What Happens When Render, Manifest, Jember & More Talk Decentralized AI + Privacy? [Full Transcript]

This is the full text transcription of our Twitter Spaces discussion featuring leaders from Render Labs, Render Network, Manifest Network, THINK, Jember, and Scrypted on Decentralized Compute, AI, and Privacy.

You can also listen to the full audiorecording here:

03:48 — Sunny Osahn (Render Network Foundation)
Good morning, everyone. I hope you’re all well and good. Just so everybody knows that, well, just so I know everyone can hear me, could you give us a thumbs up, please, guys? Just to make sure everyone can hear. Oh, there we go. Excellent. Thank you, Taco. Thank you, Shapeshifter. And think. Excellent. This is going to be an incredible spaces. Welcome, everybody. My name is Sonny from the Render Network Foundation. And this AI Spaces is focused on decentralized compute.

04:18 — Sunny Osahn (Render Network Foundation)
AI and privacy, all privacy, depending on your regional accent. Now we have a whole host of people here. We have Paul from Render Labs. have Trevor from the Render Network Foundation, Tim from Scripted, Eric from Manifest, Edward from Jemba, and obviously, Taco. So yeah, I think we could just get into this. So Taco, how are you doing first of all?

04:49 — Taco (Mainfest Network)
GM I am doing amazing. I think from everyone on the on the panel today it is is nice and morning time but I am on the opposite side of the world just finished super AI here in Singapore and getting ready to head to permission to fly to New York tomorrow for permissionless so. Wow. OK. Yeah, you definitely yeah well I I was an AI panel.

05:18 — Sunny Osahn (The Render Network Foundation)
Yesterday in Berlin and I flew back yesterday and back home in the UK. So it’s afternoon for me. I’m kind of like midway between yourself in Singapore and others where it is GM for them. But the topics we are going to cover today guys is:

• The convergence of AI and privacy as app developers build models, agents and tools.
• The untapped potential of decentralized compute and how it can power a large amount of applications that builders may not even know are possible.
• Why privacy and decentralization principles work hand in hand to protect users’ data.

That’s very relevant as well because of a recent breach that everybody got notified about. I think that was yesterday or the day before. And we’re also gonna talk about:

• Use cases across traditional FinTech finance, generative AI video, tools for filmmakers, and a whole lot more

So it’s going to be a very interesting conversation, I think. So setting the stage, let’s start with the big picture. We’ve got AI development happening on two very different fronts. On one side, the tech giants with huge resources shaping the narrative and the other a growing push for decentralization.

06:43 — Sunny Osahn (The Render Network Foundation)
Blockchain-based models that aim to flip the script. So how each of you thinking about where AI is headed in a world shaped by both of these forces? I will give this one first of all to Trevor.

07:01 — Trevor Harries-Jones (The Render Network Foundation)
Thanks, Sunny. Hey, everybody. It’s pleasure to be here. And some of these panelists are absolutely amazing. So for us, where this is going is really exciting. Today, most of AI is fairly centralized. But as we look at where it’s going and developments recently, like Apple’s liquid glass and infrastructure behind it, we see more and more of this being pushed locally and to the edge.

07:31 — Trevor Harries-Jones (The Render Network Foundation)
And really for us, decentralized is the core of that push. And it makes a ton of sense when you think about not having to have your data ever leave your device. I think it’s absolutely central to the battle that’s happening at the moment for your data in AI and one that is sorely needed to protect all customers in an AI world.

07:59 — Trevor Harries-Jones (The Render Network Foundation)
We see it very strongly shifting and supplementing in an AI direction, not replacing, but pushing certain tasks towards local and towards edge and reserving others for centralized. OK, excellent. That’s quite a balanced take there. Let’s see. Tim. Tim, can you hear me? I can.

08:23 — Tim Cotten (Scrypted)
Guys, thank you for inviting me onto the show. It’s great to be among legends. several of you I know and are friends, and I just deeply appreciate the opportunity to talk about what script is doing. And I just want to launch in with, I’ve been thinking about this particular set of problems around centralized AI, liquidity of compute, and the freedom of expression, both for humans and for machine.

08:51 — Tim Cotten (Scrypted)
For over two decades. that really began for me when I was developing AI for video games at Electronic Arts in the early 2000s. And I worked on the first generation of massively multiplayer online games. And those are supposed to be living worlds full of non-player characters. And it’s only now that the state of the art has really gotten to a point where we could see a next evolution of that. Equally, there are many lost opportunities at the moment as the giant monolithic companies are

09:21 — Tim Cotten (Scrypted)
Essentially hoarding as much hardware as possible while the rest of us are left to wonder how do we keep computing with the AI? So it is of course wonderful to have Render here and everyone else because we’re going to need to push a lot of this stuff to the edge where no one can really stop what we’re trying to do. And when I mentioned freedom of expression, I mean things that corporations

09:50 — Tim Cotten (Scrypted)
Play it as safe as possible, right? They always want to hold their reputation. And there are many things where we just as average people would greatly benefit from a more decentralized AI, especially in AI video production. And Scripted is leading the way on that. That’s really what I care about doing right now is taking our agent frameworks and taking the best parts of Web2.ai video and bringing that to Web3, because I think there’s great opportunity there.

10:19 — Sunny Osahn (The Render Network Foundation)
OK, excellent. That, again, is quite a well-balanced answer. And I like the term monolithic in describing some of these larger entities in the space. OK, excellent. think let’s have Eric.

10:40 — Eric Bravick (Manifest Network)
Hey, can everyone hear me? Loud and clear. Awesome. Thanks. Great to be here. Always great to talk to this level of panelists. So yeah, this is topic we run into a lot. And we’re really trying to develop a nuanced conversation around big tech here. There’s obviously

11:10 — Eric Bravick (Manifest Network)
Big tech has created a lot of good things and a lot of bad things. And when you’re talking about AI and the reshaping of society around AI and what’s coming, we have to keep this conversation balanced somehow. But it’s hard. We tend to get into a lot of doomerism when we’re talking about this.

11:37 — Eric Bravick (Manifest Network)
I think that’s largely around the incentive structures that corporations have to hold. And of course, decentralization is one of the nice answers to get around some of those incentivization structures. One axis of this conversation is what are those incentive structures and how does decentralization

12:06 — Eric Bravick (Manifest Network)
Breakdown maybe where big tech won’t if even if it wants to be able to have a healthy approach to AI. We all know the problem, right? We all know that AI is going to be super important, has a lot of potential good uses. We see a lot of that at Manifest, of course, you know,

12:35 — Eric Bravick (Manifest Network)
Everything from great biotech to health stuff and so many more. big tech is going to have some incentive problems here. And that’s really what we need to, I think, get right with our blockchain infrastructure. A lot of people talk about, and rightly so, talk about data privacy and

13:03 — Eric Bravick (Manifest Network)
And enabling creators and using decentralized hardware and all those things are very important. And we love all those things. But I think really this is a battle of incentives, bottom line. And I think we need to focus a lot on what are the protocol based incentives that we in the decentralized world are putting in for the direction of all of the AI tooling that we’re going to be.

13:33 — Eric Bravick (Manifest Network)
Building. Right now we’re thinking a lot about incentive structures would be the first place I would kind of talk about there. Does that make sense? Okay. Yeah, yeah, that makes sense. It is very interesting to delve down that sort of route in terms of thought. Okay, excellent. I think what we’ll do is move on to the next question unless anyone has something burning they’d like to say on that particular

14:02 — Taco (Manifest Network)
Topic. Well, yeah, no, Sonny. I think one of the things like we have we we know everyone on the panel with us today, but we have so many different communities joining us today from so many different sides. I’d like to take some time introducing each of the panelists that we have.

14:25 — Sunny Osahn (Render Network Foundation)/Taco (Manifest Network)
Oh yeah, absolutely. Why not? Everyone knows who we have today. I think we start backwards from where we are right now. Mikey from Who are you and why does I care? Yeah, sounds great. You know, I just love this conversation. Such good friends here. One of the things I started to see the power of this back in 2014. were we basically incubated a company.

