Episode 103
· 01:16:17
Not enough time is spent on strategy. Yeah. Are you working on the right things? Yeah. And actually most people are probably not working on the right things.
swyx:Yeah.
Jack Bridger:Hi everyone. You're listening to Scaling Dev Tools. I'm joined today by Sean Wang, also known as swyx on the Internet, for his return appearance on scaling dev tools. And I think in that first episode, I gave a very gushing, fan appreciation. I'm definitely a fan fanboy of, Swix.
Jack Bridger:Have been for a long time. swyx does the Latent Space podcast, which is probably the biggest podcast on, like, AI engineering right now. Runs the AI engineering conference. Well, so we've got summit, and you've got the, AI Engineering World Fair. Two huge conferences in San Francisco.
Jack Bridger:Been to both of them. They're amazing. And in the past was, DevRel at companies like Netlify, Git, and AWS, and
swyx:Temporal and Airbyte.
Jack Bridger:Temporal and Airbyte, and was, like, huge in building public. You've got a long history of, like, doing amazing stuff.
swyx:Oh, thank you. So yeah. And generally my second career.
Jack Bridger:Yeah. So you kind of yeah. I think at when we first spoke, you were just at the beginning of, like, moving into AI stuff. Right?
swyx:Maybe. I don't even remember when that was. I know that's just before the Superbase interview that you did because when you recorded the chat with Ant, I was in the other room.
Jack Bridger:Oh, yeah.
swyx:I didn't know he was recording with you. I would have popped in and said hi. Yeah. Yeah. Yeah.
Jack Bridger:Yeah. And they were 2 of our biggest episodes. So
swyx:Yeah. Yeah. Nice.
Jack Bridger:Yeah. But thank you so much for coming back.
swyx:Thanks for having me. Thanks for coming to the studio. Yeah. This is where we record Layton's space. I am now, you know, a podcaster just just like you and and, it's having fun with it.
swyx:Yeah.
Jack Bridger:Yeah. That's, amazing. So I guess, like, there were a few things I wanted to talk to you about. There's almost an endless list. But we're gonna talk about angel investing in dev tools because you've done a lot of that, and what you've learned and how you how you do it.
Jack Bridger:We're gonna talk about, how you see the future of, like, Gen AI stuff affecting dev tools, whether there's things that people should be thinking about. And I think also just well, I would say those are the 2 main topics that we wanna talk about and just obviously. Okay. Quick announcement. WorkOS are hosting a one day conference dedicated to enterprise software.
Jack Bridger:It's called the Enterprise Ready Conf, and it's taking place on October 30th in San Francisco. They have got the founder and CEO of Vanta, Christina speaking, as well as the head of product for ChatGPT. Yes. ChatGPT, Brit. So if you've got questions for them, go along.
Jack Bridger:And they've got the founder of my the founder of Work OS, of course, Michael Grinich, who's been on the show twice. And you know how much he knows about enterprise software. So if you're kind of thinking about, you know, selling enterprise software or you already do it, you should definitely check out this conference. Go to enterprise hyphen ready.com or click the link in the show notes.
swyx:You know what? One thing I do reference, about for DevRel is, like, I I I I think, like, any competent DevRel eventually stops doing DevRel for one company and then starts doing it for many companies because it's a very enough skill set that if you do well, you can sort of franchise it. And that's how I view the AI engineer, conferences. So but this is you know, instead of speaking at in one at a time for one company, you could just run the conference and have, like, a 100 companies come. So in that sense, I'm kind of applying my DevRel thoughts to how should an industry's DevRel motion look like.
swyx:That's where we are at with that.
Jack Bridger:Yeah. I I didn't think of it that way. That's, that makes sense.
swyx:Like, speaking at doing in one talk, like, it's no longer exciting. But coordinating a 100 talks is Scales. This scary.
Jack Bridger:Yeah. Yeah. And yeah. You did a very good job of that. It was great.
swyx:Oh, thanks.
Jack Bridger:Yeah. Actually, maybe that's a funner side of, like, how was it to organize such a huge conference?
swyx:So this for people who weren't there, let's give a concept so the first one was 500 people and then the second one was 2,000. The conference is modeled after the the large form factor conference, the the World's Fair version. And, Summit will always stay small, but World's Fair will will grow. The large form factor version is modeled after basically stuff I saw throughout my DevRel career, DevTools career even, which is KubeCon, Big Data London, and AWS Summit, I think are the 3 that I like. There are some others I haven't personally been to, but I'm a big fan of on YouTube, which are GDC, like Game Dev Con, and Strange Loop, which which, recently ended.
swyx:And, I just think, like, every industry needs a definitive conference for their practitioners to gather and talk. Like, it's really that simple. Like, it's not there's no big brain thing beyond once a year, let's get together and talk stuff. And people change jobs there. People find new, coworkers or collaborators or just launch startups there.
swyx:All sorts of things can happen when you get people together, and I I think that's the the goal of that conference. So those other conferences that I mentioned are between 5 to, let's say, 30000 people, and we're at 2,000 now. So we're we're trying to grow to that scale such that AI engineering, the the community that I help to name and serve, also gets to that level of legitimacy, as a as a community. So I I think, like, that's a good cause. You know?
swyx:I I like causes where it's bigger than me, and this thing's definitely already bigger than me. And all I do is, I serve one specific role that, you know, pretty much only I could I could do, which is get people together.
Jack Bridger:Yeah. Which is yeah. As you said, it's it's hugely valuable. And was it like was there, like, a moment where you thought consciously, like, this is what I'm gonna do, or is it like something that naturally
swyx:Yeah. It doesn't happen by accident. I had to so the the the launch sequence for AI Engineering Summit was the same as the launch sequence for AI Engineer itself, which is I run about in February of last year. I had a hackathon, which I had the fundamental realization that software engineers building with AI were a much different category
Jack Bridger:Mhmm.
swyx:Required different skill sets. The kind of people who are successful generative AI were different than the ML engineers. And also that software engineers would win over the prompt engineers. Now it seems obvious, rewind this, you know, February 2023, it wasn't as obvious because everyone was only talking about prompt engineering, and didn't really have a title for anything else. And so, then I was like, okay.
swyx:Like, we should develop this term. I went through a few candidates including LLM engineer, cognitive engine. There's a bunch of other people that things that people, people suggested. But, worse is better, which is a phrase that all dev tools people need to know. And AI engineers, like, the the the the worst version of something that can possibly stick and therefore, it, the world will descend towards it.
swyx:So I just kinda went all the way and claimed it. I wrote the essay, but before writing the essay, I got the conference ready so that when I wrote the essay, I also launched the conference. And then I rebranded the podcast to focus on that as well. So yeah. All those things that happened between April June last year, and then October was the conference.
swyx:So and from there on, it it kinda started. But that that launch is probably one of the most intentional things I've ever done in my life. It wasn't an accident.
Jack Bridger:Yeah. That is, I guess, not surprising actually knowing you. Just like
swyx:Oh, no. I'm I'm usually very random. I'm usually very random and some frustrated. Like you seem very Chaotic. Really?
swyx:No. No. No. No. No.
swyx:No. Like, I have principles not plans. Right?
Jack Bridger:Like Oh, yeah.
swyx:There's like there's like certain themes I like to keep coming back to. There's a certain way I like to operate, but I usually don't plan like 3 months out.
Jack Bridger:Okay. That's actually surprising. Yeah. That makes sense. You because you wrote that essay, I think, I guess, quite a lot on, like, Twitter and stuff about, like, like, you have, like, the you you have loads of, like, greater than this is greater than this.
