The Evolution of Developer Relations in the AI Era with Rex St. John

Sheikh Shuvo: Hi, everyone. I'm Sheikh. Welcome back to Humans of AI. This is a very special episode where we dive into the world of developer relations. Broadly defined, developer relations is the umbrella term for how technology creators interact with and build communities with end users. Right now, we're seeing exponential growth in the number of AI frameworks, ecosystems, and tools. Fostering and managing developer communities has never been more important. Our guest today is Rex St. John, founder of Taroko Technology, someone who's been working on developer communities long before it was cool. Thanks for joining today.

Rex St. John: Rex. Really appreciate it. Glad to be here.

Sheikh Shuvo: Can you give us an overview of how you got where you are and what exactly developer relations means to you?

Rex St. John: My background is I originally started off with a degree in marketing and then during the 2007 meltdown, I was at this startup doing cryptocurrencies, microtransactions, early form of it. Every single engineer got a job in a week, even at the worst point of the recession. At that point, I decided I was going to learn to code and then I got into consulting, global application, full stack frameworks, and then ultimately technical evangelism, building developer relations programs at Intel, ARM, NVIDIA, most recently Saga, on the web three side. I've got like a cross-section of AI, IoT, edge computing, software, everything. So that's how I got here. Along the way, I have a lot of thoughts about where things might be heading on the AI side.

Sheikh Shuvo: Can you give a couple of examples of what makes a great developer relations community?

Rex St. John: On the community side, I think the biggest problem people have is they stand up a community and all the energy dissipates and dies away because everybody's launching a new community every other day. The second somebody raises funding, the first thing they do is stand up their Discord and Telegram developer portal. Developers might originally go in there and be excited, but if you're not able to keep up a cadence of releasing daily, hourly, minutely, having constant value, people will wander off and that's really hard to keep the momentum going. If you're a startup, you probably don't want to take that on necessarily, unless you have the ability to service and create energy on a daily basis. People need to be very careful about how they anchor their community. In many cases, people are better taking a step back away from their own product and solution and saying, what's the higher good here, and saying, let's find some friends that share this common purpose and see if we can't anchor in something that we collectively can deliver value on, on a daily basis. So kind of like divide the load a little bit. I think that's a better move for most startups and early phase companies than trying to maintain a Discord where people will just wander off almost guaranteed. A quick example might be something like the Refi DAO where they based it on Gitcoin fundraising. They based it on networking. They have a wide range of different branded activities across different time dimensions. They have a lot of partners and it's organized on a broader theme versus just one product in the Refi community. I always really liked and respected how the Refi DAO did it. It's a smaller community, but they keep people engaged on a daily basis. And it's because they're rooted in a common purpose versus a single product in the ecosystem. I think the common purpose approach is a better way to anchor a community.

Sheikh Shuvo: Makes a lot of sense. With your experiences of seeing developer relations evolve over the years across multiple types of companies, looking at the landscape now, how is developer relations and how you manage these communities changing with the prevalence of AI and spike in different technology companies creating these?

Rex St. John: So, let me just recap where things came from when I got started. During the 2007 era, when I became a programmer, there was a really amazing entry point. If you were willing to put in the work, open up a terminal, test out Ruby on Rails. There were a lot of really creative and entertaining Ruby books. Node was coming along, and the iPhone had just launched. There was this really nice explosion in developer experiences, and the accessibility of being a developer, especially a full stack front-end developer, was better than ever.

I remember very distinctly, I was in Capitol Hill teaching myself to program here in Seattle. I joined an agency called Substantial, who's still around. They were willing to take someone like me who was just sitting in a coffee shop, teaching myself to code, incubate me for a little bit, get me productive, put me with some other engineers, and I could become a developer. And that's exactly what I did.

I think this era from 2007 to 2023 is over now, and it was characterized by this interplay between people doing these full stack cross-platform solutions like React Native versus the specialist developers, mobile iOS, Android, Rust, etc. I think something has shifted, in my opinion, where I don't even know if what we're going to have is, I don't even know if you could call it developer relations anymore. I think we're going to get something new.

During this era, the primary blockers that I faced as a developer, like I can't speak for everybody, but you know, if I specialized in iOS, a lot of people wanted to specialize in a particular framework because you got paid more.

