what do you see there? GLM 5.3 flash, DeepSeek v four pro. Like, you're not paying too much, but it's really good on getting things done. And what does the percentage mean? Okay. It might have to do an extra turn. It might have to do, you know, a little longer work. But, like, if your agent's off there doing stuff for you, like, what do you care if it takes an extra minute or two, but it's 10 times cheaper? So this curve, like this this cost curve, think is everything, and it's what we care about.
And we ought to be talking about when something comes out, what's the cost per task? And then rewarding, oh, wow. They just got that done for that price? That's crazy. I can't can't believe it's that efficient. Right? Because this is where Grok and DeepSeek and GLM are just killing it, is on cost per task. And so this is a cool milestone. Like, till June, we were sort of waiting, waiting, waiting to get the milestone where, like, it was good enough.
And I think GLM 5.2 was that moment in June where like, okay, now I can stop using quad. And then when you saw GLM 5.3 flash, it's I can stop using quad and wow, it's cheap. Yeah. Right? And so we've crossed both of those milestones.
Alan (Nomadic Rogues)I would I would love to see this. I think this is a great representation and it's difficult. I'd love to see it over time. Right? If you could imagine going back in time and then as they stack in the future is where things move. We had a conversation this morning and it's amazing. Everyone was excited for, you know, GLM five one. Like, my god, it's gonna be the best ever and this was around what the fable the fable mythos fable time frame. And then, you know, five one was the best thing ever and then five two came out and like 5 one's so passe, but the token cost for $5.01 started to go down in $5.02 and $5.03
and and it's so funny, the $5.01 that we thought was amazing didn't even exist in people's memories a year ago, right? We couldn't imagine. So it's seeing this change over time, where does it I I would imagine there's a there's a cohesion at some point is you'll start to get clusters of this cheap inference but really good capability, kind of like you were talking about David, that that bench there, I think gets more defined
and that's where people really start digging in and you've got really good inference for not a lot of cost. And I would love to see if we, you know, kind of starve out some of the frontier the frontier providers that that sell all the, you know, you must have mytho, you know, Fable five one or else you're not you're gonna lose out. Like, we don't care.
Bowen BelmerYeah. Okay. Fine. Think that's exactly what's gonna happen, Alan. And think that's why some of these big companies are, like, in a rush to go public and things like this, so they can make their exit before that happens. Because eventually, like, we've already gotten to the point, and this happened very quickly, where in pure and raw intelligence, the models are way smarter than the average person or really any people, ultimately. I mean, like, even if we go back to, like, ChatGPT two, may maybe this is hubris and totally not correct, but at the time of ChatGPT two coming out, I used to write a lot of blogs. I was writing blogs, like, five days a week on the real estate space, And then ChatGPT two came out, and I remember initially being like, I can still write better than this. So, like, I it still makes sense for me to manually do that. That is not even remotely close to the truth anymore for, like, anybody,
basically. And it maybe not wasn't even true then. But my point is we're now at the point where, like, that that pure, like, intellectual horsepower has just completely been eclipsed, and now the the use case will also be eclipsed, meaning that every model will be smart enough and capable enough to do every, you know, normal feasible task that the average person needs. And so the incremental increase just will not even be detectable because you'd have to be trying to do tasks that, like, you aren't even thinking of. That's kinda how I'm viewing it. It's it's almost like,
in the automotive world, horsepower just doesn't matter anymore. It it means nothing because now we have electric cyber trucks that that can tow a Porsche faster than a Porsche because of, like, electric engines, immediate torque, that type of thing. And so now, like, if you're measuring a driving experience, it has to be by something that's not horsepower because it just, like, has completely become irrelevant, essentially. You know, same kind of thing. I think that's what'll ultimately happen.
David JohnstonYep.
Bowen BelmerYep. And then they're gonna get blood dry by centralized providers.
David JohnstonYou you know, I I've I I I would love to see that. I think it's unlikely to happen because what's gonna what's gonna change, and honestly, what what already matters today is less the model and much more the product Mhmm. And the convenience and the ease of use. And here's where we're in a bit of a bubble is we have the technical ability to access these cheaper models that most people have never heard about.
