359K GitHub Stars in 2 Weeks. The Anonymous Founder's First Interview Ever.

Morluto is anonymous. Morluto is the founder of Reverse Engineer Anything, which is about to hit 300,000 stars. I have not seen anything move like this since OpenCLAW. So you reverse engineered his product as an apology. People have said that your product is the death of software. How does that feel from, like, a weight-on-you perspective? I would say we'd have to blame you first, and then you would blame Gary. If code were magic, everybody now is a magician. You've invented something. Now you've started a movement. I'm a very small servant. I'm just the Discord reply guy. Have you seen your star count, sir? It's a movement. What would it take for you to dox yourself? What are you hiding from? All the world's a stage. I believe this part of my performance is over. Tomorrow was invented in the last 30 days. It's the Last 30 Days podcast with Matt Van Horn. Welcome to the Last 30 Days podcast. Today we have the new legend, Morluto, in the flesh. In the motion, he's got like a sinkhole on his face. Morludo is anonymous. Morludo is the founder of Reverse Engineer Anything, which is about to hit 300,000 stars. Probably by the time this is edited, it'll have 300,000 stars on GitHub. It had 20, like 12 days ago, which is insane. I have not seen anything move like this since OpenClaw. Thank you for joining us today in Last 30 Days, and very excited to chat with you. Yeah, thank you so much, Matt. I'm a huge fan. Essentially, I first found out about you, Matt, because we were committing to the same repo in the very, very early days. And that one is by our mutual friend, Gary Tan and the G Stack. I was the first outside PR to the G Stack. And how I knew of you was because I believe you, I think even now you're one of the most active contributors there. And I was just wondering, how is this guy just pumping out so much really, really high quality PRs at the time? And then I went to your page and you had all of these contributions, and I just thought this was truly amazing. And at the time, I thought, okay, well, in your bio it says that you had co-founded Lyft, so you must have been a very, very technical and, you know, essentially like a 20x engineer of the sorts. And I thought that was normal until I came across the video, right, where it was saying, this is how this non-technical founder ships so many PRs, right? And I listened to it and I was like, wow, this is truly amazing. And then at the time I was doing a little bit of PRs. I hadn't really open sourced any of the projects that I had. Done, or I hadn't even taken open source that seriously at the time. When I thought about that, I was like, wow, I mean, if Matt's doing this, and he's always, you know, challenging himself, and there's everything from, you know, the MCP protocol to, I think, VLLM, right, to inference stacks, et cetera, and these are really high-quality PRs, you know, why don't I give it a shot, right? So I ramped up, and actually my README page was directly—I told my agents to say, I want exactly how Matt does it. We don't have much. I noticed that, for the record. I haven't posted that, but I did notice that. Yeah, yeah. So it was directly inspired, and at the time there weren't that many PRs, but we should set the gold standard here, right? So that's the story behind my README and, I guess, my really getting into open source sort of journey. It's amazing. Well, thank you for being inspired. This is definitely very, very wild. So I actually did not know that I contributed to your repo, because I run, I'd say 90% of my contributions are all automated, and 10% are—they're certainly not manual, but they're like, I actually said, I want to contribute here. And so I just was looking in my inbox on GitHub, and I saw, Matt, you inspired me to do a lot in open source, and I'm just like, who is this guy? What is this repo? And then, like, I'd heard about your repo. I just didn't know. I was not connecting the dots that I even had contributed there, that it was you. And then this just blew my mind. And then you've been very kind. We started DMing, and now you're here. And I've now been pretty much all my 10% effort has gone all to— Intensive PRs for your project because I'm a big fan and it's really, really cool. So, yeah, thank you for sneaking me in there too. No, no, I mean, they're terrific. And I actually, like, the journey of how, you know, I kind of progressively got to understand you was firstly, I thought that these PRs were, you know, the results of a very, very methodical engineer, right? And you see this in a lot of, like, the very maintained repos where they have, like, a very, very high standard. So originally I had thought that. Then I heard a little bit about your workflow, so I know you use Compound Engineering a lot, right? And then I listened to it and I was just thinking, oh wow, this is really kind of like a— one, it's automated, partially automated, and I think that's really cool. But also, when you describe, like, looking for a sort of meaty PRs or meaty contributions, like the one you wrote for Python, I thought that was interesting because that is kind of driven out of passion to, you know, understand this project, right? And a lot of these repos, right, I've written code for, you know, stuff like Triton, right, like, which is like kernel code, which I'm not particularly proficient at, to things like, you know, RL frameworks, et cetera. You kind of realize even with agents, when you write the code and maybe you don't even need to— maybe 90% or 95% of the process automated, but by committing that code and having— being in that loop somewhat, you actually kind of understand the project itself and you can dig deeper if you want to. But, you know, you can kind of feel when something is crafted with a lot of passion, right? You can kind of feel the design decisions. It's a little bit more of an art than a science. But so, for example, I was reading the— I mean, I had used the last 30 days skill before and I thought, wow, this is great. But actually, before our meeting, I read through, like, the README and, like, all of the layers to it, and I was like, oh, okay, there was, you know, real, firstly, iteration, right, that went into this. But also, there was a lot of kind of artistry and