14:52 — Mikey Anderson (THINK)
They were doing pretty good. They’d grown to like one or $2 million a year in revenue in the first 18 months. It’s like, okay, this is great. We said, hey, what would happen if we brought in kind of this cutting edge recommender algorithm AI? We got it working and we accelerated that revenue to like $12 million a year in annualized revenue in like a seven, eight month period of time. And so this was right around the time that the algorithmic feeds really started kicking in and social. And it became really clear, like AI is a tool that can watch the users and figure out

15:22 — Mikey Anderson (THINK)
How to aid them on a path to buying products, to making decisions. And it became really clear to me, oh, this is the ultimate manipulation tool. This is the thing that slowly over days and weeks and months can adapt people through strategies to be able to change almost any opinion. And that’s powerful and that’s scary. And as I went on, we built

15:44 — Mikey Anderson (THINK)
And I got involved in one of the first ML ops companies because I was like, this is the future. I want to be right at the ground level. We built the United Nations global data program, a lot of the research and development systems for Nike, Toyota and Merck. what we saw this full time is that even in these big companies, the ability to scale across clouds, to be able to bring all this hardware together, to be able to watch all of the different agents. Where’s your source of truth?

16:10 — Mikey Anderson (THINK)
It was actually a problem that we even had back then. And we identified that the Ethereum blockchain was the ideal place to be able to log agent actions, like hashed agent actions, so that the developer can track what’s happening, so that you can track across multi-cloud setups. And we wrote a paper on how to do that. And so this is an extenuation of something that I’ve been working on for a long time, is that we need to own our intelligence. And so we’re building the Think Agent protocol, which is intelligence you own.

16:41 — Mikey Anderson (THINK)
Thanks for having me. Awesome.

16:46 — Taco (Manifest Network)

Always great to have you, Mikey. Paul.

16:49 — Taco (Manifest Network
Introduce yourself so everyone knows who you are and what you do over the magic you make. Thank you. Good morning, everyone. Yeah, no, I spent 10 years with Google, built a lot of different ML systems there. And the conversation this morning between incentives and big tech and decentralized resonates a lot with things we’ve seen over 10 years.

17:18 — Paul Roales (RenderLabs)
Know, currently building great solutions for the community at Render Labs. is applications and infrastructure on top of the decentralized render network that will help users utilize the network, utilize these decentralized platforms, get these incentives right for themselves. And I’m excited to dig into this. It’s been a, I like the tone of the conversation this morning. Thank you guys.

17:45 — Taco (Manifest Network)
Edward, are you there?

17:50 — Edward Katzin (Jember)
Hello, are you able to hear me? Perfectly clear. Who, what do you do? yeah. So, Taco, thank you. Sunny, thank you for pulling together and assembling such an esteemed panel. I have to admit, it’s an honor to be amongst old friends and also see some faces that I recognize as incredible notables in the space. So thank you. I’m Edward, Edward Katzen. I’m the CEO and founder of Jumper.

18:17 — Edward Katzin (Jember)
I’ve been working in AI, I hate to admit, since the early 90s. back in the early days, the early 90s, early 2000s, all of the stack that we built was proprietary. We literally ran our own hardware, built our own models, ran our own infrastructure. It wasn’t until into the 2010s that I really saw the rise of big tech and the shift to really leveraging centralized and controlled infrastructure to run our AI.

18:45 — Edward Katzin (Jember)
And so what’s really exciting to me about the state of the industry, this conversation, and the assembled panelists and audience is the opportunity to really explore how this is evolving. Because what we’re building at Jember is the tools to enable either web two companies or web three projects to automate their core business workflows and their core business operations with secure AI workloads. But to do that, we have to honor privacy. We have to honor the sovereignty of the company and the people’s data.

19:14 — Edward Katzin (Jember)
And we have to do it in a way that ensures availability and reliability. So by leveraging the key essentials of what Web2 companies expect, the advances in decentralized technology, we’re really able to provide that. And we’re really excited about what we’ve been building on both manifest and render. So thank you for the opportunity. Awesome. Thank you. Thank you for being here this morning. Tim.

19:42 — Tim Cotten (Scrypted)
Yeah, guys, my name is Tim Cotten. kind of mentioned it earlier. I’ve been around in the space for a very long time. And I kind of front run, I was front running the AI agent space, even back last year in June, pitching it on stage with a 16 Z during crypto startup accelerator. And there only a few people who really understood it. And I was very gratified to see that the virtuals team, for instance, and then shot at AI6Z.

20:12 — Tim Cotten (Scrypted)
Had managed to push the entire AI agent infrastructure forward really fast. And that led us to understanding we can bring superpowers to web three, but also ask the question, what can web three do for AI? And I strongly believe that it’s not just about decentralizing the compute, although I believe that’s actually incredibly important. And I believe providing

20:41 — Tim Cotten (Scrypted)

Like liquid ways of routing for various types of compute, like quickly on demand with frictionless payments is incredibly important. But also simplifying for the user, because right now everyone’s stuck in a prompt land where everyone is becoming an amazing prompt engineer. And our key insight is from the gaming industry, nah, people like buttons, people like clickable icons, people like to mix and match. They like to be able to create content the fun way.

21:11 — Tim Cotten (Scrypted)
And so we’re building a product for that powered by agents and powered by, well, several of the networks that we’re talking to in this chat today, including the ability to decompose payments when people send them to us with X 402 rails. So I’m just hyped to be here. I’m looking forward to listening to what everyone’s working on and then how we can continue to push the space forward. Awesome. Thank you. Eric.

21:41 — Eric Bravick (Manifest Network)
Hey, yeah. So I’m Eric Bravick. I’m originally a neuroscientist and psychoneuropharmacologist. I have been in AI for about 35 years and tech about the same amount of time, helped build a lot of the core original internet out, and then got into decentralization, fast forward through a ton of AI stuff.

22:10 — Eric Bravick (Manifest Network)
Got into decentralization about January 2010 and said, yay, we’re finally here. We’re finally starting the journey to having some new tech around distributing AI workloads. So been working in that space ever since, just trying to bring together AI and decentralization. Sorry, yeah, AI and decentralized tech.

22:39 — Eric Bravick (Manifest Network)
And yeah, focus is really on AI. And at Manifest, we really are trying to build the sovereign cloud infrastructure. Essentially, you’ve got sovereign hardware at the bottom layer, sovereign cloud in the middle layer, and then that’s what gets you to sovereign AI.

23:06 — Eric Bravick (Manifest Network)
And then, of course, looking forward just a little bit, as narrow models exceed human capability, which many already have, you get kind of the beginnings of what will become an AGI there as you knit everything together. I won’t go through the technicals of that, but if you do that all

23:33 — Eric Bravick (Manifest Network)
The core idea here is if you do that all centralized, you’ve got some incentivization problems there that could be worked through but are difficult. If you do that decentralized, you get a better shot at setting up incentives to be, for example, pro-human, et cetera, et cetera. So that’s really the goal that we’re working for.

24:00 — Eric Bravick (Manifest Network)
We see AGI’s of various types emerging over the coming decade. And we’d like a lot of those to work for humans rather than for say, know, maximizing attention or something like that. So that’s our core mission. Nice. And last but not least, I actually just was talking to Trevor on the phone.

24:29 — Taco (Manigest Network)
A great human being. But Trevor, who are you and what do you do?

24:34 — Trevor Harries-Jones (Render Network Foundation)

Hey, thanks. I hope everyone knows me at this point, but for those of you who don’t, I’m Trevor. I’m on the board of the Render Network Foundation. And I’m also a COO at Otoi, the largest contributor to the network today. My background is actually building and scaling large SAS web systems. So everything about pumps being insufficient

24:58 — Trevor Harries-Jones (Render Network Foundation)
Really resonates with me. I love great consumer interfaces. For me, what’s been so exciting about this next phase that we’re moving into here is as it bridges from just AI into agentic, I’m starting to see some phenomenal things that can be done now in terms of interfaces that I think unlock consumer usage in a way that just was not possible when we started this project.

25:26 — Trevor Harries-Jones (The Render Network Foundation)
It’s staggering to watch it change day to day here.

25:31 — Taco (Manifest Network)
Thanks. Sunny, and for the last people that don’t know who you are, who are you?

25:38 — Sunny Osahn (The Render Network Foundation)

Hey, guys. So yeah, my name is Sonny. I am also from the Render Network Foundation. I take care of our grants program, and I also assist in social engagement. hosting things like this, getting to chat to some of these elite people. It’s like the Avengers of AI have assembled here.

26:00 — Sunny Osahn (The Render Network Foundation)
Coining what Edward was saying there about assembling. Yeah, so it’s amazing to take part in this and co-host this space with you, Taco. I’d like to know, so we’ve heard for years that decentralized compute would be an absolute game changer, but so far the space has moved in slow, uneven steps. What feels different this time around? Is it where AI is now?