Jack Bridger:This is greater than
swyx:Yes. Those are principles. Yeah. Right? Not fans.
swyx:Yeah. In some sense, maybe I somewhat don't believe in plans because obviously things can change and, the future is uncertain and, you know, the plan gets changed the first moment it first contact, it has a reality. But obviously, I believe in planning. Like, the the act of planning itself is preparation that is necessary anyway. The plan itself can change whatever.
swyx:Yes. So I think actually I need to do more planning rather than less. And if there's anything I'm more interested in than AI engineering, it is this idea that we could be more intentional with our lives and with the way that we win and succeed in business and in life. And I'm trying to form a philosophy around that. This is already getting too philosophical already so we can we can drop off there.
swyx:But, so if I ever started a thing that would be bigger than AI engineering, it would be something around the topic of winning or intentionality or high, high agency. To me, all those thing those three things are basically the same thing, which is how do you win on purpose? How do you, build a support group or accountability group? And a system for yourself even even without others around you that, improves your chances of doing whatever it is you're trying to do with your life. I know this sounds a little bit, like, self help y, but it is that in that domain.
swyx:For anyone who's interested in this topic, I think the the line of thread that is very prevalent in San Francisco is this talk given by Dan Gross, at YC called how to win. And it starts off with get 8 hours of sleep and it ends with, you know, some business stuff. But if you don't, for example, get 8 hours of sleep, if you don't manage your own personal health, you're not gonna do the other stuff. So, you know, get, you know, out of sleep, get good nutrition, and then work from there.
Jack Bridger:Yeah. And and where do you see people, like, falling down the most? Because is it like when when when you say this, like, about the intentionality, the things that I'm thinking is, like, I have maybe multiple things that I'm working on and sometimes, like, some of them are not, like, all moving in the same direction. And is it that sort of, like
swyx:No. How did those multiple things that you're working on arrive at your plate? Should you be working on them? I don't even care if they're moving in different directions. It's fine if if they are.
swyx:But should you be working on them at all? Should you just drop it? Because is it
Jack Bridger:sometimes it's like it was just a good opportunity. Like, maybe you needed some money or, like, you know
swyx:Right. Sure. I definitely do think that this kind of thinking is for people who have some security in their life where, you don't have to be rich, but you you can just be secure in a sense that if all the other stuff fails, I can go get a job somewhere and and submit my agency to someone else's agency, and that's fine. And then I I will survive and live and provide for my family, and that's all okay. I think this is for people who want to strike out on their own and who want to take some risk, and knowing that there there's a safety net for them if they fail.
swyx:And I definitely in San Francisco here, like, you you definitely feel the the idea that it's actually more than okay to fail. Like, having failed actually improves your employability because now you can say you're a former founder, you know, and and people like to count their companies count their employees by how many former founders they have under their wing. Right? Because then it proves that they're better than all of them. Like, Superbase says, like, we have 30 former founders.
swyx:Yeah. It's great. Great for you, Superbase. And, and so that's great. Like, I I think, like, once you achieve some some level of financial security and career security, whatever that means to you, then you should think about how to win.
swyx:And and, you know, we there's one thing we all have limited time of. It's just time on earth. And, and, we should try to do more impactful things with with that time. Whether whether or not it is I I even think, like, the the phrase strike out on your own is too much. Not everybody should be a founder.
swyx:Not everybody should be a content creator. You can do more as a group than you can on your own. And if you just find a group, team, leader that you're just like, you like working with and you do they they get the most out of you, then do that.
Jack Bridger:You
swyx:know? And those that's kinda irrelevant from, like, is this what you're made to do and is are these the most important problems in the world that you're solving, that that that you can solve? So I really like that. So for example, I'll give you a concrete example and then we can sort of move back to dev tools. I always like the format of Alcoholics Anonymous.
Jack Bridger:Are
swyx:you how familiar are people in the UK with that? Yeah. Seems like
Jack Bridger:an American. I mean, I think I've I mean, I've it comes up in, like, a lot of books as well about, like, the habit. Like, that I think that habit. There's a book about habits. I mean, I've I've read
swyx:Atomic Habits or something else? Could be. There's there's many. Yeah.
Jack Bridger:There was it was not that one. It was the other one. Okay.
swyx:Yeah. I think Alcoholics Anonymous is just a good, adherence and accountability system that doesn't have to anything to do with alcohol. So what if we had AA but for accountability or for achieving your goals? Right? So they they have all these steps.
swyx:They have they they they're nonjudgmental. They go around. They share their issues and, there's a support group element to them. There's a little bit of anonymity or at least chatting about rules or whatever they call that, which is like you don't you don't share what you learn in in those meetings. All of which I think we need and we don't have in in life.
swyx:It's just weird that it just doesn't exist and I would like to start have that or start that or take part in it. Again, this is one of those things where, like, it's a it's a mission. What what I what I define as a mission is something where you're happy if someone else does it better than you. So I actually don't even wanna start this. I just wanna join it.
swyx:It just doesn't exist, therefore I have to start it. So, so that is the mission. The winning group is is that. And but the the the one problem with one problem with Alcoholics Anonymous is, well, one is alcohol so it's a very serious topic. This is a bit less serious than than life and death and, you know, mental addiction and and illness and all that.
swyx:This is this is just work. This is less lower lower tier than that if in the hierarchy of personal issues that are important and life changing. But, I I think there are too many steps. I think they really want to reduce things to 1 to 2 steps. So, everyone, you know, in my my ideal vision of the world, everyone comes in, they're prepared to answer 2 questions.
swyx:Alright? And only talk about 2 questions and hear it from the other, everyone else attending. And 2 questions are, 1, what are the most important things in the world that you should be working on? And 2, why haven't you done it yet?
Jack Bridger:Yeah. That's really powerful. I was just almost, like, answering it for myself and it's Yeah.
swyx:I I think And then just go around. Because Yeah. Because you hear in other people's stuff also helps you with your your stuff. And anytime you're not winning, it's one of those 2 things. You're not working on good things or you're you're blocks in some way.
swyx:It's a mental block. It's a financial block. It's a physical personal relationship block. Whatever. It's like blockers and whatever direction or goal setting.
swyx:Actually people a lot of think so, you know, a lot of these, like, you can count a lot of this self help stuff in the productivity domain where, you people say that you wanna get you wanna have this productivity system, bullet journaling, whatever. Not enough time is spent on strategy. Yeah. Are you working on the right things? Yeah.
swyx:And actually most people are probably not working on the right things.
Jack Bridger:Yeah.
swyx:So myself included, I I one of the reasons I have not started this yet is I don't feel in a very secure position to teach this because I don't have it figured out. I think I I must feel like I wanna have made it first and then I can turn back and give back. Whereas this is kind of like building support system for myself because I need it. Yeah. Therefore, I don't have any answers.
swyx:It's just very hard to lead people if you don't have any answers. If you don't have any credibility. But, in a sense, like, I I am thinking about this. I I am, working on this with myself, with, ChatGPT sometimes. We are in my company working on or working on this idea of a small CEO.
swyx:We I have the domain small dot CEO. And, literally, I wanna just hand it over and and have it just be my boss and and ask those 2 questions to me every day and make sure I'm aligned on it.
Jack Bridger:That sounds great.
swyx:It doesn't it actually we I pitched this at a at a dinner with, like, some ML researchers. They were they were shocked and horrified.
Jack Bridger:Really?
swyx:They were like, well, it's it's the inverse of what you're supposed to do with AI. AI like, you you the human are supposed to be at the top level and the AI is supposed to be at the bottom level executing the will of the human. And so I'm here I'm really talking about the inversion of that. The humans are very bad at at executive function and and following through and doing things you said you were gonna do. And what if you just handed that over to to the AI and then you became this the mindless executing things machine?
swyx:So that's a little bit of a controversial thought. I made it more palatable for them by saying it's kind of like a layer sandwich of human machine human. So human at the top is board of directors. Machine is CEO. And then human at the bottom for doing the thing that you signed up to do.
Jack Bridger:Yeah. And I guess, like, it's also, like, you're not necessarily giving like, it's may it's helping you to make decisions based on what you think and it's asking good questions.
swyx:Yeah. That's standard question that any manager could ask. If you had to add any manager, it would go through the same manager playbook that any manager would have and you don't really even need them there.
Jack Bridger:Yeah. And I guess the accountability is the huge factor is that, like, as a CEO, like, you're kind of, you could argue you're not really accountable to anyone else in a sense.
swyx:Accountable to your customers.
Jack Bridger:Accountable to your customers.
swyx:Your investors, your employees. Yeah. And, the other thing I also think, like, I think some people different people work in different ways. I find that I'm a very stateless person. Like, I like to receive information, react to it, do something else, and then I I move on to the next thing.
swyx:But as a founder, as a CEO, as a manager of people, you are paid to maintain state.
Jack Bridger:Yeah. In
swyx:the same way that a lot of dev tools are paid to maintain a database. Yeah. You know, any, like, logging, observability, database, database, or any chat or and and it's just various forms of database.