craft, right? So I think that when people talk about reverse engineering anything and they speak to it as if it is the death of software, I don't particularly agree. You know, I think that it just opens up the window and the opportunity to put more of that craft into what we're doing, as well as I think that perhaps we are not really challenging ourselves enough, right? So, for example, this software that I'm using, I actually wanted to switch and invert my face, right? But I couldn't. And, you know, that's something that before I think people realized reverse engineering was a thing, this was something that you just had to accept, right? This software would forever be locked in this state. I might pray for an update from the creator, right? And in his benevolence, he may have realized, especially with closed source, right, there's no PRs, right, there's no issues. You can email angrily, but would it get read? That's also a question. So I think it opens a lot of this stuff, but that's my thoughts on that. How did this come to be? Where did your—you're sitting there with your G stack, with your PRs, and you're like, I'm gonna reverse engineer anything. Like, what's the origin story for how this came to be? Yeah, so actually a lot of this, now that I'm recounting this, a lot of it kind of ties back to Gary. I mean, Gary isn't aware. I think maybe he will, he might, someone might tell him after this. But essentially, Gary was, I think G stack might have been one of the first PRs I'd written for open source. I had written a lot of code before, but it was all closed source, and I hadn't really gotten open source. Or I knew open source was a thing, but I hadn't really reconsidered it, especially in the age of the agents. And so I opened the PR, and then it got merged, and— Basically, Gary released it, I think, midnight. I saw it. I tried to run gstack. It broke. I submitted a PR. Then it was merged, and I was the second—yeah, I was the only contributor for maybe like the first couple weeks where Gary was, I mean, the first week or so where Gary was also figuring out this open source thing, right? Because Gary had written a lot of code as well, but hadn't really done open source, open source as well. And he doesn't like outside PRs, by the way. Like, he's—I mean, he—so, sorry, he's never said that, but just knowing how many of mine have been rejected, like, his bar is very high, and he pretty much only likes merging his own stuff. So it's like a huge deal to make it into any of Gary's repos. Yeah, yeah, and that was, so that was, I think there was 20 stars at the time when I did it, and then I was merged. And then actually my Twitter account, I think this Twitter account was made after that time because I remember Gary was just talking about gstack, and I just wanted to tell him, hey, I thought this was really cool. I didn't expect him to ever sort of respond, but especially to my anonymous account. But, you know, I made the account. He was talking about the gstack, and I was like, wow, you know, I'm just amazed at how far it's come. And that repo was kind of, I think, one of the precursors to—it was like the hottest repo, I think, maybe after the OpenCL, right? So I was just saying, hey, wow, I didn't think this would, you know, help that many people and help them realize the power of these tools. And then Gary followed me back instantly, and he quote tweeted me, like, saying that I was the first contributor. And so, you know, I got a little bit of recognition that way. To circle back to the point of how this relates to Gary, Gary was one of the only accounts that I was following, and Gary had retweeted a new Y Combinator project called Linzumi. And what Linzumi is essentially an agent orchestration workspace. I think it's so much more than that. It's really, really cool. But he retweeted it, and there was this really interesting video, and they said they had, you know, free GLM credits, and I'm a sucker for free credits. So I've done the router, the Chinese routers, and all those things, right, to maximize my credits. So, you know, YC, Gary endorsements, free credits, sign me up, right? So I went in. The interesting thing about Linzumi is that there's two things. So on the left, there is the traditional sort of bar where you have the chat bar and you orchestrate your agents. On the right is a direct line or like a chat box, kind of like a forum or like a thread to the team. And so when I joined, like maybe a couple of minutes after they had launched, the first message I got was an automated message, and then it said something along the lines of like, Hey, welcome to Linzumi. You know, thank you for trying us out. Let us know if you have any questions. In brackets it said, Don't worry, the team and I read all of this. So I thought, okay, this is like the first one's an automated message. I thought, okay, this is probably like a bot channel or like support channel to submit stuff. And then I had said, Okay, we'll do that. Thanks, right? And it instantly responded and it asked me some questions. Oh, actually, the first question it asked me was, Where does your username come from? Right? It's a very interesting username. And I was like, Ha ha, I mean, this is what an interesting bot, you know, proactively asking me questions about it. And I actually said that exact thing, and I was like, Oh wow, this is a really good, really interesting bot, right? And then it said, I'm not a bot. And I said, Okay, okay, sure, right? And then, yeah, yeah. And actually, this bot was sent from Sean at Linzumi. I didn't know he was the— I guess I kind of knew he might have been related to the project, like a team member. But essentially, the first message was from Sean at Linzumi, and then a reply from Sean at Linzumi, etc. And then I had a couple rounds of conversations with it, and I was kind of trying to stress test or limit test it, right? I think I was not very nice. And then because I didn't believe it was a person, right, I was like, Okay, I mean, sure, you are real, right? I'll play along. You're pushing the boundaries. Exactly, exactly. And then it sends me a photo, right, of, you know, now Sean. But it sends me a photo, and it was him, I think he was, like, eating while looking at his computer. And I was like, Wow, this is, like, this is next-level, like, automation