26:28 — Sunny Osahn (The Render Network Foundation)
Compared to a couple of years ago? Is it new regulation, better blockchain infrastructure, or is there something else? Who would like to take this?

26:40 — Eric Bravick (Manifest Network)

Well, I have an answer that’s near and dear to my heart, if you want me to lead the charge out there. Continue. OK, so I think there is one factor here to consider is there is a really big difference between

26:58 — Eric Bravick (Manifest Network)
The theory of decentralization in AI and the practice. And this is largely a physics problem. So even back 35 years ago when everything was tiny and we were just doing basic back-propagated neural networks and things like that, like really simple stuff, really simple math, it became very apparent very quickly that you had physics problems with

27:28 — Eric Bravick (Manifest Network)
Distributing this type of computation. You’ve got to move a lot of data around. You’ve got to move a lot of memory spaces around, especially in training sets. And that’s become increasingly so as the technology has advanced. we took at Manifest, we looked at this problem and being deep infrastructure experts

27:56 — Eric Bravick (Manifest Network)
We actually solved it a different way. Rather than moving how the compute works from a physics perspective, as far as where it’s located, how it’s configured, et cetera, in contrast, MostD pin is a collection of hardware that is scattered widely.

28:25 — Eric Bravick (Manifest Network)
Throughout the world, it’s usually consumer-grade hardware. Nothing wrong with that, especially for certain job types, and I’ll get to that in a minute. But there was something missing from that equation, which is if I need the physics to work the way they work at hyperscale, which for some use cases is the efficient way to do it, do I have still a decentralized option?

28:52 — Eric Bravick (Manifest Network)
And the answer was always no. But what we did is change the ownership and sovereignty model for the people who own the hardware in those hyperscale data centers. And that gave our customers an option to do what we do with the render network, for example, which is we triage the job coming in through AI.

29:22 — Eric Bravick (Manifest Network)
What needs to be privacy preserved and run in a hyper scale type format, we put in dense data centers where you can have say a chassis with eight B200s all hyper memory connected and you can do like a super large memory job. And then we route the jobs for example that can be

29:50 — Eric Bravick (Manifest Network)
Can be unitized and put on less expensive hardware. And for example, Render is one of our partners for that. We send those jobs out to Render. And the company, by the way, who enabled that is Jember, who’s on the call today. So they’re writing all our router software. So I think that is a newer evolution of D-PIN.

30:18 — Eric Bravick (Manifest Network)
I don’t think you had anything like that in previous years that actually worked. So now that you have an example of that that actually works, you as a developer that would normally drop into AWS and just get a job done cheaply and efficiently, although some would argue with the cheap word there, but relatively efficiently,

30:47 — Eric Bravick (Manifest Network)
You now have that equivalent ability only leveraging the best of decentralized networks all under one roof, right, essentially. And I think that’s an equivalency to the Web2 world that the Web3 world has not had up to this point. Now it does, and we’re excited to roll this out and expand this over the next year. I think that will be one of

31:17 — Eric Bravick (Manifest Network)
Multiple game changers that has occurred recently. Because you’re correct in your fundamental statement. The road to D-PIN has been exceptionally slow. all of us that have been working in it have been… Heck, I’ve been with at least half of the D-PIN projects that exist in the world today in some way, shape, or form. And yes, it’s been very frustrating for all of us how slow that’s moved.

31:46 — Eric Bravick (Manifest Network)
I think part of that was the tension between decentralization and the physics of the problem in AI computation. So I think we finally crested that hill and come up on the peak. And I think we’re finally ready to make a dent in that. So that would be one answer. I’m sure the panelists have many other answers, but there’s lots of things breaking loose now in Deepin that

32:15 — Eric Bravick (Manifest Network)
That are newly breaking loose that I’m excited to see what happens with. Excellent. Thank you so much for that, Eric. Tim?

32:27 — Tim Cotten (Scrypted)
I mean, I love Eric’s perspective because it’s speaking to a like a fundamental medley of things that are happening all at once. One, we have deepened finally becoming accessible to we have organic need. And if you guys look in the audience, you’re going to see AV beans. Now, for those of you who don’t know me very well, I became not just the founder of a company, but also the godfather of a meme coin quite by accident.

32:57 — Tim Cotten (Scrypted)
Because someone read one of my blog articles back in October and took my thesis on autonomous virtual beings and turn it into a meme coin. And it instantly went viral. And then the guy tried to rug the whole community for like $20,000 or something ridiculous. And so they asked, hey, Mr. Cotton, Tim, Tim, are you a real person? Are you actually building decentralized AI things? you?

33:23 — Tim Cotten (Scrypted)
Working on AI agents. And I was like, yes. And they’re like, will you adopt us? And I said, yes. So now I have a wonderful community of people who want to find utility in decentralized AI applications. And what’s really special about this is that it’s not just people anymore creating spam or meme coins, you have people who are aligning capital in the hope that they can find

33:52 — Tim Cotten (Scrypted)
Some sort of future, whether it’s in the virtual ecosystem, or the Eliza ecosystem, or the AVB ecosystem, wherever it may be, they’re hoping that they can somehow connect their dollars with something useful. And we’re finally seeing that happen. In fact, pretty soon, you’ll see the agent commerce protocol from virtuals come out. And the idea here is, we make, we make decisions liquid, we make, like commerce between agents liquid. And what are those agents going to be doing? They’re going to be providing

34:22 — Tim Cotten (Scrypted)
Valuable services to humans and to other agents and who are they going to use? The very things that everyone here is talking about in the panel, all the D-Pen, all the decision making logic, all the routers, all of that stuff is going to be necessary for a new economic unit, literally virtual labor. Virtual labor has this potential to actually become a decoupling mechanism for crypto.

34:51 — Tim Cotten (Scrypted)
Instead of it being based on fiat, it can actually be based on an economic unit of labor. And this is the first time I think in history we have that opportunity. And this is what is exciting to me. That is a very interesting take. And yeah, I think that is a huge shift compared to previous years. Excellent. If anyone else has something burning on that question that they’d like to input. Sure.

35:21 — Trevor Harries-Jones (Render Network Foundation)
Can I go for it? So I’d love to talk just from a motion graphics artist perspective, because I find that’s always where we’re best at our core. And I think two things have held us back in the past. First, there weren’t open source models that were of the equivalent of some of the closed source models. So it’s been a bit of a waiting game for those to emerge, number one. And then as those have emerged,

35:50 — Trevor Harries-Jones (Render Network Foundation)
The communities have jumped on them. We’ve seen refinements in their abilities in a deep-seed kind of way, really pushing the bounds of what is available and on both centralized and decentralized and leveling the playing fields. without those two and the fact that with that refinement, they’ve managed to reduce the sizes of these models to the point where they actually physically can run on decentralized hardware.

36:19 — Trevor Harries-Jones (Render Network Foundation)
I think these are all sort of massive precursors to us even be able to have the conversation about running AI jobs on these devices. So it’s been a long steady wait, but there are a number of indicators that these are now possible in large decentralized networks. So for me, it’s been blocked just at a base infrastructure level. What I love beyond that though is I do see

36:49 — Trevor Harries-Jones (Render Network Foundation)
Agents only accelerating. They’re obviously much smaller and well attuned towards helping split up tasks in a way that really fits decentralized well when you listen to what the folks at Manifest are doing there in terms of AI being used to help route jobs. These are all relatively new when you think about things from an emotion graphics artist perspective. Excellent. Thanks, Trevor.

37:18 — Edward Katzin (Jember)
It’s funny if I can, I’ll jump in on the comments that Trevor and Tim have just made. Tim, your point about the virtualization of labor is spot on. So if I look at decentralization, early in my career when I was on Wall Street, we used architectures like reactive systems design to ensure availability and reliability. And we would deploy that in at least four or more data centers. So we had decentralized architectures even in the early 2000s.

37:45 — Edward Katzin (Jember)
I think the shift with the rise of decentralized and crypto in distributed ledger, there’s been significant friction and adoption. And people have talked about the lack of interoperability between open source and closed models amongst protocols, across chains. And that’s created a lot of friction, a lot of uncertainty and a lot of doubt, especially amongst Web 2 companies that want to adopt Web 3 and decentralized technologies. And I think what we’ve seen, especially over the course of the last six months,

38:14 — Edward Katzin (Jember)
With the rise of vibe coding is this virtualization of labor that Tim is talking about. So we’ve seen it realized with 10x or 100x productivity increases in the software development and the software engineering, but we haven’t seen the equivalent yet in infrastructure. So what Jember did with Render, in partnership with Render and Manifest, is we deployed workflows that leverage the Render GPUs and Manifest CPU network infrastructure.

