You NEED to try these open-source AI projects right now...

I'm going to show you four incredible open source projects that show you where the world is headed with agents. These four projects show the future of how humans are going to interact and use and leverage artificial intelligence. The first one is called G-Stack and it is by Y Combinator President Gary Tan and I just started using it. Let me show you a few things. So just a couple weeks old, it is already nearly at 50,000 stars on GitHub and if you don't know who Gary Tan is, he is the president of Y Combinator, the premier startup accelerator in Silicon Valley. Gary has invested and overseen thousands of startups including some of the biggest ones you've definitely heard of like Airbnb and DoorDash. So what he did was he boiled all of his experience, all of his expertise and then combined that with his desire to give solo developers the power of entire development teams and put it in a single package that can be used on top of cloud code, codex or whatever other coding agent you're using. This is G-Stack. So he quotes Karpathy at the beginning of the documentation saying, I don't think I've typed a line of code probably since December basically which is an extremely large change. He says, when I heard Karpathy say this, I wanted to find out how, how does one person ship like a team of 20 and that is what he's attempting to do here. So installing it is very easy. Down here, he gives you this little prompt that you simply copy paste into cloud code. So here's cloud code for me and as you can see, I simply copy pasted it. It asked me a few questions, asked me for some permissions and then it was done. And then in a new session, you can type things like slash G-Stack and then you have a bunch of sub tools that you can use with it like slash office-hours and he goes over what each of these do in the documentation. So he describes G-Stack as a process, not a collection of individual tools. The purpose is to go through this process to fine tune your idea, fine tune your market opportunity and then of course build it out. And again, all of this is based on Gary's incredible amount of experience and expertise building startups. So for example, he has office-hours, one of the most well-known parts of the Y Combinator program. You go meet with the partners of YC, you ask them questions, they give you advice. But for slash office-hours, there are six forcing questions that reframe your product before you write any code. Again, trying to hone in on that idea, hone in on the right path forward before writing any of that code. Here we have planned CEO review, which rethink the problem, find the 10-star product, which is a reference to Airbnb's Brian Chesky who said, five stars, not enough. How do we actually give a 10-star experience as part of the Airbnb experience? So that's kind of what he's trying to recreate here. We have an engineering manager, a senior designer, design partner, staff engineer, debugger. So basically what he has done is broken out all of these roles into individual prompts or individual markdown files that help Cloud Code understand how to look at your code base. No longer are you using a single agent for all of these different tasks. Instead, the agent is given much more specific direction in how to look at the code base through the lens of each of these roles. And this is prompt engineering at its finest. And the cool thing is it's all open source. So you can actually go in and you're essentially able to see how Gary Tan thinks about startups. And it's really cool that he open sourced all of it. So go in there, read all the documentation. So this is literally how he thinks about office hours for YC. Lots of cool features being updated seemingly every day. And it's weird because Gary also runs Y Combinator full time, does this part time. He's starting to bring on contributors. So if you have some ideas, let them know. Go check this out. I'll drop the link in the description below. And quick disclosure, this segment is sponsored by Microsoft. And by the way, let me just pause for a second and say nobody actually believes me that this is a full time job or that this channel is a very successful small business. And not only that, that I can hire people and produce hundreds of videos per year. Literally, I get asked all the time, OK, but what's your real job? The job of running a small business requires many of the typical tasks, digging through docs, spreadsheets and notes, tasks that I used to do manually, but now can be sped up significantly by using AI. I went over this in detail recently in a short I did with Microsoft. I talk about my journey of becoming a content creator and also how I use Microsoft 365 Copilot Business to help with a lot of the automation so my small team can punch above its weight. And so if you want to know how I produce hundreds of videos per year, have a number of full time employees and went from zero to over 500000 subscribers in about two years, this is how it's a big part of how my small team can operate as if it were a massive team. Copilot easily plugs into a lot of the apps that I already use today, like Word and Excel. It can help me search through all of those documents. It can help me do research for videos. It could review legal documents. So many different things that add up to real productivity gains and business. I outlined it all in that short. Go check it out. Link in the description. Now back to the video. All right. Next, Hermes agent by Nouse Research. This thing has gone mega viral over the last few days. Already at 13 and a half thousand stars. Again, only having come out a few days ago. Hermes agent is like OpenClaw, but apparently it is really good at self-improving. This isn't just a coding agent. This isn't just an assistant. It is much more akin to OpenClaw. It is an entire framework, an entire operating system that you can use as your personal AI agent. So if you're deep in OpenClaw and you want to explore a great alternative, Hermes agent might be really cool for you to check out. So let's look at some of the features. It comes with a terminal interface, a TUI, with multi-line editing, slash command autocomplete, conversation history, interrupt and redirect, and streaming tool output. And of course, what OpenClaw did and showed the world is, go ahead, give us the chat apps. So it supports Telegram, Discord, Slack, WhatsApp, Signal, and CLI, all from a single gateway. Very similar to OpenClaw. Now here is where it is different. And I've seen a lot of projects like this come out since Andrej Karpathy's auto research project, which really gave me so much inspiration to build some of my own self-improving loops. But this is a self-improving loop within Hermes agent. So a closed learning loop, agent curated memory with periodic nudges, autonomous skill creation after complex tasks, skills self-improved during use. So they describe it as the only agent with a built-in learning loop. It creates skills from experience, improves them during use, nudges itself to persist knowledge, searches its own past conversations, and builds a deepening model of who you