Jack Bridger:And a state in this case is like the vision.
swyx:No. Is this thing done? What what task is is owned by whom? And it's like I see. Are they overdue?
swyx:Where what else is the status update? Like Yeah. That's just the typical manager CEO Yeah. Nonsense. Yeah.
swyx:And, like, isn't, probably could be done better by someone else than me because I don't like to hold that state. I like to be stateless. In particular, you know, about the do you know the Eisenhower matrix?
Jack Bridger:I think a lot of
swyx:people I'm surprised that it's not uncommon. Most people have seen it. They don't know the name of it. It's the 2 by 2 chart. It's a 2 by 2 chart where, there's tasks that are important and not important, and there's urgent and non urgent.
swyx:Yeah.
Jack Bridger:Yeah. Right.
swyx:And so, obviously, everyone says you have to do throw away the non important things, only do important things. And then prioritize the urgent things and not the non urgent things. Yeah. There's a nonlinearity with the Eisenhower matrix where many important things in life are non urgent, but they're important. Yeah.
swyx:And sometimes they become urgent. So, a common scenario for this for me is taxes. Filing taxes. I know every year, like, the I have all the information by December 31st when I need to file for April 15th, which is the filing day in in the US. Why don't I take any time within that 3 months to file?
swyx:Because it's not it's not urgent. Yeah. On April 14th, it's very urgent. Yeah. And suddenly I'm, like, scrambling for accountants and stuff.
swyx:There's many nonlinearities in life like that where there's zero difference between whether you do it today or tomorrow.
Jack Bridger:Yeah.
swyx:But if you don't do it before a deadline, then suddenly it becomes very expensive. Whether it's like flights or hiring someone or whatever. The the actually the costless things are the invisible deadlines where you don't even know there's a set deadline but there there was one and you didn't know it and it passed.
Jack Bridger:What kind of things would that that would be like?
swyx:Sometimes it's, training a model that you you know you know there's, like, this brief window where, you know, it was possible to train, open source, like, GPT 3 class models, which is what GPT j was doing and Lamo was doing. And then now there's now that increasingly closed source and there was there was brief with nowhere. It was e kinda easy to get into the AI teams. Like, even, like, NLP or LLM teams, they were not prestigious. Yeah.
swyx:You know? It was a very small window. I've I've interviewed some of these folks where they were like, yeah. We, like, we accidentally like, I accidentally became a leading researcher in this area because no one want no one else wanted it. And I just, like, I thought it was interesting.
swyx:I just picked it because it was open.
Jack Bridger:Yeah. And,
swyx:like, now I'm, like, one of, like, 10 people in the world that has trained a large language model to this scale. And you'll continue to be so. Those those initial buck or active decisions compound. So that was a invisible deadline that nobody understood except for people who are paying attention and some people who are lucky. And, I think they're many like that where like, so for me, sometimes it was let's say the visa stuff.
swyx:Right? There was an invisible there's a hard deadline with my conference trying to me trying to get to my own conference, and get the visa in that time. But many invisible deadlines in the 6 months prior where I didn't know that I should've done something. Yeah. And then that passed and, like, that made it more expensive or caused further delays for me to Yeah.
swyx:To get my visa up to right up to the point where it was the week before the conference and my last fallback to the fallback failed because I also missed another deadline to get my visa waiver. Yeah. Many many other cases. I I think, like, a lot of the smartest planners in the world understands to clear the fog of war. Because it's one thing to know a deadline and then miss it because you were busy.
swyx:It's another thing to not even know there was a deadline there and also miss that.
Jack Bridger:Yeah. The principles, it's like thinking intentionally about what you're doing and like
swyx:Yeah. Intentionally, having agency and, you know, maybe a more jargon y of term for it is executive function. Yeah. But really just high agency. Yeah.
swyx:Doing what you actually internally want to do and want to exist for good reasons.
Jack Bridger:Yeah.
swyx:And, surprisingly, you know, people don't do that. They they respond they they much rather copy. Yeah. They much rather react. They much rather look an existing leaderboard and existing, company ladder, for example, and try to climb that.
Jack Bridger:Nice within
swyx:that. So and those are those are plenty good as well, but they're they're not exactly high agency in the sense that you you create a game that you think, is worth winning and actually pursue it. Because, I think those are the situations in which you have the most interesting mission, that people can kinda join you and support for you. Like, I think if you pick a a leader board or if you pick a company rung a company ladder, it becomes a zero sum game. Yeah.
swyx:Competing against. You're competing. And, competing is good. It's fine. I I think that's a big part of winning.
swyx:But I think it is these are situations where, it's zero sum. Your competitors don't share their information with you, all that all that stuff. But once you pick a worthy enough cause, the the whole thing changes. Like, your competitors are like, oh, yeah. Yeah.
swyx:We ran into that too. Here's how we fixed it. Because at the end of the day, the cause is more important than any individual winning and that's that's really helpful and important in life.
Jack Bridger:Yeah. So and I guess that, like, if bringing that to, like, AI and dev tools companies in a sense, so, like, if we call them dev tools, but, like, OpenAI at the beginning, no one else was thinking of doing this, like Mhmm. And they were, like, playing their own game. They weren't really competing against anyone else. It was, like, can they actually do this?
Jack Bridger:They believe this is important. Focus on that. Yeah. I was like, maybe now
swyx:it's I think I think Ilya Soskever works like that. You listen to any interview that he's ever done. He's just a been a through and through believer in AGI. And maybe the rest of the company believed it less than him. So in some sense, I feel like OpenAI potentially got lucky with a few hires that they made that skewed them in that way because otherwise they would have just been another research lab that didn't go anywhere.
Jack Bridger:Interesting. And do you think now it's become this, like, it's much closer to, like, this, like, ladder situation? Not ladder, but like
swyx:there's a lot of internal policies in OpenAI. I know a few friends who've been effectively managed out and fired, and, due to internal politics issues and that arises in any, you know, 1,000 plus person company. I I think the question before OpenAI now is, like, how much how well are they actually committed to the AGI goal instead instead of being like a product company, like a like a new Google effectively. I think the people who are very pro a like, obviously, any outcome like that is amazing and worth pursuing and whatever. Right?
swyx:But I think the AI purists are disappointed in OpenAI for basically selling out to monetize their APIs, to monetize chat gbt, to for them to, to to effectively stop thought leading on what, Frontier AGI should look like. It's it's debatable. Like, we're talking we're recording this in early September. They're likely to announce GPC 5 early October. And we'll see what GPC 5 looks like.
swyx:We'll see. I I I mean, I like, this is one of those things I I tried to debate less about what the shape of AGI is because it's very hard to quantify and, like, there are definitely people that are way smarter and more dedicated than to it than I am. I just think that though, like, it's worth understanding that the people who are most worried about safety are also the people who take AEI more more seriously than anyone else. Like, the the the very reason they got here in the first place is they took AI more safely than anyone else. And because the AI purists tend to disregard safety.
swyx:I don't think they take AI as seriously as they think. Because if you did take AI seriously, you should actually care a lot about safety and you should care about missing. Like, you should care about you should care about erring on the side of being too cautious because the other side is catastrophic.
Jack Bridger:And so I guess, like, Ilya is like the type of thing.
swyx:Potentially very, very like, he blew up his own company. He blew up OpenAI to try to make it safe in that field. So now he's doing his own thing. But, you know, OpenAI is definitely on its on its own path now. It's it's completely rid of all the safety.
swyx:Like, there there's no longer any safety division. There is a safety systems person, Lilian Wang. She's she's amazing. But there's no longer a super alignment team. Those those guys are gone because they went to Entropic or they started their own company.
Jack Bridger:Yeah. That's, yeah, that's very interesting. To bring it to dev tools
swyx:Yes. I can tie the mission thing to dev tools.
Jack Bridger:Yeah. No. No. No. No.
Jack Bridger:No. It, but it was, like, actually, like, I feel like it's kind of related that someone so yesterday, I was talking to, said why are you interested in dev tools? Because devs are gonna be obsolete. What's the point of building tools for devs? Like
swyx:What was your answer?
