The Evolving Role of a Software Engineer with Forrest Brazeal
Forrest Brazeal, Head of Developer Media at Google Cloud, joins Corey on Screaming in the Cloud to discuss how AI, current job markets, and more are impacting software engineers. Forrest and Corey explore whether AI helps or hurts developers, and what impact it has on the role of a junior developer and the rest of the development team. Forrest also shares his viewpoints on how he feels AI affects people in creative roles. Corey and Forrest discuss the pitfalls of a long career as a software developer, and how people can break into a career in cloud as well as the necessary pivots you may need to make along the way. Forrest then describes why he feels workers are currently staying put where they work, and how he predicts a major shift will happen when the markets shift.
Forrest is a cloud educator, cartoonist, author, and Pwnie Award-winning songwriter. He currently leads the content marketing team at Google Cloud. You can buy his book, The Read Aloud Cloud, from Wiley Publishing or attend his talks at public and private events around the world.
Announcer: Hello, and welcome to Screaming in the Cloud with your host, Chief Cloud Economist at The Duckbill Group, Corey Quinn. This weekly show features conversations with people doing interesting work in the world of cloud, thoughtful commentary on the state of the technical world, and ridiculous titles for which Corey refuses to apologize. This is Screaming in the Cloud.
Corey: Welcome to Screaming in the Cloud. I’m Corey Quinn and I am thrilled to have a returning guest on, who has been some would almost say suspiciously quiet over the past year or so. Forrest Brazeal is the Head of Developer Media over at Google Cloud, and everyone sort of sits there and cocks their head, like, “What does that mean?” And then he says, “Oh, I’m the cloud bard.” And everyone’s, “Oh, right. Get it: the song guy.” Forrest, welcome back.
Forrest: Thanks, Corey. As always, it’s great to be here.
Corey: So, what have you been up to over the past, oh let’s call it, I don’t know, a year, since I think, is probably the last time you’re on the show.
Forrest: Well, gosh, I mean, for one thing, it seems like I can’t call myself the cloud bard anymore because Google rolled out this thing called Bard and I’ve started to get some DMs from people asking for, you know, tech support on Bard. So, I need to make that a little bit clearer that I do not work on Bard. I am a lowercase bard, but I was here first, so if anything, you know, Google has deprecated me.
Corey: Honestly, this feels on some level like it’s more cloudy if we define cloudy as what, you know, Amazon does because they launched a quantum computing service about six months after they launched some unrelated nonsense that they called [QuantumDB 00:01:44 ], which you’d think if you’re launching quantum stuff, you’d reserve the word quantum for that. But no, they’re going to launch things that stomp all over other service names as well internally, so customers just wind up remarkably confused. So, if you find a good name, just we’re going to slap it on everything, seems to be the way of cloud.
Forrest: Yeah, naming things has proven to be harder than either quantum computing or generative AI at this point, I think.
Corey: And in fairness, I will point out that naming things is super hard; making fun of names is not. So, that is—everyone’s like, “Wow, you’re so good at making fun of names. Can you name something well?” [laugh]. Absolutely not.
Forrest: Yeah, well, one of the things you know, that I have been up to over the past year or so is just, you know, getting to learn more about what it’s like to have an impact in a very, very large organizational context, right? I mean, I’ve worked in large companies before, but Google is a different size and scale of things and it takes some time honestly, to, you know, figure out how you can do the best for the community in an environment like that. And sometimes that comes down to the level of, like, what are things called? How do we express things in a way that makes sense to everyone and takes into account people’s different communication styles and different preferences, different geographies, regions? And that’s something that I’m still learning.
But you know, hopefully, we’re getting to a point where you’re going to start hearing some things come out of Google Cloud that answer your questions and makes sense to you. That’s supposed to be part of my job, anyway.
Corey: So, I want to talk a bit about the idea of generative AI because there has been an awful lot of hype in the space, but you have never given me a bum steer. You have always been a level-headed, reasonable voice. You are not—to my understanding—a VC trying desperately to prop up an industry that you may or may not believe in, but you are financially invested into. What is your take on the last, let’s call it, year of generative AI enhancements?
Forrest: So, to be clear, while I do have a master’s degree in interactive intelligence, which is kind of AI adjacent, this is not something that I build with day-to-day professionally. But I have spent a lot of time over the last year working with the people who do that and trying to understand what is the value that gen AI can bring to the domains that I do care about and have a lot of interest in, which of course, are cloud developers and folks trying to build meaningful enterprise applications, take established workloads and make them better, and as well work with folks who are new to their careers and trying to figure out, you know, what’s the most appropriate technology for me to bet on? What’s going to help me versus what’s going to hurt me?
And I think one of the things that I have been telling people most frequently—because I talk to a lot of, like, new cloud learners, and they’re saying, “Should I just drop what I’m doing? Should I stop building the projects I’m working on and should I instead just go and get really good at generating code through something like a Bard or a ChatGPT or what have you?” And I went down a rabbit hole with this, Corey, for a long time and spent time building with these tools. And I see the value there. I don’t think there’s any question.
But what has come very, very clearly to the forefront is, the better you already are at writing code, the more help a generative AI coding assistant is going to give you, like a Bard or a ChatGPT, what have you. So, that means the way to get better at using these tools is to get better at not using these tools, right? The more time you spend learning to code without AI input, the better you’ll be at coding with AI input.
Corey: I’m not sure I entirely agree because for me, the wake-up call that I had was a singular moment using I want to say it was either Chat-Gippity—yes, that’s how it’s pronounced—or else it was Gif-Ub Copilot—yes, also how it’s pronounced—and the problem that I was having was, I wanted to query probably the worst API in the known universe—which is, of course, the AWS pricing API: it returns JSON, that kind of isn’t, it returns really weird structures where you have to correlate between a bunch of different random strings to get actual data out of it, and it was nightmarish and of course, it’s not consistent. So, I asked it to write me a Python script that would contrast the hourly cost of a Managed NAT gateway in all AWS regions and return a table sorted by the most to least expensive. And it worked.
Now, this is something that I could have done myself in probably half a day because my two programming languages of choice remain brute force and enthusiasm, but it wound up taking away so much of the iterative stuff that doesn’t work of oh, that’s not quite how you’d handle that data structure. Oh, you think it’s a dict, but no, it just looks like one. It’s a string first; now you have to convert it, or all kinds of other weird stuff like that. Like, this is not senior engineering work, but it really wound up as a massive accelerator to get the answer I was after. It was almost an interface to a bad API. Or rather, an interface to a program—to a small script that became an interface itself to a bad API.
Forrest: Well, that’s right. But think for a minute, Corey, about what’s implicit in that statement though. Think about all the things you had to know to get that value out of ChatGPT, right? You had to know, A, what you were looking for: how these prices worked, what the right price [style 00:06:52 ] was to look for, right, why NAT gateway is something you needed to be caring about in the first place. There’s a pretty deep stack of things—actually, it’s what we call a context window, right, that you needed to know to make this query take a half-day of work away from you.
And all that stuff that you’ve built up through years and years of being very hands-on with this technology, you put that same sentence-level task in the hands of someone who doesn’t have that background and they’re not going to have the same results. So, I think there’s still tremendous value in expanding your personal mental context window. The more of that you have, the better and faster results you’re going to get.
Corey: Oh, absolutely. I do want to steer away from this idea tha