215: Staying Technical as a Manager
Welcome to Python test. Matt Makai is here. I met Matt many years ago. So, but you're you're up to new things now. So, Matt, can you introduce yourself and let us know what you're up to?
Matt:Yeah. Sure. My name is Matt Makai. I am the VP of developer relations and developer experience at AssemblyAI. Before that, I was, 9 plus years at Twilio where I was also in developer relations.
Matt:And then my entire career before, either AssemblyAI or Twilio was as a heads down software developer. So I started out in, Java Java World, J2EE, a lot of, you know, big iron, software projects, that were closely associated with, you know, usually, like, mainframe transition type stuff. And then I got into Python, and I was like, oh my gosh. Like, I can hack on stuff in Python for, like, an hour at night, and I feel more productive than 9 hours a day in Java. At least at the time, the Java ecosystem was, kinda slow moving.
Matt:So, yeah. So that's that's my my background. Also run full stack Python and a new side project that I can I can talk about as well?
Brian:Well, I'm glad that you tried Python because it really could have been any language that you tried after Java. Yeah. Like,
Matt:and I I did try Ruby. Ruby on Rails was super hot back then. This was, like, before the days of Node. Js. It just didn't, like, quite click with me.
Matt:I thought the, x, Ruby is more implicit, Python's more explicit, and I just that kinda clicked with me for some reason. So you gotta kinda pick the programming language that matches your brain.
Brian:Yeah. Yeah. When early on, Pearl matched my brain. I'm glad that Pearl doesn't match my brain anymore. But
Matt:That's incredible. That one that one also did not stick for me, so I'm I'm impressed.
Brian:I do wanna talk about your your new site project, but and maybe this will shoe in. But, staying technical as a manager, and this is, actually, this is this is very close to me because I'm I'm a technical manager. But, Matt, you are you in a management or leadership role now?
Matt:Yeah. Absolutely. So, currently, I have a team, of 7, which actually I just handed off a few folks. So, while I was at when I joined AssemblyAI about a year ago, I was running marketing, but we really were going for a very developer focused DevRel bent, and sprinkled in a little of, like, demand gen and, content marketing. So, was recently managing a team of 11, but before that at Twilio, I had a team of 35.
Matt:So I had managers that I was managing, for the individual teams. So, been managing either in, you know, sort of the, the player coach model, which I typically think of as, you know, if you're managing, at most, like, 3 to 4 people versus a sort of a pure manager where you spend the majority of your time managing, which is, like, often 8 to 12 people. So I've been manager since, I don't know, probably 10, 11, maybe 12 years now. Oh, wow. Yeah.
Matt:And, and and it's similar situation to you, at many of those times where, you know, I'm I'm both coding and and doing a lot of code reviews and things like that as well as, managing as well, which is a a tough balance.
Brian:I'm a I've been in management for probably about the same time you have, really. It is always a fear. And one of the reasons why I'm interested in this topic is to even if I were to to, like, leave leave a lot of the technical stuff to other people, that at some point, there's gonna be a shrinkage and they don't need as many managers, and now I'm not one of the top coders anymore. So, yeah, that's one of the fears of being a especially a lower level manager is that, you're gonna let everybody else pass you technically, and then you're not one of the obvious people to keep around when layoffs come. So, yeah, that's that's and then also trying to find another job.
Brian:So, it's difficult finding a job because people there's a there's a lot of times where people are like, oh, well, you're a manager. We don't have any manager roles. I I can I'm okay with being an engineer too. Well, yeah, but you're not right now. So, things like that.
Brian:So how do you stay technical as a manager, Matt?
Matt:Sure. Well, let's let's just take us a little bit of a step back, which is just, like, obviously, is an incredibly pertinent topic as we've had just this hundreds of thousands of layoffs across the software ecosystem over the past, like, you know, 18 to 24 months, and there probably will be more the first half of this year. It's it's it's a very diff I would I would I would say that most people I've talked to have said that this is much more like a 2,001 kind of dotcom crash, situation for for layoffs as opposed to even great financial crisis wasn't necessarily even as bad, as it is right now. So that's kinda saying, like, I think in a lot of cases, like, you really can't prevent getting laid off. A lot of times, it's really just like a numbers game.
