Come join us as we discuss everything open source with guests that are pillars in the industry. Welcome to The Secret Sauce.
Welcome back to the show. We're about to learn the secret sauce. Excellent. Omoju. Hey.
Bdougie:I am super happy to have you here having a conversation. I'm a big fan of your work and what you did at GitHub. We worked at GitHub at the same time. No longer work at GitHub. And, do you wanna introduce yourself to the audience of what you do, who you are?
Omoju Miller:My name is Omoju Miller. I am the CEO of a startup called Fimeo, and we work in the data space. Initially, we started working with Web 3 because our my thesis initially was Web 3 needed lots of security tools for developers. And my area of expertise is machine learning. And I knew that Ethereum, the ecosystem, had majority of the smart contracts.
Omoju Miller:So it felt like we should build some security tools around the programming language solidity. We built it. We shipped it. Unfortunately, the market is kind of like shifting underneath us.
Bdougie:Yeah.
Omoju Miller:And there are not that many active developers. So we're in the middle of a pivot, and we're pivoting into something else around data as well. And we begin to look into all kinds of different things. One of the areas that truly interests me is end to end debugging of applications in production, It's basically distributed systems. Yeah.
Omoju Miller:Like, how do you actually do that? It's very difficult. Almost virtually impossible. And then even in thinking through all of that kind of stuff, something else that's coming up is data provenance and data lineage. All of these things are basically the same kind of thing.
Omoju Miller:It's basically tracing the path of something through a very complex system. Yeah. And that's what I'm doing now.
Bdougie:Cool. Yeah. I'm looking forward to following that journey. But I wanted to actually start from earlier days Mhmm. At Moju.
Bdougie:I know you you'd spend some time were you're a PhD candidate?
Omoju Miller:I got a PhD.
Bdougie:Oh, you got a PhD? Okay.
Omoju Miller:I did.
Bdougie:At Berkeley?
Omoju Miller:Yep.
Bdougie:Okay. So what's your like, what's the background that you led into working machine learning and data?
Omoju Miller:My so my undergraduate degree is in computer science, and I got really in love with artificial intelligence, the idea.
Bdougie:Was that also at Berkeley?
Omoju Miller:No. Well, this was not at Berkeley. This was at University of Memphis
Bdougie:Okay.
Omoju Miller:Years years years ago. And so it was basically like symbolic, symbolic systems. And I really got into it, but it didn't really it was too early. It didn't have, like, a commercial application outside of, like, the military defense complex.
Bdougie:Yeah.
Omoju Miller:So it was like a you'll get expertise in something pretty much going nowhere. And in the intervening years of doing that work, I became a stay at home mom. I stayed at home with my son for 4 years. And in those intervening 4 years, the world shifted. The iPhone was launched.
Omoju Miller:We now had ubiquitous data. We had Google Summer of Code. Python became like a real real language with lots of packages. So all of and then, most importantly, search, Google search really won the search space. So statistical based approaches to symbolic systems were the thing that was winning the day, probabilistic, statistic based systems versus what we were doing.
Omoju Miller:And so it was very clear to me that, all right, I guess I actually now can go back because the things I'm interested in, actually, I have a commercial application now. So I came to Berkeley, UC Berkeley to do almost like a cognitive science PhD, how people reason about technical material. I did it in the graduate graduate division. It was a computer science education PhD between the education department and the computer science department. I did all that.
Omoju Miller:And if you really wanna know how people reason about technical material, the person with the largest dataset is a company called GitHub. They have the largest dataset of how people actually actively work in the technical spaces and learn technical material. And so it was very clear to me that I found really serious about that kind of work. It's really good work at GitHub. And I went to GitHub and was part of the founding ML team, the machine learning team.
Bdougie:Yeah. What year was this? 2017. Okay.
Omoju Miller:So the founding ML team, the previous CEO of GitHub, Da Funkt
Bdougie:Yes.
Omoju Miller:At the previous year during give up GitHub Universe, they said he went rogue on his keynote and started talking about ML and AI and how they should use those systems and their data to build the interesting things on the platform, which is what led to them creating an ML team. And it worked out perfectly. I joined that ML team. We built recommendation engines. We built few things here and there.
Omoju Miller:And that but it was so it was so new that it was like a uphill battle of actually trying to get ML into the systems because people didn't think about it at all. They were just doing, like, pull requests or or something like that. And then we started doing this research project, Hamal Hussain, who was a colleague of mine at the time, semantic search. It was like using natural language to search for functions or natural language like, oh, if I could say this, could Python write me a function? So doing all this research.
Omoju Miller:And it kind of like, this model seemed to work a little bit. And so we released that as an experiment in one of these GitHub, universes. And one thing led to the other. I shifted roles from being the ML person to being the technical adviser to the CEO, who was then Nat Friedman.
Bdougie:Yeah.
Omoju Miller:And Nat was very ML savvy. So I guess that wasn't makes sense why, you know, I was technical adviser.
Bdougie:Super into AI at this point.
Omoju Miller:Super into AI. He's very ML savvy. And so one of the things that we ended up doing was that work with OpenAI and Microsoft and actually, like, collaborating GitHub, Microsoft OpenAI to bring GitHub Copilot, which is basically, like, truly advanced semantic based search, natural language search. Copilot was a fantastic success, was one of the first LLMs that actually proved super intuitive.
Bdougie:Yeah.