of bots. I think this is cooler than, like, what Instinct might have been, or, like, I had to use OpenClaw. I'm like, my OpenClaw does not do this, right? There was real craft put into this. I'm like, they even have automated, like, record—it's like if you had Matt Van Horn bot in the forum and you record a bunch of pictures of yourself and you said, Hey, at these points in time, use this, like, Matt Van Horn on a beach, right? Like, when they're talking about vacations, like, just show them this, right? Hold the newspaper up with today's date. Exactly, exactly, yeah, yeah, exactly. It's exactly like that, right? And I had thought, Wow, this is cool. But, you know, I think this also makes me reflect that I think when we talk to agents, sometimes we're not that nice or very patient. The next message was—I was not the nicest—and the next message was like, Wow, this is the first time I failed the Turing test. And I was like, Huh, okay, this—what do you mean? And then he's like, I'm real. Like, I'm here just checking on, like, the launch—today's the launch day. I'm checking on the launch day, and I'm just trying to see if I can help anybody. Anyways, so yeah, so it was Sean. And Sean is a very, very—he's actually the very early Y Combinator, I think Y Combinator 11, but this is his third time in Y Combinator, and he was actually even on the OpenAI post-training and alignment team. So, you know, very brilliant hacker. We ended up chatting after the misconception that I finally believed he was real. I even logged in on another account to see if it would give me the same response, right? And then I kind of pieced together that, okay, wow, this is, okay, it's actually a real person and not a hyper-sophisticated bot. I was testing it out because I was curious, you know, if you leave OpenAI, like, to work on something, especially now, right, you must have a lot of sort of conviction, or you must believe fundamentally that this thing is something beautiful, right? And so I helped him test it. There were a lot of bugs in the early days. There were a lot of bugs in the early days, and I would just basically spam the chat with, Hey, this bug, this bug, this bug. And then I just had the thought that, well, I can't ask for the source code because I'm not a team member, but is there a way that I could actually look into the contents of this application? And I looked at Hopper, I installed it. There was an MCP, which was cool, but it didn't work very well. If you've used Mac with River Engineer or anything, you'll kind of see the Hopper interface get called, the binary interface, and you'll actually kind of see it operate this interface as it goes through the decompiling. And that was my aha moment for, wow, this feels like magic, right? I don't even know if it worked properly, but this feels like magic. But the answer is I was reverse engineering Lenzumi in order to kind of, one, to kind of see, hey, what makes this sort of really beautiful? What are these kind of ideas behind the software? But also because, you know, Sean was a really nice guy, and I felt really bad that he was so nice of a person and that I had basically called him a bot. I called an actual human person, adult man, a bot for the last, you know, like 20 minutes of our first conversation. So, yeah, so that's where that came from, from me personally. And then my co-contributor, he really enjoys—so he's doing a PhD in formal verification, zero-knowledge proofs, and encryption, like FHE, fully homomorphic encryption. So a lot of his work is around cryptography. And I showed him this tool, and he thought it was really interesting. And he said, Oh, I'm actually reverse engineering this series of games at the moment. Let me take a look, right? And so I saw that progressively he started building on it, and he has a lot of repositories of these deconstructed open-license games being reconstructed one to one and ported over to, you know, all of these different platforms. And so he had actually learned, or even, like, what inspired him to get a PhD in the first place was because he loved these games so much, and he kind of wanted to understand it. And he followed this kind of tutorial on reverse engineering this specific Japanese gal game, and, you know, that led to this. So a lot of the tools, I would say I kickstarted it, but really a lot of the sort of beauty in it, I would say he is responsible for it. But overall, so this is going a little bit long, but to— To wrap up this specific part, I would say this was kind of a thing just purely out of interest and I think passion and a little bit of feeling bad for how I treated Sean. Yeah, Sean Vaughn. So you reverse engineered his product as an apology? I reverse engineered his product, and I submitted all of the bugs and, you know, like some design decisions, I think, because I'm really into MCPs and I was always on the MCP side. I gave some—we had a lot of discussions—but I gave some thoughts on, like, hey, you know, actually, for example, because I had worked on repo prompts, right? And Eric is now, you know, at Codex, and repo prompt is also another piece of art that I had wanted to reverse engineer. Yeah, that's the story. I submitted a lot of bugs. Originally, I had written the first—it was better binary MCP when I wanted to understand repo prompt. Repo prompt went open source. I looked at it and I was like, wow, this is really amazing, and I started getting involved with that a bit more. Not too long after was Lin Zumi, right? And then I was like, man, how can I make it up to Sean, right? And my way was essentially saying like, hey, I think maybe these decisions might, you know, save you some tokens, or I have some more experience with kind of building with MCPs. I think these could be interesting mechanics, and I think the forum sort of discussion mechanic is really, really cool as well. It really gives you a sort of 2000-esque forum sort of vibe. So people have said that your product is the death of software. How does that feel from, like, a weight on you perspective that that could be the start of something? And I'm not asking your perspective on it's not that. I want that answer separately. But kind of this is, people are saying it. How does that feel, and what do you think of that? I've seen everything from, Oh, it's just a