Matt:And so I think a lot of times people take it really personally. I do think there's things that you can try to do to stack things in your favor a little bit. And so this goes to, like, your point of, like, you know, if you think that you're gonna, like, lose out in coding to someone who's, like, newer, I actually think what matters, a lot more, and I say this as someone who's, like, gone through layoffs myself, is, like, is somebody, like, just wildly valuable to the business. And, obviously, skills coding is one of many skill sets. And, typically, if you combine multiple skill sets, like, this is what, you know, developer relations is is a lot about.
Matt:It's like you are not only coding applications, but you're also writing technical tutorials or you're creating YouTube videos or you're giving a talk at PyCon. It's the combination of those skills that are rare. And so if a company really needs that for its business and knows that it's gonna be extremely difficult to replace you, that gives you a little bit more job security. But, again, it sometimes, they just decide to cut teams arbitrarily. So, I do think that, like, coding I would look at coding as, like, one of many skills that you can, you know, kind of try to stack the deck in your favor.
Matt:But, you know, it's been a tough time in the industry. Like, I don't think anybody should ever take it personally that they got, like, laid off. Like, I got laid off from Twilio. It actually turned out to be awesome for me, but I'm very fortunate that I just happened to be in, you know, a little bit of the right place at the right time, when it happened, and I was I was ready for it. But it's, you know anyway, so I think coding is one of these things.
Matt:So, like, going back to your original question of, like, you know, how do I stay up to date with coding? The big thing for me is, like, I work on in my in my spare time, and I have a family and everything. Like, you know, like, I I I don't have, like, a ton of spare time. I tend to, like, do a little bit of this in the mornings before, like, my son wakes up. But, like, I work on generic versions of a problem that I face.
Matt:And so what that allows me to do is, like, think about things in a way that, are separate from the particular details that I face, like, at the company. So let me give you, like, 2 examples because, like, both of my side projects have essentially been in this model. So I've worked on full stack Python for about 10 years, taken a bit of a break from it. I will go back to it eventually. But it was like, I wanted to make sure that I really understood the full scope of the Python ecosystem as a when I was a developer evangelist at Twilio and still a software developer.
Matt:And so I would just write, you know, pages and and find the best articles on a new web, frame web application framework that was coming out or a new, database that had come out. Like, the time between 2,007, and 2,000 even even up until recently, the explosion in the number of technologies around, like, web development in in the Python ecosystem was, like, it was really hard to keep up. And so I just used my side project as a forcing function that I would keep up with all of that stuff.
Brian:Yeah.
Matt:And, and the coding part of it was, like, you know, Jinja templates, and I was using Pelican to create a static site. And I would like, you know it wasn't super coding heavy, but it was enough to, like, keep me, you know, really familiar with my tools. And I think that's that's one component to it. It's like, if you wanna still be able to code even as you go into management, like, yeah, you're gonna have to spend some time on the side just like most software developers do, in order to, like, make that possible. And if you work on something that is gonna make you better at your job, but is, like, separate from your job so you can think about things in a little bit different way and doesn't you know, you don't wanna just be working on work as your side project.
Matt:You wanna be working on, I think, like, an abstract problem. That's kind of how I I tend to approach it. And there's a bunch of, like, tactical stuff we can talk about as well, but that's, like, my conceptual framework. And I have a a new side project called Plushcap. It's plushcap.com.
Matt:It allows me to see the landscape of every developer focused company that's out there and how they're investing in their docs, how they're writing blog posts, how their YouTube channel is doing. So I can see every single company and how they're doing from publicly traded company all the way down to a pre seed company. And then I can say, oh, that tactic is working really well. Maybe we should experiment with that on my team. So you can kind of you can self sort of learn from others, and I do that through the tool that I created.
Brian:We may need to have you back on and just talk about Plushcap at some point, because that sounds fascinating, and I don't I think I've got, like, 20 questions around it. So
Matt:Okay.
