Ep. 25 | Developer Advocacy With Annie Sexton  ## Metadata - Author: [[Overcommitted]] - Full Title: Ep. 25 | Developer Advocacy With Annie Sexton - Category: #podcasts - URL: https://share.snipd.com/episode/4d5f1d30-fb73-4cdd-aa5f-09906cd83cda ## Document Notes ## Highlights - **Unexpected Path To DevRel Stardom** - Annie accidentally fell into developer advocacy after earlier YouTube experience and a Fly.io role shift to video. - Her first Fly.io video went viral and transformed her career into a DevRel focus. Transcript: Annie Sexton It's good to be here. I'm a developer advocate, and my flavor of developer advocacy is mostly YouTube content. I actually secretly have done this for many, many years, not related to tech at all. I had a couple of channels that I tried to get off the ground, had some progress. I learned a lot of things, learned what not to do. And when the opportunity arose at Fly, I jumped at it. Did you work at Fly before doing the YouTube content creation or do you? Yes. So it's actually a little crazy. I hired as my title was what is it javascript specialist whatever the hell that means to write blog posts about javascript and to get the javascript community excited about using fly And then last summer we decided to do a shift to video and because i had had that experience building those YouTube channels, it was the most like hold my beer moment of my entire career. And my first video went viral. And it was great, which is like, I cannot say that all of my videos meet that kind of success, but it was it was very hair flip kind of. It felt really good. And it's been really, really fun to do. I feel so lucky that I just accidentally fell into this role. ([Time 0:00:50](https://share.snipd.com/snip/173a3268-937a-4759-b9f8-c74d6354db5e)) - **Learning On The Job In Osaka** - Annie learned web development after college by taking a job in Osaka where she barely knew how to program. - She learned on the job, drank from a fire hose, and built skills through hands-on mistakes like rolling her own features. Transcript: Annie Sexton I did not go to school for this at all my major in college was international studies which is was useless um so i took a really strange path to get here i taught english in japan right out Of college. I had a horrible experience. I love teaching, but I had a horrible boss. It was like a really manipulative, abusive work environment. And a friend of mine who I'd studied abroad with in Japan, he knew that I had had a little live journal blog. I knew a bit oftml knew how to make things bold and italic and he was like hey you had a live journal in college do you want to we're hiring for a web developer role do you want to like learn The rest of it and I was like oh okay sure so bless his heart he taught me the very basics of css and JavaScript and helped me get my very first job as a web developer in Osaka. And I barely knew how to write a for loop. My interview was all in Japanese with a Spaniard. It was mostly it was a localization company. So most of the most of the people who worked there were foreigners. But the lingua franca was Japanese because the few Japanese people they did employ didn't really speak English. Everyone else, I mean, my Japanese was like, I thought my Japanese was okay. And then I come into this localization company where everyone is fluent beyond words. And yeah, anyways, so it was a really, it was a small and scrappy company. It is near and dear to my heart, even though the working conditions were kind of terrible. I actually took a 30% pay cut from being an English teacher to get that first job. My friend who helped get me this job just on a whim, this was when Ruby on Rails was getting really popular. And he was like, well, we need a new corporate website. So let's try Ruby on Rails. And I was like, so do you know Ruby? No. I barely know how to do programming. So we just kind of learned together. We did everything the wrong way. Like we built our own auth. We built our own pagination. We built our own, like they had gems for these things back then, but we were like, no, no, I'll just make it myself. It was bad. It was very bad. But I learned so much. It was drinking from a fire hose every single day, but it was, it paid very, very little. And I lived in the smallest apartment you could ever imagine. I've seen walk-in closets bigger than what I lived in back then. But it was delightful. And it was such a good experience for me. And then I moved back to the States after about a year of working at that job, a little less than a year. And I just started working at random little agencies, got into WordPress. I feel like a lot of people honed their skills in the WordPress days. A lot of love to WordPress. I will never touch it ever again, but I learned a lot of things. And that's kind of how I got into things is just through that one weird job in Japan. ([Time 0:03:50](https://share.snipd.com/snip/9b5dfc8d-459f-4532-a8fd-428b045e250c)) - **DevRel Is Education-Driven Trust Building** - Developer advocacy blends community building, education, and marketing tailored to developers. - Building trust through high-quality education and authority is key to effective DevRel. Transcript: Annie Sexton Familiar with that and um what's your day-to look like so the question of what is a developer advocate is hard to answer because