Omoju Miller:And then OpenAI OpenAI. Like, everybody's using ChapJPG. Turns out that it wasn't just, Copilot is amazing, but it's a developer tool. The average person is not going to use it. If you're not a developer, you don't need to do natural language search.
Omoju Miller:If you put that same model in a chat environment, anybody can then use it. We all used to chat, and the thing took off like a rocket.
Bdougie:Yeah. Do do you think Copilot, its success helped this sort of wave of new because, like, I was at GitHub during the Copilot launch.
Omoju Miller:Mhmm.
Bdougie:I knew of OpenAI. Mhmm. I was like, I kept seeing all these ML things that everyone tried. It was, like, cool little party tricks or cool things that you could show. One of the things I actually I would show all the time is when you go to, any GitHub repo and you go slash contribute, there's a whole blog post about this, but it shows GoodFirst issues and documentation.
Bdougie:And it's not just about just the one label, but it can give you an onboard ramp, and that was ML powered. Yep. And, like, we would ship things like that at GitHub, and then, like, we'd move on to the next thing. Yeah.
Omoju Miller:People there wasn't, like the challenge was you'd ship things, but there wasn't a product manager whose job was Yeah. There wasn't an ML product manager. And now I'm sitting in the role of a CEO that I do products, that I also lead engineering. You need a product manager to drive that product to its full Yeah. Fullest extent.
Omoju Miller:And it's also nontrivial to create ML product managers. ML is still relatively new, even though it's been around forever. But commercially, only a few companies were really, really building ML. So what does it mean to be a product manager? So I think I think it was just so new, people just did not understand.
Bdougie:Yeah. Yeah. It's it's it's fascinating because, like, there is even, like, the were you part of the GitHub trending stuff that so, like, what's the story behind the GitHub trending? Because it feel like there's no innovation since it shipped. Like, it just sort of exists.
Omoju Miller:So, I mean, I think trending has been around before the ML team even came about. Yeah. It's just like whatever was trending right now, it's a it's a simple query type system. But even so it's, knowledge discovery. Like, you've got all this stuff.
Omoju Miller:You wanna know what's really hot. What are people using? And I think initially it was based on stars Yeah. Which we all know is not enough of an indicator, but it gives you some kind of signal. And whether stars are good as a a signal or not is irrelevant because trending became a thing that changed people's lives.
Bdougie:Yeah.
Omoju Miller:If you trended on GitHub, you could write your career check.
Bdougie:Yeah.
Omoju Miller:And so then we wanted to actually do it better. There was a Nigerian developer that I got to know, Unicode developer. That's I mean, he has a real name, but Prosper.
Bdougie:Prosper.
Omoju Miller:Yes. Prosper was doing this Laravel stuff, and all of a sudden, he built this Laravel thing and started trending and, like, woah. You know? I think it was one of the first things that you saw that the Nigerian developer ecosystem is really, like, punching above his weight.
Bdougie:Oh, yeah.
Omoju Miller:And so you're seeing all this, and we were like, way makes sense if you could even refine it. So you could search by country, you could search by language, you could search by this, you could search by that. And then we built and shipped that.
Bdougie:Yeah. Yeah. And I just met Prosper for first time in person in Nigeria for the Oscar Fest. Fest.
Omoju Miller:Oh, you went to Oscar Fest? Yes.
Bdougie:Yeah. Oh, nice. I was one of the me, Angie Jones, and a few others keynoted. And, Prosper, I got to hear his story. Because because of Laravel I'm not in Laravel, so I didn't know who Prosper was.
Bdougie:I knew Unicode developer and folks inviting him to speak other places. But now he's, like, working on Novu. He has a whole ecosystem around him, other developers who are growing up. And, like, the turn the time during Nat Friedman at at GitHub, like, we spent some investment in Nigeria.
Omoju Miller:Yes. We did.
Bdougie:Because there was clearly trends in emerging markets. And when I say emerging market, not just, like, developing countries, but, like, new countries that are doing open source and doing it well.
Omoju Miller:Mhmm. Mhmm.
Bdougie:So that was also part some of the work you did under That
Omoju Miller:was that was the well, I think that is actually how I got my job.
Bdougie:Yeah.
Omoju Miller:Because I was always very I'm a Nigerian American, and I go to Nigeria very often. And I always Nigeria has so many people, and I met so many intelligent people. And then it's also like, if I believe I'm intelligent, then everybody else is like me is also intelligent. What are we doing? And so the best part about working on the data team at GitHub is you work on the data team at GitHub.
Omoju Miller:You have all the data at your fingertips and you have production access, which means you can be a sleuth. You can just live in that data and be looking for things. Yeah. And that's what I used to do. And so I would write all these queries because I was just curious, who is doing what where in the world?
Omoju Miller:Yeah.
Bdougie:So
Omoju Miller:I would write all these queries because I wanted to see who was doing what, and I wanted to slice data in multiple ways trying to figure out, is the thing I'm thinking about my hypothesis, is it really being validated? Because I had all this crazy hypothesis. Like, I I'm one of those people that truly, truly deep down believes that technology is one of the best things that has ever occurred to humanity. And that if you give people true access to education and technology education in particular with capital in a permission less way, we can actually change the world for the better. Yeah.
Omoju Miller:So therefore, my whole goal was let us reduce the barriers to entries as much as we can. And data is a great signal. If a place is trending and people are doing stuff, let's go talk to them. Is there something that they need? Is could we help them even get better?