Brian:One of the ways places where I I looked at, trying to stay technical, and I and I've heard this before, is to have it be something completely different than your your normal day job. Like, trying to, I don't know, do a a forum for one of your hobbies or something like that. But I love this idea of, generic versions of a problem that you're working on your company. It's a it's something that when I was, when I started writing about testing, the I was definitely thinking of of problems I had at work and trying to solve them, but I wanted to do it open in the open, and I couldn't do that. I can't write the problem down.
Brian:Well, 1, like, that I'm facing at work because it's, it's it's the intellectual property. I can't share that.
Matt:Right.
Brian:But also, it would be boring for everybody. It's a it's it's usually, like, super deep detail into something that only we care about, or a few other people do, but that's not the part that's not the point of the problem. The problem is some sort of, software, architecture problem that has nothing to do with the domain. It's like, like, for instance, the on with the with testing, the the early on, I I had a need to have multiple checks. So I I wrote a plug in called Pytest check that I still maintain that is, normally in in all tests, you can only assert once.
Brian:If it fails, it stops. And I needed to have more. We needed that in work for waveforms, because we've got, like, you know, both a a bandwidth and frequency and other things you need to check. And I don't want to stop checking if one of them fails. I wanna see the whole picture.
Brian:But I knew that that was gonna be generic enough for other people, so trying to solve that in public. Also, I didn't know how to solve it. So solving it in public and writing blog posts and everything helped, have other people come on to to correct me and say, oh, you have you thought about doing this? And that's how I started getting in touch with some of the core contributors of Digest, because they were the some of the best people to say, oh, yeah. This doesn't work because of this, and you should look at this internal variable, and that'll help.
Brian:So when you're writing full stack, did you have a a lot of feedback from people? Would would people contact you and say and you're you're in left field, man. You you you're wrong about this. Or did you get very many corrections?
Matt:Yeah. I mean, quite a bit. I mean, I think the thing that helps there is, like and I think, basically, this is it seems like it matches your your, experience as well. It's like, if you just try to approach things in a way that, like, hey. I'm I'm doing this, and I'm I'm also learning at the same time.
Matt:Like, I'm not the definitive. You know, like, I have enough experience to be able to write in a neutral, but still, like, enter entertaining way about a topic, like, informed enough, but also with the humility that, like, hey. Like, I actually don't know everything, and I very much, like, welcome feedback. I think that is often what helps to differentiate projects where people will give you, like, valuable feedback versus just being kind of a jerk. You know, because if you say if you're kinda putting yourself out there like, I know everything already, like, then I think you just open yourself up a little bit more to, like, people that are, like, you know, disagree with you.
Matt:So yeah. I don't know. I think on on full stack, I definitely, you know, got a lot of feedback along the way. You know, a lot of what was really helpful was actually, like, I I, you know, I work on it every single day. And then, you know, sometimes you get some interesting feedback through social media, but, like, a lot of times it's like going to PyCon and just, like, talking to people and being like, oh, I built this thing.
Matt:Like, what what do you think about this? And then, like, they would take a look and email me back, like, a a couple weeks later and be, like, hey. I read through your stuff, like, and give me feedback. So I think it's a combination of, like, I I personally really like the fact, like, you get some feedback online. But, also, if you bring that into conversations in an appropriate like, in appropriate conversations, like, people also give you really good feedback, through that as well.
Brian:Yeah. I'm would you So you you started full stack, when you as you were in a full time engineer. Right?
Matt:Yeah. I did. Yeah. Back in 2012. Yeah.
Matt:In fact, like, the reason the reason why I did was because, like, I was home for the holidays, and, this is when I was, like, still pretty young and young and single. And, no, like, family or anything. And I just, like, went on a, coding, like, vendor because I had just, like, done all of this, like, Python stuff, like, deployment related stuff. And the full stack Python started out as, like, almost completely deployment related information. And I really just wanted to get it out of my brain.
Matt:It was like a way of, like, me relaxing during, like, the holidays. It was just like, I have all this stuff that's rattling in my brain. I'm just gonna put it down on on paper and publish it online, and then it just kind of, like, snowballed from there. So, yeah, I was a I was a full time developer when I actually started it.
