Ex-Head of Eng at Instagram: Career Regrets and Learnings | James Everingham

28:52Even though I came from a smaller startup and went into Yahoo and I went from 15 engineers to 1 ,200 or something, I had previously ran large organizations, you know, at Netscape and, you know, Tell Me was even pretty large. So I had that background already. I interviewed Ryan Olson. He's one of the original. I know Ryan well. Okay, great. Yeah. And so we were talking about Instagram stories and he mentioned something really interesting in that project, which was that actually when the project started, you immediately kind of cut the size of the team. almost in half or significantly, just down to like Ryan Olson, a few other people, so that the project could move faster.

29:38And he said that was a big part of the project's success. And you were the head of engineering at the time. So I was curious if you had any thoughts on what did you see that made that a good decision at the time? You know, Instagram stories was an interesting thing. Now, like your natural instinct is like, Like, hey, Snapchat's gaining traction and, like, we need to go do what Microsoft did, turn the whole company against this. But really, this is where your entrepreneurial skills come in. It's like, hey, what we need is a small team that can move really fast. So, like, you want to set it up so that you have really capable people.

30:19You have the right types of engineers who are very proactive. And then you remove any roadblocks from them. And so you give them a separate area where it's not dependent upon like deep APIs and blocking points into other organizations. So you set it up so that they can move fast. And so we did that purposely, right? And I think it was like three months. We had – it was a three-month project. We went from zero to releasing stories. And it was a pretty big success. I think maybe a lot of managers might naively think, oh, competitors coming. We got to move fast. That means throw a lot of engineering resources at it.

31:06So here's a 20-person team. Go, go, go. As opposed to, it sounds like you have a different philosophy here, which is a small talent-dense team. That's right. With the founders leading it directly. Kevin and Mikey were heavily involved and brilliant product thinking and execution at that level as well. When you get a small team that's capable and you give them clear ownership and you remove dependencies, that is one way to increase speed. So if you get a large team, they tend to be interdependent. You start federating large pieces of architecture and pretty soon those dependencies end up slowing you down.

31:51So I think the small – if you – sometimes it's better to start with something small and build on it rather than trying to build the big thing at the beginning and evolve it over time. And that's sort of what this was. It's like, let's get it working. Let's get it stood up. Let's get this thing working and then plug it into the product and we'll scale it instead of like, what does this look like at large scale? And let's build that from assuming that. That ends up often being a way less efficient way to get something done. And it's definitely a lot slower. Mikey Krieger had like, and it was the name of my conference room.

32:32It was called this value that he said, do the simple thing first. I don't know if you recall this, but like, hey, you know, like if we're going to go off and we have a thesis, like before we go off and just pour the whole company in it, what's like the simplest, hackiest thing we can do to go prove this thesis true? And, you know, that was sort of part of the values of Instagram is like, what's the simplest thing we can do to go prove this thesis before we scale it? You were hired as the head of engineering. And at the time, Mike Krieger was still the CTO there. What are the differences between that CTO role and the head of engineering role?

33:10And how did you partner with Mike Krieger? And what went well, what didn't? Yeah. So Mikey brought me in. The engineering was hitting about 100 people. It's a common place where teams start to break. And Mikey's brilliant. He was actually a really gifted manager too, but he just didn't want to do that. And he wanted to focus on the technology and even coding. That was his true joy. So he brought me in. So the initial confusing things to the team was like, what are our different roles? Like, you know, who do I listen to? I'm used to getting directions from Mikey. So you have to be pretty clear.

33:51So the first thing that we did is like we wrote our job descriptions and we published them to the org. And so, you know, and classically a CTO is like works with customers, new technologies, works on the product and the technology roadmap, experiments, all of these things, builds the external engineering brand. And the VP or the head of engineering, it's usually a people management thing. It's like hiring, execution, org structure, performance reviews, all that fun stuff, right? And so think technology roadmap for CTO. Think process people, staffing, execution with the VP Avenge. So we published that.

34:38And we even came up with like flow charts. Like, hey, I want, you know, we put it out to the team. We're like, hey, I want five more headcount. Go do this project. And it's like, come to me. Hey, you know, I want to use this new technology. I think it would be cool. Go to Mikey. So, like, we also put, like, a roadmap out there. And then it took some evangelism, you know, like, hey, you know, to get people to learn where to go for what they needed. People often confuse this, by the way. You know, it's like people confuse the difference between a CTO and a VP of Edge. Like I get called into a lot of companies saying, hey, we need a new CTO.

