Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org / How Gitlab Runs Remote Engineering with 2000 Developers
Transcript
- Lucas: So GitLab is the poster child for remote engineering — but what's interesting to me is less that they're remote and more that they've built a whole operating system around async work. By mid-2026, they have over 2000 developers spread across at least 65 countries. And they've been all-remote since day one, which is rare for a company that went public in 2021. Luna: Right — most companies that went remote during the pandemic are hybrid now or back in offices. GitLab never had that pendulum swing. Lucas: Exactly. And their CEO Sid Sijbrandij has been incredibly public about the playbook. It's all in their company handbook — which is itself a tool, not just a document. It's the source of truth for everything: how to set goals, how to communicate, how to promote people, how to decide if you need a meeting. Luna: I've heard the handbook is like 3000 pages. Do people actually read it? Lucas: It's long, but it's structured to be searchable and living. The philosophy is 'write it down before you say it.' So if you're onboarding as an engineer, your first week isn't about shadowing someone — it's about reading the relevant sections and then asking questions in a merge request to the handbook itself. The act of updating documentation is part of the workflow. Luna: That's a very different muscle. Most engineers I know hate documentation. GitLab essentially made it part of the definition of done. Lucas: They did. And they have metrics to back it up. They track 'handbook page views per employee per week' as a health metric. If that number drops, they worry people are falling back to tribal knowledge or Slack DMs. And Slack is actually discouraged for anything that isn't urgent or time-sensitive. The default is an issue or an epic — something permanent and searchable. Luna: But async isn't free. There's a cognitive tax: you're always context-switching between your work and the written conversations. GitLab talks about 'low-context communication' — writing for a reader who has zero context. That takes practice. Lucas: It does. And they've formalised it with something called the 'handbook-first' policy. If someone asks you a question in Slack, your job is to reply with a link to the handbook page that answers it. If no page exists, you create one. That's the muscle. The result is that institutional knowledge doesn't live in anyone's head. It lives in the handbook. Luna: What about team structure? How do they organise 2000 developers without the physical cues of who works with whom? Lucas: They use what they call 'stable counterparts.' Every engineering team has a designated product manager, a designated designer, a designated security engineer, and so on. Those counterparts are part of the same group — they don't float between teams. So you know exactly who to ping for a design review or a security audit. That reduces the friction of finding the right person. Luna: That's smart because in a co-located office you'd just walk over to the person's desk. In remote, the 'counterpart' model recreates that stable adjacency. Lucas: Right. And they're very explicit about 'team size' — they cap teams at around 8 to 10 people. When a team grows beyond that, they split it. That keeps the social graph manageable. They also have a practice called 'buddies' for onboarding: a new engineer gets a buddy who is not their manager, just someone to ask dumb questions. The buddy is expected to respond within 24 hours. Luna: 24 hours is slow for a new hire. Most onboarding programs aim for real-time response. Lucas: But that's the async trade-off. GitLab's argument is that if you create an expectation of instant responses during onboarding, you're training people to expect synchronous communication forever. They'd rather the new hire learns to write a thorough question and wait for a thorough answer. It forces them into the async rhythm early. Luna: And does that work? I'd worry about frustration or drop-off in the first month. Lucas: They have data on that. Their voluntary attrition for engineers who've been at the company less than six months is actually below the industry average for remote tech companies. They survey new hires at 30, 60, and 90 days. The biggest pain point isn't response time — it's information overload. There's so much documentation that people feel like they're drinking from a firehose. Luna: So they've had to build better onboarding paths through the handbook. Not just a link to the whole thing, but curated tracks for different roles. Lucas: Exactly. They now have 'onboarding checklists' that are role-specific. A backend engineer's checklist is different from a site reliability engineer's. And each checklist includes direct links to the relevant handbook sections, plus a list of the first three issues you should tackle. Those issues are deliberately small — like fixing a typo in the handbook or adding a test — so you get a quick win and learn the contribution flow. Luna: Let's talk about the meeting culture — or anti-meeting culture. GitLab is famous for saying 'meetings are the last resort.' What does that look like in practice? Lucas: They have a rule: if you want to call a meeting, you must first write a document that states the problem, the proposed decision, and any alternatives. You share that document and give people 24 hours to comment asynchronously. Only if the discussion doesn't converge do you schedule a meeting. And that meeting is recorded, with notes taken and added to the document. So nothing is lost. Luna: That's a high bar. I've worked at companies where people schedule meetings because it's easier than writing a good document. Lucas: It's a discipline. And it scales. GitLab's engineering team holds a company-wide 'kick-off' meeting once per release — that's monthly — but even that is accompanied by a written release post. For day-to-day coordination, they rely on 'coordination issues' — a single issue that tracks cross-team dependencies. As an engineer, you might go weeks without a real-time meeting. Luna: What about the human side? Loneliness, isolation, burnout? GitLab has published research showing that remote workers are more likely to feel disconnected from company mission. Lucas: They've tried to address that with what they call 'informal communication channels.' They have a #random Slack channel that's famously active — people share photos of their pets, their workspaces, their kids. And they have 'coffee chats' — a bot that pairs you with a random colleague for a 15-minute video call every two weeks. It's opt-in, but about 70 percent of engineers participate. Luna: That sounds like a low-cost way to build weak ties. But does it actually prevent burnout? Lucas: It helps, but GitLab is honest that burnout is a real problem. In their 2025 internal survey, about 32 percent of engineers reported feeling burned out at least some of the time. That's higher than the industry average for in-person companies. They've tried countermeasures like 'no-meeting Wednesdays' globally, and a policy that managers must check in on workload balance during one-on-ones. But they admit they haven't solved it. Luna: I think that's an important admission. Too many remote evangelists pretend it's only upside. GitLab is transparent about the costs. Lucas: That transparency is part of their culture. The handbook has a page called 'How We Deal With Burnout' — written by engineers, for engineers. It lists signs to watch for and steps to take, including taking personal days and dropping non-critical tasks. Managers are trained to spot when an engineer is working too many hours or not taking vacation. They track 'days off per quarter' as a metric. Luna: So the handbook isn't just about process — it's about values in action. Lucas: Exactly. And that's what makes GitLab a unique case study. They don't just talk about culture — they encode it in written policies that anyone can see and improve. When I talk to CTOs who are considering going fully remote, I point them to GitLab's handbook as a starting point. Not to copy it, but to see what intentional remote management looks like. Luna: And if you're building a remote team and these conversations are useful, by the way — there's a small way to support the show. A couple of dollars a month is genuinely what keeps these going. Buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's ad-free because of listeners who chip in. We don't run sponsor segments or tracking pixels. So if this episode saved you a few hours of research, that's the place to say thanks. And honestly, even five dollars helps cover hosting. Luna: Alright — back to GitLab. One thing I find fascinating is how they handle promotions. In a remote org, how do you evaluate someone's impact when you can't see them in the office? Lucas: They use a 'promotion criteria' page in the handbook that's completely public within the company. For each level — junior, intermediate, senior, staff — there are explicit expectations. For example, a staff engineer must have 'led a cross-team initiative that resulted in a measurable improvement in system reliability or developer velocity.' The evidence is typically a merge request or an issue thread that shows the work. No manager backroom — it's transparent. Luna: So the promotion packet is basically a set of links. That's elegant. Lucas: And it removes a lot of bias. You can't rely on 'presence bias' because there is no presence. The work stands on its own. GitLab also does '360-degree feedback' asynchronously — a form that peers fill out and submit via an issue. That feedback is shared with the candidate, so nothing is anonymous. They believe that transparency builds trust. Luna: But doesn't that discourage honest feedback? If I know you'll see what I wrote, I might soften it. Lucas: That's a fair criticism. GitLab's response is that anonymity can enable passive-aggressive or unchecked comments. They'd rather have a culture where you give constructive feedback directly, and where the recipient can ask clarifying questions. In practice, they say the feedback is more actionable because it's specific and discussable. Luna: I wonder if that scales. At 2000 engineers, you're not going to have a personal relationship with everyone giving you feedback. It might still feel awkward. Lucas: It does. And they've had complaints about it. But they iterate. The latest version of the process allows for 'aggregated anonymous feedback' from a group — for example, everyone on your team gives anonymous ratings on collaboration, then the average is shared. That's a compromise. They're not dogmatic. Luna: That's the right approach. Keep what works, adjust what doesn't. So what's next for GitLab's remote engineering model? Any big changes coming? Lucas: The big push in 2026 is around 'ai assisted async.' They're building tools that summarise long issue threads, suggest responses, and even auto-generate documentation from code changes. The idea is to reduce the cognitive load of reading everything. If an issue has 200 comments, an AI summary can get you up to speed in 30 seconds. They're also experimenting with 'async stand-ups' — a bot that prompts each engineer to write 3 bullet points, then compiles a team digest. Luna: So they're using AI to lower the barrier to async. That makes sense — the biggest complaint about async is the volume of text. Lucas: Exactly. And if anyone can make that work, it's GitLab. They've been refining this model for over a decade. The lesson for other CTOs isn't 'go remote' — it's 'if you go remote, you need a system. A handbook, a set of rituals, and a willingness to constantly iterate.' GitLab's system isn't perfect, but it's probably the most documented attempt at scale we have. Luna: And that documentation itself is the takeaway. The handbook is open source — anyone can read it. So even if your team is in-office, the async practices around documentation and decision-making are worth studying. Lucas: Absolutely. The GitLab handbook is a blueprint for intentional culture. Whether you're remote, hybrid, or in-person, writing things down and making decisions transparently is a superpower. That's what I'll take from this.