Brian:Well, you you brought up one of the things of keeping keeping up, trying to make yourself less easy to lay off is is having more than one skill. Having skills in not just coding, but also, communication and public speaking and, and teaching, write writing. Do you think that your involvement with with full stack Python and with speaking at conferences and things like that led you to be somebody that would make a good developer relations person?
Matt:Yeah. Definitely. So I and and, that was one of the reasons why, like, Twilio originally hired me. I was I was developer evangelist for them. Twilio was still pretty small.
Matt:They didn't have a they had a few developer evangelists, but that was kind of their primary marketing, effort in order to market to developers to use the API. So it was, yeah, I was, like, you know, doing a lot of writing and and speaking, but I was still, you know, 90% of my time was spent coding. It's kind of the perfect, match for what you want in someone who is a potential, like, developer evangelist, developer educator.
Brian:Those skills, I I didn't think that I would value at work as much, but it but I do. The, I knew I wanted to be able to do public speaking more and, to write better. So, blogging, podcasting, speaking at conferences, something I wanted to do on my side projects, and just to get involved with Python. But it it turned out that it has helped me at work. I'm a better I'm a better speaker at work.
Brian:I'm I'm not I'm not afraid to talk at a at a meeting or anything like that. It doesn't even matter who's there. The some of the the large managers can be there, and I can stand up and talk without sweating too much. I think that that's comes from a lot of my involvement with the open source community. Yeah.
Brian:And I don't know. Were you before you were a manager of, like, up to tons of people, were you somebody that was comfortable with, with leadership or with public speaking, or was that something you've learned?
Matt:Certainly the public speaking, but I think, you know, to be a really good manager, you have to be, just constantly intentional about, wanting to be a really great manager. And to be fair, like, I actually didn't particularly want to be a manager. Like, I thought I would just be an individual contributor. I just I just mostly like, once I had the opportunities, I was, like, you either have to grab ahold of them or not. Right?
Matt:And I think, for better or worse, like, a lot of managers kind of go into management because that's kind of where they're pushed, but they don't really wanna take that opportunity and be amazing at it. And I still have a ton to learn. But, like, I am extremely intentional about the way that I manage. And, you know, I I I want to be a really good manager in addition to still having all of my technical skill set. I'm not sure I necessarily see that same desire from every manager.
Matt:I think some of the, you know, unfortunately, some of the the the weaker managers that I've I have personally had in my career are the ones that think that because they got the title, now they make all the decisions, when in fact, like, that's typically, like, not how things work. Right? Which is why, like, I've I've been very intentional about, you know, studying servant leadership and what are the right ways to manage folks that are particularly technically talented and and, creative. You know, there's different management styles, but you also have to, like, kind of know your own personal preferences and things like that and kind of come up with something that works, for you and the people that report to you and the team that you have at that time. So, all there's, like, so many variables in management, and I think that is a one big thing that people have to recognize is, like, there's, you know, there's different ways that you need to approach this problem just like you need to approach different architectures, you know, technical architectures or whatever.
Matt:And these are not like, I am a manager. I know how to manage. Like, you don't you may know how to manage in, like, one for slice of a situation of a team and completely not know how to manage in in other situations. So there's, like, definitely a self awareness component to it as well.
Brian:Well, I wanna bring up the the a management style that, grates on me is maybe it's not a style, but the managers that actively try to make sure they look like the smartest person in the room either just take credit or just dismiss the some of the accomplishments of the people that are working for him, because they want they wanna be looked at seen as the superstar.
Matt:Oh, and I think that's actually how you become a better manager if you want to be a better manager is you just think about those traits and why they were so grading, and why they actually caused you as as an individual contributor to do worse work. Like, not act it has the opposite effect. Right? I know that when I highlight the accomplishments of folks that report to me or different teams or things like that, that makes them wanna do, like, more of that. It makes them want to be better.
Matt:Right? Because they they get recognized for their accomplishments as opposed to feeling like somehow someone else is gonna take the credit for that. But you also have to have the right culture. I mean, some companies have such, you know, sort of, like, toxic cultures where even that type of, thing is not possible. But I don't know.