Omoju Miller:And so we went to Nigeria. I planned this Nigeria trip. GitHub had already been going to Nigeria because it was already trending.
Bdougie:Yeah.
Omoju Miller:And then I planned the GitHub Nigeria trip for the CEO for and his team. And we went there and we met a bunch of developers. And the best part is a lot of those developers did not know that each other existed.
Bdougie:Yes. This is true. And this is how Asuka got started.
Omoju Miller:It's because they didn't know that collaboration. Existed. And so our event was one of the first time they all met each other in the country.
Bdougie:Yeah.
Omoju Miller:They didn't have to go somewhere outside to see each other, and that was a very powerful thing. And the crazy thing is we would go to, like, a cafe. I remember we got in, went to this cafe for espresso, and that was like, oh, who knew? Snap. Like, looking over someone's shoulder.
Omoju Miller:This person is writing code and pulling creating a pull request on GitHub.
Bdougie:In the cafe in Nigeria.
Omoju Miller:Right there. Yeah. So you just see things like that. Everybody is trying to be a developer of some kind or the other. People are using these tools.
Omoju Miller:It made a lot of sense to go over there and go talk to them and see what tools do they actually truly need. Is there anything that we're doing that is holding them back? And also, can they give us feedback on the paths we should take with the different kinds of products that we're going to launch? And one of the things that came out of that was our trending. Like, we need to improve this thing to be country specific.
Bdougie:Yeah. Yeah. It's amazing because even, like, the made in Nigeria, I forgot his name. Idowale. Adewale.
Bdougie:I've sorry. I'm, like, butchering your name. I'm
Omoju Miller:so sorry. Wale.
Bdougie:Yes. He built this made in Nigeria, which is all projects open source made in Nigeria so they could showcase stuff. And that was one of the projects that also started trending
Omoju Miller:Mhmm.
Bdougie:After it it shipped, like, 2019, 2020. And, the the one thing that I noticed because, we did work with the Black the Cat at GitHub. So on the the tails of this trip, we held
Omoju Miller:a a
Bdougie:hackathon, CodeNyja. We had Black the cats in country doing open source hackathon. So by having all these, like, sequitur events that happen, this connection, now you see, like, SHEDRACK. You see Sans and Gotti. You see a lot of folks, like They
Omoju Miller:blew it up.
Bdougie:Yeah. They're all coming, and they're participating, like, the Linux Foundation events and and projects. And now you have Chaos Africa, which now has a community across country, across continent, to be able to participate and have in country events but in multiple countries. And I got to see it firsthand of, like, I got to meet everybody. First time in Nigeria was couple months ago for me.
Bdougie:We got to also see that sort of momentum. But I think when you make that investment, you you can sort of it's not about reaping the benefits, but it's, like, sort of stepping back and watching Mhmm. Like, the impact that that makes. And, I don't know if you know about this, but, Figma also did a very similar tour.
Omoju Miller:Oh, no.
Bdougie:Yeah. So Dylan, the CEO of Figma, went to, Nigeria as well. And the result of that is there are so many new product designers
Omoju Miller:that
Bdougie:came out of that.
Omoju Miller:That is one of the things I've started to notice. I see this thing on Twitter all the time, on x all the time. I don't even know if it's true, but a lot of people truly believe that some of the best designers are in Nigeria.
Bdougie:Yeah. Wow.
Omoju Miller:They feel like they have a certain kind of flare that is better than anybody's eye. So people believe that. And maybe all of that is why Dylan went, or Or maybe that's because Dylan went, it can I don't know? But some designer stuff is happening down there.
Bdougie:But I will say that the product designers are are down there and growing up in this this ecosystem. They're using Figma. And I think the benefit of him being there and being talking to the folks and teaching them the basics of the tool, it isn't completely changed the generation. There's, like I wanna write a blog post around how developer tool startups, they all they do the wrong thing where they focus on the experts and get the convince the experts at the big fangs to use your product when the best thing you could do is go build college curriculum with your tool.
Omoju Miller:Oh, okay. Alright. So I'm taking notes.
Bdougie:I'm taking notes. Like, you build and I think we we had
Omoju Miller:a conversation about this. And I've been I've been, like, acting it out. It's true. But, you know, when people are sitting here in the valley, the average person you meet has, like, most of my friends have PhDs because I have PhD. Right?
Bdougie:That's pretty nice.
Omoju Miller:So no. No. But it gives you very it's not nice because you have a very skewed view of the world.
Bdougie:Yes. The correct.
Omoju Miller:Yeah. Your view of the world does not scale. Most human beings are not going to get a PhD. So if you listen to a bunch of research scientists doing AML as your feedback loop, you are toast. Those people who don't know how to make products, they're great researchers.
Omoju Miller:It's very, very different. So being intentional about truly creating a diverse ecosystem of spec of people around you and listening to them to build a product is so important. I'm gonna be going to Nigeria next week.
Bdougie:Okay. Wow.
Omoju Miller:And
Bdougie:you go back home or you you do work out there too as well?
Omoju Miller:So I wanna work from Nigeria. I haven't been since the pandemic. I my plan was always to go back every year and maybe even go for, like, 8 months, 12 weeks, like, do, like, long stints of work, but the pandemic put a kibosh on that. And now that everything is open back up, I wanna start going again, but I wanna do, like, this first pass of what is ecosystem like? Because one of the challenges that I noticed is a lot of people have moved out.