35:14And I'm like, why? What's breaking? And they almost always describe VP of Inge problems. I'm like, I don't think you need a CTO. I think you need a VP of Inge. It's a very different person. As we kind of wrap up this Instagram leg of your career, is there anything that you think I might have missed or kind of like lessons that you learned from this part of your career? Yeah, no, you know, I think that the one thing that I remember is like, you know, when I came into Instagram, it was a different company. Everyone's nervous when they go into a new job. And they brought me in right when engineering teams start to break, but like, these weren't breaking.

35:53And I was very confused at the beginning. And I was starting to feel like I was a dinosaur. I'm like, wait, you know, this is all working. This isn't breaking. Like what kind of management voodoo is going on here that I don't understand? And then about 150 to 200 engineers, it started breaking. And I'm like, oh, wait, these people are so smart. They're able to IQ their way past like that normal barrier. Because what usually breaks is communication lines. Like you have to hold a lot of stuff in your head. Use face familiarity with people. It's called, you know, Dunbar's number, all that stuff. But like the people were just so gifted and smart at Instagram.

36:31they were able to IQ past that. Plus the great thing, the reinforcing thing about Instagram was the best software is built by people who use it. And like everyone in Instagram was a power user, right? Like, and even if you go back to Boomerang, do you know, do you remember Boomerang? When it first came about, it was, I think it was an intern, Tim Leonardo, I think was the intern. he went and built it as a hackathon in like three days. Like one of the most engaging features the Instagram released was a hackathon built by an intern in three days. Like it's just amazing to work in a team where you're the primary customer and you're building stuff for yourself.

37:13So anytime you can do that, like that's awesome. Moving to the next part of your career, I saw you. I remember when I joined Instagram, all these very high profile people were leaving Instagram to work on Novi. It was the hot project at the time. I was thinking, oh, wow, this new crypto thing must be really big. And so, yeah, what drew you to working on Novi or the crypto project at Meta? And what was the impact you saw on it? So inside Meta, I had met David Marcus. David was running Messenger at the time. And we had just met, but he's like, hey, I want to run something by you. And I went and he was like, don't tell anyone, but I'm going to leave Messenger and this is what I'm going to do.

37:59And this is my first meeting with David. And he's like, I've heard good things about you. Would you consider coming and running the engineering aspect of this? And he was selling me this crazy crypto story. Like, and I loved Instagram, right? And I was like, you're insane. Like, that's a crazy thing. Why would I go join that? And like, I just couldn't get it out of my head. And then like the next day it started hitting me like, oh, wait, this is an area for impact. And so I went back the next day and David was kind of grinning and he was like, I knew you'd be back. He's like, I call this the mental virus.

38:38He goes, it takes about 24 hours to infect people and then they get it. And I was like, darn you, David. And long story short, I'm like, okay, I'm going to come start this. I was heavily invested in the mission, which was financial inclusion and bringing safety and low-cost money movement to people around the world that needed it. And I loved that idea. And I remember telling my product partner, Kevin Wheel, at the time, hey, I think I'm going to go do this. And he – I explained it to him and he was kind of like – had that same look I had. And then like the next day he throws a 30-minute meeting on my calendar and I'm like, oh, he's going to try and talk me out of this.

39:27And he comes in and he goes, I think I got to go do this too. What? Like, oh, this is going to be tricky. So I connected him with David and I'm like, great. This is going to be a tricky one to navigate. But it worked out well and we went over and started the Libra project at the time. How did the project go and what were the pivotal points in it? Oh, man, I think it's pretty public how that project went. Like we weren't able to ship it, right? Like it was – there was a lot of pressure on Facebook at the time, just political and regulatory. And like they weren't excited about us, what their view was as thinking we're going to rewrite currency.

40:06It's like not what we were doing. But like there was not a lot of strong understanding on the government. They're not the best, deepest technologists. So – but they didn't want us to release this and it was a shame. David tried to do this like the right thing. He's like we announced what we were going to do, that we were going to work with regulators and everything. In retrospect, we probably should have just released it and asked for forgiveness. But that sort of put us into a tailspin where we had to keep re-architecting the project to be regulatory compliant. And it took us years to do that.