They don't even know it exists, let alone how to access it. And so the challenge I would put to the Morpheus community, in the next year like, we we we've had three years. What's the next year? My hope is that we're able to make the leap to an incredible UI UX experience where somebody can get the benefits of those better cheaper models with none of the technical hurdles. And if we can get there, then yes, we can challenge
Quad and OpenAI and groups like that, but having the models alone ain't gonna cut it. Yeah. What I You have to compete on product convenience, bringing the person's context, and just making it just work. Right? It's the old little apple. It's like, it just works. And that's ultimately what people want. And usually that traps them in these gated walled gardens. Right? And our our hope and our vision and our goal ought to be how can we get them that convenience
without sacrificing their privacy or their agency, and still the value flows to the user. Like, I think we can do that now. We we we have we didn't pull it off in web two, but I think we can do it now. Yeah. I was gonna say, do you think that happens in in decentralized
Bowen BelmerAI in a different way or or more easily than it did with, like, Web three and DeFi and blockchain? Because that was kind of the shared concept, right, is that, like, hey, get it get it in people's hands so it just works. They don't really need to know how it works under the hood as long as it works, and somebody who knows what they're doing can verify it. But but in my opinion, like, blockchain still hasn't really at least DeFi hasn't really gotten there. It has more than ever. Like, it's definitely heading in that direction, and there's people using apps and and businesses that run on top of chains and things like this. But it but it still hasn't really broken that, like, everyday person zeitgeist.
I think this has the capability of doing it more more, like, straightforward, but I'm curious to hear what you guys think. Well, it's it's because we have the AI tool itself.
David JohnstonYeah. All of a sudden, like, I can make a beautiful website. I can have my agent run a marketing funnel and do all of those things. Like, doing the protocol, great, then building the network, then building the products, but getting all the way to the user interface. You don't need a team of a thousand people anymore. You don't need a 100,000,000 or a billion in venture capital. Like, the cost of building these things just keeps falling.
So it is the tool itself that is making it realistic first time.
Alan (Nomadic Rogues)I I agree. The the simple stuff gets solved, like you say, in the the front end, the UI, and that allows the humans and the agents to work on the really hard problems. Right? So the good engine for Morpheus is is its crypto, you know, background. And that's tough. That's a friction for a lot of the mass adoption. So that's where I'm spending a lot of time on is is how do I mask that? How do I make that a non issue? Because that's what maintains the privacy, the sovereignty, the the anonymity that, you know, all that kind of stuff. You need that baked in, but how do you mask that or how do you how do you make it easy? And I you know, Bo, to your comment on on web three,
I don't think people really tackle that. They're like, look, you gotta you gotta be, you know, you got to be one of those guys and understand how to, you know, set up a node and and do that stuff. I think to David's point, we have the horsepower in a smaller more concentrated area. So I can spend an afternoon and go, alright, give me this UI. But now I'm gonna spend a week on how do I mask the complexity that makes it friction and remove the friction and that will solve this and I think there's that flywheel that's happening. Have all the work that I'm doing, the work that I see the rest of the community doing is absolutely all of that is that that accelerate accelerated
flywheel because of the tools you have. We've taken Moore's law and applied it at such a greater aspect. It's not just to compute, it's output has been is doubling
David Johnston18 days. Every 18, not eighteen months, every eighteen days. It's just amazing. Yeah. And and just to build on that, Bo, did you see the launch of this new Linux client? I think it's pronounced Omaki. Mhmm. But it's been all over my x my x feed, this is a famous developer, d a h, that did like Ruby on Rails and a bunch of other stuff and companies that people use all the time. And he's like, hey, Linux never got to mass market,
but with an AI it can. Yeah. Because all of a sudden you just tell your agent what you want and it will magically fix your setup versus only being people who are developers and will spend a hundred hours optimizing their setup. Like, this changes the game. And so I think we're gonna see this ripple across a bunch of stuff, where it's like there were a bunch there was a lot of friction to do x. Now we can rip out all that friction. How do you reimagine
the operating system or the company or the whatever with this new reality? So personally, I'm I'm pretty excited because there's a lot of things like, oh, it should be x. Like, this is how things should work. And then it doesn't in reality because of all that friction. You can rip out that friction. Like, you have to really reset your assumptions about how things are gonna work in the world with this new building primitive of an agent that can take out all that friction.
Bowen BelmerThat makes sense. Yeah. That just reminded me. I had not seen that that new version of Linux OS. That's awesome. But it reminded me that there's this company, this startup called Monaco Glass, and they're making a pair of, like, AR glasses that run on Linux so you can run open source software. And then you'll be able to have a terminal open and talk to your agent that way. And I can't wait. I have one on preorder. I don't think they're shipping for a while, but it made me think I forgot about it. It just made me think of it. And so shout out to those guys because pretty soon, I'll have my agent.