Bdougie:Yeah. Well, Well, that's a challenge. I we actually I had this conversation a lot with folks because when you get to learn the Figmas and you learn to GitHub and you get the connection to ecosystem, like, Prosper being one of them, like, he actually has a place now outside Nigeria. So, like and that's where he spends the most of the time with his family where he can grow and be established because there's there is a disconnect. There's, like, almost 2 different economies Yep.
Bdougie:Building right now in Nigeria where you have folks working for Western companies, and then you have folks who work in Nigeria. Yeah. And there's a huge disconnect between them.
Omoju Miller:Disconnect. So I wanna go and see what that looks like now. And just to and I think Nigeria I think of Nigeria as a place that sharpens your knives. You know? Because you go there, you come back, you're just sharper.
Omoju Miller:Your brain it just, like you need a fine tuning. And I think Lagos, for me, is my fine tuning. It just helps to, like, reset my brain. Like, don't take anything for granted, hit the ground running, be smarter, be it just makes me better.
Bdougie:Yeah. A 100%. And, like, so you were talking about you you have a lot of friends who are doing PhDs, and sometimes you it's it could be a detriment, because everyone now is at a level that she can't understand to sort of beginner mindset or sort of introductory mindset. Do you think that was the issue until OpenAI for ML and and AI in particular? Because for me, nothing made sense to tell Copilot when it came
Omoju Miller:to AI.
Bdougie:I was like, oh, wow. This actually really cool. What else can we do? And I expanded my pilot to go try all these new tools and and try to mess with ML data models. Models.
Bdougie:But I had no interest prior because I felt like it was just always above, like, above my weight. Mhmm. So did OpenAI basically just sort of break that chasm?
Omoju Miller:I think they did. They truly democratized access to AI for builders. Yeah. And I think people OpenAI is not paying me. I think people truly vastly underestimate what has actually happened.
Bdougie:Yeah.
Omoju Miller:Because now prompt engineering, people are, it's not engineering. But you can go and write sentences, queries in natural language, not in SQL, in natural language. You could write what you want your responses to be.
Bdougie:Yeah.
Omoju Miller:You could fit that to something. And if you don't know how to fit into something, they will show you how to fit it into something. If it can do a 100 of these, you can build a model that can then do something. So the barrier to entry now is the level of literacy. Yeah.
Omoju Miller:If you can read and you can use a computer, you too can build a very sophisticated, state of the art ML model. That is life changing, and I think people just don't understand how powerful that is. You don't have to have a PhD. You don't even have to be technical. You have to be a human being, which already comes with preloaded problem solving capacity.
Omoju Miller:You just have to think up the thing and go do it.
Bdougie:Yeah. And it's like the the concept that I've been learning with this fine tuning that OpenAI has been been working on. I know other folks are building this as well and there's other frameworks and tools, but, I mean, my gateway right now is OpenAI until it's it's not. But there's a the the actual consumer prompt, but chat prompt, but then you have the system prompt. Yeah.
Bdougie:So the system prompt is, like, the idea of, I want you to give me links of every restaurant that has a Michelin Star in San Francisco, but I also want phone numbers. I want linkable, like, clickable and a markdown table, that I could just, like, quickly ship over to my EA and say, okay. We need the here's our bookable restaurants. Like, that system prompt of, like, how I want them to be linked, how I wanna be structured, or what charts I want is absolutely magical. Because, like, the thing that we're doing right now is kind of like what I saw at GitHub happening, like, with the trending and seeing that that sort of process.
Bdougie:And, GitHub is kinda like refocusing efforts, and they have they have their their road map. But what I did is I taken Open Sauced outside to then do that trending, but expand it quite a bit. Because, like, we we built this thing, on the DevRel team, so we're 2 people for the longest time. And Don Goodman Wilson, built this thing called Cerebro, And it was basically Looker on top of GitHub's data.
Omoju Miller:Mhmm.
Bdougie:And we would go into places like Brazil, which is another emerging market for for GitHub, and identify top contributors in JavaScript, Python, etcetera. And then every time we did an event down there, we'd go reach out to those folks and say, hey. We're GitHub. We know you're doing great work because we have the data. Let's have a coffee chat.
Bdougie:Let's do a meetup and a talk. And that was extremely powerful
Omoju Miller:So that, being able to now do that, building Cerebro in 20 minutes.
Bdougie:Yes.
Omoju Miller:I think that is the that is the change. It's not I think it's so powerful because it's not just a consumer tool. You could use that as a consumer tool. But it's also a developer tool. And it happens to be a very good developer tool because the documentation is great.
Bdougie:Yeah. So I
Omoju Miller:think for all of us who are in the developer tooling market, you should actually study the go to market of OpenAI Because the developer tools are superlative. They never scare you. You always feel empowered that you could do something. And they give you so many examples. I think it's a I think it's a shift.
Omoju Miller:How it shifts, I don't know. And then but the interesting thing is the definition of who's the developer also changes.
Bdougie:Yes.
Omoju Miller:And have you started seeing that in the data that you're looking at? Because does a developer actually have to be truly, truly super technical?
Bdougie:Yeah. I mean, my answer, no. Yeah.
Omoju Miller:They just have to build stuff that works using software. And what if now all the building does happens often mostly natural language?