40:42and ultimately we got it to the point where even the government said this was the platinum version of Staplecoin, but they still didn't want us to release it. And I think that's when we gave up. And we were a pretty big org by then. I think we were like 3 ,500 people. And yeah, they put all payments and everything under David at that point. And I think in a funny moment, David came to me. Oh, well, I'm like, I can't keep doing this. I'm going to go do something different. And I loved David and felt like he was a friend. I didn't want to let him down. So it took me a couple of weeks to get the courage up and practice my resignation to him.

41:26And I was ready. And I'm like, I'm going to go in our one-on-one. Here's my story. I'm going to give him all the time, but tell him I need to move on. I go in there and I sit down and David goes, well, before we get into whatever you want to get into, I have some news to share with you. I'm like, what? And he's like, I'm going to leave. That's what I was going to tell you. And I'm like, well, what are you going to do? And he's like, I think I'm going to go get into my entrepreneurial roots. I'm like, how? And one thing led to another. We kept talking and I ended up leaving Meta for a couple of years.

41:57And I went and started LightSpark with David because I was invested in this mission of, and that's what LightSpark's doing. And I love it. It's just an amazing mission of like, you know, the way I think about it is finance is the last area for disruption, you know, but I do think like the cost and latency of moving value globally needs to go to zero. Because when it does that, when you can move money like packets of information, like you can stream money, you can move at low cost areas that need it. Like it's just going to be a revolution. I have no doubt that that's going to happen at some point.

42:30And so you were working in the crypto space and now you're working in AI, what made you leave the crypto space if you know there's impact there? Yeah. First off, I'm like a crypto skeptic. I never was a buyer into all of this stuff, but I do think that it's really good for money movement and as a base layer. So what happened there was I was working and I co-founded LightSpark and David wanted to do it in LA. So I figured I'd move down there. I commuted for two years down there and, uh, and I formed a team and build it. And I was kind of like at a point where the team was formed, I'm commuting three days a week.

43:13I didn't feel like I was adding a lot of value, uh, until we were at the next level of scaling. So I talked to David and, you know, David and I agreed like, Hey, yeah, fine, I'll step back. And, you know, we're very good friends. So it was like an easy conversation. I groomed a successor and stepped back. And at that time, Meta learned I was free. And they were like, hey, you know, we have a role we want to run by you. We hear you're free. And I'm like, I don't know what that role would be that I'd be interested it in. And they're like, well, what about dev infra, you know, like running all the developer infrastructure at Meta?

43:53And I'm like, oh, hold on. Like I'm, that's my background. Like I'm a developer tools person. That was the first 10 years of my career. Like, and it hit me like, oh, wait, you know, what was I even doing at Instagram and all these things? I'm a developer tools person. Like, that's really what I want to do. And with AI coming into a picture, it was like, awesome we get to go write science fiction stories about how developer productivity is going to be and go try to make that happen with unlimited resources i'm like sign me up for that so i was so i came back and uh i love that team dev infra is amazing and such a senior team like a thousand engineers building all of the developer tooling for inside meta it was just amazing i love that job, did it for almost two years.

44:41And the only reason I left is because I had that observation and I had to go start this company. And what was the observation? Well, so, you know, we built something internal called DevMate. DevMate was an agentic platform that allowed engineers to sort of like go learn from each other and then scale these agents. And these agents were working on tooling, not just coding. They were driving tools and doing all types of things. And it went viral internally. We couldn't keep up with the demand. Within months, it needed an entire dev server per dev made instance. And there were 40 ,000 plus engineers at Meta.

45:25And pretty soon, everybody wanted an instance. And some of them wanted three instances. So you're like, how do we allocate all the dev servers? But it was such a strong pool. You're like, oh, I see the dynamics at work here. Like 40 ,000 developers is fun, but it would be fun to go do this for 40 million. And so we took a lot of these lessons and we're like, yeah, we kind of think we see around the corner what the world needs for this to work inside their companies. And that led to the formation of Guild. When I look at the AI space, it feels like a bloodbath or it's incredibly competitive. When you saw the space, I guess, what gives you the confidence that you will compete well or what is the niche that you're going to focus on?

46:15First off, I don't have the confidence, but I have the drive. And I'll tell you that much. It's like I deeply believe that what we are building is needed. So there's that. But like I'm always, you know, I'm always looking out of the corners of my eyes of like who else is building this and, you know, who might not execute. That's just what an entrepreneur does. So if you look at a lot of the startups and a lot of the companies, you know, I think a lot of them are building for like the future past. And what I mean by that is they're building for things that fill in gaps of what LLMs will probably be able to do.