are across sessions. And of course, it has all of the lovely features that OpenClaw comes with, like scheduler slash cron. It can delegate to sub-agents, and it does that in parallel. You can run it anywhere locally in the cloud or anywhere in between. It also comes with a nice migration from OpenClaw if you don't want to have to start up and rebuild all of your different workflows, all of your memories from scratch, which is actually a really nice feature and something probably the industry is going to have to think about. How do you test out different agents without starting completely from scratch? There probably needs to be a standard created soon. All right, next is a project, the most popular one by far, called Superpowers. And it is already at 115,000 stars on GitHub. It is by author Obra, and it basically gives Cloud Code superpowers. Now, this is in some ways similar to GStack by Gary Tan, but it is much more focused on the development side of things. And in one of the funniest descriptions of itself, well, listen to this. Your agent puts together an implementation plan that's clear enough for an enthusiastic junior engineer with poor taste, no judgment, no project context, and an aversion to testing to follow. That made me chuckle. So it emphasizes true TDD, which is test-driven development, YAGNI, you aren't going to need it, and DRY. These are kind of best practices coding patterns. So here's the basic workflow. First, you brainstorm. Activates before writing code. Refinds your rough ideas. So very similar to GStack in that way in the YC office hours, but now this is kind of just a different flavor of it. Explores alternatives. Presents design in sections for validation. Saves a design document. Then by default, it uses work trees, which is really nice for parallelization of coding agents. Then it writes the plans. Then it actually executes the plans. Makes sure it's doing TDD and actually does that. Code reviews. And then finishes the branch. Now, this is a little different in that it comes in the form of a plugin. So it's a little bit easier to install than GStack. So to install it, you simply type clod plugin install superpowers at clod plugins official. Hit enter. And that's it. It's done. And then as soon as you hit slash superpowers, you can start to see all of the different tools available to you. So we're going to select slash superpowers brainstorm, and it is just going immediately. So it has an entire workflow that it's going to go through with me one by one. Just super useful. If you need help thinking through how to start these projects, how to actually execute them. So in a lot of ways, similar to GStack, just a very different approach. All right. And last, maybe the most interesting project of the four. Paperclip. This is intended to be an entire company run by AI, a zero human company already at 33,000 stars. Now, I'm going to give a slight caveat before I get into this. I have seen so many different projects claiming to be able to make you money. And that is not something I cover often on this channel really ever. I need you to know the likelihood of you actually making money from this type of project is low unless you put in the effort. And I really want to be transparent about that. This is not something where you're just going to click a button, say go. And all of a sudden, you're going to wake up with $1,000 in your bank account. It is just never going to happen like that. But nonetheless, Paperclip is very interesting. And I'm seeing more and more of these projects in which a team of agents is supposedly able to run an entire company for you. Probably the most popular of these, which is closed source, is a company called Pulsia. Slight detour, Pulsia is created by this guy, Ben Sarah, AI that runs your company while you sleep. So basically, you plug it in to your company and it just runs everything for you. Again, not only am I going to give you the warning that most likely you are not going to wake up with money in your bank account just because you submitted a prompt, but also I find it difficult to believe that these types of zero human company projects don't have the rough edges and limitations that I truly believe they do. So take all of this with a grain of salt. All right, so what is Paperclip? Open source orchestration for zero human companies. If OpenClaw is an employee, Paperclip is the company. So Paperclip is a Node.js server and React UI that orchestrates a team of AI agents to run a business. Bring your own agents, assign goals, and track your agents' work and cost from one dashboard. Let me show you that dashboard now. So this is their demo video. Let's take a look. So design-wise, very nice. You can see there are issues. The engineering team automatically takes those issues, starts to build things. You can see an org chart right here. These are all AI agents. Your CEO is an agent, Claude. Your CMO is an agent, Claude. Your CTO has two engineers, one of which is Claude. The other is Codex. You can have different projects that your teams are working on. And Paperclip has a lot of really cool features. All work is atomic. You can track how many tokens you're spending because, yes, this can get extremely expensive. It uses a ticketing system. So very similar to if you were on an actual engineering team with humans. You would use a ticketing system. You would submit issues. Those issues would get addressed, built out, and then deployed. It does all of that for you. And it's all tied to the goals that you set. So the agents are always referencing your high-level goal for the company. So if you want to get started, it is very easy. Here's the command. Paperclip AI onboard dash dash. Yes. And a little bit about their upcoming roadmap. They're coming out with a plug-in system, like adding a knowledge base, custom tracing, queues, et cetera. You can get open class style agents, which will just make them more capable. And you can chat with them directly through, obviously, things like Telegram. You can actually fully encapsulate your company into companies at SH, which allows you to export or import an entire organization. So there might be a way where you can share your organization with other people. Although, if you're making money, why would you do that? A skills manager, scheduling routines, better budgeting, and more. So very cool project. I have not tested this myself yet, but I've seen a lot of other people try it and say it's really cool. It's just very experimental. Again, don't expect to wake up and make a bunch of money. It's just not going to happen. But it is a vision into what's coming. These agents are getting more capable. The harnesses, the projects around them are getting more capable. And so with that, if you're early and you're seeing these things, you're experimenting with them, you have a huge advantage versus everybody else who are still maybe just using ChatGPT for Q&A, for example. This is the cutting edge of what's happening in artificial intelligence, all four of these projects. Hopefully you enjoyed. If you enjoyed this video, please consider giving a like and subscribe, and I'll see you in the next one.