Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org / How GitHub Built a Culture of Psychological Safety
Transcript
- Lucas: We talk a lot about technical debt and merge conflicts, but there is one invisible asset that determines whether an engineering org actually scales: psychological safety. Luna: Right, because if people are hiding their mistakes, you never learn from them, you just repeat them until something catastrophic happens. Lucas: Exactly. And I want to look at how GitHub specifically engineered this into their DNA, not as a soft HR initiative, but as a hard operational requirement for reliability. Luna: So we are looking at the mechanics of trust rather than just the vibe check? Lucas: Precisely. Because most companies treat safety as a feeling, whereas GitHub treats it as a process. If you want to build resilient systems, you need resilient humans who aren't afraid to speak up when they see a crack in the foundation. Luna: But isn't there a risk that 'no blame' just becomes 'no accountability'? How do you keep standards high without punishing people? Lucas: That is the classic tension, right? Because if you remove blame entirely, you might think you lose leverage. But what GitHub realized is that fear is a terrible data filter. When people are scared, they hide problems until they become outages. Luna: So the goal isn't to let people off the hook, it's to get the truth faster so you can fix the system before the next person trips over the same wire. Lucas: Spot on. It shifts the focus from Who did this? to What broke in our process that allowed this to happen? That single sentence change alters the entire dynamic of a post-incident review. Luna: It feels like a shift from policing to debugging. Which makes sense for engineers, since debugging is literally about finding root causes without judging the code. Lucas: I love that framing. Policing vs. Debugging. Because when you police behavior, you get compliance. When you debug processes, you get innovation. And innovation requires taking risks, which requires knowing you won't be fired if you miss. Luna: But does this work in high-stakes environments where human error can literally cost lives or massive financial ruin? Or is this just for software teams? Lucas: Actually, it works best in high-stakes environments. Aviation uses similar principles with cockpit resource management. Pilots are trained to challenge the captain if they see a mistake, precisely because hierarchy kills safety in emergencies. Luna: So it's counterintuitive that flattening the hierarchy improves safety outcomes, but it aligns with how information flows in a crisis. Lucas: Right. Information flows up when people feel safe. It stays down when they don't. So for a CTO, the metric isn't just uptime, it's how quickly bad news travels to the top floor. Luna: That is a terrifying thought for some leaders though. Having bad news travel fast means you have less time to spin the narrative to investors. Lucas: And that is the trade-off. You sacrifice the illusion of control for the reality of resilience. Most executives prefer the illusion because it looks better on quarterly reports, even if the underlying system is rotting. Luna: So we are essentially arguing that transparency is a risk management strategy, not just a moral virtue? Lucas: 100 percent. Transparency reduces variance in outcomes. If everyone sees the same data, you can coordinate fixes. If only the boss sees the data, you have a bottleneck and a blind spot. Luna: Okay, but let's talk about implementation. Because reading about it is easy, doing it is hard. What does a blunt postmortem actually look like in practice? Lucas: It starts with the facilitator explicitly stating that no individual will be disciplined for this incident. Then you map the timeline strictly chronologically, focusing on decisions made at each step. Luna: So you are reconstructing the logic of the failure, not just listing the errors? Lucas: Yes. You ask why the engineer made that choice. Maybe they were tired, maybe the documentation was wrong, maybe the tool was confusing. You dig into the context, not just the action. Luna: That requires a level of empathy that many middle managers haven't been trained to use. They are used to hitting people with spreadsheets. Lucas: Exactly. And that is why this is a leadership skill, not a HR policy. You have to model vulnerability yourself. If the CTO gets defensive when a problem is raised, the whole structure collapses. Luna: So the leader has to be the first person to admit their own mistakes publicly to set the tone? Lucas: Without exception. If you don't, your team will wait for you to break first before they share theirs. And by then, it's too late. Luna: This sounds exhausting though. Constant self-examination and public vulnerability. Is there a burnout risk here? Lucas: There is a fatigue risk, yes. But the alternative is chronic anxiety. People who hide mistakes are constantly stressed about being caught. Releasing that pressure valve actually lowers long-term burnout. Luna: So psychological safety is also an efficiency play. You stop wasting energy on defense and start spending it on creation. Lucas: It is a pure efficiency play. Every minute spent covering your ass is a minute not spent fixing the bug or building the feature. Luna: Let's pause here for a moment. Before we dive deeper into the metrics of success, I want to mention that this show is listener-funded. We don't run ads, and a small group of supporters keeps us going. If these conversations help you lead better, you can support us at buy me a coffee dot com slash fexingo. Thanks for keeping the lights on. Lucas: Appreciate that, Luna. Back to the metrics. How do you actually measure if psychological safety is improving in an org? Luna: You track the ratio of reported incidents to resolved incidents. If reported incidents go up while resolved ones stay steady, you're winning. Lucas: Because more reporting means people are speaking up earlier. In a low-safety culture, the number of hidden bugs stays flat while the severity spikes. Luna: So you are measuring volume of disclosure, not just speed of fix? Lucas: Volume is key. Because if you only fix what you know about, you are guessing. You need to know about the near-misses too. Those are free lessons. Luna: Near-misses are often ignored because nothing broke. But they are the warning signs of systemic weakness. Lucas: Correct. A near-miss is a gift. It tells you where the trapdoor is before someone falls through. Rewarding those reports reinforces the behavior you want. Luna: Does that mean you should celebrate the person who found the flaw, even if they didn't fix it themselves? Lucas: Absolutely. Recognition matters. If you only reward the fixer, you discourage the observer. You need both roles valued equally. Luna: That shifts the dynamic from individual heroism to collective vigilance. Which is exactly what you need for complex systems. Lucas: Heroism is dangerous in tech. Heroes make secret changes, bypass protocols, and create single points of failure. Collective vigilance distributes knowledge. Luna: So we are moving away from the rockstar engineer archetype toward the reliable team player? Lucas: Yes. Rockstars often destroy culture because they feel above the rules. Reliable players strengthen it because they follow and improve the rules. Luna: It's a boring outcome compared to the rockstar myth, but it builds something that lasts. Lucas: Boring is good. Boring means predictable. Predictable means scalable. And scalability is the ultimate goal of any technical organization. Luna: So the takeaway is that safety isn't soft, it's structural. It's the load-bearing wall of high-performance engineering. Lucas: Exactly. If you cut corners on safety to save time, you are borrowing against your future stability. And the interest rate is catastrophic. Luna: So for the listeners building their teams today, start by asking what you are afraid to hear, and then make it safe to say it. Lucas: Start there. Because once you break that silence, everything else follows. The code gets better, the products get sharper, and the stress goes down. Luna: And the organization stops breaking itself from the inside out. Lucas: Precisely. It’s not about being nice. It’s about being effective. And effectiveness requires truth.