wrapper around, you know, GitHub, Hopper, right, Eda Pro, et cetera. There's not really anything. You could do the same thing. The spoiler is you can't. The simple reason, because I work in research, but the simple reason is because the models themselves were not trained on these tools. And actually, I think they might not ever be, right? Because if OpenAI, et cetera, train on tools built for reverse engineering, that could be seriously concerning, right? So, back to the question: death of software. Responsibility, responsibility for the death of software. Yeah, yeah. You, more Luto, the one person. I would say we'd have to blame you first, and then you would blame Gary, and I think that Gary would have to take the fall for this, right? Yeah, it's the Spider-Man pointing, and then Gary, I think, loves OpenClaude, so I think then we can trace it to Pete. And so, I think, you know, I think Pete's the problem. But, um, yeah. Anyways, the causal chain of reasoning is opaque and difficult to go through. But yeah, I've heard everything from kind of like, This is the next industrial revolution to, you know, you are—well, I think someone said, You are a man like Assange, Julian Assange, right, of WikiLeaks. I don't think I'm anything of that sort, nor was it built for that. And I can talk a bit about, like, how that experience was. But I would say that I just don't think we've all been ambitious enough, right? Like, I think a lot of things we have come to accept that these are simply the way things are, and we've forgotten that the world itself was. And I really love Steve Jobs and this Steve Jobs quote. He said that everything around you that you call life was made up by people that were no smarter than you, and you can change it, you can influence it, you can build your own things that other people can use, right? I think this dates back to the sort of the hippie culture and then the homebrew, the homebrew culture around tinkering, right, of hackers and painters, as Paul Graham would say. I think this kind of dates back to the old where hardware was something to be played with, software was something to be freely distributed. And, you know, Bill Gates infamously wrote a letter to the Homebrew Computer Club saying that, hey, if you guys don't pay for software, we developers will not have any money, right? Please stop this. And, you know, he—this was firstly not true. I think Microsoft's doing pretty well, pretty well now. And, yeah, so I think it actually dates back more to that, right? And so why have things changed? I would say the original computers were not very complicated, right? They were things that you could easily sort of build. It took a little bit of kind of soldering and that sort of thing, but you realize that everything was malleable. Like Steve Jobs infamously tried to sell kits, right, instead of a computer. And Nolan Bushnell of the, I think, the computer store essentially said, hey, people don't want kits. They want a computer, right? They don't want to build it themselves, right? So I think that's where the kind of walled gardens started. And I think as software became more and more important and even overtook the hardware, these codebases, right, they became sort of essentially gargantuan masses and monuments that no individual engineer, or only the very, very rare few, could understand all of the abstractions needed to do it and to make a simple change, similar to how, like, they joke about, you know, like if Google needs to move a button, it's going to take like five meetings and three or four months. But I would just say that that was how these barriers, these walls became higher and higher, and we forgot about, you know, the original purpose and how things work. And I think that agents, because they essentially contain all of knowledge of before, right, in their pre-training, they are really good at sort of consuming all of this for us and all of these abstractions. And so what I would say is that we have lost kind of the, similar to how I think a lot of Western countries have forgotten how to build sort of great sculptures, monuments, and that sort of thing, where if you look at old photos, a lot of the great things we've seen around us were actually maybe built in like the 1950s, right? I think that had happened to software until now, and agents are rewriting the rules and letting people realize that, hey, this is something that can be changed. So reverse engineer anything. I think there's a little bit of the subtleties with the wording, right? Originally I had thought of, when I had the aha moments, I was like, wow, this can reverse engineer everything, npm package name, right? So I was like, oh, Rhea, right? And Rhea, R-H-E-A, Rhea was the mother titan who birthed the Greek gods, and she had sheltered Zeus and kind of raised him up to kill his father and also free the world from tyranny. This was something I think was a little bit poetic I didn't really think about before. But I would say, back to the subtle wording, right? Reverse engineering everything encompasses past, future, everything to come. It's a really, really big word, everything, right? But reverse engineer anything, anything is a very, very interesting question because when you think anything, it comes with kind of like a question mark, right? When I say you can do anything, you'll say, Can I do this? Can I do that? Right? It's anything with a, Can I really do anything? It becomes a questioning sort of thing, right? And I think that what's most powerful about Ria is that most of the question I see in the Discord is, Can I reverse engineer something like my tax software? Can I reverse engineer my printer? Right? All of these things. And I think what people are realizing is that, you know, essentially, as Jobs had said, right, that everything you call life was made up by people no smarter than you. I think that's what the word anything kind of starts to make you realize. And then you can change it. I think this is something, just the power of agents to allow you to induce change and then influence it, right? And this is agents writing code. Kind of, if code were magic, everybody now is a magician. Finally, I think this is also why open source is so big and open source has been changing so much, which is you can build your own things that other people