46:55Like look at Cursor. It's autocomplete. Now, you know, LLMs are starting to get to be able to completely write code without an IDE. It's like, okay, you wrote, you build a product on a bus stop, not where the end of the line is, you know. So what's that end of the line? And like what is needed regardless of all of that? So what we're building is a control plane. And it's a layer of infrastructure that allows you to manage these agents inside a company. Like, you know, your agents, if they're rolling around in your infrastructure, you want to treat them like employees, right? Like some of them, some of them you want to have access to things.

47:32Some you want to restrict. You want to report on what they're doing. You need to observe. You don't want one agent just like tanking your token budget. You need circuit breakers. You need a governance layer for that. Regardless of like where this ends or which model company wins, you need this, right? So we believe that we're building something that's needed. And in any scenario where Anthropic or OpenAI or Meta or somebody else wins, you just need this layer. Let me paint you a vision of where I think things are going to go. So I think a few years, like right now inside a company, you have a bunch of internal web applications.

48:17You have your HRIS systems and, you know, your intranet and like developer like CICD tools. And, you know, you have a lot of internal apps. I think there's going to be a layer of like thousands of agents that will sit on top of this and help drive them. Right. For example, you know, we build an agent that we use on our own system here that basically doesn't make duplicate bugs possible. Right. It's a really simple thing. It's like you submit a Git issue, you know, it fires a webhook. Our agent wakes up and goes, let me look and see if there's one that's like this. And it just marks a duplicate and goes back to sleep.

48:56That's different than a coding agent, right? So you can imagine anywhere where there's a tool integration, prompts, and evals, you can go into design. You can go into product. You can go to HR. It's just basically workflow management in a company. And that's the thing where we think it's going. But if that's the case, then, you know, you're just like people, you probably don't want your engineers having access to your finance systems. You don't want your finance people having access to your code base. Like, you know, all of these things you're going to need to set up policies around. And even like regulatory policies, you're going to need that, especially in a large public company.

49:37So you need layers to be able to enforce that. AI is non-deterministic, right? Like it's like you can't guarantee what the outcome is based on any inputs. If you're building a stable infrastructure layer, you need it to be deterministic. So you need to put a deterministic layer on top of this non-deterministic technology. It's the only way right now. Okay. So it's almost like this harness that will, yeah, control plane is a great way to put it. Control plane for the agents within maybe a corporate environment. Centralize it. Like right now, you see people using these AI agents. We had one customer that they have their developers running agents on their laptops.

50:23They don't know where they're running. This is one that blew through their token budget, one agent in 12 hours, and they didn't know. So you centralize that. okay here's a central runtime you you make it so agents can work together in a secure way so you put bubble wraps around them um you don't want these agents just yellowing at your infrastructure right so like you have to put control around them you know i i i i use this analogy like there's this movie gremlins you remember gremlins i haven't seen it but gremlins yeah it was like easy spielberg movie where these little furry creatures started multiplying quick and they just were causing chaos in town, taking over the electric power grids.

51:04Like AI agents are like that, right? It's like they're just like, once they start multiplying and they're like, you can't predict what they're doing, like that's a problem. So you need to be able to put a layer there to provide some predictability and control around that. That's what we're building. Yeah. At this point, I feel like you have so much management experience. So I wanted to discuss this high-level management philosophy that I see in the industry, which is that I see that this Elon Musk style of management is kind of like getting in the details, micromanaging. And I hear a lot of people who are against that as well.

51:44And I'm curious what side of that you're on, and maybe we can discuss the pros and cons of each of the approaches. Yeah, sure. Well, look, I think that there's different tools for different jobs. And so what you're describing when you talk about these two different management styles, they're in two very different environments. And so I'm not a micromanager, but I've also been fortunate enough to have incredibly capable people to work with. So, you know, you can think of an org as a top-down org, like where you have a leader with a very clear vision and you're hiring a workforce. He or she is hiring a workforce to help achieve that mission.

52:23Or, you know, you can be the opposite of that and consider yourself at the bottom of the org chart. And you're trying to act as a tool to help that team achieve their vision. That's the one I am. Now, that being said, there are times where if your team isn't performing and I've worked – I've inherited some teams that were not working well and things like that where I've had to go in and micromanage but not for long. You want to only use it as a tool in my opinion to fix things and then step back out of that. I think that our job in technology is a creative – this is a creative job, right? Like you don't know what it looks like often.