Bdougie:Yeah. And this is something that I I what I really enjoyed when working under Nat when he was CEO is, like, we did reframe what the developer is. And as we started count, like, over a 100,000,000 developers on platform. But before I learned how to code professionally in 2013, I was writing scripts and doing jQuery. Was I a developer?
Bdougie:Probably. But in my mind, pre 2013 when I was doing sales and had a finance degree, I was like, oh, no. I just do this on the side. And, like, I I run scripts to play my first person shooters and mod the game and stuff like that. But, like, you're manipulating and developing some software to do win in a game.
Bdougie:Like, you're a developer.
Omoju Miller:You're a developer.
Bdougie:Or if you're a designer using Figma to convert CSS, like, you're a developer. So, like, the the scope has brought in. And I think where I there was always a sort of push and pull on, like, Stack Overflow's 15,000,000 developers versus, at the time, GitHub's 50. It's like, Stack Overflow thought a developer was this.
Omoju Miller:Very specific. Probably somebody that earned a living with a title called software engineer that worked for a company. That's not that that is a very specific idea of what a developer is. That's not what a developer is.
Bdougie:Yeah. It it it it's the scope has broadened so much more. And now as we're introducing more prompt engineers, like, it is a gateway into, well, what if I can go one layer deep? What if I knew what the Python was that was outputting this and take this back to the product I'm trying to build?
Omoju Miller:Yeah. And I think that ad is actually an amazing thing. It is, it's not very often that you get to live in a world where your dreams come true. And for me, I feel like I'm living in a world where my dreams are coming true because my dream ultimately has been, let us drastically reduce the barrier to entry to technical work. And we're seeing it happening in real time.
Bdougie:Yeah. Yeah. So I I going back to the the the thing you you said, which is those who have the biggest dataset in ML are the ones that win, and GitHub being the one that has the biggest dataset. So I'm I'm curious what are other datasets that folks are, at least in this sort of developer tool ecosystem that we're looking at, that we might be also paying attention to pretty shortly?
Omoju Miller:I think it's actually kinda very hard to find one outside that is that is a language. What what you said about what developers do and all of that, what about all those things? Who's the best person? Who's the designer? So I think the scope of what you should look at is what are all the expertise roles needed to ship a product?
Omoju Miller:Yeah. People who build, like, development is one kind of thing. It's not enough to ship a product. You need all these different, You need a designer. You need so where do those people sit?
Omoju Miller:What where where can you find the best? Where can you curate? I don't know. And then and and and it's because of, are they doing artifact management? Is there something like version control for images?
Bdougie:Yeah. Yeah. I mean, they're I all I know is ImageMagick, which is not version control, but it's more of a manipulation of images. But I think there's yeah. I mean, there's a lot of opportunity there.
Bdougie:Because, like, you see, what is it? Midjourney Mhmm. Is, like, the tool we use for all of our blog posts. We generate a Midjourney image. It goes as a social card.
Bdougie:And, like, we now have a dataset of, like, what what prompts have we used previously to get what we're doing for the next image, for the next blog post. And, like, our brain is all word-of-mouth. Like, how I learn how to code. It's like I asked the person next to me, what did you just run there? Tell me that.
Bdougie:Like, whenever you do a prompt in Midjourney, tell me that prompt, or let's share these prompts at a place that it's repeatable. And I I think there's, like, this world that we're sort of, like, building where we're building, like, this bigger brain Yeah. Or knowledge. And I think there's, like, I I'm curious your thought on this. The opposite opposite end of the spectrum where Twitter and Reddit have now locked down their dataset because of the sort of consumption of all these new LOMs that are they're working.
Bdougie:So do you see, like, the a push and pull of, like, private data? Now we have our AI thing. Stack Overflow being another one that just launched their well, they actually haven't launched. They just announced their Overflow AI.
Omoju Miller:I think everybody is, realizing that, oh my god, data is truly king.
Bdougie:Yeah.
Omoju Miller:And maybe I've been giving my data away for free, and so I should lock it down. And then I should be the only one to benefit from that dataset.
Bdougie:Yeah.
Omoju Miller:It's a rational thought. I think it's very shortsighted.
Bdougie:Really?
Omoju Miller:I think it's very shortsighted, because the best comes when you bring all of it together.
Bdougie:Yeah.
Omoju Miller:So if you just build, like, something that only works for Stack Overflow or only works for Twitter, now you're getting to this very specialty based thing. And maybe you make money and then you're good, but you don't invent the future. The future is gonna come from something that is way more interdisciplinary. And that's gonna so maybe they put a license on it and you can get access to that data. People need a way to get access to that data.
Omoju Miller:If what they're talking about is I need to figure out a way to benefit from giving you access to my data, then fine. But literally locking it down that only I get to use it is naive.
Bdougie:Yeah. Yeah. It's it's and I I guess I want I'm curious, like so we have OpenAI, where we we spend a lot of this conversation on. But then Facebook or Meta
Omoju Miller:Meta has to say.
Bdougie:The the llama v 2, which is open source. It's licensed. Interesting. But it's still the community can train it. Mhmm.
Bdougie:And we all can learn from using the LAMA version. So does does Meta win?
Omoju Miller:Meta is gonna win in the long run because number 1, we're all using this chat chippity, and everyone's liking it, and you're happy with it. Most enterprises so, like, like, let's take the usual average consumer out of the picture. Enterprise level consumers, companies who need to use this thing for something way more sophisticated. Many companies have challenges around using proprietary models. They are not as trustworthy.