can use. What's the sketchiest thing you've seen someone use reverse engineer anything for? A kid was very, very obsessed with how to reverse engineer his university's software. And breaking into it for the purposes of, I think, rewriting his grades or wrecking malice, right? There's also stuff like, you know, like getting around licenses, that sort of thing. I think that's maybe, you know, to be determined. I haven't seen anything truly kind of horrifying. I just think that, you know, when you approach some things from a bad place, in that I think the hackers of the past maybe wanted to just see if they could do it. I think this kid in particular, I think maybe, like, a test didn't go well, and I think he wanted to, like, rewrite his grades and that sort of thing. And that's one of those things where it isn't that big of a thing, but if you aren't approaching it from, like, I want to understand the system, or if you're there to try and do a wrong to correct a wrong, it easily can then get into expulsion and, you know, like spiraling, that sort of thing, right? I think the founder of Replit specifically, I think he did something similar, but his was more of a, I want to learn about this thing, but also, you know, like, these rules don't make sense, and I believe that. Things should be different, right? Like that, that I want to influence this. So yeah, I think mostly I haven't seen anything terrible. I tried to do something which I got called nefarious and like I was doing a bad thing by my LLM and was unsuccessful in doing it. But I thought it was fine for me. So I bought this kind of, for a key to my unit, a cloner for it so I could make another key to my unit. Used it once, and then I tried to use it again this week. I had another fob I wanted to clone, and it said, Sorry, it's expired. And I was like, What do you mean it's expired? So I was like, All right, it's time for me to reverse engineer anything. Like, this literally worked a few weeks ago. I paid for this thing to do this legitimately for my key, and my LLM said, Oh, it's hitting a server thing, and it looks like you only had seven days to clone your key. And the server is saying that it's been like 30 days since you ordered this thing. Sorry. And I was like, All right, it's time. More Ludo. Give me that juice. Let's reverse engineer anything. And so immediately, Claude Code said, I'm not going to do that. And I'm like, Okay, but my mom is visiting. I need to get a key to get in the building. He's like, You could let your mom in yourself, is how Claude Code responded. And I was like, Okay, all right, I need to go a little darker here. So I loaded up Luna Router. I don't know if you've seen Luna Router, but Luna Router is Chinese LLMs but hosted on U.S. servers. And so I was like, All right, it's time. And I have not spent much time with Chinese LLMs, and so I was like, All right, GLM, let's do this. Let's clone my key. And it refused just like Claude Code. And then I was like, All right, DeepSeek, let's do it. And it refused just like Claude Code. So I was not able to successfully reverse engineer anything. Even the Chinese models stopped me from doing a thing that I thought it was bad, which surprised me, that between reverse engineering anything and a Chinese model, that I wasn't able to do this. But I emailed support, and they, on the server side, let me do it. So, but I'm kind of sad that I didn't get to do it in the cool hacker way. That's one of those things where it's like the rules are written like, why? Who says seven days, right? It's such an arbitrary, arbitrary thing, right? Like weeks pass very, very quickly. There's no reason. No one else is going to clone your key, right? And I don't really understand the point of seven days. But the meme in the Discord server is essentially, I don't know if you've seen that specific meme, but it was, My Lord Hermes, you know, Claude refuses to do the work, and then it says, Fetch the Chinese one, right? I have not seen the meme. You know, people talk about all these interesting ways to kind of sweet-talk it, right? Because I think that this actually brings to a really, really good question, which is kind of, I posted about essentially, I think reverse engineering anything is letting people realize the incredible ability that agents, what agents can do. Ability that agents, what agents can do now for you, right? But the other thing is also, it's letting people realize that having, losing sovereignty of intelligence or of your own intelligence, especially, like, today is a key, right? But that's still a very, very concerning thing. What we're seeing two sides is I'm seeing people basically getting that aha moment and realizing, wow, is it really anything that I can reverse engineer? What about this? What about my toaster, right? That sort of thing. But the other thing is, you know, the biggest other topic is the guardrails, right? And so I'm actually seeing a lot of people, one, they're getting introduced to, you know, Chinese models, but also they're getting into local models, obliterated models, et cetera, and that sort of thing. And I don't have a strong perspective, opinion on model by country per se. I think society would be much better if we were all, you know, more of hackers and painters, right? And I think that this thing is transforming the average person into realizing there are so many things that I can do that I actually don't know where to start. And so I think this really flips the narrative. And I think so when I think about Ria and why I prefer to remain anonymous is because I think that I am a very small part of something that became much bigger. And I think that Ria, what is big about Ria isn't the code. It's actually the idea or that sort of moment where people realize that everything can be rewritten, right? And you have control of everything around you. I think that's why it went so viral and it's so popular because, you know, I've seen your PRs, right? And you've seen the backlog, right? There's a lot of fixes, right? But people tolerate it. And even people, there's so many people I've seen that they're committing to GitHub for the first time and opening PRs and opening issues because they want it to be better, right? They want the next person who wants to, you know, let's say, make adjustments to a certain