38:44 — Edward Katzin (Jember)
But the initial setup of that took a lot of scripting. It took your typical DevOps effort. And what we’ve seen now is over the course of the last couple of months is the ability to bring all the advantages of vibe coding and being able to leverage the AI enabled workflows to manage, deploy, monitor infrastructure. And we’re now starting to use that literally with manifest and render. And that’s been a huge unlock. And that’s now enabling us to really

39:10 — Edward Katzin (Jember)
Use the decentralized compute in a way that honors the requirements of very risk adverse companies. And that’s been an amazing evolution over just the last six months.

39:22 — Sunny Osahn (Render Network Foundation)
Wow. Things are changing at a fast pace. seems like there’s been a lot of unlocks recently. Go on, Teko. My bad, Sonny. I saw Mike’s hand go up too. I know he wanted to touch it. Yeah. my co-founder has a saying, everything happens for a reason. And that reason is incentives.

39:39 — Mikey Anderson (THINK)
And if you really look at D-Pen, all the changing regulations that are happening in both blockchain and AI have created a situation in which it really favors either the big guys or the people that don’t play that by the rules, the pirates. And so really what we’ve been waiting for is technology to get to a spot where actually the rebels identify a clear need and start building off of it. like, honestly, this is one of the reasons why I’m so pro-render at this point in time is

40:07 — Mikey Anderson (THINK)
Render actually went out and found a user group, Hollywood based VFX people. created a product or they have a partner that has a product that sits on all of those desktops and they actually made it happen. Like they’ve been the rebel sitting there, not pumping bags, not doing all the kind of normal stuff that more of the pirates would do. And they’ve actually been providing value on and on. And I think that that’s the pattern that when we look at connecting a user to value that is the thing that’s been missing and

40:37 — Mikey Anderson (THINK)
Partially it’s just because it’s too easy for the pirates to make easy money to actually go and compete against the billion dollar folks. now that we’re seeing like, you know, it’s kind of like when you start to tear cloth or whatever, once you get a little bit going, it’s really easy to go. And I think we’re starting to see that like this whole space is moving like compared to any other technological revolution. This is moving so much faster than the internet, so much faster than mobile. And this group’s at the center of it. And so it’s cool to be part of the good guys.

41:06 — Sunny Osahn (Render Network Foundation)
That’s awesome. It’s, I think when it comes to it, know, people who are doing the building and continually building, they’re going to be recognized. I think that’s definitely going to happen. But when it comes to AI agents, now that’s a term that’s been one of the biggest buzzwords or buzz terms this year. So let’s cut through the hype. What is actually possible today or I can’t see if anyone’s put their hands up. So whoever has, please go for it.

41:57 — Mikey Anderson (THINK)
I mean, I’m happy to jump in here. So AI agents can be thought of as a class of intelligent applications. an agent, you know, like if you have agency, it means you have the ability to do something. You have the freedom to do something. And so already you can see agents being built into larger systems, right? So when you use chat, you’d be to your any of these systems, they’re sending them out to many agents who are breaking up the task and bringing them back to you. Now there’s kind of an information theory problem here is,

42:27 — Mikey Anderson (THINK)
Trying to operate in a closed environment like that, it actually makes it really hard because they have to have a human that’s actually in there making sure that every single one of these agents is doing their job. And it’s the same problem that socialist or communist countries get into with information theory, where you can’t really control millions of people centrally. Where instead, when you build from the ground up and you actually build agents that

42:55 — Mikey Anderson (THINK)
Have actual agency informed by the blockchain, right? So they’re actually protected using math. The software inside of those, like if you put Eliza inside of that or Lang chain inside of that or anything else, and you start getting agent workflows where these agents are creating economic or even intellectual value together, that’s the spot where we’re going to see emergence. And we’re already starting to see this, like when you see what virtuals is doing in Eliza, it’s kind of like the shape of things to come where they’re forming the memetic agent world.

43:25 — Mikey Anderson (THINK)
But then the tools are being built on top of that, right? So like yesterday we have, we have got a builder’s group of about 50 builders that are all building AI agents together. And we just included on the beta, that’s kind of like visual agent builder, where you’re able to connect in all your web two and web three tools into this agent. So you can, you know, connect it to any LLM. You can connect it to, you know, a calendar, a calculator, like any kind of tool. And so, so right now we have the ability for, you know, everyday people to be able to,

43:55 — Mikey Anderson (THINK)
Together their own agent, kind of like building their own lightsaber, right? And we anticipate that as we get, you know, go from 50 users and we open up our beta to all 2022 people that pre-ordered one, that we’re going to start plugging in render, we’re going to start plugging in manifest, we’re going to start plugging in all of these different tools from all the different partners in the independent AI Institute, and now we’re going to have a large group of builders that are actually taking deep in

44:21 — Mikey Anderson (THINK)
And making agents with it. And I think once we see that, think things are going to start moving really fast because when consumers can just tell you what they want, there’s so many people ready to build.

44:33 — Sunny Osahn (Render Network Foundation)
Yeah, that’s a very good take as well. Take it away, Hando. yeah. Well, because I got to follow up on Think. mean, like, I love this so much. I think I’m to say some words, and they’re going to sound really complicated, and I hope to break them down really simply. And this is just a direct reflection on the current state, where we are seeing the nascent skill economy for agents starting.

45:03 — Eric Bravick (Manifest Network)
Where they do need to use things like the render network. And what Mikey’s talking about is he has the lightsaber system, right? Assemble your own agent. And what I’m predicting is that because of the liquid payment rails, especially in crypto, but also through Stripe to hit real world finance now, there’s a bridge. We’re going to start seeing unsupervised self replicating utility functions that are agents.

45:32 — Eric Bravick (Manifest Network)
And what I mean by that is probably a human first, but eventually no need for a human. Someone funds an agent with a utility function that says maximize my gains. And once a 10 % threshold is reached, split the money, change the code and try a new utility function with the two sub agents I just created. And again, that means that the agents can pay for their own storage, their own computation.

46:01 — Eric Bravick (Manifest Network)
Their own compute, I mean, like their own crypto wallets. And at that point, you have a Cambrian explosion that is essentially like just more economic units paying for things through crypto. And it suddenly spikes the prices of things like Ethereum for settlement layers. It suddenly increases the throughput of L2s. Suddenly, we’re seeing a ton of economic activity that’s not maximal extraction of value, but actual creative value.

46:30 — Eric Bravick (Manifest Network)
Like the actual creation of value. And for me, that’s not very far away. We’re talking months, not years. It’s a trivial thing to do today if you really just wanted to do it, even though it might not be very popular and it is limited. For instance, stock picking ones just don’t work very well, right? And I think everyone chasing quantitative agents is competing against the best of the best with very limited information.

46:59 — Eric Bravick (Manifest Network)
But there are opportunities like, like what render provides. For instance, if an agent manages to figure out all the basic viral meme things humans love, like kittens jumping off Olympic swimboards into a pool, there’s no limit. They can just decentralize out the comp, like the actual video rendering, put some cute sound effects on there and throw it on TikToks and boom, you have influencer agents and they can make more of themselves. That’s coming.

47:28 — Eric Bravick (Manifest Network)
Who’s gonna get the value from that? I’m curious.