Omoju Miller:They don't feel safe. They wanna use something that they can train themselves. And if they wanna use something they can train themselves, the only thing they can use that they can train themselves is an open source model.
Bdougie:Yeah.
Omoju Miller:So they want an open source model that they can train, that they have the they they own that model. They can do this and that. It's part of their intellectual property. So open source models are not going away. They are so necessary in the context of enterprise.
Omoju Miller:Average human being, then you use something that is commercially available. But, like, a huge corporation, they're probably not gonna use this thing. They they will want their own version of its maybe it's on prem. They want their own in a private cloud. They want it to be they literally want guardrails over everything.
Bdougie:Yeah.
Omoju Miller:So if OpenAI adopts a position where they do that, then maybe they can win. But you're going to need to be like, this is completely mine.
Bdougie:Yeah. Yeah. And I see I see a world where OpenAI does expose a bit of maybe out of date models to the community to see what to to bring their version up to the next. Yeah. I don't I don't know if they if they win in a world where they they're a closed source forever, But I do, I do see a space where
Omoju Miller:But if they if they overtake it if their role is now we're just a better version of search, we're a better version of doing some, software, this and that. They will then form a market for themselves.
Bdougie:Yeah. Yeah. They'll they'll for sure figure it out. And but the thing is, like, we're GitHub 1. So, like, defunct and, and cofounders of GitHub, they took an existing protocol, which is Git, and they built an interaction around that.
Omoju Miller:Mhmm.
Bdougie:Well, we all had, like, a centralized place of how we do version control, and we have a, like, a way we could take that step to okay. Now we can collaborate and have, like, this social coding, social graph experience. So I guess what I'm getting at is, like, when I look back at GitHub and, like, what were the the sort of pillars that or they're even the stepping stones that made them be so successful and have the biggest dataset, but also look at SourceForge, which was, like, the exact opposite where it was, like, way more closed and, like, a different business model that didn't quite scale as the world shifted away from ads. Mhmm. I think it's it's the same thing for for Meta where Meta is willing to take chances.
Bdougie:Like, their and their their CEO is is fighting other CEOs. So, like like, literally, physically fight Mhmm. Other CEOs. But
Omoju Miller:And then also their CEO is a technical CEO.
Bdougie:Yeah.
Omoju Miller:He's a technical founder CEO. So the same person that founded the company is still staring the company Yeah. And that person is very technical.
Bdougie:Yeah.
Omoju Miller:And that person is also very business savvy. So they can learn from the history of the past. Yeah. Open always wins. You just need to give people a collaboration layer and a way of doing discovery.
Bdougie:Yeah.
Omoju Miller:And then let's write better licenses over the things we create. So that there's a commercial application for it. You can still use it for commercial application, but figure out some other kind of thing that you need to put around it.
Bdougie:Yeah. So so in the area that you're that Fimeo is in around the data layer, and, like, we're now seeing a world where, like, there's, like, the democratization of data. Mhmm. But then there's, like, the data that you own Mhmm. And protect, like, your sort of personal identifiable information.
Bdougie:Is there a world where I'd I I'm signing up? I I guess this was like the blockchain where or Web 3, you sign up, you're part of the ecosystem, but you're able to take your keys home.
Omoju Miller:There is. And this is why we started off in web 3. So what really, really interested me in even founding Fimeo was open source. Yeah. That's the thing that I really get excited about.
Omoju Miller:And one of the things I realized that we needed more in open source was a way of actually doing coordination around organization. Who gets to do what? Like, just who is the maintainer of this thing? Is this person compensated? Are they okay?
Omoju Miller:Are they happy? Do they have like extra people that can help them? If I'm really using this, I don't wanna wake up one day and then this person has decided, you know what? I quit. I'm tired.
Omoju Miller:I don't wanna do this anymore. And so then for me, Web 3 just was like, oh my god. This is the next evolution of open source because now you could have a a DAO, a distributed autonomous organization, an organizational framework that you can then slap on top of a project. And so you can then have some kind of, at least, governance around this. Like, who owns what?
Omoju Miller:Who gets to triage an issue? Who gets to review this? You can codify it in a way and then have it run through code. Have the thing and then make sure that the things that you all decide is immutable and it's all transparent. So inspiration less is immutable, it's transparent.
Omoju Miller:That idea got hijacked by scammers for coins and stuff like that. And so when people think about it, all they think about is, like, scammy coins. But the infrastructure and the governance mechanism is very, very powerful. And so I'm beginning to start thinking about all of this stuff is we need a way, and this is what Fermilio is gonna be doing, to bring that to data because it's about you need some kind of fingerprinting of some kind of trace of ownership in a very complex ecosystem. A blockchain allows you to do that.
Omoju Miller:If the blockchain is open and it's immutable, you can do that, But you need something that is privacy preserving. Maybe you don't want everybody to see. But these are very sophisticated ideas, and we wanna play around with it to see if they scale.
Bdougie:Yeah.
Omoju Miller:Do they actually truly scale? Or are they gonna fall apart at the end?
Bdougie:Yeah. And this is something that I I don't know if it's similar to, like, what we're seeing with this, Blue Sky, and now, Jack Dorsey's, like, Web 5
Omoju Miller:Mhmm.
