Platform Engineering: Asking "Why"? with Evelyn Osman

Mauro Morales and Dimitris Karakasilis explain how Kairos turns any Linux distribution into an immutable, A/B-upgradeable OS shipped as an OCI image, and where it fits alongside bootc and other immutable system efforts.

Flatcar maintainers Thilo, Mathieu, and Chewy explain the immutable A/B partition layout, Ignition provisioning, systemd-sysext for Kubernetes, LUKS/TPM disk encryption, and how Nebraska coordinates monthly fleet updates without SSH.

Evelyn Osman joins David and Laura to defend AWK and Bash, argue platform engineering must start by asking "why", and pick apart whether every workload really belongs on Kubernetes. Also covers decoupling CI from CD.

Laura quizzes David on how he uses Claude as a CLI coding partner across the Rawkode Academy monorepo, why he calls it smart coding rather than vibe coding, and the trade-offs around productivity, skill atrophy, and the environmental cost of AI tools.

Adriana Villela explains why developers, not observability teams, should instrument their own code. We cover OpenTelemetry's auto vs manual instrumentation, sampling and cost trade-offs, single-pane-of-glass backends, and the environmental impact of telemetry.

Hazel Weakly unpacks open source sustainability through Eleanor Ostrom's commons principles, covering reciprocity, governance, VC funding pitfalls, maintainer burnout, and why web browsers should be treated as international public utilities.

Pepe Amengual, a core maintainer of Atlantis, joins David and Laura to walk through the Terraform pull request automation server: how it runs plan and apply from PR comments, OpenTofu support, the path to 1.0, and the security model around credentials.

Otavio Santana joins Laura and David to dispel myths about modern Java on Kubernetes. They cover JVM container awareness, GraalVM native compilation for serverless cold starts, and frameworks like Quarkus and Spring.

Mark Fussell, CEO of Diagrid, explains how Dapr's building blocks (pub/sub, state, workflows, virtual actors) tame microservice complexity, and how Dapr Agents extends that same actor and durable-workflow model to agentic LLM systems.

Koray Oksay (Kubermatic, CNCF Ambassador) joins David to discuss why Kubernetes adoption still trips up teams: skipping container fundamentals, picking between managed services and self-hosted clusters, navigating tooling choices like Flux vs Argo CD, and structuring multi-cluster environments.