But at the same time, while you were specializing, it seemed like all these platforms and startups came along that were incentivized to try and attack the value of your specialty. So like React Native came along and they said, well, you don't need a native iOS developer who's a specialist. They're too expensive and hard to hire. Why don't you just hire web developers and they've got React Native and they'll do your mobile development and they'll do it cross-platform. So it's kind of like an interplay between the specialists and the cross-platform people and the web people, just kind of went back and forth. Web was very unstable. It continues to be very unstable. Everybody's complaining about how hard it is to choose how to go about building a web front end, multiple device targets. I mean, you have hundreds of Android devices with different screen sizes, plus iPads and everything and different and like figuring out how to handle that on a native and web-based, you know, like that's been a huge annoyance. It continues to be. And workspace fragmentation, I think this isn't even one that I was really cognizant of until I learned about Repl.it and GitHub workspaces. It takes hours for a developer to set up, set up and configure their workspace on a weekly basis in some cases. A new developer might take a long time to get up and running as part of a team. They have to install all kinds of stuff. And workspaces across developers on the same team might be wildly different. So Repl.it kind of came along trying to solve this problem.

So I think like this kind of era of developer relations, I don't know if this is the future anymore and it was pretty typical what I did, you know, I'd done this a dozen times at global scale for APIs at Mashery, Web3, Asaga, you know, Silicon, Edge, Edge AIoT at ARM, Intel and NVIDIA. And as you throw up your documentation portal, you've got some code samples. You do, you know, have the ADHD Asperger's developer relations person like me go and do lots of talks, maybe some developer influencer programs. And it was very technical in nature. And I don't know if this is, um, I just don't like, I just spent a lot of time thinking about this after seeing, you know, Amazon CodeWhisperer, GitHub Copilot, Repl.it is now giving away. Uh, AI tools for free, basically, and I, and I just playing around with Stable Diffusion. I don't know that we're, I don't even know if I would, I want to call it developer relations anymore. Uh, I want to call it creator relations because. These tools allow the concept of a full stack developer to just expand dramatically. It's like I could be a full stack developer plus a business person and artist. Um, so I think I, my theory is kind of developer relations becomes creator relations or maybe ecosystem relations or something like that. Um, because like the original division of skills for developers. I mean, this is probably not the most accurate thing in the world, but it was like, you had the full stack generalist doing the front end, back end.

You know, on the left, you had the full stack generalists doing the front end, back end, and then you had the specialist developers like game engine, cybersecurity, embedded systems, kernels, robotics. And now I think this model, which, you know, this is loosely how I would think between 2007 and 2023, I want to displace it with this model, which is a full stack creator. And this is a designer, artist, musician, movie producer, some type of creative professional. And this person has been like fed steroids. It's like, you know, you can do prompt engineering, low code systems. You have all these generative AI tools to do videos, open source AI. And it's like, these people can just create a million times more stuff than they could previously, at a million times better quality. And that's only going to increase.

And then in the middle, a full stack business creator, it's like I was doing a research report on cybersecurity with ChatGPT 4.0 with being enabled, I could just do web scraping and it just like, I was asking it. Give me the top 50 cybersecurity startups, rank them by fundraising and put links to their websites and do a short comment. And it just did it. I'm like, well, as a business person, that makes me a million times more productive.

And then you have a technical creator and I think this persona maybe is what developers turn into where it's like, you still look at the code, you still need to know how the code works. Because a lot of times I like, here's an example. It's like, no matter how good AI is, if the AI takes a look at your chest x-ray and says, pretty sure you have cancer that, that, you know, we give you six months and that's going to cost you 2 million. I don't think any humans can accept that without a second opinion, you know? And the same is true of code. It's like, you can AI generate all kinds of stuff, but someone's gonna have to check it at some point and that that person that checks it might be the end user and they have to report it to someone that says, Hey, there's something messed up here.

So I've got this technical creator and I think this technical creator might take the place of the developers and maybe developer relations people, ecosystem people. Take over these different personas. I've kind of talked a lot. Do you have any comments about this?

Sheikh Shuvo: I love the term creator relations there. I think it's very accurate and almost makes it a lot more accessible to with diving into that term a bit more, looking at the full cycle of creating. Obviously, there are many ways that different industries do this looking across the entertainment industry in particular. Are there any things from other industries and how sort of talent is managed and encouraged that you'd love to see included in a technical ecosystem?

Rex St. John: I think you're pulling on a thread that is just huge because these new personas, as in the 2007 era when coding bootcamps became a thing, you saw a lot of people like lawyers and doctors, or Wall Street finance people who were like, "This is awful. I want to be a programmer," and they would go through a bootcamp, learn to code, and then they would come out as this new type of worker, which is like this full stack developer. And a lot of people went through that. I did that. My background was in marketing. I was like, "I don't like marketing. I can get paid more to be a developer. I'm going to do that." So I think we're in the exact same kind of period right now as that 2007 to 2010 period, which is the learn to code era when that was so leveraged. I mean, when I learned to code, I was able to increase my salary like 30 percent a year for year over year because it just kept being a great skill.