everyone's going to give you you ask 10 different developer Advocates are going to give you 10 different answers. But I think the themes you will hear are it's about community building. It's about education and it's about building relationships with your target audience. And I do say target audience because let's be honest, it's a marketing role. It is a marketing role. I think a lot of developer advocates kind of shy away from that term because they're like, no, we just want to be buddies with our customer. Like, that's not not true. But like, we're paid money to sell you things like that's that's not not true, I think. So that is part of it but you can't market to developers in the same way as other people not to say that like we're so special but but truly like you can't it's a much harder to bullshit us And we're very i would say untrusting of a lot of sort of you know sparkly kind of marketing words um i think there's a reason why we'll get to ai in a second but i do think that i i would be surprised If the ai marketing tool is actually working on developers specifically maybe other customers right but like uh developers are a lot better at smelling b. And so the way you basically appeal to them is by gaining their trust, right? And showing that you are an authority on a particular subject, and you educate them, and you peak their interest, right? And that's really at the core of really good developer advocacy. And I would say good marketing in general. Like, I think that's marketing done well in a non-shmarmy way, you know? What I do is very specific. Some developer advocates give a lot of talks at conferences. Some of them write blog posts. I do YouTube. Again, I just sort of fell into this role. It's kind of crazy and I'm pretty good at it. I am a big believer in education as a tool to gaining trust in users. And even though I really didn't like that particular English teaching job that I had, I have always been a teacher at heart. I love, love, love teaching. I have infinite patience for it. I would definitely consider myself a jack of all trades, except for when it comes to teaching. There's very few skills that I'm like, yeah, I'm stellar at that. My ego is way up here on my teaching abilities. I feel very confident in that, which is really nice to have one thing that you can actually feel that confident about. And it's something that really brings me joy. And I think a part of that passion for education comes from the fact that I had to learn all this stuff myself. And respectfully, developers are terrible writers and terrible educators. Not all of them, obviously not all of them, but many of them are. And I found I had so many questions that were just never addressed. There was so much assumed knowledge. Obviously, you have to assume some amount of knowledge. Like if you've got this repository on GitHub that you want people for a library, there's an assumption that you know how to use Git, right? But to get the files, you know? So there's like some assumptions that are understandable, but I think more often than not, people do not cater to the appropriate level, in my opinion. And I would much rather cater to a lower level like not even lower level lower level of experience than is necessary because it makes it easier for other people to understand right everyone Loves a explain it to me like i'm five kind of thing everyone like no one doesn't enjoy those things even if they already understand the material i think so i think giving people a you know An entertaining format to learn things sometimes things that you didn't even know you wanted to know um is a really great way of building a relationship with the audience that you want To target and in the case of fly i mean i do make videos more specifically about check out this cool product like i literally just an hour or two ago i published a video on our new managed Postgres service you should check it. Or that one was more just like, look at this product. Look, isn't it amazing? Right? Just sort of that's more like basic. Look, we're selling you a thing. But many of the, in fact, most of the videos are me explaining a weird thing about the platform that honestly you don't even need to know to use Fly. It's just kind of like, whoa, look at this weird thing we did with the CLI. We created our own IP protocol in user space so that we can proxy this, that, and yet it was very, very, it's very, very cool stuff how we make things behind the scenes. So I like to do a lot of how the sausage is made kind of videos because for the most part, when it comes to company videos you know you kind of have to minimize the number of videos that are All about your platform as weird as that sounds because largely people don't care you know and i think that piquing people's interest and that like that those are going to be the people That will return to these videos that will return to the channel there's plenty of people who don't even use fly yet that are subscribed to the channel because they just like to learn These weird little things that we do and they're super fun explainer videos so i think focusing more on education more on entertainment is a really great way of building an audience That really values your content and doesn't just feel like, oh, here's another video trying to sell me a thing. So that's been my approach, and