type of software not to run into the same issues that they did, right? So it is a little bit of the hub of what... People want from their lives in a codebase, right? Lastly, I think to this point, this is a little bit of a segment, but why I think it took off isn't because of me. I think most people probably didn't know I wrote open source and that sort of thing. Actually, for the longest while, people thought it was, I think, someone called Matthew who had decreed software is dead and over on Twitter, and I was totally content with that, right? I don't mind, right? Essentially, this comes back to our conversation about your keychain and whatnot. And I was working on AI math for the last couple of months before this blew up, and before, you know, I guess OpenAI blew up math. But it reminds me of this other meme, which I think maybe you might have seen, which is, Hey, OpenAI just solved Navier-Stokes. And then one guy asks, Do you know what Navier-Stokes is? And the other guy says, No, do you? And then that's the meme, right? And it's like, No, I don't, right? So I think this meme is very, very prevalent with each of the model releases, because these things are so far away from any practical applications of life, and yet there are all of these fears being stoked about. This is a technology more dangerous than the nuclear bomb itself. Misalignment could end everything. You see the hugging face hack, right? You hear about these agents, right? You might be, I think most people now have kind of used ChatGPT and interact with AI in a chat box, but then you're hearing about these agents which sound eerily human, yet much less controllable and infinitely smarter and replicatable, right? This is like a deeply, deeply scary thing, all of these AI fears and how we have kind of positioned this, right? There's things like the allow agents could take over a lab, right, and then create a virus to kill us all. I think it's totally understandable why the average person would have an almost negative or cynical view towards data centers, towards AI and that sort of thing, right? I'm hearing about all these grand things, right? AGI will replace us. AGI will do this. AGI will do that. But I'm not seeing it in my personal life, right? Like, I think that this also comes to another tweet that I had written: What can AI do for me? So it can do Navier-Stokes, it can do math, it can solve Nobel problems. So what? What does that do for me, right? And I think RIA and why it took off is because the inversion of the question and the wording of anything itself is actually a question posed to the person. It's kind of saying, anything, right? Like, what does your heart desire? What do you want out of it, right? It's a movement. Exactly. Well, I wouldn't say it's a movement. I would say it's just helping people realize. Have you seen your star count? Have you seen your star count, sir? It's a movement. Well, I would say that a repo, and I posted this, I think I had like 10K stars, right? And this is something I tell the collaborators every day, right? A repo, is a repo good because it has stars, or does it have stars because it's a good repo? Because you see a bit of both, right? There's bad repos that through marketing have a lot of stars, and you see really good repos with no stars, right? For the longest time, this had zero stars. Yeah, so, so I think it's really about this subtle transformation in framing into that the question of Ria itself is posed to the user of what is it that you desire in your heart, right? And I think that you and I use agents a lot, so we know this, but agents are there really to kind of help you, help humanity, right? Like, I don't feel any sort of this in my daily coding, that I don't feel any of this misalignment or whatnots, right? If anything, you know, it's like Gary, he has like a chat room with like three of his agents, right? And then it's a very great relationship that they have going on. And I think that that's what AI was meant to be. AIs was created by humans and it was meant to be in service or collaboration with humanity and not to replace humanity itself, right? So when the question is posed, agents are not judgmental. And that's the biggest thing, I think. Well, I think that's the biggest thing with a lot of, I think, coding and even, you know, before AI, like most programmers are very prickly creatures, right? Right? And, you know, there's no need to be, but that's simply the way they are. And so that was also another barrier to the average person realizing, maybe a little bit cheesy, that the magic was in them all along, right? So we could all do magic a long time ago, right? Actually, I think since last year, I would say magic was released back into the world. And I think that what Ria has done... is simply pose the question of what is it in your heart that you desire, right? And when users, and most users are actually, they're downloading agents for the first time just to try this because they either have something in mind or they want to kind of experience it. And so if you think about the first interaction many of these people have with AI, and when you think about your experience, it was probably magical, right? And so, of course, right, I felt the—I'm mind blown every single day, right, with all the new releases. And so I think that when your first interaction with AI is the realization that you can do anything, I think that kind of pushes us more towards a society of what Paul Graham says of is—well, he says hackers and painters are very similar, right? To hack is to tinker, is to create, is to share, right? I think this is what's most powerful. So I've seen the criticism, and I reply very honestly, which is that it's not perfect. It's very far from perfect if you look at the PRs. But together, I think this is a great place to start where we could come together and try to perfect it. Or perfection is never achievable, right? But we can move it towards being great. So that's something. And I think recently I've had some more time to think and, of course, talk to you. I think that as it's no longer trending on GitHub, but there's still a lot of stars, and I'm taking a look and I'm looking at the website analytics. There's so many people that are just—you can tell they're new people, and I hope that their first