Jack Bridger:My answer was, like, that I don't know if devs are gonna be obsolete. Honestly, like, I don't know. Like, it it's every step has been surprising how much how how useful tools like Cursor now are, but people still need infrastructure. And so this hard, like and and that could become way bigger because, yeah, it's gonna lead to technology doing a lot more things. And so maybe things that help developers you know, maybe it's helping less developers, but they're doing they could be doing a lot more and control of a lot more infrastructure.
swyx:Yeah. Possibly. Yeah. I should highlight one thing that's a common conflation that I used to have, maybe still do, which is you conflate the dev tools and infrastructure. There's a world in which they're the same thing.
swyx:There there's a world in which they're different things. But it's interesting to think about why they're different and where they're different. There's definitional issue issue there. But let's let's go back to the the core of the arguments. So my standard answer to this topic of will AI replace engineers.
swyx:Right? And it's it's Change. Is that AI engineers will be the last job, not, not the not the first job to be replaced. Because AI engineers are the job that replaces the other guys' jobs. Right?
swyx:Like, you need AI engineers even more than you need ML researchers to do the last mile of automation for AI lawyer, AI teacher, whatever. So if you believe that AI job it will take away jobs, which I don't because human humans are very good at creating new jobs, then, you still want to be an AI engineer because they will be the last job that, that does all those things. That's the pet answer. I think the the second level answer for the dev tool people is that there's a very strong interplay between systems engineering and AI or systems, where AI is, like, very good at translating natural language to some format that it's trained to do. But it's not very good at systems design and not precise and not fast.
swyx:All these things are issues that systems, like, we're we're just much better at for the last, like, 50, 60 years. So, again, like, there's there's there's a there's a world in which, yeah, I can model all that and do it eventually, but, humans will comparatively have an advantage there because of the precision that it that is required, and the amount of thinking and planning that goes into making a database or making an Internet or whatever. Making designing a chip. Although chip design is is one that is being tried by, being tackled by LM. I I have an investment in that area.
swyx:But I I I think, like, it's definitely assisting humans rather than replacing them. This is something that people that's a story that people like to hear a lot, that, yeah, we'll not replace this. It will help us. It's a very touchy feely good thing. Like, wins wins for everybody.
swyx:And, probably true for up to a point. Like, we it's just we're all just talking about timelines here. Right? Like, when classic example that peep that people say, like, you know, automatic teleper seems, ATMs. When they're introduced, did the number of bank tellers go up or down?
Jack Bridger:Do you know? You would you would assume down.
swyx:Went up. Went up. For the next 20 years and then went down. Why did it go up? The number of banks continued to grow.
swyx:I know.
Jack Bridger:So it was like a bigger trend outside.
swyx:Exactly. Like yeah. But, yeah, eventually, they they go down and nobody regretted it. There was no systematic, like, crisis in society because there are no more bank tellers. They found other jobs.
swyx:Like, we were very good at coming coming with other jobs. Like, are you so cynical of human needs and human creativity that you don't can't imagine that there will be new jobs? I don't think so. And, I'm I'm completely fine with that. I don't have any issues.
swyx:I think that I think dev tools are have a lot of value to them. And dev AI dev tools have one thing going for them and the other all the other industries do not, which is they can have the user and the developer in the same loop because they're the same person. So they have this kind of natural self acceleration loop that doesn't exist in the other industries of, like, AI lawyer, AI teacher, whatever. You always have to, like, ask someone else, is this what you want? Oh, it's not what you want because it does some super obvious thing that you learn in, like, 1st year of doing the job.
swyx:And then the engineers have to go back in their room and code for a bit.
Jack Bridger:Mhmm. So the feedback loop's much tighter because Obviously. Developers are like, no. This sucks. Yeah.
Jack Bridger:Let me check. Obviously. Yeah. And so I guess there's, like, what you think. And there's also, you know, how you're how you're acting and where you're putting your money and stuff like that.
Jack Bridger:Yep. So maybe that's a segue into, like, your angel investing, and the the things that you believe are are gonna be important.
swyx:Yeah. Interesting. Give me more prompting questions around that.
Jack Bridger:Yeah. So I know there's some companies that we've had on the show that you've invested in. So you've invested in SuperBase. Could you talk a bit about, like, why you think that Supabase is, still gonna be, like, a really important thing if we don't have regardless of what happens, I guess, is
swyx:Okay. Yeah. That that AI lens is unnecessary. Like, Superbase is just a very good post grads company. Yeah.
swyx:It's enough. You know? Yeah. When I invested, there were 3, 4 guys coming out of YC. And, by the way, Thor put me onto the investment.
swyx:Like, I I wish, like, I wish I was that smart, but no. I just piggyback on Thor. So Thor I think you have Thor Thor to thank for Was
Jack Bridger:Thor already working? I
swyx:think he was about to join. Okay. And he was, like, so convicted. Like, he he does this. He tell he tells his friends to invest.
swyx:Yeah. And then we're like, okay. Like, you're you're smart. Like, you you know what you're doing. But I really actually hesitated a long time for Superbase because, they're, at the time, very, very thin wrapper around Postgres.
swyx:And you're just like, okay. Like, why don't I just run Postgres on its own? And I think that I underestimated the sheer desire for better Postgres. Mhmm. And that's effectively what they've done.
Jack Bridger:Yeah.
swyx:Then the the the even, like, the fire base for Postgres angle came along afterwards. So you invested before? Yeah. Then just didn't have
Jack Bridger:any of that. Still sounds.
swyx:Yeah. Really? Yeah. Yeah. If they had said that to me, I would be like, ah.
swyx:Okay. But I don't think it existed back then. I I think sorry. I I think that's a lie. I think I think that someone had, I think if you ask Paul, he would have said that someone posted super base with that title while they were in y c before they were ready to launch and it went, kinda viral on Hacker News.
swyx:And it was like the first inclination that they should go after the the firebase angle. But I didn't I wasn't paying attention to that, initially.
Jack Bridger:And it wasn't like the hero It wasn't the hero. Section.
swyx:Not yet. I think I think it was hero shortly after because, they've been writing that for a while. I think it's good. Like, yeah. This thing that sucks that that is gonna be killed by Google someday, We're doing it better.
swyx:It's open source. Like, it makes sense. Yeah. You know? So this is one of those things where, like, a really like, there doesn't have to be any AI in it.
swyx:Just have a really good system that that people can spin up, that you know, AI coding tools and dev tools can use if they want to. Like and there's so many I'm pretty sure GCP 4 knows the super base SDK inherently and just can't just code it. So it's it becomes a much better interface to Postgres than AI's could write themselves. And that's great for both Supabase and for AI. Then they can have this great persistence layer that's kind of intuitive and and, and open source and manageable, like, in in any environment that they want.
swyx:It just works that, I I think a lot you find this pattern a lot that, like, what works well for humans happens to also be good in the in the AI centric world because then they just provide a better layer for something on that that humans and LMS can use, except for some tools, which we'll talk about eventually, where that started to diverge as well. That, human needs do differ from L and M needs and there are some companies spinning up. Some of them are here. Here being Solaris which is the AI co working space that we're recording in, where they they actually specialize in that difference between LLMs and human needs.
Jack Bridger:Yeah. Actually, it'd be great to know.
swyx:Yeah. We've been talking about this on on my podcast as the LMOS. The the class of tools that LLMs basically always need. So they need to search the web. So, you know, you can plug them into the dot.go API or the Google SERP API or the whatever Bing has.
swyx:Or you can plug them into some more dedicated LM focused search engines like Exa or Tavoly. Or you can build your own. There's many people building their own. Or you can do browser automation like BrowserBase here is doing. All of these are very subtle different at scale requirements for plugging the results of a web search into an element.
swyx:All of them, for example, they need some kind of, like, proxy rotation. They need some some indexing, some different kind of caching or whatever. But for example, a very simple example is oh, okay. That that's one example. There there's another tool in the toolbox for code interpreter.
swyx:So OpenAI has its own code code interpreter, but for everyone else, there's e two b, another early investment, which, by the way, I invested before they were an AI tool. And one simple example is that your error codes are no longer for humans. There are prompts that are fed back into the LLM so you can fix the code that generated the error code. Right? So so of course you would do it.
swyx:Right? Like, right now, we we we take whatever error code we got and we chug it back in and hope the LLM figures it out. But what if the error code was a prompt that says specifically what you needed it to do? It does not for to be super clear, that is my dream of what e to b can become if they were that focused.
