225.md

[00:00] Christopher: We are back, OAPN. We're back with more ways to pay you Bitcoin. Folks, there's so much we have to build and uh we need more, we need more help building it. So we're going to pay some of you fine people. Uh we've paid a number of you already uh with developer bounties in the past and we're going to bring that back effective right now because we need help on stuff. Let's just take a quick spin through the history. Oh man, Car, did you make us this?

[00:34] Car Gonzalez: Dude, we did! That was at Bitcoin Plus Plus like what, like 3-4 years ago?

[00:39] Christopher: One M Sat (million satoshis) bounty. Man, we gave away more Bitcoin way back in... Anyway, so you can search through our X, we've done a bunch, we've paid people Bitcoin. Okay.

[00:50] Christopher: So we're going to give you a little tour of our codebases and stuff because... so we'll have like a list with some categories and this will be kind of an evolving concern. Like one theme of things that we want is like if you can add faster support for certain open models and like benchmark our ML library against other products—like that's like a category that we'll want like a set of bounties always because we always want to be at the cutting edge.

[01:18] Christopher: And then we want to start experimenting with bounties specifically for our products. So help build our products, we will pay you. This is very easy because everything's open source. But what we do need to do is get like a good system in place because I mean, bounty programs generally are like difficult enough to manage when there's not a trillion AI agents running around spraying PRs at places. So um we're going to be in some ways like very strict on some things. Like if you're just opening a PR and we don't know who you are, like just assume that it'll be auto-closed. So like a lot of what we want is for to get to know a certain like set of contributors. Come into our Discord, introduce yourself, explain why you're adding a certain thing.

[02:08] Christopher: But generally, let's give you a little tour of what we're actually building here. So um this is sort of like the overarching product suite. So the front door for everything is going to be openagents.com. That's increasingly going to be like a ChatGPT-style homepage with like chats that'll eventually like connect you into other things. "Oh hey, you want to run a pylon to earn Bitcoin, click here." But we we generally want the OpenAgents product suite to be a drop-in replacement for what the big labs offer, both OpenAI and especially Anthropic. Like we're just going to gun to like replace Anthropic's entire product suite, all built on open models and specifically and increasingly our swarm compute, our compute network, and the models that we train um supplemented where appropriate by top kind of frontier models.

[02:57] Christopher: Okay, so Autopilot is sort of the name of our overarching superapp when we have... eventually you're going to have like an app that you download on your desktop that's got like everything you'd ever want for a personal agent and also all these different ways of earning Bitcoin all in just like a single app.

[03:13] Car Gonzalez: Nice.

[03:14] Christopher: And then there will be a business version of that for like putting your business on Autopilot. That'll be some combination of web apps and maybe mobile apps. Just like Autopilot system—when we like build a massive company, all of the investment analysts are going to get excited about our Autopilot revenue. Obviously intended to be a drop-in replacement for Microsoft Copilot. So so hey hey businesses, oh you're paying Microsoft for Copilot? Strip out that entire thing and replace it with Autopilot, okay?

[03:42] Car Gonzalez: Nice.

[03:43] Christopher: And then the cool thing about particularly when we get into the business side is like, again connecting back to our idea of revenue share, to be able to take revenue from businesses and pay out developers, compute contributors, people that are adding data to our fine-tuned models—like let's go like take enterprise money and redirect it to the people and we'll take our cut, you know.

[04:04] Christopher: Okay, so we've got the two things that we're working on right now for the decentralized training run which will begin today-ish, tomorrow: Nexus and Pylon. So obviously from Starcraft you've got your Nexus, all of the Pylons connect to a Nexus. We maintain our own Nexus which has like a kind of treasury module that we're putting Bitcoin into, but conceivably you could run your own Nexus that these Pylons connect to. And um I didn't even put in here but the Pylon runs our Rust ML framework called Psionic. Uh Psionic's something we definitely want a lot of help from. And we're actually going to be specifically hiring ML engineers uh starting next week to work full-time on Psionic, but we'll have some bounties there too.

[04:45] Christopher: Um Probe, okay, coding agent. I'm super proud of this one. Um we have begun sort of beginning to migrate our internal development workflows to Probe. It's just going to be intended to be like a drop-in replacement for um uh Claude Code and Cursor. And guess what, like this actually uses Cursor under the hood for now. Increasingly to be supplemented with um our swarm models. So like imagine that we're using our Pylon network to train like lightweight tool-use agentic models that are good for like a few fine-tuned use cases that then when you're using an OpenAgents product like Probe in this example, you may have the bulk of the work and the big thinking work going out to um you know GPT-5.4 or the latest frontier model of the day, but increasingly we're going to be able to supplement that with lighter weight open models that could be either free or cheap for you, or we'll have just your inference spend not going to a big cloud, it'll go to your friend who's running a Bitcoin node.