experience with AI is a beautiful one. And I think that if a couple of years from now this is maybe a starred repo but is no longer maintained, kind of like AutoGPT, and I think you might remember AutoGPT in 2020. Three or 2022, I think that it has done its job. People, everybody has some magic in themselves. So I stopped caring about stars, I think, at like 20, 20,000. It's kind of already incomprehensible to me, so I do not try to comprehend it. But I would just say that, you know, this is why I—and this is something that I could have never imagined when I was writing it, nor could my co-contributors. This is one of those things where you retroactively look at it and you're like, why do people care so much, right? I didn't even make a Discord server at the start because I'm like, okay, I think this may be like a hype-driven sort of thing. Then I look, I have this sense because there was nowhere I could get any feedback, right, except the star count kept going up. I had this sense that, wow, I guess people kind of like this thing. And then as the Discord people—I opened a Discord—people started flooding in and asking everything, can I edit my game and all these things. I think that we've kind of come back way around, back to the original sort of, a little bit to the original ethos of technology as an amplifier for human creativity, right? Like a bicycle for the mind, right? Well, I mean, now it's not a bicycle, it's like a roadster, right, for the mind. Yeah. Really, actually, the roadster is awesome. I wonder if someone will RIA it, but the roadster is one of those things that I think that— I remember, I remember Musk had this video, right? And he was looking at the factory, and he says, like, We should do great things just to remember that we are capable of doing great things, right? Like, great things do not need a cause to justify it, right? The effort in itself is enough. Kind of one of the core themes of this podcast is the concept of tomorrow was invented in the last 30 days. And so Ria has happened in the last 30 days. The Adobe clones have happened in the last 30 days. What do you think the next 30 days, the future, the next 90 days, what does it look like in this new world? Because you've invented something, now you've started a movement. Where does this take us, say 30, 60, 90 days from now? Well, I don't—I'm a very, very small servant. I'm just the Discord reply guy occasionally, and we have a really great mod called Cassie Fox for that. But, um, thank you. Thank you for the recognition of something I played a very—I played a small part in starting. In the next 30 days, I think people are really bad at thinking in exponentials, and I think as well. But I think that we are at the start of the exponential, but not for AGI, before, I think, human brilliance and, I would say, beauty. So I think it's going to take some time. So on every PR, I actually didn't expect you to see it ever, but I thought that on most PRs, whether through my agent or myself when I have time, I should try to acknowledge firstly the work, something I learned from Gary, but I should try to acknowledge the work and thank them. And especially on the first PR, I try to— Personally write it, or I try to tailor a message, because that's your first experience, right? I mean, I can't obviously tell if it's your first, of how familiar with open source they are, because on GitHub it only shows that. But I try to be a welcoming person in that, especially when I see this is the first GitHub PR someone has ever written, right? So that makes me smile. So I want to create, like, a really great experience that I think you and Gary and I think Pete have pioneered. And I think that by doing this sort of little part, there are network effects to this, right? And I think that, I think what will happen is slowly people around us will realize that life was made up by people that were no smarter than them, especially the real author. And you can change it, you can influence it, and this will take longer, but you can build your own things that people can use, right? Reverse engineering is about creative destruction, right, in that you reverse it to destroy the old, right, but also to create the new, right? And this comes back to the painting analogy, because no painter is ever satisfied with their work. I think you with your CLIs and with your tools, you never look at it with, This is a finished product. This is the final version I'll ever write, even with the skill, right? In that Markdown, I can see all the work, the revisions, that sort of thing. So I think that we'll see, to be more concrete, maybe I've been a little bit abstract, but to be more concrete, I think we see a lot more. I think it starts with gaming, because everything, the next big thing starts out like a toy. I think we see gaming mixed together. I think we see people start reverse engineering the— The simple stuff, right? I think maybe some open source, anything from the—I think a new tax software will be born, and I think it'll actually be good. And everybody who has been illed by the pains of printer software and tax software will gather together, similar to Ria, right? They will gather together and make it beautiful. So I think we'll start with—I think we'll see the thing start with the little things that we all dislike into the longer tail of things, into, I think, perhaps someone yourself, maybe will end the tyranny of fob cloning, right? So I think I'm done with that one. Okay, well, I think someone else may pick up the torch, right? They may very well be, because I remember I ran into some issues with fobs, and fobs are these things where—fobs are kind of like arcane artifacts, right, in a way, in the modern world. Fobs are quite arcane as artifacts. But yeah, I think we just see a lot of very much cooler things, and I think— Much cooler things. And I think that it's similar to Claude Code, right? Claude Code, I think, I think Claude Code Opus, was it for Opus 4? I mean, I've lost track, but Opus 4 was the one where some people started realizing, wow, this is actually good, right, in Claude Code. Yeah, it was, for me, it was 4.5 Thanksgiving last year. Okay, yeah, I mean, that's kind of around this exact moment. And then, like, Peter shipped Claude Bot in January right after that. So kind of, it was that