Matt:I think that also feeds into, like, I wanna work at places where I can recognize other people and they get recognized for for the things that they do. And everyone feels good about that situation. Like, I can deflect any sort of credit onto the people that truly did the work. And and that is how, you know, I think you create a much better company culture culture overall where people are actually, happy and productive.
Brian:Yeah. So, is does Plushcap, do the technologies behind that line up in any way with what you do on a day to day basis? Or
Matt:sure. So I do, I would say yes. So it's it Pluscap is built in Python Django. I you I deploy it to to, you know, just a virtual private server. I use Ansible, a lot of, like, sort of, like, tried and true tools.
Matt:So my opinion, when I and and, again, like, everyone kinda has to develop their own style. But when I use side projects, I actually tend to use, like, the same tools that I know. So I'm using, you know, my code editors, VIM and TMUX, which I've been using for 15 plus years. You know, Django, which, amazingly, I've been using for 15 plus years. Still a unbelievable framework.
Matt:Obviously, Python, I continue to increment on the versions. I've gone from Python 2 to Python 3, you know, all that. And now I'm taking advantage of, like, Python, you know, 3 dot 10, 3 dot 11. So so incrementally, I'm learning through, some of these tools as they, you know, like, Django just went, what, 5, 5 o or whatever it is. So I'm still keeping up with the tools, but they tend to be the same tools that I use consistently.
Matt:So it's I think that in some ecosystems, this might actually be a really difficult thing to do. So, for example, like in iOS. Like, I actually think you need to spend a lot more time, learning Swift and iOS and keeping up with it that may make it actually really hard to do on the side. But for me, for Plushcap, it's a Python stack, and then I add tools in that I think are gonna be uniquely relevant, and and often do, correspond to my job. So, for example, there's a tool called Ollama.
Matt:It's a way to run, large language models, on your local system. And I run, like, the, the Dolphin version of, like, Mixtral, 7 8 times 7000000000 parameter model, locally. And I actually use that to create, summaries, for, blog posts. So Plushcap basically takes in all of the public data and information, blog posts, site structure, YouTube videos, all sort of things, and basically creates a profile for various companies. And so and part of that, I take all the blog posts.
Matt:So I I have a database of, like, tens of thousands of blog posts that companies have written, and I'll create summaries for each one. So I actually run those through a local large language model, and create a summary out the other side. So you can take, like, a 5,000 word blog post and just distill it into a single paragraph that will give you the gist of that that whole thing. Right? So that way, you're actually able to consume I I can actually consume, like, the fire hose from 100 of companies and actually understand what they're working on and what they're doing.
Matt:So that's how I, like, mix in tools that are relevant, that are new, that actually give me, like, an understanding of, like, kind of, like you know, they actually help do help me at work to be better because I understand how, you know, this this particular LLM, is is valuable for certain things and not for others.
Brian:That's pretty cool. Is Plushcap just a private thing or is it available for
Matt:other people? It's available. It's it's plushcap.com. Okay. And, it's you'll just see it's like a database of all these different companies, you know, and you can, sort of sort them by different categories and click into them and see their blog content and things like that.
Matt:So, for me, I use it actually mostly to I use it, in part because I that's actually how I operate my teams. I can say, like, you you know, hey. We wanna publish, like, x number of blog posts and, you know, like, it it basically gives me enough metrics, of what I call input metrics, like, input into, like, how much our team should be producing, that then I can combine that with our, you know, proprietary metrics and really make sure that the team is, like, operating, according to, like, the way that we we need to run the business and actually drive the business results. So yeah.
Brian:Any, any sneak peek at some of the answers? If I if I've got a limited amount of time, should I write blog posts? Should I go on YouTube?
Matt:Oh, for actually doing developer relations stuff?
Brian:Yeah. What's the what's the biggest bang for the buck right now?
Matt:Okay. So this is my personal opinion and, you know but I think each company really only gets to be known for, like, one thing. And so you gotta figure out what your thing is as a company if you wanna be an exceptional company. So for Twilio, it was developer events. I went to 86 developer events in 2014, my 1st year.