47:39 — Sunny Osahn (Render Network Foundation)
I like that line of thought. I’m curious to hear what Render Labs has to say. Paul. Yeah, I agree. in that, you know, every successful product service throughout history has delivered a lot of consumer value for the amount of value they capture. you know, and so I agree with the sentiment there that…

47:56 — Paul Roales (RenderLabs)
As we develop really useful products on top of these decentralized platforms, the consumer is going to capture most of the value, as it should be. We should be creating great things for the world. In terms of why is it taken so long? I don’t want to get back to that topic too much, but I think everything happens pretty slowly than all at once. I worked at Waymo for

48:25 — Sunny Osahn (Render Network Foundation)
Six years and the project’s been around 18 years now, right? And so, and then all of sudden everyone sees, you know, it’s overtaken Lyft in San Francisco, right? It has bigger market share than Lyft in a major city. And you see, you know, tons and tons of cars on the road. And so everyone’s like, oh my God, it’s here already. It’s like, happened overnight. Well, it’s an 18, 19 year project, right? And you know, many of us have been working for many, many years on

48:53 — Sunny Osahn (Render Network Foundation)
The decentralized platforms, the things we believe in, the reasons that we work so hard. But now we’re going to see it come to fruition really, really quickly. And we’re seeing lots of things in AI happen really, really quickly after a decade plus of work, a decade plus of development. And so it is an exciting time, but it’s also a time to think about why we’re doing this and making sure we are getting those incentives correct. I’ve loved…

49:22 — Sunny Osahn (Render Network Foundation)
How much conversation we’ve had about incentives and those are the ultimate drivers of, you know, where things will end up. And so, you know, as we’re going through these times where things are growing really quickly, adoption’s really picking up all of a sudden, you know, we do need to think about maintaining those core incentives and why we’re involved in it. And so, no, I mean, that’s why I like this group. I mean, the group that we, you know, have here today, I mean, kind of all gets that and really understands

49:52 — Sunny Osahn (Render Network Foundation)
And keeps that perspective and is not going for that quick bag or something along those lines. But yeah, that’s the view from Render Labs. So just to kind of expand on the question then, what do you think needs serious work before it’s real? What’s a big stretch for now but feels inevitable in the next five years? Yeah, I think the

50:22 — Paul Roales (RenderLabs)
The friction and the, you we were just talking about stable coins on Stripe a second ago, right? Stripe is incredibly good at making things very low friction, right? Adopting their payment platform is sometimes just a few lines of code or a few clicks where, and, you know, I think, you know, one of the things we’re focused on at Render Labs these days is like, okay, so it’s,

50:51 — Paul Roales (RenderLabs)
10 steps to run an AI job on the render compute networks there or something. How do we get that down to three? How do we like make this very, very easy? How do we have the right abstraction layer like that we had with, you know, generating graphics where it’s right inside blender. It’s just, just, you know, a few clicks inside blender and you’re running your, you’re generating your images on the render network. so finding those right abstraction layers, removing a lot of the friction. think a lot of the core underlying

51:21 — Paul Roales (RenderLabs)
Decentralized D-Pen infrastructure is there. It’s just getting it to the user and delivering it in a way that makes it really easy and productive and delivers them a lot of consumer surplus because it is so easy.

51:35 — Edward Kazin (Jember)
Awesome. Thanks, Paul. And if I can, I’ll jump in on that as well, Sunny. And the comments that Paul and Tim made basically come down to the following. I think what’s inevitable in the next five years, as Tim rightly pointed out, is autonomous agents. And we’re starting to see them. I think a slightly different perspective that I have from Tim is I’ve been watching the rise of these agents within corporations and the world of Wall Street, the world of insurance, the world of finance.

52:03 — Edward Kazin (Jember)
And these things have been coming to life in closed systems since the 2010s, especially in the world of high frequency algorithmic trading and the advanced trading bots. And the point that you made about their ability to interact in a broader ecosystem is essential. And in order to have the autonomous agents fully realized, we need to nail what you just described there, which is the interoperability and the abstraction layer, So we need to allow these agents to interact. And I think what is

52:31 — Edward Kazin (Jember)
Part of the ignition of this Cambrian explosion that Tim rightly called out is MCP. And what we’ve seen is just by using MCP over the course of the last four months, we’ve been able to significantly integrate and expand our agent workflows in ways that we couldn’t a year ago. And so if we solve for the alignment of the incentives and the interoperability across these systems, I think the rise of the autonomous agents is inevitable. Yeah, I totally agree. I think we’re saying the same thing.

53:00 — Paul Roales (RenderLabs)
I was phrasing it more in like reducing friction for end users that are humans, but MCP greatly reduces the friction for the agents themselves, right? And so agents don’t want to be doing 500 lines of code to generate a call, right? They want to be generating five lines of call. so, yeah, mean, I think as friction goes down, usage goes up.

53:27 — Paul Roales (RenderLabs)
Most likely, and we have a lot of tailwinds to produce that, but yeah, go ahead, Eric.

53:34 — Eric Bravick (Manifest Network)
Yeah, yeah, I’d echo that for sure. I think I’d roll up the conversation so far as like integration and operations, right? So there’s a lot of narrow model agent solutions right now that do their task just fine, but don’t integrate well with anything else, right?

54:04 — Eric Bravick (Manifest Network)
If I were to summarize all the great things everybody just said, it’s integration and operations, right? Is the thing that’s to me feels inevitable, but isn’t quite there yet. And the reason I think is that we have to remember from an AI perspective fundamentally, we usually go from narrow to general. And so if you think about it in that context,

54:33 — Eric Bravick (Manifest Network)
There’s a lot of agentic AI that does its narrow job passively well to very well, but isn’t really aware of context or the bigger picture that the human is still putting together. That’s what’s going to change over the next five years. That’s what feels inevitable to me. And if you listen to all the context content,

55:01 — Eric Bravick (Manifest Network)
Sorry, comments so far in that context, it’s really integration and operations. In other words, generalization of agent intelligence. Does that kind of make sense?

55:15 — Paul Roales (RenderLabs)
Yeah, I agree, Eric. And I think, you know, right now we’re still hand assembling all the kind of playground, right? So we’re saying, oh, use Stripe. Okay, here’s how you use SMS. Here’s your Twilio endpoint. Here’s

55:30 — Paul Roales (RenderLabs)
Here’s your email endpoint with SendGrid, right? And you can imagine in a few years that something like the render image generation endpoint is advertised and you don’t have to set that up. You don’t have to like go set up your billing account and like deposit some token there and everything else. It’s just like, the agent will realize that, hey, I want to generate an image at this point and this is a great endpoint that I know about and this is how I pay with it.

56:00 — Paul Roales (RenderLabs)
Like it reduces that operations standpoint. that the human in the loop is not having to set the table for everything to happen. The table can be set by the agent themselves as well in an automated way. I think we’ll get there. That’s coming. And as we get there, the agents will adopt a lot of this infrastructure themselves. And yeah, we’ll see some nice curves probably. Yeah, that’s exciting.

56:27 — Eric Bravick (Manifest Network)
Yeah, that’s exactly right. mean, we just dialogued this week in a big roundtable with the Jember team about the future of business operations, since we’re all like old business people, really. And we envision a world very shortly where collections of agents can run the average American business end to end.

56:57 — Eric Bravick (Manifest Network)
With all of the legal compliance operations, finance, accounting, taxes, you know, all the normal stuff, probably end to end within a few years with the general, just with the general input of the user, right? The general input of, want a business that works like this, right? And that’s what’s not there yet, right?

57:26 — Eric Bravick (Manifest Network)
But in another set in another way. Yeah, you’re exactly in the right direction there. I mean, as we generalize the operations disappear deeper and deeper and deeper into the nest of AI and the humans largely, and this could be unfortunate as well, won’t understand how anything actually works or is actually put together.

57:56 — Eric Bravick (Manifest Network)
Right? Yeah. Thanks for Which reduces friction, but may have a downside. That’s a different conversation.

58:00: Sunny Osahn (Render Network Foundation)
Thanks so much for that, Eric. I did notice a few people had their hands up, but I’m very conscious of time. We’ve got about six minutes left. So I’m going to go ahead and ask. I will start with you first, Tim, because you had your hand up for quite a while. It must be aching now.

58:26 — Sunny Osahn (The Render Network Foundation)
For anyone building in AI right now, what’s one breakthrough tool or insight that they may be overlooking that could totally change the game for their productivity or cost? Something they might not even know exists, but once they do, they won’t want to work without it.

58:48 — Tim Cotten (Scrypted)
Oh, this is gonna sound insane then and I’m Yeah, you’re right. I have my hand up and I ache there’s so many things to talk about. this particular one is also very near and dear to me as someone who grew up in the 80s and 90s. There are so many good books that were written before we had deep learning about agentic systems and about the future of what various AI technologies could bring.

59:15 — Tim Cotten (Scrypted)
We always refer to these as the older AI ML systems, and they’re not even truly ML systems. They’re more like just, I shouldn’t say expert systems either. It is a gamut of various technologies that you can use that weren’t really real. They were great demos at the time. Shardloo comes to mind, SHRDLU, for object manipulation in 3D spaces through natural language parsing, written in 1976 or something like that.