I really enjoy doing video. It's also a great place for me to, I have a lot of different skills. Like I said, jack of all trades, and it's nice to have an avenue to use all of my skills in one place. ([Time 0:07:20](https://share.snipd.com/snip/71a49019-27dd-496c-98f2-a06b58b492cd)) - **Teach From Lower-Level Foundations** - Cater explanations to a lower experience level and build ladders before jumping to complex topics. - Use entertainment and how-things-work stories to attract and retain an audience beyond direct product pushes. Transcript: Annie Sexton And I found I had so many questions that were just never addressed. There was so much assumed knowledge. Obviously, you have to assume some amount of knowledge. Like if you've got this repository on GitHub that you want people for a library, there's an assumption that you know how to use Git, right? But to get the files, you know? So there's like some assumptions that are understandable, but I think more often than not, people do not cater to the appropriate level, in my opinion. And I would much rather cater to a lower level like not even lower level lower level of experience than is necessary because it makes it easier for other people to understand right everyone Loves a explain it to me like i'm five kind of thing everyone like no one doesn't enjoy those things even if they already understand the material i think so i think giving people a you know An entertaining format to learn things sometimes things that you didn't even know you wanted to know um is a really great way of building a relationship with the audience that you want To target and in the case of fly i mean i do make videos more specifically about check out this cool product like i literally just an hour or two ago i published a video on our new managed Postgres service you should check it. Or that one was more just like, look at this product. Look, isn't it amazing? Right? Just sort of that's more like basic. Look, we're selling you a thing. But many of the, in fact, most of the videos are me explaining a weird thing about the platform that honestly you don't even need to know to use Fly. ([Time 0:10:18](https://share.snipd.com/snip/f59d733b-c610-44d4-8357-44b97fd9041a)) - **Ask Stupid Questions, Then Verify** - Ask the 'stupid' questions and dig into points of confusion while researching and teaching. - When confused, use AI to probe explanations, but always fact-check with human-written sources and peers. Transcript: Annie Sexton Right. I think a lot of it comes down to really knowing who you're talking to and not assuming that they know very much. This is going to sound a little harsh, but I always think of targeting my videos towards a like, can a mid-level, maybe even like early career JavaScript developer understand this? No shade to JavaScript developers. I am one. But I just think of just like your normie. I know some Next.js kind of JavaScript developer. If they understand what I'm talking about, then I've done a good job. I think too often we will try to jump to the complicated stuff without building a ladder for people to follow you. I think I've just been very good at having to ask the stupid questions along the way. This is where I'll admit, so I use a lot of AI in my research. I always start with human written articles, books, things like that when I'm learning about a particular thing. I always, always, always start with that. And then as I'm reading, if I get stuck on something, I'll go to like ChajiBT and be like, hey, I'm reading this article. It's saying this thing about connection pooling or whatever. But what is XYZ? This sounds like this other thing. What's the difference between these things? I'm really, really good at identifying my points of confusion. And I know that if I'm confused about a thing, someone else is too. I'm trying to say this in the most delicate way. Sometimes I think that I don't always like whatever amount of like stupidity I have, I mean, I'm not actually trying to like be self-deprecating here, but it's actually a good thing When you feel confused about something because that means that, especially as a teacher, it means that, oh, a question there. And that's a question I need to answer for the people I'm teaching, right? And so I think in the past, my inexperience, my, you know, all the imposter syndrome that you, you know, have to work through as a developer, everyone deals with it. That's actually to your benefit is don't ever try and gloss over something that you don't fully understand dig into it dig into it dig into it dig into it and and don't i don't i really try Not to gloss over a lot of things because i think when you do that then you're like you're missing out on some questions that people will so ask the stupid questions ask the stupid questions Because they're probably stupid. They're not actually stupid. And I think that is where you get a better, like, I want to say curriculum. I'm not making curriculum, but you know what I mean? It's easier for people to follow what you're talking about. ([Time 0:14:35](https://share.snipd.com/snip/3954e440-f0c2-493e-b937-62adcb30246a)) - **AI As A Research Multiplier, Not A Source Of Truth** - AI acts like a patient, on-demand senior engineer for research when combined with human sources. - Fact-checking and peer review remain essential because models can