Jack Bridger:Yeah.
swyx:They are not even they're doing well, not even doing that. They're just they're doing well for other for a variety of other reasons. So, for example, what e to b is often, could be could be viewed as, like, kind of an ephemeral coding environment, ephemeral sandbox like you like you would a fly. Io, or a repllet or something like that. But, instead of human needs in humans, we need a few of these, dozens of these, maybe.
swyx:Whereas agents, they can kinda spin up 100 of them simultaneously. They just they're different they're the same needs but order of magnitude different, where it just had different requirements. Like, it can be order of magnitude in terms of, like, parallelism or order of magnitude difference in terms of, like, the latency requirements because suddenly there's there's no human waiting time. It's machine waiting time. It's different.
swyx:There there's a bunch of, like, all these, like, really long tail things that if you just specialize in solving the problems for an LN customer instead of a human customer, you start getting different products specs.
Jack Bridger:Yeah. That is actually, like, something I hadn't really thought about. That's very interesting. Yeah.
swyx:I think it's very convenient for DevTools investors to think about it this way because, it basically means that every thesis you ever had is gonna be, the the next few orders of magnitude and product requirements, like, must be 10 times faster, must be a 100 times faster, must be 10 times more parallel, whatever. They're just augmented by AIs. Like, they they, like, they they will be run-in much larger scale than humans can because obviously there's only one of me and, like, can run 10 AIs at once and not really care.
Jack Bridger:That makes sense. Those are really cool, actually. Yeah. It's got me thinking about, like, every everything you're working on, like, what does it look like if just an if an LLM is using this Yeah. Rather than
swyx:I even think, like, API keys need to fundamentally encode this concept. So, like, when when you issue the API key, you typically expect the developer to use it. Yeah. But what if you had a different key for an LM? Like, what different permission levels do they need?
swyx:What human in the loop, review do they need for for anything that's nonreversible? Stuff like that.
Jack Bridger:Yeah. Because maybe you would flag things differently if you saw a usage that's like
swyx:Yeah.
Jack Bridger:In a A friend of mine,
swyx:Fuad Martin, He worked on indent.com for 5 years which was basically just human in the loop approval process and he was like, I think this is something that people want for security and compliance reasons. And suddenly, AIs come along and they're like, oh, I think we want humans to review AI, you know, tasks. Like, you know, suddenly it important thing. But he had worked the previous 5 years on this problem
Jack Bridger:And then suddenly it's
swyx:like without going super viral or anything. It was a decent, like, profitable company. It just wasn't, like, huge. It was, like, 3 people company. And then 2 months ago, OpenAI acquired them.
Jack Bridger:Wow.
swyx:For for this. Yep. Very very like, if you just happen to work on dev tools and you you work on things that, like, scale well for AI's and humans, like, this this not this they're they're the same thing. They're not they're not different problems. They're just different skills.
Jack Bridger:And so maybe, like, some tweaks, but just, like, potentially I
swyx:don't know if I don't know how much he's tweaking. He's he can't say anything to me now that he's he's working for them. So Yeah. Yeah.
Jack Bridger:Yeah. Okay. Cool. What other investments are you excited about?
swyx:So you're asking me to So is this asking to name your Okay. Wait. Maybe I can call you. Children? I can tell I can name stuff that has
Jack Bridger:Fireworks seems very exciting.
swyx:Yeah. I can name stuff that has been, like, popular in in recent media. So fireworks, I'm not an angel actually. Fireworks, I am an adviser. I'm I I do a little bit of dev rel advising and, fireworks pursued me for, like, half a year and then I was like, fine.
swyx:And they are they're basically the former PyTorch team from Facebook and service I think the the original vision was basically spin out and just offer PyTorch as a service. And now it's more closer to, an inference, engine similar to Together, AI. Yeah. They they serve, the full stack of, like, fine tuning to, providing GPUs to, providing, like, like, serverless versions of what they do to, like, full, on prem stuff. And and, like, there there's more stuff I can't say.
swyx:Okay. Like, cool. But, but, you know, out of the, the many, many GPU as a, as a as a service providers out there. It's basically together in Fireworks. In my mind, they're the 2 leaders.
swyx:Mhmm. And this is the one that Sequoia picked. So, like, I think it I think they're doing fine there. They do need more, like, work in terms of their DevRel which is which is where I come in. They are hiring DevRel if if anyone wants to wants to join a a top tier.
swyx:Definitely one of the top executing, AI infrastructure companies I I've come across. And, surprisingly no head of DevRel yet because I refused to to to join a company. So so I'm kind of like interim head of DevRel while we while we figure out who who actually runs DevRel there.
Jack Bridger:Yeah. Yeah. That sounds like a very good opportunity.
swyx:Yeah.
Jack Bridger:Just because, oh, yeah. So then I wanted to ask you a bit about, like, actually getting into angel investing and if someone wants to, like, invest in dev tools.
swyx:Yeah. So You wrote a
Jack Bridger:great post, I should say, that people should definitely read the article that you wrote about us.
swyx:Yeah. Because I don't remember it anymore. You don't get into angel investing with the expectation that it'll make you money. Okay. Because you probably won't And even if you do, you won't see it for 10 years.
swyx:Right? If you're successful, the the the the company won't exit for a while because it it will just keep going up and up and up, which is what's happening with Supabase. Yeah. And that's fine. That's great.
swyx:You you have to be happy happy and cheerful for them. I think the so the the first thing obviously is you have to have some money. I thought I was richer than I was because of Netlify. And because when so I counted my Netlify net worth and I counted my temporal net worth, I was like, oh, I have some, like, spare cash that I'm not doing anything with. Like, why not
Jack Bridger:Yeah.
swyx:Start investing? And then and then the SERP ended. And I realized that the paper money doesn't mean anything. And so now I'm now, like, less rich than I thought I was. Anyway but, you yeah.
swyx:You start investing with some of your own cash and eventually what which you do what I'm doing now, which is that you invest other people's money. So as many investments as I as I seem to have here, it's not my money. It's actually money that other backers have given me to invest for them. And you can do it in 2 ways. You can either raise a rolling fund or a syndicate or you can be a scout.
swyx:And I chose the easier way, which is the scouts. Because rolling fund syndicate, you have to do a lot of paper, lawyer work, tax work on your own and scout stuff is all basically all taken care of for you. And you don't realize anything until the company exits, which is great. So yeah. I have a I have a good deal with, with CRV that does it but every firm every major VC firm has them.
swyx:They all want to build good relationships with you. If you're prominent dev tools, DevRel person, just hit them up. They'll they'll probably say yes. And, also then so then then the real question is why do it? Right?
swyx:Those are the mechanics of how to start. You should also by the way, you should also have, like, some amount of good industry networking. Right? So if you are a good podcaster, Tweeter, Linkediner, whatever, like, you'll get that. Others, you can kinda have more more or less have to come from their own professional work history.
swyx:So why do it? I think, one, like, there's a very obvious one which is that is basically my CV. So if I ever become a VC someday
Jack Bridger:Yeah.
swyx:I'll just go look and then they'll be like, good deals. Yeah. Like, that's the job. Yeah. I don't have to apply it.
swyx:I don't have to fill out any application form. I just show them this and they're like, yep. That's what we would want Yeah. In someone that we work with. What else do you need?
swyx:Yeah. And then, 2, you also get to work closely with the founders for anything that they need. So typically, my contract with founders is that if you need some DevRel advice, particularly before launch, if you're hiring DevRel, I, you know, I'm I I can help there. And through that, I can observe how they do what they do and, you know, all all a lot of these are, like, gonna encounter problems in in ways that, I would never have thought of before. And I think that's preparation for me becoming a founder myself.
swyx:Yeah. And going through those those exact exact same journey. So it's it's really good to basically be friends with these people who are, like, a few years ahead of you Yeah. If you intend to to go that route. I'm still not sure I'm gonna exactly go that route, but I definitely have some of these that I would now count as advising me more than I advise them.
swyx:And it's just a kind of like a two way street there Yeah. Eventually.
Jack Bridger:That's that's actually amazing. So one of the things that you that I read recently from you was the your post on DevRel.
swyx:Is DevRel dead?
Jack Bridger:Yeah. Is DevRel dead? Oh, good. I know we didn't talk about discussing this, but
swyx:Yeah.