53:15You just know what outcome you want, right? And if you're going to do that, you have to free people. You don't want to hire smart people and think for them. You want to just focus them on the outcomes. You want to like clear their – what's blocking them. You want to be able to probe and ask great questions and course correct them. But if you start micromanaging really smart people, you're going to lose them. Now, if you look at some of these classic examples that you mentioned, like Elon and stuff like that, I think there was a lot of – lost a lot of smart people that weren't willing to work in that environment.

53:51But I didn't understand the environment. Maybe it was necessary. I can't speak to that. I've built my career around hiring really smart people, much smarter than me, and getting out of their way and just supporting them, you know, trying to make them, give them the space to be the best that they're at. And so I think, you know, not saying there isn't room for this micromanagement, like get down into the details. You need to do that every once in a while. But like that's not if you just that's your only mode of operation. That's your only tool. It's only good for a very specific thing. When you look back on your your whole career, is there any career regret that comes to mind that you think other people could learn from?

54:37Oh, man. So, yeah, there's a lot. actually. And I wouldn't say they're regrets because I'm pretty happy where I'm at and everywhere led me here. So you have to also look back and go, I'm not going to punish myself over mistakes if you're happy where you're at. I'll tell you things though that I wish I would have known or I wish I would have appreciated when I was younger is I'm not an intern age anymore. I'm a little older and time is my scarce commodity now. So back when I was younger, it was okay to like beat my head against the wall for six, seven years. I'm like, I have all the time in the world.

55:15I would like to have that time back and spend it well. So like, be sure you're doing what you love and be sure you're having good days because that's really your scarce commodity, you know? So that's one thing. Go where the smart people are. Like I learned this by accident. Like I was fortunate enough to get recruited into Netscape and Borland, like where all the smart people were. Going there really helped build me a network that's just people, the early engineers at Netscape and Borland are kind of like owning the Valley now and Mark Andreessen. So that's also good. And that's where I got the best mentors and learned the most.

55:58And I'd say one simple tactical, and if you're just starting at a company, this might sound silly, but I would say, I wish I would have just maxed out my 401k on day one. Like just go max out your 401k. It sounds like you're going to get a reduction in salary and you may not be starting out at high, but trust me, you'll forget it. But like when you're my age, that is going to be a giant amount of money. I wish I would have done that earlier. So that's the other tactical thing I would do. If you could go back to the beginning of your career and give yourself some advice, knowing everything you know now, what would you say?

56:34I would speak up more. I remember being afraid to speak up because I was worried that people would think I was an idiot. But it turns out that those are the people now that I notice are the people that even if what they say is off, they're not afraid to speak up. So speak up, like be curious. Like I wish I would have networked more and asked people like more about what they were doing. And, you know, one simple thing is like it didn't occur to me as sort of a version of go where the smart people or like I was seven years in Pennsylvania working. That's where I'm from. It didn't occur to me to like go to like I got recruited into Borland.

57:16But like in retrospect, it was like if I were an actor and like it didn't occur to me in seven years to go try Hollywood. It's like, you know, it made all the difference was getting into the valley where there was so much opportunity and resources. I wish I would have done that earlier. That makes sense. Awesome. Well, thank you so much for your time, James. I really appreciate it. It was great to meet you. Yeah, great. Thanks for taking the time. It's a fun conversation. Thank you for listening to the podcast. It's a passion project of mine that I really enjoyed building. Another passion project that I've been working on kind of in secret is building an ergonomic keyboard that I wish existed.

57:54And I finally have a prototype. So I'd love to show you what we've built. It's ultra low profile and ergonomic. And I couldn't find anything like it on the market. So that's why we built it. I'll put a link to the keyboard in the description. You can take a look and learn more about the project there. We could definitely use your support. Also, if you have any feedback for me about the show, I'd love to hear it. Comments on YouTube have led to guests coming on like Ilya Grigorik and David Fowler. I wasn't aware of them until someone dropped a comment. Also, feedback in the comments helped me learn to reduce the number of cliffhangers in the intros.

58:29So your comments definitely make a difference. Please keep letting me know what you'd like to see more of in the show, and I'll see you in the next episode.