hallucinate. Transcript: Annie Sexton So AI gives me a lot of anxiety. I'm gonna, I'm not gonna lie, but I have thrown myself into learning about it. And that has really helped my anxiety. It has made me feel a little more, probably not in control of things, but on top of things enough. I'm going to start with the good things. AI has drastically improved my career for the better. Like my career would not be the same without it. Excuse me. As I said, the way I found it useful is after learning about a thing from some human written content that could be documentation, that could be a book, I will then go to some LLM and... Can y'all hear that in the background? I'm sorry if you can. Whatever. I'll go to some LLM and I will pretend like it's just a it's a senior engineer who can answer my questions. I usually if I can copy and paste something in, I will. I don't assume that it knows everything. If it's like really ancient knowledge, right, about like IPv6 addresses and like how that came up, like that's pretty standard. It's set in stone. There's not a lot of opinions about when IPv6 came out. Like it's just, it's a fact, whatever. I'll still fact check that kind of thing. But usually what I'll do, let's say I'm reading an article about a particular subject. I'll copy that article into Claude or Chashi BT. And I'll say, here's this article that I was reading. I don't understand this one section. It says this, but that sounds like it's contradicting XYZ, or, you know, whatever my question is. Pasting in the article i know that it has the context that it needs and it's not just going to be like pulling things out of its ass right um i mean technically it's all hallucinations but Whatever so it i have found it incredibly useful because i can dig i can dig i can dig and what's good is that a lot of the things that i'm digging into are proven facts these are not opinions Right This is how a kernel works, right? This is what kernel space is. This is what user space is. This is what a VM is. This is what a hyper-advisor is. There are just straight up answers to these. It's not ambiguous. And so I feel a lot more confident leaning on AI for the type of research that I do. And I also get my coworkers to check my work. So anytime I'm making a video, I'll have one. I mean, part of that is because if I'm talking about how the platform works, I need to make sure that I'm not lying. And so I'll give it to one of our engineers. Thomas usually writes a lot of stuff on the blog. He reads a lot of my scripts and is able to help me a lot. So I do my own fact checking. I have other people do fact checking. Fact checking is so important when it comes to using AI. ([Time 0:23:39](https://share.snipd.com/snip/70c0f4bb-a9e1-4b51-b364-e802cea3e4bd)) - **Use LLMs With Source Context** - Paste source text into an LLM and ask targeted questions to resolve confusion and dig deeper. - Use AI iteratively to narrow down misunderstandings, then confirm answers with humans or documentation. Transcript: Annie Sexton I have been able, in the past, when I don't understand something, what do I do? I either go and research it myself. I try and hone my Google foo. Or I talk to a coworker or someone who's an expert. But you can only ask like you can ask the best questions to a real human. You won't always get the best answers because people there's varying degrees of people who can explain things. People are more patient than others. And even if they are very patient, you know, if you have a lot of questions, which I always have a lot of questions, there's sort of a limit to which I feel comfortable asking them my questions. And Chachapiti will never get annoyed with me. And so I can just be like, but what about that? Wait, I don't understand that. No, that doesn't make sense. That sounds the opposite of what you're, what? Why? Why? Why? I can just be this like annoying little kid and my ADHD brain could just ask all the questions that it wants and that's how i learn is i just be like absolutely obnoxious with all of my questions Sometimes i don't even like have a very specific question and i you can still be kind of vague with it and it'll it'll start to put you in the right direction absolutely phenomenal highly. Highly recommend. Highly recommend. ([Time 0:27:52](https://share.snipd.com/snip/82250553-9261-4682-a5f8-d9c5fd0b2299)) - **Micro-Expertise Beats Surface Familiarity** - When teaching AI topics, aim for 'mini experts' by focusing on one concept at a time. - Software engineering knowledge helps little for ML; build math-backed intuition through approachable resources. Transcript: Annie Sexton Sure so when I'm doing any videos about AI or machine learning, so I always, all of my videos happen to be for a technical audience. Oops. All of my videos happen to be for a technical audience. And so, but that doesn't mean that they know about machine learning, right? In fact, I assume most of them don't. I'm still a student of this. Like I'm not a, also the term AI engineer has been commandeered by software engineers because it used to be you did machine learning. If you're an AI engineer, you were training models. You were working on the algorithms. You were working with transformers. You are an AI engineer. Now an AI engineer is somebody who uses the open AI API key, right? Like that's, you know what I mean? All you're