[1/2] AI, DevOps, and Kubernetes: Kelsey Hightower on What's Next (Category PlatformEngineering)
I saw it. interview Kelsey Hightower with JetBrains team on state of the industry 2025 year. I remember years. 7 I studied Kubernetes from his repository. Kubernetes The Hard Way, which was not easy, but helped me pass the labs to get the nameplate CKA (Certified Kubernetes Administrator) First in the company. That was when my colleague Stas and I were together. (guest release Code of Leadership podcast)Andreem (guest 43 Code release ...) and Anton (guest 17 Code release.) Think about how we will move to Kubernetes from virtual games:)
But if you go back to Kelsey, he has already completed an active career at Google and can now philosophically reflect on devops and more. I singled out 5 I was very interested in this discussion.
1DevOps: Evolution or failure? Kelsey is critical of what DevOps has become at many companies. - T-shirt instead of skill: Many companies simply renamed system administrators to DevOps engineers without changing the essence of the work. "Oh, now I'm a DevOps engineer, will I get a T-shirt for that?" - Proper implementationDevOps was conceived as a "drawing" (blueprint)It involves expanding competencies. Sysadmins had to learn how to program, and developers had to understand how the operating system worked. (For example, tying JVM under the Linux kernel). - Experience problem: Kelsey mentions people who have20 years of experience consisting of the same year, repeated 20 once (20 years of one-year experience). These are the ones who just fix servers without trying to automate or change the approach. - Platform Engineering: This is not something fundamentally new, but the evolution of DevOps. It’s a transition from “I’ll fix the server when it breaks” to creating a product. (platform) for internal clients.
2Kubernetes and “boring” technologies - Kubernetes is boring. (And that's good.): For stateless. (web) Kubernetes applications have become boring 6-7 years ago. Kelsey likens infrastructure to flying on an airplane: Airplanes are only interesting when they don't do what we expect. If people clap when landing, something has gone wrong. We just want to get off the plane and not think about flying.” Infrastructure should not be admired.: If your infrastructure causes you strong emotions, then you are struggling with breakdowns or friction. Delight should be caused by the product that you build on top of it. - The future of Kubernetes: If 20–30 For years, we're still going to be talking about Kubernetes, it's going to be a failure of the industry. We have to come up with something better. Kubernetes should be a mere implementation detail (like BIOS or machine code)It is hidden under more convenient abstractions.
3️⃣ API, Silos (Wells.) teamwork Kelsey makes a counterintuitive statement: I like silos (isolated)", but with a clear API. - Air ticket analogy: When you fly to another country, you don't go into the cockpit to discuss the route. You buy a ticket. The ticket is an API. (contract). You sit in a chair, fall asleep and wake up somewhere else. - API as a contract: The platform team and developers don’t have to sit next to each other and talk all the time. They must work together through a clear contract. (API)I need so much memory, so-and-so, so-and-so, so-and-so. When communication is needed: Talk only when you want to change the API or add a new feature to the platform. For a routine deploy, communication is an unnecessary overhead. It is funny that about the same moments about the interaction of teams we discussed with Stas Halup. First Edition of Code of Leadership.
The remaining topics about AI and the importance of soft skills will be discussed in the continuation of this cool interview.
#Management #Leadership #PlatformEngineering #Software #SoftwareDevelopment #Architecture #Processes