Jack Bridger:Yeah. I think I've
swyx:I did one interview one one podcast with, ChangeLog on that. So people want, like, an hour's work of is DevRel bid? They can go there. Yeah. But we can do a short one here.
swyx:Yeah. Yeah. What what interests you?
Jack Bridger:So I've heard a lot of founders kind of, like, saying, you know, they don't need DevRel. You hear people say things along the lines of, you know, oh, this person wanted, you know, 250 k to write to do list apps or something. And, like, this kind of, like
swyx:That's exactly what some DevRels have. Yeah. That's that's their salary, and that's what they do.
Jack Bridger:Yeah. So I guess but it but then you I also see, like, examples where there's DevRels who are very, very valuable. Mhmm. You think of, like, Superbase. They have, like, loads of great DevRels, for instance.
Jack Bridger:And it's it's to me, it's clear that the skill set of people who understand the technology, understand, like, the real challenges that developers have, and then are very happy to create, you know, whether it's content or, like, experiences or, like, fill in the gaps that customers have, that aren't gonna be met by, like, they often just go on that, I guess. That seems important. Yeah. Whatever you call that. Yep.
Jack Bridger:Yeah. How do
swyx:you What's the question?
Jack Bridger:The question is, like Is DevRel dead? Is DevRel dead?
swyx:So it it it has died off. I think I tried to quantify in that post. It's, like, down 30%. So 30% is not a 100, but it's not 10. Mhmm.
swyx:So it's yeah. 30% dead. It's dangerous. The and the fact is that many many companies never needed DevRel. And many well, there are many roles which you don't need if you can do within the team.
swyx:You don't need HR. You know? Like, is HR dead? No. No.
swyx:It's just not required at some companies because founder does everything or or, like, you you achieve, hiring through some other means. And, like, I would say the same thing for DevRel and many other functions in a in a company. It's a it's a choice whether or not you want a dedicated role to that or you wanna spread out that responsibility to the rest of the team. But it is a job to be done,
Jack Bridger:I
swyx:think which is why you're kinda arrived at. Like, marketing and, you know, you can't build products and expect people to know about it or care about it unless you tell people about it and probably like a sort of there's there's a chart, of deluxe surface area chart. Just Google deluxe surface area chart. You see this? You need to do things and tell people about it.
swyx:Do things, tell people, do things. And so, most normal dev tools companies over index on doing too too many too many things and they are terrible at telling people about it. So effectively, what hiring dev role is is you hire people who whose job is to yap all day long to to to to artificially extend that. Some of them are good at it, many are bad. And, part of DevRel is also just kinda having the emperor's new clothes moment of, like actually there are many people with, like, 5, 10, 15 year careers in their work who are horrible at their jobs.
swyx:Horrible. And they should not be doing that job and they should find something else to do. But people pay for because they they need the job title to be filled more than they need that person. And because those person those those good people are very hard to reach. That's one thing there.
swyx:I also mentioned some other thing that, comes in this sort of dev relented category, which is that bottoms up and open source has been very challenged in the growing a company, success criteria in the sense that that's what DevRel is focused on, bottoms up and opens and particularly open source. You don't have to be open source, but bottoms up definitely.
Jack Bridger:Yeah.
swyx:As opposed to top down where you kinda reach into select big name companies and then sell directly to the CTO or whatever. I would say that the attribution issue for Dorel, same attribution as any marketing Yeah. Is you don't know who watches the YouTube videos and you definitely can't tie that to dollars. Yeah. So when you hit a recession or or the company misses its goals, like, the rail gets becomes the first thing that gets laid off because they can't prove their existence, the the validity of their existence.
swyx:And, I I do think that that is a reckoning that needed to be done and and is happening. That, like, you actually don't need that much bottoms up. You actually don't need to be loud on Twitter. You actually, can just ship products to a small set of big name customers and then have everyone else show up later. I can't name, like so temporal was like this.
swyx:Like our second customer was Snapchat. You know? And we didn't do a whole bunch of, like, be loud on Twitter and launch week 5, launch week 6. We didn't do any of that. Just just build what your big customers want.
swyx:And then the big customer likes it. The slightly smaller customers, like, look up and they go, like, oh, yeah. It's good enough for them. It's good enough for us. We'll pay you, you know, x amount of dollars.
swyx:And every one of those customers is worth 10,000 Yeah. 20,000 of the smaller ones. So why bother? Yeah. So there's a there's a lot of that reckoning going on with with, is devolved dead.
swyx:And I think I think the what I what I really try to phrase phrase it as, like, the the zerp devolved is dead. Like, the the zero interest rate, like, travel around the world giving giving talks and, really just seeing seeing the same people every talk. That that is basically that.
Jack Bridger:Yeah. But at the same time, I guess, traveling to traveling around Traveling
swyx:to San Francisco. Traveling Yeah. Conference is not there.
Jack Bridger:Yeah. Yeah. Exactly. Sign up. The link's in the bio links in the description.
Jack Bridger:Links in biology. Yes. But, like, you know, it almost becomes, like, almost like a sales, like, targeted is is that is that
swyx:sort of
Jack Bridger:it's like a very here are our targets. Like, you're an extension of the marketing team.
swyx:DevRel has basically always been an extension of marketing team. It's it's very much a lot of denial of by DevRel people who think they are better than marketing to to to say that that there are otherwise. You know, typically, you have bullshit. Like, DevRel is part of the product team. We give feedback from users to products.
swyx:And all you have to ask them to burst their bubble is how many of the recent features that were shipped were from your direct feedback. And usually, it's a small amount. Okay. There are very well functioning companies that that actually do respond to user feedback. But either they're not that busy, you know, your engineers don't have that heavy of a road map.
swyx:Because guess what? Like, at at a at any hyper growth company that is serving large customers and small customers, guess who wins? The large customers.
Jack Bridger:Yeah. As in they will focus on the features if if
swyx:They will often do. Because the other ones come with, like, a 100 x more money than the other ones. Yeah. Yeah. So it is a huge problem and challenge of founders and PMs to balance those things.
swyx:And it should not be a surprise to anyone that the, the larger checks, larger customers usually win. Yeah. That's like through an all fault of anyone. It's normal capitalism. Yeah.
swyx:That's just how it works. Yeah. And so then as a DM as a as a dev rel, my request get backlogged, shelves to the next quarter. Next quarter, I come around and say the same thing again. My my users are I I I'm out there in front of my users every day.
swyx:Talk to them. They say they want this. We don't we we fall short to other competitors. When can we have this? Backlog again and Yeah.
swyx:It takes, like, 3 quarters of that to just give up. Yeah. Yeah. Which is a normal frustration. I I think, like, I don't want people to hear this and go, like, DevOps is a is a joke.
swyx:Like, don't bother. I think you should fight for your customers. I think you should just be realistic that it is not 50% of your job. It is 1% of your job.
Jack Bridger:And most of your job is
swyx:getting with new customers. Getting getting your getting your company in front of people who've never heard about it before that could potentially be users of it.
Jack Bridger:And I suppose you have to somewhat consciously or subconsciously have to be more effective at that than sales teams.
swyx:Yeah. It's typically the people who are not good at the main marketing job that then they're like, oh, yeah. Their fault was never really about that. It was about the other stuff. Yeah.
swyx:Because they're retreating to to things that they can actually do better at if they suck at the other stuff.
Jack Bridger:Mhmm. Okay. And and sorry. And in that case, the other stuff would be like?
swyx:Direct sales. Let's say customer success. Like, that you're already a customer. I'll make you more successful as a customer by by teaching you how to use the tool you just bought. Bought.
swyx:That that happens a lot. Workshops, seminars, webinars, stuff like that. Basically, things further down the funnel. So you know TOFUMOFU Bofu? Top of funnel, middle of funnel, bottom of funnel.
swyx:They're really should live at top of funnel, but, eventually, you know, everyone's gonna retreat a little bit to middle and bottom of funnel because that's where more money is. It's it's more attributable. It's also safer. You don't have to it's like lower chance of failing because if you're already talking to people who paid an annual contract to use your tool, then they're kind of captive. So you know that you're gonna have some reception there.
swyx:Just, like, less risky. It's, it's good and bad.