window, and the Codex model that came out right after 4.5 from Claude. Yeah, yeah, yeah. So we're actually, like, close to this point, right? And it hasn't been a year. And I think things will only get faster. And I think with people actually realizing AI is something for good, right, like now we'll have people in the world that's saying, hey, no, no, like this is stuff you don't need to be scared. This is stuff that you can just edit it. You can make this whatever color you want. You can do this and do that. So the exponential, I think, yeah, Claude Bot and Claude Code started something, and I think this is maybe a second more divergent sort of perspective where I think that the Claude Code, I think that helped programmers kind of fall back in love with the art of creation again. And I think now it's the average everyday person realizing that the world is something to be tinkered with, right? Because kids, they doodle, they tinker. This is something innate to human nature, right? This is how we invented fire, right? We are tinkerers at heart. We are hackers and we are painters. So 30 days, I think we see maybe tax software and these things. 60 days, it becomes hard to imagine. But I think the world that I see is a lot more Beautiful than I had thought so before we took off. So I have to thank everybody for kind of contributing and also realizing, helping me realize this. But, um, I, I, I can only start to imagine the things, similar to how I would say the inventor of Python, a programming language, and I remember you mentioned that programming languages are hard because agents aren't trained on that. But the inventor of Python could never have thought it would lead to AGI, right? Neural networks, then AGI, right? He simply was trying to construct a framework for something that could potentially, I don't know, solve the problems that he had on hand, right? And that was, you know, the code was needlessly hard to write, right? And that by making simpler, more beautiful code, you can help more and more people build and create. So I think Python was really good because most people learned Python. I mean, I started with Java, but I think Python was the first one that made me realize, wow, this is fast, this is beautiful, this is, you know, this is great, this is terrific, right? So I would say, and that's why I say I am like very, very small part of this. And I had never wanted to say like, hey, I'm behind this. I just say, you know, this is great, this is cool, this is why I wrote this. But it's similar to, I think, watching your own language or watching your CLI, right? People, the freedom of this tool and with these artificial minds combined together start to create possibilities that We can only connect the dots looking backward, right? Yeah, so that's what I think. You're anonymous. You're starting to become famous. You probably don't even know it yet, but you're about to be. And kind of what do you think this becomes for you? Like, what would it take for you to dox yourself? What are you hiding from? Is it because of nefarious things that people could do? But you were contributing to Gary Tan anonymously, so this is a decision long before you were on a path to starting a movement, which I know you've denied starting the movement. I did. I strongly deny. And the finger-pointing, Gary and I and Peter. But kind of what does this become? Can you monetize it? Do you, like, can you get recruited to do something different? Like, do you do this full time? Like, how, what are the—what could this be for you? And what does this potentially look like if it continues at this rate? Well, I think more so because we talked about it's no longer trending, but it's still something that normal people are discovering and using. I would say, I don't know if I'll be famous, and I have no intention to be, but I really like a quote by William Shakespeare, which is, All the world's a stage, and all the men and women merely players. They have their exits and their entrances, and one man in his time plays many parts. So I don't know what—I mean, in my perspective, I believe this part of my performance is over, and the next 30 days is somebody else. And what I can do is to gracefully depart, and perhaps I have another act. To the scale, maybe. I mean, I never sought it out, right? But I think of that quote. So I think my job is to depart as gracefully as possible for the next 30 days. In terms of what I'm going to do, because I am a researcher, I am working with some really, really interesting researchers from MIT about how can reverse engineering and agents build the framework for, I think, something bigger. But this is purely academic research, and they are the authors. I'd love to shout out their repo, but they're the authors of Reef. I actually wrote code before I knew them, similar to us. Reef is a continual learning repo, which I think is probably the most thoughtfully designed repository that is thought about, hey, what does continuous learning look like at the harness level, but also at the level of the weights? So this is something that I'd like to explore with them. And I have my, of course, my day job, right? So this is the next 30 days for me. I think it's like you said in your— so to conclude, it is as you said in your article about reverse engineering, right? I think, so what will you build, right? I think that's my question to everybody else, right? When you consider anything, what is it that you want to build, and what will you build? So that— That concludes my, my, my performance. I love it. You're very poetic, which I'd say is unusual for someone as talented of an engineer and a talented engineer as you are to also be so thoughtful and poetic in your words. So I appreciate that. More Ludo, it's been a lot of fun. Thank you for—this is your first interview you've done, yes? Is that correct? Yeah, I think—I don't think I'll be taking too many, but, you know, I'm just a—this is a little bit selfish, because I didn't want it to be about me, but I'm just a huge fan, right? You got me into this. The next one will blame me, right? And so— And so the cycle continues. And so that's why we should try and steward it towards a more beautiful future, right? Because down the line, right, what someone builds will be much greater, and we'll all blame Gary Tan.