59:46 — Tim Cotten (Scrypted)
Um, now we have the cognitive glue thanks to deep learning, but everyone tends to over index on deep learning only by itself and spend tens of millions of dollars training more and more models when you can actually unlock way more superpowers from the models through clever systems engineering. And I think a return to hybridization of systems knowledge from

01:00:12 — Tim Cotten (Scrypted)
Like literally go grab the introduction to artificial intelligence by Philip Jackson Jr. Right. Just go start there. It’s an amazing book. Start there. And suddenly you discover, Oh my gosh, I can solve this completely hard problem now. And neither system by itself can do it. So that hybridization is really important to me. And I do love MCP. I think there are hard boundaries in cryptography and mathematics because of deterministic proofs for like decentralizing all AI.

01:00:41 — Tim Cotten (Scrypted)
If anyone wants to talk with me about that later, I have an entire paper on this and how to solve it. But I’m going to let this part of the conversation end and just say, if anyone wants to see how we’re kind of starting to leverage Render Network and how we’re starting to really get into those image and video generators and making that decentralized, just check to thread. There is an agent called Hollywood in there. And you can just talk to her on Twitter and she’ll make you videos. Go for it. I’ll get right on that, actually, after this call.

01:01:11 — Sunny Osahn (The Render Network Foundation)
But I want to have Trevor answer this question as well because I know he’s got to shoot off very soon.

01:01:21 — Trevor Harries-Jones (The Render Network Foundation)
Yeah, thanks. Thanks, Sunny. Yeah, for me, I stay stuck on this motion graphics flow. I see the path that is currently out there in terms of AI. It’s just not being the ultimate path here that we’re going to end up on. There’s just not enough consistency and control in the way it’s done now. And I’m starting to see some amazing systems. Tim and team are really

01:01:50 — Trevor Harries-Jones (The Render Network Foundation)
Pushing the boundaries of what can be done with a horde of agents to get over some of those hurdles, but I still see them persisting. And I think we’re going to end up in a position where there is that same hybrid that he’s referencing in motion graphics creation involving both traditional and sort of the new generated workflows. And I honestly believe that will be achieved in the next five years.

01:02:18 — Trevor Harries-Jones (The Render Network Foundation)
And the achievement of that will lead to real-time photorealistic landscapes and interfaces that today are just not technically possible. you know, as we’re Star Trek folks at heart, we’re still marching towards that holodeck type experience. you know, when I think out just what’s unblocked by a lot of these new elements, I start to get really excited that in that five-year window,

01:02:47 — Trevor Harries-Jones (The Render Network Foundation)
Holographic real-time type experiences are going to be much more reality than ever before. That’s exciting, Trevor. I cannot wait for the Holodeck. I think that’s one thing that keeps me very motivated and excited. I think what we’ll do now, TACO, let’s have some kind of lightning round. Maybe we can go on another 10, 15 minutes. Definitely. Before we hop into that, one of the things that like,

01:03:17 — Taco (Manifest Network)
I have so much to tell you about basically a holodeck that was presented here at Super AI. But you touched on my Star Trek roots. And one of the really cool things that happened at RenderCon in April was the release of, I don’t want to call it fan fiction because you literally had Captain Kirk, William Shatner having a hand in this.

01:03:47 — Taco (Manifest Network)
But you guys, Render made a Star Trek with Leonard Nimoy’s son, a goodbye piece. How has that done?

1:03:57 — Trevor Harries-Jones (The Render Network Foundation)

Thanks, Taco. Well, I mean, it’s been amazingly received. I think the last time I checked in, was well over 25 million views, putting out the most viewed content that isn’t sort of one of the original movies out there in this franchise. it’s been phenomenal. Just I wouldn’t say.

01:04:15 — Trevor Harries-Jones (The Render Network Foundation)
Render created as much as the OTOY team, but it was rendered on the Render Network. It’s been amazing to just see the fans get hold of this. I hope we can create more similar content over the coming years because it really shows what can be done with all these amazing tools in a way that maintains that Hollywood visual element and creates amazing content.

01:04:45 — Taco (Manifest Network)
Awesome. Yeah, no, I’m not gonna say that 1 million views was me, but there was a lot from me as well. As we hop into this lightning round, we’ve heard a lot of deep thoughts on everything that people are building. I wanna roll into this, know, the top of the mind stuff because a lot of people have questions.

01:05:12 — Taco (Manifest Network)
And we have actually some questions from the audience that want to feed in as well. But this first question is to Mikey. Given your perspective of on decentralized compute market, what opportunities and challenges do you foresee for the future?

01:05:29 — Mikey Anderson (THINK)

Oh gosh. So last Thursday I got invited to speak at this big event here in Seattle where a lot of the biggest AI companies and a lot of the biggest institutions had kind of like their forward deployed engineers, their change management people, and then the managers responsible for AI. And for some reason they invited me there, which kind of turned the whole thing on its head because, you know, as they were talking through, were talking about their challenges, they were talking about like the internal friction within companies.

01:05:58 — Mikey Anderson (THINK)
And one of the things that kind of everybody like saw for the first time was that a corporation actually already is an information processing system. Like this by design, the org chart, the way that information flows, the way money flows, it’s a control system built for a specific reason. And the smart people in the room had kind of like this matrix moment where they realized, oh, I’m the battery in this. And when I’m implementing AI, I’m actually replacing myself. And so I think it’s actually the stories.

01:06:28 — Mikey Anderson (THINK)
And the social infrastructure for people to be able to go into the unknown and realize that right now, and like people on a farm or people, you know, kind of at pre-industrial realize the steam engines replacing them or the train is replacing them. We are in a place where people have to understand that those corporate agents that are there and kind of all the culture that they’ve developed over time, the self-protection mechanisms, the immune system of the corporation,

01:06:57 — Mikey Anderson (THINK)
These are the things that are the biggest friction right now, because there’s tools in place. The problem is just how do you actually get people to embrace this and not just be replaced? When people understand that this is their lightsaber, this is their next evolution of themselves, once they really come to grips with that and go through the grieving process of I have to change, that’s when the positivity happens.

01:07:23 — Mikey Anderson (THINK)
And I’m excited because I’m seeing a community and a movement here. Like, how many of us are cooperating towards a shared outcome? Like, that’s what ends up winning. And just like Paul said, everything moves slow until it moves all at once.

01:07:43 — Taco (Manifest Network)
100%. Tim, I know we got to touch on this when we were having dinner last year in Bangkok. But and then we also talked about earlier this week, we see a lot of the big thing on all the headlines is a need for compute. Yet we see hundreds of protocols popping up, stating that they are decentralized compute in some way, or form there, you know.

01:08:10 — Taco (Manifest Network)
Cloud network pieces. How do you think the imbalance or balance of decentralized compute resources is affecting innovation and adoption in the space? My friend, like I had to solve this problem for ourselves. So this is like, this is top of mind for me. So just like quick background, the product we’re building right now to prove out like our thesis, which is like,

01:08:36 — Tim Cotten (Scrypted)
Like what No Code did for programmers, we’re doing No Prompt for everybody else in AI. There’s way too many models. There’s way too many different things that each of them do differently and do really well and different ways of attaining their compute, whether through direct APIs from like Vertex API from Google for VO3 or for hosting some of the models on FAL or actually doing what Render is doing with the amazing endpoints that they have.

01:09:05 — Tim Cotten (Scrypted)
There’s so many different things that not necessarily one network can handle them all in a perfect way. They all just do some things differently. So our challenge was how do we leverage decentralization the most? Because it often turns out to be cheaper. It actually turns out to be the one use case for decentralization I have found that can be cheaper than the centralized version. And because it’s a great way of coordinating the liquid compute. And it turns out…

01:09:35 — Tim Cotten (Scrypted)
What I think Martian did really well for LLMs, we had to invent the same thing for our system under the hood, which is generative AI, images and video routing based on intent, based on the composition of the scene requirements. If someone says a video where they need two characters to speak and then smash cut into one of them, throwing the other over the shoulder, those are two different renderers to us.

01:10:04 — Tim Cotten (Scrypted)
And we do have a good bridging system for ourselves. For instance, you might call Higgs field for the special effect. You might call renders image node to create the initial scene. You might call Hydra for the actual conversation. All of those things have a cost. They have to be decomposed. And then they have to be reassembled. And being able to hand that work out agentically through a decentralized network, and more importantly, being able to route

01:10:31 — Tim Cotten (Scrypted)
To the right things based on the right decision-making frameworks, that has been our core challenge at Scripted. And I think we’ve well solved it and we’re getting ready to show the world what that looks like. And we’re very excited about it, but you can get a preview right now, like I said, from the Hollywood agent, which we power their tech as well. But the short answer from me is that intelligently routing demand is like the thing to do.