Matt:We did 500 events that year. So, like, Twilio just became known for, like, you meet a Twilio developer evangelist at PyCon or at a Meetup or wherever. Right? You know somebody personally from the company. Stripe became known for their, like, exceptional developer docs.
Matt:Eventually, like, Twilio actually became even better known for, like, a lot of the developer content, particularly as we, like, enter the pandemic. Assembly AI is known really well for our YouTube channel. So there's actually a list of all the top YouTube channels by subscriber count and view count, on Plushcap. And AssemblyAI is, like, you know, basically, like, the only companies that have more are, like, JetBrains has a little bit more of subscribers, the big three clouds, OpenAI. That's basically it.
Matt:Like, all the publicly traded developer focused companies don't particularly have, great YouTube presences. And so, like, you just have to pick the thing that you wanna be really great at and orient the team to make sure that they're driving the results that you want out the other side with with that tactic. And then you can layer in other tactics. Like, we do a lot of work on our docs. We do a lot of written content, but we're really, really well known for the YouTube content.
Matt:That's what, like, I'm constantly talking to other companies about. Like, they're like, how do you how are you doing, like, this YouTube content? It's like, well, you know, there's there's a whole bunch of stuff, but that's what we wanna be the best at.
Brian:Yeah. Well, like, for a while, I don't know if they're still like this, but DigitalOcean was into, like, user documentation, for a long time.
Matt:That was, yeah, that was actually the example that I I meant to bring up as well. So they actually mentioned their, tutorials in their s one filing when they IPO'd. They said, you know, we have 3,000 plus tutorials. They drive x amount of traffic, and this is actually something that powers our business, which is kind of interesting considering they, like, I believe laid off that whole team fairly recently.
Brian:I wasn't gonna bring it up, but I was like, oh my god. That's that's, that's why you're known right now. Why would you.
Matt:Exactly. Exactly. So, yeah, I don't know. They're they're struggling a little bit.
Brian:Yeah. Still still and I don't wanna diss them too much. I don't know what the internal decision on that was for, but yuck. Anyway, okay. So, very interesting.
Brian:So trying to be known for one thing. So, well, I actually, for individual, like a small company, just 1 person or a handful. Mhmm. Actually, usually just 1 or 2 or 2 friends or something doing something. You can't spread yourself too thin.
Brian:So I like that idea of, like, just pick what you're gonna do. Is it gonna be, you know, focused on videos? Like, a good example is, Fast AI or Fast Fast API. Sorry. He was just really great at awesome documentation.
Brian:There you shouldn't have a question because it's Sebastian's gonna have the answer right there. He I don't know if he was doing YouTube videos. If he was, I wasn't paying attention. And, he does show up at, like, conferences and stuff, but having the the just the nail documentation Right. Was part of this growth, I think.
Brian:So
Matt:Yeah. Especially as an open source project. Like, the the docs are exceptional. People are gonna notice that. Right?
Matt:Because a lot a lot of projects, unfortunately, don't always get that right.
Brian:Yeah. Anyway, writing, you you did your you you're still writing some. I imagine it's hard to stop writing once you start writing. Are you still, like, posting or blogging or anything, lately? Or
Matt:Not as much lately. I will get back to it. Frankly, I've just been spending so much time coding and really enjoying that, and it basically absorbs all of my my my own time. I think the other thing too is, like, we are, you know, undergoing this dramatic shift in what is possible with software right now by using large language models, even things like Assembly AIs, like, automated speech recognition, ASR. Like, the just the what is possible now is, like, so different than it was 5 years ago.
Matt:Like, the only thing that I can even compare this time, this the excitement of this time too is, like, when, like, Ruby on Rails and Django kinda came out, and I was like, oh my gosh. I can build web apps, like, so much easier. This is amazing. Right? Or maybe, like, when, like, iOS, you could start building, like, mobile apps and things like that.
Matt:That was, like, amazing. But, like, now, like, large language models truly are creating, like, a very different, computing model. And so I just spent a lot I spent a lot of my time, like, evaluating LLMs, learning about prompting, frankly, like, watching others. Like, sometimes you have something to share when you want when you feel like you have enough that's, like, known and other people aren't talking about it. And that's why I, like, wrote full stack Python originally.
