Uploaded June 2026 | Updated September 2026, 1 hour ago
In this solo AMA session, Viktor tackles a wide range of questions spanning Crossplane, AI's impact on DevOps, and modern tooling best practices. He dives deep into how Crossplane handles XR deletion, explaining the differences between cluster-scoped and namespace-scoped resources in Crossplane v2, and shares his personal workflow for developing and testing compositions — including his approach of developing functions independently in KCL or CUE, embedding them into YAML at build time, and using Chainsaw for TDD-style testing instead of crossplane render. He also weighs in on the Helm vs. go templating debate, firmly advocating against go templating for structured data, and shares his preference for Tekton over Argo Workflows for self-managed CI environments.
Viktor also reflects on the broader question of whether DevOps and platform engineering will be replaced by AI, offering an optimistic perspective: rather than disappearing, the role will evolve into managing and extending AI agents — essentially becoming a technical manager for automated systems. He touches on AI tooling hype, cautioning that many cloud-native companies are simply wrapping existing products with AI branding rather than rethinking solutions from scratch. Additional topics include advice for breaking into the Kubernetes job market, the value of open-source contributions, Kubernetes certifications, the future of software engineering, and a comparison of ECS versus EKS.
▬▬▬▬▬▬ ⏱ Timecodes ⏱ ▬▬▬▬▬▬
00:00 Intro (skip to first question)
05:56 How does Crossplane handle XR deletion before MRs finish?
08:15 Recommended way to ensure XR waits for MR deletion?
10:05 Will DevOps die in the next few years due to AI?
15:38 Should provider configs be created via XR compositions?
17:49 Too many tools marketing with AI - are they any good?
23:51 Confused about compositions development and delivery workflow
25:17 How to debug Crossplane compositions locally without render?
29:12 Why are AI ops tools all reactive, not proactive?
32:38 Chainsaw testing recommendations for 20+ cluster upgrades
36:00 What to look for when hiring DevSecOps engineers?
40:52 Should you use Helm instead of go templates for compositions?
42:32 Tips for the current DevOps job market?
43:36 Can you explain how Chainsaw works?
46:00 Worth buying SpaceX stocks or investing in space data centers?
49:00 Have production experience but can't pass interviews - advice?
52:20 Is Argo Workflow still a valid option in 2025/2026?
53:59 Which agent harness is most promising for Kubernetes?
57:14 Where do you see software engineering in 5 to 10 years?
58:49 Mindset shift from ECS to EKS - what does EKS introduce?
01:01:48 Is it normal for product teams to only do basic Kubernetes?
01:04:26 Do you use loops in Claude Code and what are they good for?
01:05:23 Deploying Envoy Gateway CRDs separately when too big for Flux secrets
In this solo AMA session, Viktor tackles a wide range of questions spanning Crossplane, AI's impact on DevOps, and modern tooling best practices. He dives deep into how Crossplane handles XR deletion, explaining the differences between cluster-scoped and namespace-scoped resources in Crossplane v2, and shares his personal workflow for developing and testing compositions — including his approach of developing functions independently in KCL or CUE, embedding them into YAML at build time, and using Chainsaw for TDD-style testing instead of crossplane render. He also weighs in on the Helm vs. go templating debate, firmly advocating against go templating for structured data, and shares his preference for Tekton over Argo Workflows for self-managed CI environments.
Viktor also reflects on the broader question of whether DevOps and platform engineering will be replaced by AI, offering an optimistic perspective: rather than disappearing, the role will evolve into managing and extending AI agents — essentially becoming a technical manager for automated systems. He touches on AI tooling hype, cautioning that many cloud-native companies are simply wrapping existing products with AI branding rather than rethinking solutions from scratch. Additional topics include advice for breaking into the Kubernetes job market, the value of open-source contributions, Kubernetes certifications, the future of software engineering, and a comparison of ECS versus EKS.
▬▬▬▬▬▬ ⏱ Timecodes ⏱ ▬▬▬▬▬▬
00:00 Intro (skip to first question)
05:56 How does Crossplane handle XR deletion before MRs finish?
08:15 Recommended way to ensure XR waits for MR deletion?
10:05 Will DevOps die in the next few years due to AI?
15:38 Should provider configs be created via XR compositions?
17:49 Too many tools marketing with AI - are they any good?
23:51 Confused about compositions development and delivery workflow
25:17 How to debug Crossplane compositions locally without render?
29:12 Why are AI ops tools all reactive, not proactive?
32:38 Chainsaw testing recommendations for 20+ cluster upgrades
36:00 What to look for when hiring DevSecOps engineers?
40:52 Should you use Helm instead of go templates for compositions?
42:32 Tips for the current DevOps job market?
43:36 Can you explain how Chainsaw works?
46:00 Worth buying SpaceX stocks or investing in space data centers?
49:00 Have production experience but can't pass interviews - advice?
52:20 Is Argo Workflow still a valid option in 2025/2026?
53:59 Which agent harness is most promising for Kubernetes?
57:14 Where do you see software engineering in 5 to 10 years?
58:49 Mindset shift from ECS to EKS - what does EKS introduce?
01:01:48 Is it normal for product teams to only do basic Kubernetes?
01:04:26 Do you use loops in Claude Code and what are they good for?
01:05:23 Deploying Envoy Gateway CRDs separately when too big for Flux secrets