I think we're in the learn to create era where it's like, if I'm a company or an agency, or I'm running a typical Web 2 era company, I'm looking at my entire workforce, let's say I've got 130 people, and I'm going, "The profile of these people is probably going to need to be different, like the people I want, if I've got 30 people, 10 people, five people." And I think there's a new profile emerging, which is kind of like this. It's like the full stack creator, either with some business or technical or art skills, or music skills, and I think companies are going to need to radically, like the companies that are successful, are going to be staffing a radically different profile. And those people are going to have to go through some kind of training, I would think, because there's hundreds of tools. Like I've got lists I was looking at with thousands of tools. A lot of them are junk. Some of them are good. Some of them are worth learning. And if I'm a company over the next three years, trying to plan, I'm like, "My whole workforce has got to get upgraded."

So I don't know the answer, but I think everyone's going to go through that pretty soon.

Sheikh Shuvo: That's fascinating. And as it becomes easier to become a great developer, what do you think changes about how that developer stands out? Is there a change in what's posted on GitHub repos, what portfolios look like, how job interviews occur? Any thoughts on that?

Rex St. John: That's such an interesting point. When I look at Hugging Face and Civitate, or Civit AI—I'm probably going to mangle this here—let me try and pull this up. So, I've been spending some time with running stable diffusion locally on my Mac. And when I look at Civit AI, this is Hugging Face, and they're like GitHub for artists. In my opinion, they're holding repositories of these models. I go on there and I can download some asset. This is a Laura, which is like, I can't remember what a Laura is exactly, it's like a thing you can add to stable diffusion to add a different styling to how things look. So there are whole repositories of these things and they're all files to download, and they all have versions. You can hover and see how this was constructed. This is not that far from almost like a Ruby on Rails manifest file, where you'll see this LoRa specified, and this just reminds me of Ruby on Rails, where you'd specify a gem. You'd say, "Oh, my Ruby on Rails project depends on this login gem." So, you're seeing artists, the workflow for artists is more and more resembling Ruby on Rails. In my opinion, these manifest files are going to, I think you're going to see a replication of what happened with Node.js where you've got a whole bunch of Node.js packages in a repository and you have a manifest file and you specify all this stuff. Someone is probably going to build a Ruby on Rails framework for this. They probably already have. I'm not aware of it because you're going to want to version this and pull in different packages and then refine it. This is primitive to where I think it's going to go. But I think being an artist is going to look like Ruby on Rails pretty soon.

Sheikh Shuvo: Going back a bit, you mentioned the prevalence of different types of tools and wanting to potentially retrain employees if your company wants to keep up. Do you have a list of tools that you'd recommend people invest their time to learn?

Rex St. John: I'm in a very heavy research phase. The more I look at this space, the more astonished I am. Rewinding your question, if I'm a knowledge worker right now, what is the single most productive use of my time? I don't think the best use of time is to roll up your sleeves and go download Stable Fusion and start pushing to their GitHub or becoming or getting a PhD in AI. By the time you've done that, it's like three years from now before you ship anything.

I think now's the time to just be in hyper research mode and just be curious. Go and follow a bunch of people on Twitter that are posting this stuff. Spend 30 minutes a day just lightly looking at this stuff and rebuild your Twitter feed, your LinkedIn feed, follow the right people and just let them inform you because you're going to need constant training on this just to stay somewhat informed. So, I think the single best use of time is to rebuild the people you follow so that you're just getting it organically every day.

And I'm trying to figure out how to use my time right now. The only thing I can come up with is that this is a period of time where research is super leveraged. There are so many new developments that I'm tracking. It's not even complete. I think I would have added multimodal models to this as the new one. These are all the things that I'm watching right now. I just compiled this deck and I just sent it to some people. I'm going to keep tuning it and updating it as I see new things.

In terms of things that I think have high potential, like ChatGPT-4 just added the ability to run code directly in ChatGPT and you can upload files to it. So, I'm wondering, is this going to displace IDEs altogether?

Like, you're just going to do coding. It's kind of nice because it's like pair programming with ChatGPT, plus you can run the code and upload files. So, I don't know what this is going to do. This is going to probably have a major impact. I don't know if people are going to use IDEs anymore or they might use IDEs in a different way. So, this is a huge development and this is recent, within the last couple of months. We're seeing revolutionary changes to how productivity as a developer happens in real-time. And it's hard to even stay on top of this.

There's a lot of discussion of prompt engineering. Some people say it's dead and only a temporary time period. Others are like, "Oh, you can make 200 grand a year as a prompt engineer." I don't know where it's going to fall. I kind of like the idea that maybe it ends up being more similar to Ruby on Rails than anything, but probably a much more productive version of Ruby on Rails.

These agents and multi-agent systems, this is new and it's a conversation that's moving incredibly fast. Microsoft just introduced this auto-gen thing. The idea is that one of the shortfalls of ChatGPT is it doesn't have context awareness and it's like scheduling and a few other things. You're kind of constructing a primitive little brain here by having a memory and context. Even I'm seeing people do things like generate different personalities, like I'm going to generate a board of advisors and one person is an engineer, one person's a philosopher, the other person is a designer, and they're going to give me advice about how I should spend my time every day.