Matt:But right now, like, I'm still spending so much time learning and actually doing this stuff myself, until I eventually get to the point where, like, I I have more to share on the topic. So I think, you know, like, kind of like what you talked about earlier. Like, you know, you kind of have to do the work for a while before you feel confident in that you're what you're sharing is, like, the right the right thing that you have something of value, to provide to other people. So, yeah. So short answer is, like, right now, spending a ton of time on with LLMs, working on Plushcap, tons of time coding, coding plus LLMs.
Matt:And, eventually, I'll shift back into a mode where I'm writing a little bit more. I think that'll actually be a little bit more on the developer experience angle because that's actually where I have much stronger opinions right now versus, like, pure programming, which was like, sort of, like, full full stack Python side. I think it's fascinating to just, like, sign up for developer services, particularly like these, there's a lot of these machine learning platforms. Like, just like going through and just, like, studying, like, how do they do developer experience? The good, the bad.
Matt:You know, there's a lot of lessons to take away and learn. That's actually probably where I'm I'm starting to get a little bit of the itch to share a little bit more. And, you know, it's obviously something I think about a lot at leading developer experience, for AssemblyAI. So, that's hopefully where, and I maybe it'll be written, but it might also be YouTube stuff because I have this incredible team doing YouTube content, and I don't do, anything myself. So I feel like for credibility with the team, I gotta do a little bit more, YouTube content.
Brian:How, I hope you don't mind me asking, but how how much, like, large language model stuff goes into creating content? Are you using it for, like, generating first drafts for blog posts or anything like that?
Matt:I don't think anybody on the team does. You know, a lot of it just comes down to the fact like, large language models are good, can be good at, creating outlines or, like, brainstorming, and they're good at summarization. But they are not particularly good at, or at least, like, it takes a lot of effort to get them to be good at, like, the more to be technically accurate and creative. Right? And I still think that, like, eventually, they'll probably get there.
Matt:But, like, the type of content that, we're doing on developer relations is very, like, you know, step by step developer tutorials, or it's extremely deep dives into theoretical concepts. Like, I have a, you know, PhD in in theoretical mathematics on my team. Right? Like, he goes into machine learning topics in, like, 6,000 word blog post, theoretical theoretical explanation that is distilled from academic papers into something that a developer can read and start to understand. Right?
Matt:That type of stuff is, like, I would say beyond the capabilities of LLMs being able to do in a way that is, like you could probably get something out the other side, but it's not gonna be, like, it's not gonna be amazing. And, like, I think right now, there's, like, a flight to quality. Like, you can tell when you're reading. Yeah. Well, you can tell when you're reading, like, boring stuff.
Matt:It doesn't matter what's created by written by a human or created by an LLM. Right now, I've not found that LLMs, other than, like, creative, creative writing, like, fiction stuff, I haven't found LLM output to be exceptionally engaging when it comes to, like, nonfiction topics or technical topics. Now maybe I'm wrong. Maybe I have read it and I just haven't, you know, been, like, wowed. I didn't know that I was being wowed by something, but that that's kind of still, I think, ahead of us.
Brian:Well, also, just I think good writing is rare still. So if you're reading, the odds are it's bad writing.
Matt:Yeah.
Brian:Decent to it.
Matt:But Yeah. I think so. And, like, if you want your company to be differentiated and exceptional for particularly for developers who are an extremely discerning audience, you have to, not think about how do we publish a 100 blog posts, per month, but think about how do we make sure that every single blog post that goes out has tremendous amounts of value to someone who is gonna, like, read that. And there's obviously different types of content that you need to produce, and not everything is gonna be a 6,000 word blog post. But, that I think is actually what is going to differentiate companies now because it's just so easy to use an LLM to produce like garbage.
Matt:And the differentiation on top is really what sets sets really great companies apart.
Brian:Yeah. Well, so I we kinda talked about a lot of stuff, but I wanna maybe try to summarize a few things for people to to look at. If they're if they're in management, they wanna stay technical. One of the things you said was, on the side doing generic versions of problems that you use at your company anyway. Yeah.