doing is you're shipping off some contacts to an LLM hosted in the sky. Like that's more or less, I mean, sometimes it's a little more involved in that, but like it has changed, which is why I do think nowadays, if you're talking about like the old school AI Engineer, you probably are gonna say ML, right? A machine learning engineer. So I've read a couple of books on machine learning that's given me a good foundation. One of them is called Deep Learning, a Visual Approach. Fantastic, very little math, which the fact that I was able to get through talking about like convolutional neural networks without talking about hardly any math is incredible. Just a testament to, I think his name is Andrew Glassner, who wrote this book. Highly recommend. It's a beast. It's like a chonky, it's a chonky boy, but it's such a good book. And I made my way through that. And that gave me a really good foundational knowledge. But I assume when I'm doing videos on these kinds of topics that people's experience is that they've used an LLM before. And perhaps and they know what an LLM is. I don't even know. I don't even assume that people know what like a diffusion model is. I don't assume that people know what transformers are. I don't. I assume people are only slightly more technical than a non-technical person using ChaiGBT, right? Maybe these people have worked with the OpenAI API or the Anthropoc API. Maybe. But that's about it. And then I assume nothing, right? I also think that when you're a modern day AI engineer, you need to know a lot less. You know, you need to know context window. You need to know what tokens are. But you don't even need to know what tokens are. That's the thing. You know what I mean? Most people understand tokens to be your text gets broken into little pieces. It's not exactly words. It's like parts of words, usually. That's it you know your text gets broken into little pieces but people don't often know like i'd be willing to bet that most people don't actually know what those tokens actually are And that's okay and so like that's a good place to start that is the position that i like to take when i'm explaining these things i mean the way i've been doing this, the reason I started That second personal YouTube channel to talk about all this was because I knew I wanted to become more of an expert on this stuff because it's the future. And part of it's, you know, I don't want to get left behind, but more like this is an incredibly powerful and also potentially very scary tool. And I want to make sure I know what I'm talking about and I know that I can see through the hype. And so what I do is I just become like a mini expert on one thing at a time, right? Just like a little baby. Like I'm gonna get like baby expert on the basics of what a neural network is and what byte pair encoding is, what this other thing is right and i think i am a huge believer in becoming a mini Expert just like a little micro expert that's my whole job and it's great so that's how i like to approach teaching about machine learning because i don't machine also here's the other Thing i'm mostly talking to software engineers when i'm making these videos right software engineering some people are going to disagree with me on this. I think software engineering knowledge is basically useless with machine learning. It's basic because machine learning is all math. Not like a little bit, all math, all. I'm going to say it again. It's all math. Like, and I never took or took calculus. I never took statistics. And it's, that's all it is. You know, aside from the idea of like input output. Yeah, nothing about software engineering has been helpful in learning any of this stuff. So it's easy to make these kinds of educational videos because I am my target audience. ([Time 0:30:20](https://share.snipd.com/snip/1ec6713b-abe3-4961-928f-12a2a3e2176f)) - **Comics As Creative Lifeline** - Outside tech, Annie creates comics and self-published a ten-copy run of her first graphic novel to prove she could finish it. - Working with an editor on a large comic series has been healing and sustaining through a difficult year. Transcript: Annie Sexton Track comic books so if you if you're interested here i'm gonna do a plug so i'm on blue. My tech one is at AnnieSaxon.com. But if you just look up Annie Sexton, like the other Annie Sexton on Blue Sky is all of my drawings. I'm working on, I actually published, I say published. I printed 10 copies of my first graphic novel this year called Peach Boy. It's not for sale. I'm sorry. But showing myself that I could finish that was incredibly motivating. And now I'm on to my next comic book. This a whole series it's gonna be like 500 plus pages i'm working with a real editor that i pay human dollars with it's amazing i cannot tell you how much writing this story working on the Sketches has been healing for me like there's so much existential crisis is like the term for the year uh in 2025 it's like on so many fronts and this little story this dumb little story Is just keeping me afloat it's just the it's the love of my life i wholeheartedly think people should make more stories it is i don't understand how anyone can find anything more joyful. I just think it's so delightful. So that's what I do. ([Time 0:39:57](https://share.snipd.com/snip/df6dddce-5a1f-4f31-a5c7-28b4babdc892))