This kind of multi-agent system is moving incredibly fast. Chaining prompts, combining prompts together with agents, is that going to be a visual interface? Kind of like Node-RED. I'm seeing a lot of tools trying to do these visual interfaces. A lot of the ones I tried were not that compelling.

The ability to template prompts, this is kind of like Ruby on Rails again. One of the big deliverables of Ruby on Rails is it let one developer be a front-end and a backend developer at the same time. And it let you template your web UI and now people are templating prompts in the same way that people were templating web UI before.

Low code and no code automation, like Zapier and n8n.io, stack AI, vellum. I'm seeing a lot of these where you can just kind of chain stuff together. This is just enabling that business person to be way more productive, like skip the engineer if you want to do everything.

Huge developments in AI and generative coding on a daily basis. Repl.it just enabled free access to Repl.it AI, whereas some of these others are charging for it. I think that's free now. That just happened like last week. That's a huge development that's just enabling people to develop whole applications with a single prompt. We're seeing that now, like, "Hey, make me a snake game." And there's a whole working snake game.

And then with ChatGPT vision, you could show it a picture of a calculator and say, "Code this for me," and it will do it. You could show it a picture of Wordle and say, "Code this for me," and it will do it. So, I have no idea where that's going to lead.

Sheikh Shuvo: Going back to your recommendation on really just investing in the research side and making sure you're curating the people you follow, could you share some examples of the people and news sources that you follow yourself, any particular podcasts or newsletters that you'd recommend?

Rex St. John: I just click follow and mass follow on X. I don't even know who's in there. It feels like there's a million people doing this, and they're all over Twitter, all over the place.

Sheikh Shuvo: Good to know. Changing gears a bit, tell us more about the projects you're up to at Taroko. And where does the name come from?

Rex St. John: The name comes from a gorge in Taiwan. Taiwan has a huge number of mountains, over a hundred that are like 10,000 feet or higher. It's a small island but incredibly mountainous. A long time ago, they decided to put a train or a road through the middle of Taiwan, which involved a lot of dynamite. A lot of people died doing that. It's still incredibly unsafe. In the middle of the gorge, they built this waterfall temple to commemorate all the people that died. My logo is based on that. It's about the willpower people are willing to put towards doing hard things.

So what am I working on? Right now, I'm doing heavy research and development on what the future of developer relations is going to be. I've learned a lot about AI, IoT, web three, and I continue to learn. I'm advising CEOs of highly innovative companies on how to adapt their marketing, developer marketing, creator marketing, thinking six months, 12 months ahead through these challenging problems. Narrowing it down to specific recommendations. We're at a time period where the amount of change is so shocking that people are going to need research like this to make sound decisions.

Sheikh Shuvo: Sounds awesome. Along those lines, if I'm really fascinated by this space of creator relations and looking to switch career directions, how would one get started in the field?

Rex St. John: That's a good question. I was a software developer, so I have some of the skills. But then you add into it this whole mix of AI-specific skills. I don't think just a business professional needs to know how large language models work in detail, like an iPhone chip and security. It's about using the apps to do productive stuff. Testing out ChatGPT and trying to integrate it into your workflow slowly is a good start. Pay the fee for ChatGPT 4, turn on Bing and plugins, and you're already way ahead. If you're a developer or an artist, it depends on what you want your character class to be. For a business person, ChatGPT does almost everything if you've got the plugins enabled. It's a hard question, but that's the base thing.

Sheikh Shuvo: Okay, I'll ask you that question again in a couple of months and we'll see where we land.

Rex St. John: From what I'm seeing, a lot of people seem convinced that the open-source AI tools and frameworks and libraries are going to expand exponentially. I think that's a good bet. We're going to see a race here between OpenAI and the closed models and the open-source community. With AI, people are expecting the open-source versus closed-source dynamic to repeat itself, like the Microsoft versus the Linux. I don't know if that's going to apply to AI because it feels like if OpenAI releases this next-generation model, I think they're calling it Arrakis, and they get to a point where they can automate everything far faster, it might change the dynamic. Maybe in AI, because of just the nature of AI, whoever gets to the next step first wins it all. Whereas Microsoft got caught by all these open-source people who innovated faster. Who knows? Interesting.

Sheikh Shuvo: For any listeners who were inspired by this and want to connect with you, what's the best way to find you online, Rex?

Rex St. John: I'm on Twitter at Rex St. John, R E X S T J O H N, and on LinkedIn at the same thing. That's pretty much all I use.

Sheikh Shuvo: Well, Rex, this has been a super informative and great conversation. Thanks for taking the time to join and sharing about your world.

Rex St. John: Absolutely. Thank you for having me on.