Jack Bridger:Yeah. And and for founders, just to be, like, in kind of, like, what should I hire, you know, Deborah? I know this is, like, gonna be, like, it depends. But
swyx:Yeah. I know. Yeah. So, there the we have, like, a couple posts on DX tips. So, by the way, DX tips is my outlet.
Jack Bridger:Yeah. That's great.
swyx:Because I
Jack Bridger:No one should subscribe.
swyx:I, so it's dxdot tips for people who want to see it. It's basically because I no longer no longer identify as DevRel anymore. I care less about it but I also like like to vent on it every now and then. And people do ask my questions, my thoughts and I think I find it better when you just write them down as one resource that you can keep sending around
Jack Bridger:Yeah.
swyx:As opposed to just responding every single time custom. You see, I have more concrete thoughts. Like, I have a thought I have a post on how to do hackathons well. I have a post on how to do conferences as well. And so this what's the post that we're talking about again?
swyx:How to hire Deborah.
Jack Bridger:To it.
swyx:Yeah. I have 2 I have 2 on those because it's such a such a big big deal. But TLDR, try to do the job yourself as a founder because if you don't it's just like anything. If you don't do the job yourself, you don't really know what you're looking for. You don't really know what works.
swyx:And when you know what works, you can have a repeatable playbook to head off to to someone else. The bigger question comes, like, when you have decided to hire regardless of whatever whatever prep work you've done before or after, when you have to decide decide to hire, what kind of person do you hire? I like to I would like to advocate for people to hire engineers who have paid for your tool at at one of your lighthouse customers. That is the sweet spot in my mind. So what what am I not saying?
swyx:I'm saying do not hire content creator divas who are big on YouTube and expect a MPM install audience on your company. That usually fails. And then do not hire engineering leaders and then do not hire other DevRel who primarily are free tier DevRel. Mhmm. All these are, failure traps.
swyx:All these are traps of hiring for DevRel.
Jack Bridger:Mhmm.
swyx:There's there's one I mean, one element of DevRel which I really like is that they are at least half engineer. Mhmm. All DevRel should be able to code. Actually, that's too strong a statement. There are DevRel that don't code and they they do well, but they are very rare.
swyx:They're in the minority of them, and they usually work at the bigger companies like Google, Amazon. For the smaller start ups, you should probably want DevRel to to code. And one out, if you make a miss hire, is that you can just convert them to engineer. And a lot of them love it Yeah. When they when they go from DevRel to engineer and maybe engineer back to DevRel.
swyx:Yeah. So give them that option. Make them go through the engineering loop and then also do the content creation stuff. Yeah. Because, do you have a stronger dev person, employee in general?
swyx:And maybe they do a little bit of DevRel, like, sometimes, maybe Not. But just just give them an out as an engineer. Yeah. I think that that makes sense. And then the paid versus free tier DevRel is something that I I hear a lot of confusion about but to me it's so obvious.
swyx:Like, so many DevRels are just they get up on stage. They're like, use our stuff. It's free. They never ever and obviously, that gets a plus because you're giving away software that's valuable for free. Who doesn't like that?
swyx:Yeah. That's not challenging at all. What's harder is making me convincing me to pay for something. Yeah. And people who've done that in their companies have done the bake off with you versus all the other tools, at risk.
swyx:They put their lines their jaws on the line to a boss and, you know, risked rejection, denial, failure, whatever, and they still believe in your tool, hire them. Yeah. Because, you know, basically DevRel's tend to replicate themselves. They'll they'll they'll use the words that appeal to them, then that will attract other people who for whom those words also appeal and those happen to be your customers that you want.
Jack Bridger:So free tier, you don't want replicating of, like, free tier reasons to come check it out?
swyx:No. I don't want only free tier. Like, obviously, free tier is important and good. That's the reason people offer free tiers. And I do believe in freemium.
Jack Bridger:Yeah.
swyx:But free tier only is bad. Yeah. We did a lot of that in my previous job. Yeah. Okay.
swyx:Yeah. So so a a lot of this also is kind of like therapy for myself. Yeah. Working through mistakes that I make that I would not wish on others and I would not want to repeat at my own company. Yeah.
swyx:So I I, you know, I I I definitely feel like people maybe mistake my authoritative tone as judgmental. It's not because I did that. Yeah. Yeah. And so I I'm just trying to learn lessons and not repeat them.
Jack Bridger:Yeah. I think I I think that all of the things that I read that you put out come across that way and they're like, it's your lessons and it's like your beliefs and stuff. And I think there was some, like, some people were like upset about, like, the Devil is Dead thing, but I don't think it was, like, it was your views on, like, how things and your experiences and your personal, like and I feel like you're trying to actually, like, help people and, like, it was, like, why deny the truth in the sense of, like, there are challenges here? And I think you gave, like, really good, like, perception or, like, a really good, analysis of the situation, and then also that, like, there are companies hiring DevRel, and here are the, like, the ways that it can be important. And, and people that adapt to that are gonna actually succeed rather than, like, denying that, like, there is this contraction in, like, DevRel at the moment.
swyx:Yeah. Yeah. I I I think people should not care basically about the contraction in DevRel. This is a industry wide, sector wide thing that almost certainly means nothing for your individual company. Yeah.
swyx:Yeah. Your com individual company situation, you should examine that from first principles. Look at the people that you have. If you if they're missing that, you might wanna hire it. Yeah.
swyx:If not, and if you're the the box office of you as a founder or or company leader, product leader, then you take that job. Yeah. And and those skills and those jobs to be done are timeless and I'm I'm happy supporting that. Whether or not you have the official job title, you could care less. I do think that it could be easier for people.
swyx:So I I think in that post, I also mentioned the concept of the full stack DevRel. Yeah. Most DevRel in in, like, companies under a 100, maybe even under 200 are full stack DevRel in the sense that they basically operate as independent autonomous units kind of like you are right now. You book your guests. You decide what your your marketing is.
swyx:You decide what your topics coverage is. And you publish. You edit. You you do everything. Yeah.
swyx:In larger companies, like, no media organization works like that. Yeah. Right? They have editors. They have, back office people that that do the processing.
swyx:The journalists and the and the reporters, like, they do the front end of, like, interviewing and deciding what's relevant and to stay in the flow. And the media organization takes their job seriously enough to invest in the back office so that the front office people can do their jobs Yeah. Well. And, what effectively all DevRel is is a media company operating as a tech company. And the sooner that a tech company realizes that they are running a media operation, they should maybe learn lessons from real media organizations.
swyx:And maybe not operate at the same scale, but, the functionality and grouping of jobs probably makes sense in some sense. What else should I say there? Yeah. So the, I have concepts about the front, middle, back office of of Dervahl. People can read that on the x tips.
Jack Bridger:Yeah.
swyx:There's no I also I also mentioned this in my book by the way. Book plug. But that that that goes that goes back to my previous KodingCare. KodingCare handbook, talks a little bit about the front middle office for for developers. But I think for DevRel, that is necessary and, increasingly important as people want to systematize the way that they work with their community.
swyx:I don't know. Like, I wish more tools. I I could recommend more tools. You know? Like, they used to be like Orbit Yeah.
swyx:Or Common Room or Yeah. Stuff like that. They all none of them are working out super particularly well. I think Orbit failed.
Jack Bridger:Yeah. Yeah. Acquired by a postman or was that another one? I don't know.
swyx:I don't know. Yeah. And, so so, like, I there are other sort of community services. There's a lot of, like, draft dot dev used to used to be a thing. Iron iron something used to be a thing.
swyx:Basically, like, turn would turn a podcast into a blog post or the other way, blog post to podcast or videos or whatever. And I I wish there were more tools like that that were more successful so I could point to them as examples so you can use this. But all of them are experiments at this point. And, I think that the problem is, like, there are all these mechanics of doing the job. Like, produce one blog post a week and then like, you know, one webinar a month, whatever.
swyx:And there's no taste. There's no love. There's no, engagement in the community. Yeah. I think that's what people really need.
swyx:That that engage that authentic engagement. They can't really, systematize yet. I I think I think that someone out there could, but right now it's it's all always these, like, thought leaders that is that that lead things. And that's great if if you have a thought leader. If you don't, then you're shit out of luck.
swyx:And, I always wish that part of my my job with the X tips is to try to kinda demystify that process so that more people could kinda engage in in that conversation because they just never been taught how to. It's not a thing that you learn in university or anything.