01:11:00 — — Tim Cotten (Scrypted)
Just intelligently routing the demand to all of you who are offering services.

01:11:07 — Edward Kazin (Jember)

Nice. This next one is to Ed. Because I see you as such a solver in things. In an ideal world, how would you like to decentralized compute markets evolve? And what steps can be taken to achieve this vision?

01:11:29 — Edward Kazin (Jember)
Yeah, great question. And it’s been a huge challenge. What I’ve seen, especially from the corporate world or call it traditional centralized web too, is the recognition that even though you may have your infrastructure on AWS, GCP or Azure, you still have a critical failure point. Like when AWS goes down, we watch significant portions of the internet go offline and major corporations are impacted. So, any project, any company that cares

01:11:59 — Edward Kazin (Jember)
About availability, reliability, resiliency, having a truly performant transaction infrastructure, can’t rely on a single cloud services provider. So the need to be able to distribute the compute resources, the networking resources, all the other essential elements of the infrastructure is essential. But the coordination, the interoperability, the management, the monitoring of that is a huge challenge.

01:12:24 — Edward Kazin (Jember)
So I think there three layers that we’re addressing on that. One is a no code approach to infrastructure. So what vibe coding is done for software engineering. Bringing that to the world, as Eric alluded to, it’s about interoperability and operations. So how do we really enable the AI workflows to automate that? And then as Tim rightly pointed out, the agents have to be autonomous and the agents have to be able to deliver their own utility and they have to be able to provide for their own economics.

01:12:52 — Edward Kazin (Jember)
So they have to be able to interact with the crypto systems and the fiat systems to work seamlessly across them. And I think what we’re seeing come together now is the second layer, which is the interoperability around payments, both in crypto and as was mentioned, Stripes on ramps. And then I think the third piece now is the interoperability of the protocols. So that as Eric mentioned, now when the agents are operating across decentralized systems, they actually have the context.

01:13:20 — Edward Kazin (Jember)
And they have the economic means to cross those networks that compute that memory space. And we’re seeing that really come together and where we’ve implemented it for real is in what we’ve built out with render and manifest. So if you look at how we’re able to distribute our AI workloads onto render GPU and then leverage the manifest infrastructure for our sovereign computing, so where we have to ensure absolute security, absolute privacy, absolute regulatory compliance.

01:13:48 — Edward Kazin (Jember)
We’re able to prove that and by leveraging distributed ledger technology, it’s immutable, it’s transparent, and it meets all the requirements that traditional Web2 have, plus delivering the performance and the benefits of decentralized and bringing the capabilities of AI. And this isn’t some far-flung future. Like we’ve literally been watching this evolve over the last 18 months. And the willingness of traditional corporations to step into this space.

01:14:16 — Edward Kazin (Jember)
Has increased exponentially. I think that moment that Tim talked about, when people have this personal realization and they’re stepping into the unknown, now they’re more willing to do so than they ever have before because they realize it’s an AI or die moment. And if they don’t adopt and leverage and use these technologies, their competitors will.

01:14:33 — Taco (Manifest Network)
All right, yeah. And that is that it’s the shipper die piece that we’re all in at all times. feels like Paul. This falls really heavy into what render has been able to do of taking. And I think Trevor said it once when when it asked him at RenderCon what the what the cost was. And the simplest answer for when it comes to render is just electricity because,

01:15:02 — Taco (Manifest Network)
Moved from centralized GPUs to individuals hardware that was doing this. How do you think decentralized compute market compares to traditional centralized compute markets in terms of supply and demand? Yeah, I mean, there’s certainly a huge volume of supply that an individual may have bought.

01:15:27 — Paul Roales (RenderLabs)
A very nice GPU for gaming, for their personal rendering work at home. And then, like many cars or many other assets, it sits unused 23.5 hours of the day or something. so the marginal cost to them is the electricity plus the very small delta, right? And so, yeah, I think we haven’t touched on it here, but one of the big advantages for me

01:15:57 — Paul Roales (RenderLabs)
Of this decentralized approach is that it’s incredibly green to be reusing computing cores that have already been manufactured, that are already out there, that are already installed, and utilize them to their maximum capability. And so, like you said, this converges to the price of electricity, and even less than that because there’s this very interesting concept of megawatts, right?

01:16:26 — Paul Roales (RenderLabs)
When electricity prices actually go negative, when you get paid to consume electricity. And this happens all throughout Texas all the time because of the amount of solar and wind being produced there. And so, you know, there’s actually instances throughout the day, every day globally where electricity, they’ll pay you to consume it, right? And you have a GPU that’s being unused on your desktop. This becomes very, very exciting, right? We talk about information is too cheap to meter.

01:16:55 — Paul Roales (RenderLabs)
They’re about information where they literally pay you to consume and generate knowledge. This is something that only decentralized can do.

01:17:04 — Taco (Manifest Network)
Nice. Yeah. And it’s really interesting how it sort of puts it. People that thought that we’re just using it for a half hour a day are now able to get more out of it as well. Eric, this is this is to you because you touched on pricing, hidden pricing, I should say a little bit earlier. What role do you believe market dynamics such as pricing and competition
play in shaping the supply and demand of decentralized or sovereign compute resources.

01:17:41 — Eric Bravick (Manifest Network)
Yeah, of course an interesting and core question. This really is a sub question of incentives. pricing dynamics will be important as this develops. They’re important now, but if you really, really want to unlock the

01:18:09 — Eric Bravick (Manifest Network)
The market dynamics of price, you need really, really slick behind the scenes arbitrage, we haven’t, that’s something that’s under development. It’s not quite there yet. There’s some good examples of it, but I think we’re very early in those dynamics. Where you get to

01:18:37 — Eric Bravick (Manifest Network)
Agentically as everything that we’ve talked about today gets rolled out, what you build is kind of a, you know, you’re building the base of the pyramid, right, and you’re building higher and higher and higher. So you’re building Chapel on the Hill here. Where we get to pretty quickly once all the

01:19:01
Once all the basic elements are in place and they’re controlled through AI is you get you get a lot of routing of compute workloads based on price, right? So you’re going to get less of that now because the physics of the problem kind of

01:19:26 — Eric Bravick (Manifest Network)
Restrict that a little bit and you can’t add complication for the user. That’s the one thing we know, right? I think we’ve all touched on that a little bit here is if you make things hard and complicated, people won’t do it even to save money. But as you see AI take hold in infrastructure, in DevOps, in job routing, price will be a core tenet of that.

01:19:55 — Eric Bravick (Manifest Network)
As AIs get smarter and smarter about routing and cost saving. the question of electricity actually is a really good one here to springboard off of. At hyperscale, how electricity works is really the primary limiting factor for rolling out.

01:20:21 — Eric Bravick (Manifest Network)
New data centers and new AI capabilities right now. electricity has become the largest limiting factor. And so dynamic routing of jobs to where the power is, which by the way, interestingly, we mentioned solar and wind here as well, might be where the sun is shining or where the wind is blowing at the time. So you’ll see compute

01:20:50 — Eric Bravick (Manifest Network)
For example, migrating with the sun across the globe, right? All of that will become more of a reality. All of that is actually based on price. whenever you, or price is how we measure all of that, if that makes sense, right? Price is the metric by which we determine what efficiency is. So I think that’s,

01:21:18 — Eric Bravick (Manifest Network)
A really interesting thing to watch over the next few years as we focus more AI on operations and more AI on efficiency of operations at scale. What you’ll find is you’ll see all of that sort of stuff. All of that, the metric is price and the market dynamics of price, right? So it’s probably another way to think about that.

01:21:48 — Eric Bravick (Manifest Network)
Does that, did that make sense?

01:21:53 — Taco (Manifest Network)
It does to me because but Sunny, any any missing points on that that you think Eric missed? You know what Eric? I think you hit the nail right on the head. It was exactly what I was going to Perfect. And last over to Trevor. I’m actually going to not ask you the question I had because it sort of feeds into the question that one of the listeners put.

01:22:22 — Taco (Manifest Network)
And so I’ll just sort of switch that up. Within businesses that use large compute, large amounts of compute, how do you convince decision makers of say, instead of Netflix, Hollywood to go from AWS or their own data centers to a decentralized solution?