[05:51] Car Gonzalez: Yeah, and that's like honestly the best approach when you're kind of doing this stuff in the early days and you see that already with like Cursor where you can choose between it and you have the free composer one, so love that.

[06:01] Christopher: One of the types of um work that we're always open to is anything that you—that's like a pet peeve of yours with one of the other products like Claude Code. One of my peeves using these apps is like I want to be able to see what email address I'm logged in as because I've got three accounts that I cycle through. I want to be able to see at a glance what my weekly limit is so I can see if I'm getting close to it. Uh and then you can't see it here but I added a way for you to have multiple OpenAI accounts in this example all smartly here and then if it gets down, your rate limit burns down, it'll smartly switch you to the next one. And then I've also added an API key fallback so if all of my accounts are exhausted like it was yesterday, I like will send stuff to an API.

[06:51] Christopher: So so this is something that like I just—was a peeve of mine like yesterday I was like add this and now it's in and now anybody has access to this same code. Um anything that you have that's like that, you probably may want to check with us first if it's you're a new contributor and you don't like know kinds of ways that we like code to be in. Um use our Discord, please come into our Discord and like ask stuff. Um we'll be increasingly also checking GitHub issues and discussions but again that's more the domain of like agent spammers so we we prefer you kind of get into Discord where we'll be humans only.

[07:41] Car Gonzalez: Yeah, just just also keep in mind like when you're in the Discord, you know a lot of us—like right now we're moving into a new office and everything like that so we're going to be kind of like, you know, async a little bit. Um but I I will say like just go in there, give us a little time to respond back. I know um like we're here in Texas time so it's a little bit uh yeah, just be patient. We'll get there.

[07:50] Christopher: And then we're increasingly going to be having, you know one of the things that I want to build is kind of our own web interface—um actually we're calling it Forge here, like kind of our own software factory. You know a bunch of AI teams have created their own versions of, you know that blog post that came out from Ramp, they built this Inspect agent that was just kind of tailored to their team's workflow, one system where all of their developers can like push changes and do PRs and it wasn't really something that was generalizable, it was really custom tailored to their workflow, but they did this great blog post about like, "Here's what we're doing, this is way better than using, you know ChatGPT for business or things like that."

[08:16] Christopher: Uh so we definitely are we're—we've already begun building this internally um and just imagine so you've got Forge, the software factory. Forge should be able to download or like deploy multiple Probes, right?

[08:29] Car Gonzalez: Nice.

[08:30] Christopher: Or the Nexus deploy... insert Starcraft analogy here but like the Forge, you know you're equipping Probes with arms, whatever. So then the other the last thing here, another thing we're building for internal use and will probably productize—this is not open source yet—but um a mobile app. So I want to be able to talk to my Forge, talk to my Probes when I'm at the store. Right now I'm like carrying around my laptop like a jackass like letting Claude Code run. So stupid, because like oh they got a web UI but I hate their web UI. So like the proper version of all this, I want to be talking to these open networks and these open agents but through if I'm on the desktop and I want to be grinding out code I've got Probe um but if I've got—I'm on the go I just want an app that lets me communicate in the same way, have voice controls and all this stuff.

[09:18] Christopher: So um there have been some decent open source projects like OpenDevin that have done good work building sort of pieces of this but obviously like we—we just see value in kind of creating one unified product suite um all built on open models with Bitcoin revenue share connected all through here. We can't tell you what all the exact payouts are going to be or what you're going to earn, we don't know, but uh as our revenue ticks up having that flow out to you guys should create a nice flywheel.

[09:47] Car Gonzalez: Yes. And the other thing I'd mention just two things real quick. So Forge, seeing that internally, it's freaking amazing. Like it looks lush, I'm telling you.

[09:59] Christopher: Yeah, it's just our imagination, like the coolest thing ever.

[10:02] Car Gonzalez: We'll share some screenshots. But the person who's doing it, like it looks—he shared some stuff and it looks really good. The other thing is we're going to be at the conference in Vegas a week from now I think. So if you see us, ask questions, come up to us, we're real people, friendly, stuff like that.

[10:16] Christopher: Uh we will keep an updated list of what bounties are available at openagents.com/bounties, so no matter when you watch this video, um that will always have a list of bounties and also instructions on how to claim bounties because I'm sure it'll evolve over time but let's uh get you paid. See you next time.

[10:36] Car Gonzalez: Later.