Jack Bridger:Yeah. This is, this is really helpful.
swyx:Thanks, I guess. Yeah. I I don't know what to say about that apart from, like, happy to answer questions on on the excerpts. I tried to timebox that because I think I need to my, like, next level of success, you know. Like, it's again, with with going back all the way to the initial conversation about winning, is this the thing I'm supposed to do, like, advising on DevRel?
swyx:I don't think so. I don't think, like, people, like, super need it. Like, there's many other qualified advisers out there. My advice isn't gonna be, like, that much game changing compared to, like, other people. Except except maybe in AI engineering.
swyx:Like, obviously, that's the that's the place I carved out. Yeah.
Jack Bridger:Practice. I think you've I think you're one of those, like, people that's, like, good at many things. And I think, I would still consider you as, like I think that you have a very, very rare ability of, like, communicating concepts and thinking in systems. And I think that you probably are, at least, that I've come across in the kind of, like, DevRel space, the best person. Like, you have like a I think your your purpose in the AI landscape is like is is is bigger probably.
Jack Bridger:So
swyx:Yeah. I think so. I I think, like, I love what I love about it is I can repeat myself and not feel bored. Yeah. That's that's that's a sign that you've, like, you have something you believe in.
swyx:Yeah. You know? And I've done that. I've done many many podcasts just evangelizing the idea. Yeah.
swyx:People get jobs on it and sometimes they don't even know that I was involved in it and that's great. Yeah. GitHub recently launched a Hugging Face competitor and right across their front page was AI Engineers. Yeah. And, it's
Jack Bridger:just this
swyx:market that, like, it's it's it's growing and I I think, you know, if we serve it, well, I think it's it's own industry that's gonna emerge that I I had a had a part to play in and I think that's beautiful. The the other thing now is to build services rather than thought lead. And that's something that I historically haven't done before and I'm I'm doing now. So I'm I'm building AI news, which which you know about.
Jack Bridger:Yeah.
swyx:And seeing what it could become. Mhmm. Just had a chat with Andrej Kharpathy this morning about, yeah, casually casual name casual name drop there. Woah. But he is I I would consider him a mentor of mine.
swyx:I've considered trying to work with him more, but he's very independent minded and obviously doesn't need any money or anything like that. But there's a general desire for, general purpose summarization, and teaching. For him, he wants to make the best course in the world for AI. Yeah. For me, it's, you know, there's a lot of in generative AI, there's a lot of, like, put in this prompt and then it becomes, like, 10 different things.
swyx:Mhmm. There's a lot of signal that becomes noise. Yeah. And I want to take noise and condense back to signal. And I think that too many people are doing the thing where, like, signal noise.
swyx:And then not enough people are doing noise signal.
Jack Bridger:Yeah. And you that's kind of what you do as well. Right? Like as a person, like I feel like that's one of the things that you've been pretty good at.
swyx:Thanks. I try.
Jack Bridger:I think I think so. Right? Even when we're talking about, like, taking this, like, condensing things into, like, kind of principles or, like
swyx:Yeah. I like that. Yeah. I do like that. I wish I was so I my problem is I am too much of a hoarder and completionist.
swyx:And the people who are best at at signal to noise sorry. Noise to signal are minimalist. Do you want the who's the other guy? Jack something? The visualize value, you know?
swyx:Jack Butcher. You seen him?
Jack Bridger:Oh, yes. Yes. Yes. Yes. Yes.
swyx:The you would only be the kind of guy with, like, 2 words in your bio, you know, and, like, his he he, like, just tweets an image where it's like that and it's like that and it and then it's like you get 10,000 likes. No. I'm not that kind of guy. I I'm a word cell. I like, I just I I have so much words.
swyx:So to I think that that is too simplistic and I don't like that. And sometimes it's very inauthentic in this in how the minimalist take the most engaging, rage inducing headline and just tweet that. That's it. So I like nuance, which is a flaw because nuance is also noise for some people. It is.
Jack Bridger:Yes. But I feel like people resonate to, like, like, are drawn to their own You don't like, amount of, like Yeah. Abstraction. I think
swyx:you should you should have tunable noise. So, what I what I would like, I don't know if I don't know if we're we're gonna build this, but I would like an basically, an urban dictionary. Have you seen urban dictionary?
Jack Bridger:Yeah.
swyx:Dictionary where you can kinda tune your your explanations of any everything anything to you, to what the the depth of nuance that you would want. So I want I literally have it in the spec sheet somewhere. I want the 2 word version or something. I want a one sentence, 1, 2 sentence, 1 paragraph, 3 paragraphs, 1 essay, 1 book, something like that. And just constantly update that to and and then you also want apart from length, you also want to vary by, like, thinking hat.
Jack Bridger:Yeah.
swyx:So you want the funny version. Right? Because there's always there's a little bit of truth and humor. Yeah. You want the cynical version because there's always a bit of, truth in, like
Jack Bridger:alright.
swyx:If you strip away the hype y bullshit, like Yeah.
Jack Bridger:What are
swyx:you really doing? Yeah. And then you also want the hype y version. You want the you want the, like, what is the best possible version of this that could possibly be even if you fall short. Yeah.
swyx:And somewhere along the line is truth.
Jack Bridger:Yeah. But
swyx:you're never gonna if you only have one version of truth, you're never really gonna hit it because you're it's impossible to represent truth in in one definitive statement. You'll probably need, like, 6. That's something we're ideating on me and my cofounder. I don't know if he'll ever see the light of day, but kinda do want it to exist in the world where we have, you know, rotten tomatoes for everything.
Jack Bridger:Yeah. Yeah. It it the levels of kind of abstraction, detail reminded me of those that talk at the first AI Engineer Summit.
swyx:Yeah. Amanda, Amelia Wannberger. Wannberger.
Jack Bridger:Yeah. Wannberger.
swyx:Yeah.
Jack Bridger:Yeah. That that was the first time I saw that concept where you're using other lens to have, like, different levels of, like, detail.
swyx:Yes. Yep. So, she was working in Adept. Adept, failed shortly after. But she still continued to do interesting work with, email.
swyx:Like, just today, I was look I was obsessing over the the latest product that she's working on. She's she's back at GitHub now. She used to work in GitHub Copilot. Now they're basically at the GitHub Innovation Lab doing stuff
Jack Bridger:like that. Cool
swyx:stuff. Whatever they want. And, like, yeah. What I love about the conference is, there's a forum for people like her to present their work because her work would not be eligible at NEURIPS or ICML or any other of these, like, academic conferences because it's not academia. It's product.
swyx:Yeah. It's engineering. And that's what we're focused
Jack Bridger:on. Yeah. That's amazing. I think we're probably at At your time? At your time.
Jack Bridger:Yeah. That's it. But this was amazing. Yeah. And I enjoyed it.
swyx:Hopefully, there was something that you
Jack Bridger:were great. It was so great.
swyx:Okay. Yeah.
Jack Bridger:From my perspective, I hope everyone listening and enjoyed
swyx:it as much as I did.
Jack Bridger:A little bit rant.
swyx:I was worried that we spent too long at the start philosophically, but whatever.
Jack Bridger:Well, people can tell us if we did. From my perspective, we didn't. Yeah. Thank you so much. I'm sure
swyx:on this this Thank you for your continuing to grind. Like, you are one of the more authentic voices out there for for, like, dev tools. And, like, you know, I think your your personal intellectual curiosity, like, shines through. And, you know, I hope you never lose sight of that.
Jack Bridger:Thank you. And, and where where would you point people to? We've talked about d x dot tips. Oh, boy. Small would we Yeah.
swyx:So, like, probably the highest, most relevant thing is latin space. So go to latin dot space. It's the podcast that I run with Helessio in this studio every week. And that's, the most polished version of what with everything I have going on. And then twice a year, you can come to AI Engineer.
swyx:That's ai.engineer. And then I'm working on a startup called small dot ai. And right now we we do a daily newsletter that is definitely not polished but is endorsed by some of the top names in the field.
Jack Bridger:Andrei Kapafes. Yeah. Yeah. That's amazing. Well, thank you so much, Sean.
Jack Bridger:Thanks everyone for listening. Thank you.
Listen to Scaling DevTools using one of many popular podcasting apps or directories.