Bdougie:Block thing that they're working on. I'll let you know from Angie Jones. She's done some couple presentations, and I've watched a couple of demos of, like, you bit you when I explained, you get to take your keys if you wanna move them here or if you move Yeah. There. Like, you're able to, like, remove yourself from the system if you choose, but you could also be a part of the system and the sort of the borg of the learning.
Omoju Miller:So it's it's like that. And we already, to a certain extent, use that because Git is open. Yeah. Git is a tree. It is a ledger.
Omoju Miller:It is a tree. And it's a historical record of everything. So in a way, it's kind of immutable because you can see the trace of everything. So it doesn't mean that you can't roll something back, but it's kinda hard to tear that tree apart. You can see everything that has occurred in that tree and then it scales.
Omoju Miller:And if you decide you don't actually wanna use GitHub as the coordination lay and all that, you could still use Git. We could just spin up your own server and do it. The average person does not. But if you wanted to, you could just take your stuff and then go spin up your own server and then build all the tools around it yourself. Blue Sky and all those things, they're based on protocols.
Omoju Miller:They're based on these things, and you can just take your keys with you. If you don't wanna sign into this server, you can sign into another server and all your followers and everything goes with you. The challenge is key management. The average person doesn't really understand encryption and how do I keep my private keys and all this stuff. So we have to build better UI UX around that to make it trivial.
Bdougie:Yeah.
Omoju Miller:If you're a developer who does back end development, you do this all the time. We SSH into a EC 2 instance on Amazon. The only way that it knows the results because you need trust. Yeah. We have a PEM file.
Omoju Miller:It has a public private key. We do the handshake. Oh, yes. It's actually you. You can but we need to do that in a way that is so trivial
Bdougie:Yeah.
Omoju Miller:That nobody has to be having RSA encrypted. Like, you don't need to even know. It just needs to work. And we're gonna have to figure out how to do that very, very well around data and to have this fingerprint along with the data so that wherever the data goes, you know that this is actually your data, my data, their data so that we can figure out how to do profit sharing. Because ultimately, that's gonna be what it comes down to.
Bdougie:Yeah. Yeah. And this is something that, so, like, the GitHub CLI when I was, being developed and launched, there's, one command which is gh auth, and it does that handshake for you. It's not perfect, and sometimes it gets it wrong. You don't have to override files.
Bdougie:But, like, to save your at least your SSH key to be able to push up to the the Git server, that is trivial. And for and for years, it was always a thing I had to go Google or look at
Omoju Miller:the box. Again?
Bdougie:And I it it is sort of like developers can figure this out and understand how to go the way to it. The consumer or the average consumer is not going to
Omoju Miller:So we have to bring encryption tooling down to the consumer level. It already exists. We know it exists. We use it all the time. It works.
Omoju Miller:But it's a developer tool, and it's not even a great developer tool. Because everybody still has to go, like there's always a fair, like, is this thing the public one or private one that I just exposed? There's some anxiety, like, oh, I can't remember. They have to, like, go back to your notes and be like, okay. I did this.
Omoju Miller:I did that. We need to bring that down to the consumer level. So once we do that, bring it down to the consumer level. And there are many people who are probably working in this already, I'm I'm assuming, I hope. And if not, well, we will.
Omoju Miller:And but once you bring that down, then how do you tie that to your data? And what does it mean to actually be your data? Is it a data that you have to do we put something on top of your browser that what before you even log in, you have to do the handshake so that everything that flows out has your fingerprint? So these are the things I'm beginning to think about in a very clear way of how do how does this become a product that is not invasive that everybody can use?
Bdougie:Yeah. Yeah. And I was gonna, I was gonna make a tongue in cheek joke when I missed the opportunity where when I used to explain GitHub pre Microsoft, I used to joke that we're a collaboration tool built in an open source, blockchain protocol, which was Git. Yeah. So the there's a lot of nuance within Git and a lot of, like, that similarity to what the blockchain is or what blockchains are, that I think gets missed a lot of times.
Omoju Miller:Oh, it's so missed. I gave a talk at GoToChicago, where I actually, like, tore it apart, about what does what's the difference? It's like 5 separate things that it must have. Immutable, distributed, permissionless. Because you you could it may not be distributed.
Omoju Miller:It it has to be permissionless. So there are, like, all these different things that must exist. And when all of those things can exist in tandem, then you have a system that is truly, like, trust certified and you can't pretend to be not who you are. Because at the end of the day, it's about trust. We need to figure out how to solve trust at scale.
Omoju Miller:And today, the way we do it is your browser does it for you. It tells you, like, oh, the certificate. This is we still issue certificates, but the average consumer does not know anything about a certificate. But there's a certificate on every computer, and it's stored in somewhere. And then when you log in, the browser does, ah, the certificate for this website.
Omoju Miller:That this this is, it's not certified, and it gives you that warning. It's like, don't don't don't move forward. I think they may be scamming you. So we do it all the time, but we need to do it in a way around data. All this tooling that we're talking about was not designed for data transmission, like, for data as an asset.
Omoju Miller:Yeah. It was just like communication protocols or or something. But no. No. Now we wanna do it around, like, literally data as an asset.
Omoju Miller:And that's why I'm interested in the whole idea of I'm thinking about maybe a very good use case or a place to play around with is product designers because they actually have data artifacts.
Bdougie:Okay. But what do you mean by that?