Brian:I think still, like, writing technical, blog posts is a good thing. And if if if you haven't done it before, then I mean, anybody that hasn't written openly before should try it, I think. I think it's a good way to stay technical. And I think even myself, I prob I I write way less open blog posts than I used to. But the experience of doing it, at least in the past, has made it so that that's how I think now.
Brian:I I think in terms of, do I know it enough to explain it to somebody else, whether it's YouTube or, or or blog post or something? And then side projects of actually do I know this enough to actually have a completed thing that that other people can use also? Or even if it's just that I can use. There's nothing wrong with side projects that you're the solo user for. So, I think that's good.
Brian:Exactly. The other thing was was making sure that you're not just thinking about coding skills, but also thinking about things like, like, writing and speaking and, just communication skills and people skills and management. You can always be improve your style as a manager. Anything else that we should throw out there is,
Matt:I had a I had just a few quick tactical tips, which is, like, I think you also have to, like, know yourself. So if you're constantly, like, bashing your head against the wall, like and you're frustrated because you're like, I wanna be a better writer, but you just don't wanna do it. Like, maybe don't put the pressure on yourself to do that. Maybe there's something else that you wanna improve. So I think improve.
Matt:Like so for me, I think the the big thing that whether you wanna work on some side projects or you wanna be, you know, coding in addition to managing is, like, just having something every day that you can accomplish. So for me, like, I really enjoy getting up in the morning. I tend to wake up pretty early, and I just write a little bit of code. It's the first thing I do because I'm not distracted. And, sometimes I write code for 2 minutes and sometimes I write code for an hour, you know, before my son wakes up.
Matt:And I've just I'm done for the day. So I think just having something every day that you can accomplish. I think I've actually also just to get back to, like, the the real technical angle, there is a tremendous advantage in having a really quick deployment pipeline. So I deploy Pushcap, like, changes to it every single day. And it's because it's, like, one command.
Matt:It's, like, deploy, deploy, deploy. Right? So if I have an idea and I have enough time to code something in the morning, I can immediately deploy it and go send it to a bunch of people. Because there's people that I didn't have as friends or or colleagues that use the tool. And so for me, like, that's a like, an almost like a little bit of, like, an adrenaline rush.
Matt:It's like, can I finish this little feature and, like, send it to people before, like, the start of the day my, like, actual day starts? So that only comes with having a really easy deployment pipeline. And then the one other, tip I have for people, particularly if you're, like, a manager and you wanna keep up with your coding skills is, like, I actually take all of my notes, in VIM. And that allows me to just have the muscle memory. So I have, like, all my notes for, like, my one on ones, to do lists, like, anything is, like, in markdown in and I just use them for everything.
Matt:Right? So when I'm, like, going and coding, like, I don't have to, like, like, there's no, context switching. Like, I'm encoding and writing to me are, like, almost to some extent interchangeable. So that way, like, I'm just I'm always fluent in my tools. So I think that actually is, like you just wanna make sure there's, like, no impediments to you just being able to do this stuff, on your own.
Matt:Right? Like, if if you have to overcome 5 hurdles in order to work on a side project at night, you're you're probably gonna, most of the time, fail at the 3rd hurdle. So just don't have any hurdles. Just do it.
Brian:I don't know how I overcame this, but it used to be that, like, Outlook would close a window if you hit escape. And, I think that maybe I've overwritten that or something, but being a v VIM user, or VI I so I haven't actually used VI directly for well, I use it, like, if I'm SSH ing into something. Yeah. But, normally, it's either Versus Code or PyCharm in VIM mode. But, but, but, yeah, I I I all of my writings in markdown and and VI too.
Matt:So So you're you're already doing it?
Brian:Yeah. Yeah. Definitely. So, we could talk on this subject for a long time, but I wanna wrap things up. Thanks a lot, Matt, for, for coming on the show.
Brian:And, Thanks
Matt:for having me, Brian.
Brian:We'll talk to you later.
Matt:Sounds good.