We'll be leveraging, you know, new Linux, and then I'll be able to just talk to it on the move, and it's gonna be awesome. Cool. I don't know what I'm gonna do with it, but I'm gonna do something. Tell you what. No. That's great. Okay. There's one other news story I wanted to make sure we covered. I know we're at top of the hour here so we can rip through it, and it kinda ties into a second story. But the the big news was that the US government sided with OpenAI against the New York Times, the DOJ saying that in
training an LLM on copyrighted works does not violate copyright law, and treating it as an infringement would hurt US science, prosperity, and national security. Not really sure. They just throw national security on the stuff because it sounds cool. But the this is big, actually, I think, for AI, and it's it's a big win for everyone in AI, centralized providers and open source alike. It comes at the cost of a win for the centralized providers as well, But it it does, you know, support the idea of, like, freedom of intelligence and information and speech and being able to access that information,
you know, on a uninhibited basis, basically? Well,
David JohnstonI'll give you my hot take. Copyright for so long has been abused way beyond the original fourteen year, you know, for for patents and copyrights. Now it's, a like, hundred years or whatever, largely thanks to the Disney Corporation wanting perpetual royalties on the mouse. Yeah. I'm glad that's dying. Right? Copyright has been so abused that it locked up what should have been public works now for generations. And so that's finally starting to break down. So I think in general, that's a good thing. Yeah. You know, it's like Voldemort
winning a suit, and you agree with the outcome, but it's Voldemort. Yeah. Yeah. Right? So, you know, scam Altman, like, not a fan, but it was probably the right call by the court. Because think about the analogy for you. You read a book. Mhmm. Are you able to talk and write about your reading the book? Yes. Obviously. Yeah. Right? And it's hard to say that it's different. Well, I had a picture of my reading of the book and I used this calculator. It's what an LLM is. It's a word calculator
to read the book. Okay. Well, it's just an extension of you. Right? So I think it's a good thing because we want information to be free, open, transparent, knowable, and not all to be lost. History. Do I like that they're buying old books, copying them, and then burning them? No. Not a fan of Yeah. That's not great. You know, that that's not that's not ideal. But it was it was probably the right call by the court, and probably sets a good precedent for open weight models and freedom of information. So Yeah. It's also just a battle that cannot reasonably be won.
Bowen BelmerI mean, because it just, you know, we see this all the time with with China and other places. Like, they don't they don't abide by US copyright laws. Like, that's not a thing for them, you know, and it's gonna continue to not be. And to your point, David, it's funny how when something becomes more accessible, all of a sudden the tightening of the grip happens. Like, if I was watching a Disney movie and then I was using binary and coding my own rendition of Mickey Mouse doing something using binary,
no one would think twice about it. It would take me twenty years to do it, but no one would think twice about it. But now that I can spin it up instantly, it's a problem. You know? And so it's, yeah, it's funny how that works. But the yeah. We definitely agree on this for sure. And it is one of those things where it's like that meme that's like,
David Johnstonyour your least favorite person makes a great point. It's like it's like the fake headline, you know, from whatever, the onion or whatever it is. But I just wanna say, Bo, I really like this new format. I like that we can pull up websites. I love that we can show things off. Yeah. And I think it's a lot easier for people watching me like, oh, go there, click this.
Bowen BelmerSo Yeah. Having the live chat is great too. I didn't I mean, I knew it existed, but I didn't this is our first time people, like, really experimenting with the live streaming as opposed to recording. So, yeah, it'll keep improving. We'll make it better. Awesome. The only last thing I wanted to to mention is with the win for OpenAI against the New York Times, it is expected that Astra, which is their upcoming model, will be released imminently. Knowing us, it was released, like, literally while we were talking. I don't think so. But I don't know. One hour ago, they did make a post. No. It's not it.
So, yeah, we'll see what that brings. I mean, again, though, this harkens back to the conversation we were just having. It's gonna do all of the things better and be really expensive. Like, that's that's what it's gonna be, I'm sure. So we
David Johnstonwill we'll see. We will work hard to make sure as many people as possible can benefit from AI, own their AI, keep their keep their value that they like Keep intelligence free. Yeah. Freedom of intelligence. 100%. Cool. Cool. Well, thanks, guys.
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...