Omoju Miller:Their job is to create little bits and pictures of things. So and at least the human can recognize. Yeah. You can see the thing because data is such an ephemeral idea. It's hard to figure out what what is data.
Bdougie:Yeah. This is this is fascinating. There's a lot there's a lot. I'm we're excited to see what you what comes out of Vimeo, because I think you're onto something. And as far as, like, these datasets and for folks who who owns the data is king at this point.
Bdougie:And, I think a lot of companies and incumbents are are aware well aware of that. And I see, like, where years ago, we saw, like, Foursquare where you check into a coffee shop, and then that data was then added to the the sort of Foursquare, which I think is Foursquare still around?
Omoju Miller:Yes. It is. I am a loyal user. It has broken up into 2 separate things. Okay.
Omoju Miller:They're not paying me. They're not paying me. It's Foursquare and Swarm. Yes. And I still use it religiously, especially when I travel because I'm tuning it to my taste.
Omoju Miller:So if I land in some country I do not know, it already has my history. I always need to know where am I going for coffee and what restaurants am I going to. I want somebody to solve that for me, and I I use it for that.
Bdougie:Wow. And that's I mean, that data layer that's obviously, people are still using
Omoju Miller:Mhmm.
Bdougie:Is super valuable, especially country by country, city by city, but bought business to business. And I think the one thing I'm super fascinated about this new trend in AI is, like, we're actually learning how to classify interactions Yes. Recommendations. So, like, the one thing I did wanna talk about in this this sort of trend the last 10 years, we saw where it was all user generated data, and then we had algorithmic data. So, like, things like the the home feed on on x.
Bdougie:And then now we're moving into AI generated data Mhmm. Where we had to do the the the steps before to get where we are today. And we could always go back to the cycle of, like, okay, user generated, AI generated, or, sorry, algorithmic generated, but then AI that feeds to everything.
Omoju Miller:Mhmm.
Bdougie:So do you feel like we're are we have we hit the sort of, what's the not synergy, but, like, singularity?
Omoju Miller:No. No. We have not. So right now in the MLAI world, everybody's excited about foundational models. Yeah.
Omoju Miller:The foundational models itself and their capabilities. Oh, this model has a 1,000,000,000 parameters. It can do this. It can do that. Yes.
Omoju Miller:It is. It is very powerful. But for that thing to truly become effective, it needs to be put in applications.
Bdougie:Yeah. Truly
Omoju Miller:cutting edge applications are few and far between. What I see is copycats. Everybody has their agent, their little AI assistants. AI assistants is just one avenue of this. So I don't think we've not even reached the singularity at at all.
Omoju Miller:Like, we're not even nowhere near it. And then what we're talking about is only in the realm of language, and that's it. Only in the realm of language. It's not even fully giving you full on reasoning capacity. It's just language and it's very good at problem solving, but it's language.
Omoju Miller:Most of human intelligence is not limited to language.
Bdougie:Oh.
Omoju Miller:Vision. What about vision? What about vision is a hard one. We still don't have full self driving cars. Like, full self driving.
Omoju Miller:It's very, very complex problem, and it's not about language. It's about problem solving a complex world, dynamic changing adaptive world. It has nothing to do with me talking to the car and this and that. Humans do so much in the subconscious. And that's because we do so much subconsciously, Most of us are not aware of how truly staggeringly powerful human problem solving is.
Bdougie:Yeah.
Omoju Miller:How many people I am certain there have been so many humans who have had experiences of maybe something hijacks your consciousness. And you're not drunk, but maybe something exciting or something just hijacks your consciousness. And you find yourself, you've driven home, and you can't remember how. And you're just like, oh, snap. I wasn't even thinking about but somehow, you almost you drove home on autopilot while your mind was thinking about something else.
Omoju Miller:It even, like that is hard. Yeah. We these systems can't do that.
Bdougie:Yeah. Yeah. And there's still a ton of work to do.
Omoju Miller:There's so much work to do. I'm very excited about all the things we've hacked with, speech and language.
Bdougie:Yeah.
Omoju Miller:But there's a whole world of all the things that need to be done.
Bdougie:Yeah. Yeah. I'm looking forward to the stuff that folks are doing in computer vision and, like, things like video. Like, not just generating video, but reading and understanding and comp comprehending. So, I know a couple of folks I've chatted with around the Open Source ecosystem that are working on that in particular.
Bdougie:And, yeah, I I think there's just a lot that's gonna go on this this sort of area of the industry, and I don't think we're we're nowhere even close to being done.
Omoju Miller:Oh, nowhere. Nowhere. There's still a lot more work. We're even fully mapping out the pathways that humans human themselves. Like, you know, how I eat this, and it affects this, and it affects that.
Omoju Miller:All of those are, like we're not we don't know any at that.
Bdougie:Cool. Well, Emonju, thanks so much for the conversation. I'm, like, super excited about everything you're working on in this whole arena. So I'm looking forward to continuing the conversation offline. And, I'll have to I'll have to share some stuff that we're working on open sauce too as well because I think, I I would love your feedback on it.
Omoju Miller:Alright. Thank you for inviting me, and please stay saucy.
Bdougie:I think we should have done that. Secret sauce of the podcast produced in house by Open Sauced, the open source intelligence platform providing insights by the slice.
Omoju Miller:If you're
Bdougie:in San Francisco and interested in being a guest on the show, find us on Twitter at saucedopen, and don't forget to check out open sauce at opensaucedot pizza.