01:22:48 — Trevor Harries-Jones (The Render Network Foundation)
Yeah, great question. You know, we’ve tried on a couple of them and in many cases, you know, it’s an impossible sale. You’re selling up against a company with global scale stability, SLAs and more that just deliver a product that is hard to match from a decentralized basis. And it’s almost to me, one of the

01:23:18 — Trevor Harries-Jones (The Render Network Foundation)
One of the misunderstandings that folks seem to have here about where compute will come from is they believe the Mid Journeys of the world or the open AI should just flick a switch and use render. And what they’re missing on this is it’s a much more nuanced sale. There are elements of their workflows that are well suited towards decentralized, but it’s not a head to head. It’s sort of picking off at the edges.

01:23:48 — Trevor Harries-Jones (The Render Network Foundation)
And it’s supplemental in many ways to what they’re doing. You know, as you see orchestration and platforms like MCP and others unlock AI automation of routing, it becomes a little bit more of a reality that it could be done, but still one that needs a way to go before it’s proven out. And honestly, I think it will be proven out first on Web3 projects and projects like our partners here.

01:24:18 — Trevor Harries-Jones (The Render Network Foundation)
As that scale is achieved and as it shows actual success and usage, I think you’re on a better footing for selling against large, large players. But, you know, it’s really a question of proving out these technologies that are very nascent before being able to sell head to head against a behemoth. So there’s a path, but we’re at the very, very onset of it now. Yeah.

01:24:46 — Taco (Manifest Network)
It’s the competition of incentive models, think as well, that will eventually get sovereign and decentralized compute there. As we start to close out, want to, one, thank everyone for joining us today, but I want to do the final round robin of calls to actions that anyone might have. And this has been an amazing panel. Thank you everyone for joining us today. Sunny, this has been an honor to host this with you.

01:25:16 — Taco (Manifest Network)
I feel like you’re a voice actor and we might touch on that at a different time or sometime, but any calls to actions from anyone.

01:25:26 — Mikey Anderson (THINK)
If you’re a builder, I can go first. would love to have you go ahead. Oh yeah. If you’re a builder, we would love to have you join the Think community. Cause right now we are, you know, we’re on kind of a 30 to 40 day sprint to protocol launch. And right now all the builders are starting to get first access to some of these authoring tools. And we’re kind of in that test mode. So if you want to be part of something from the very beginning, and if you have a vision and you want a community to support you, come join the Think agent community. Yeah. Tim.

01:25:57 — Tim Cotten (Scrypted)
Mine’s very easy. In the threads for each of the posts for this particular space, and since we’re on X, go find Hollywood agent, please follow her, and then give her some prompts. Talk to her. Create some content. the more you do, the more we get to push towards networks like Render, the more we can use things like Manifest, the more we can find reasons to integrate with Think. There’s lots of reasons to do this. So push us forward.

01:26:25 — Tim Cotten (Scrypted)
And then stay tuned with me. Please follow me because we’ll be releasing our newest no prompt solution for generative AI video and images very shortly and I want some testers. Thank you. Awesome Ed. Yeah, this has been an amazing panel. Thank you taco. Thank you Sonny and thank you for everyone participating. Yeah, so with Jember AI we’re in the midst of building out our infrastructure on manifest and render.

01:26:53 — Edward Kazin (Jember)
And in doing that, we’re provisioning the tools that are automating our DevOps infrastructure provisioning and management, because we need them for ourselves. Manifest and Render Community need them. So we’re building that out. And so anyone that’s excited about enabling AI orchestration, choreographing advanced autonomous AI agent workflows across Render and Manifest, definitely check out Jumper.ai and reach out to us. And we’d love to work with you. Awesome. Thank you.

01:27:23 — Paul Roales (RenderLabs)
Yeah, know it’s a great conversation today, guys. Yeah, let’s keep the conversation going. So my invitation would be that if you’re interested in these topics, if you want to talk more about the topics and how we develop together, because none of us here are going to develop a complete solution that’s going to solve everything that we’ve envisioned and talked about today. And so, yeah, happy to keep the conversation going. And my DMs are open and happy to chat with you.

01:27:53 — Taco (Manifest Network)
Eric any last calls to action?

01:27:57 — Eric Bravick (Manifest Network)

Yeah, so we’ve got a ton of stuff going on at the manifest network. And a lot of those projects are very exciting. I think the probably the biggest thing to call out right now is we’re headed for our first public listing of the token. We’ve got a large private.

01:28:19 — Eric Bravick (Manifest Network)
Token economy right now, but we’re going public with that on July 15th. So we’ll be out on our first exchange by then. So would love community support to come out and support the token listing, of course. Beyond that, we’ve got a ton of great networks that we’re integrating with. We’d love to have some more integrations going on there. So if you’re

01:28:49 — Eric Bravick (Manifest Network)
If you’re running an AI project or you’re an L1 and you want to help or you want help running AI workloads as in the ways that we’ve described here, we’d love to talk to you. We’ve got a really strong integrations program going right now. So that’s going really well. And yeah, there’s that was so much more I could talk about, but I’ll hold it there. Awesome. Trevor.

01:29:19 — Taco (Manifest Network)
Did Paul cover it or any other?

01:29:23 — Trevor Harries-Jones (The Render Network Foundation)

No, I got two. One, think you guys can tell we’ve got some amazing launch partners here. But if you know any other projects that have a use case that would fit with offline decentralized compute, we’re always open to talking. But I think the bigger part guys is, we’re in the midst of sourcing and rolling out the decentralized network here. In all my tenure at the project,

01:29:48 — Trevor Harries-Jones (The Render Network Foundation)
We’ve had a closed wait list. So folks, this is your opportunity. For those of you who qualify, obviously it’s a very limited initial first project, but we’re looking for high-performance nodes that are on Linux, that are available and can be utilized for long periods of time. And I know that’s a small subset of our wait list, but really want to just call out here that we’re actively sourcing those we’d love.

01:30:18 — Trevor Harries-Jones (The Render Network Foundation)
You guys to be involved and we’d love to get through this initial test and roll this network out at scale to the larger wait list. Awesome. Yeah, it’s awesome to see how Deepin grows one person at a time that way. Sunny, I like hosting with you, man. This has been an amazing time. This has been pretty good. There’s a couple of things I want to say as we close out. One is,

01:30:46 — Sunny Osahn (The Render Network Foundation)
Tim, used the Hollywood agent while we were on this call. I asked for a Lambo floating in space and it produced it. I love it. It’s pretty cool. I will be using this again for damn sure. The last thing I wanna say is I mentioned that I look after our grants program. Essentially, if you are someone in the space who is looking for GPU power for serious rendering of…

01:31:14 — Sunny Osahn (The Render Network Foundation)
Any of your 3D animations, any work that you may be doing in Hollywood, et cetera, definitely reach out. But also if you want to play around with some AI generative tools that we also have on the Render Network, you can also use some of the grant for that as well. So you can all DM me, I’m more than happy for that. Or you can reach out at grants at renderfoundation.com. I have to get flowers where flowers are due.

01:31:42 — Sunny Osahn (The Render Network Foundation)
I know there’s a community member that created a bot that tracked a lot of what happened on the Render Network being both on the burn side and the use case side. And it sort of got to a burden part for him. And he was like, I want to hand this over to the community who can help. seeing the grants program and render just, this is, you’ve done an amazing job. We would love to support this. So the grants go so much farther than just that. I want to give you flowers for that, Sonny. Oh yeah. You know what?

01:32:11 — Sunny Osahn (The Render Network Foundation)
Jay and there’s many other community members that I can, that I’d be able to name. They’re doing a fantastic job. We have one of the best communities here in Web3. And it’s amazing seeing some of these initiatives that they’re just doing off their own back. know, costs got a little bit high and obviously we can take care of that. That’s no issue whatsoever. It’s just so cool to see people developing and doing things and, you know, being proactive.

01:32:39 — Sunny Osahn (The Render Network Foundation)
Right. That’s a key thing that I love to see. So yeah, this has been an amazing space. We may just do this as a bit more of a frequent, regular thing. What do you say, Taco?

01:33:09 — Taco (Manifest Network)
I’m down for that. I think, yeah, I think I’m down for that. Excellent. Excellent.

01:33:03 — Sunny Osahn (The Render Network Foundation)

I think we will be in each other’s DMs. Let’s figure this out. Definitely.

01:33:08 — Taco (Manifest Network)
My call to action is follow everyone on this panel. Every speaker follow them. All of their projects they usually have in their bios somewhere. See what they’re doing and do not be afraid to ask anyone in a DM for help and learning on the next step. Everyone has their DMs open, so this is an amazing time of helping others help others.