Ship It Conversations: Justin Garrison of Sidero Labs on Kubernetes, Platform Engineering, AI, Golden Paths, and Knowing What to Say No To

This is a guest conversation episode of Ship It Weekly, separate from the weekly news recaps.

In this Ship It Conversations episode, I talk with Justin Garrison of Sidero Labs about Kubernetes, platform engineering, bare metal, AI, golden paths, and why knowing what to say no to may be one of the most important skills a platform team can develop.

Justin is Field CTO at Sidero Labs, the company behind Talos Linux, and co-host of Fork Around and Find Out.

We start with the evolution of Kubernetes and how managed services like EKS and GKE made Kubernetes easier to consume while also pulling teams deeper into proprietary cloud ecosystems. Justin explains why on-prem and bare metal are getting renewed attention, especially as teams look at cloud costs, data sovereignty, and the operational overhead that comes with constantly optimizing cloud environments.

We also get into where Kubernetes helps and where it becomes self-inflicted pain. Justin talks about abstraction, cognitive load, and why teams tend to use familiar tools for problems they were never really designed to solve.

A big part of the conversation is platform engineering and golden paths. Justin argues that every organization needs its own path, but platforms become dangerous when they try to centralize everything. He shares why one of the best decisions his team made at Disney Plus was simply saying no to stateful workloads.

We also talk about what really belongs in a platform: security controls, logging, monitoring, software supply chain visibility, and cost management. Justin explains why centralization can help in those areas, but can become a bottleneck when applied too broadly.

Near the end, we get into AI, security, tooling dependency, and engineering culture. Justin makes the point that people have always formed strong attachments to tools, and AI is another version of that. The challenge is knowing where AI actually helps versus where it becomes another dependency teams stop questioning.

The big takeaway: good platform engineering is not about supporting everything. It is about understanding what should be standardized, what should stay flexible, and what your team should explicitly refuse to own.

Highlights

• Why Kubernetes has become increasingly productized

• Why some teams are moving back toward on-prem and bare metal

• Where cloud cost optimization starts to become its own operational burden

• Why Kubernetes helps with abstraction and cognitive load

• Why familiar tools often get used for the wrong workloads

• What golden paths actually represent inside an organization

• Why platform teams need to know what to say no to

• What should and should not be centralized

• How AI changes engineering workflows without changing the need for judgment

• Why finding work you actually enjoy matters for avoiding burnout

Links

Sidero Labs: https://www.siderolabs.com

Talos Linux: https://www.talos.dev

Justin Garrison: https://justingarrison.com

Fork Around and Find Out: https://www.forkaroundandfindout.com

More episodes and show notes: https://shipitweekly.fm

On Call Brief: https://oncallbrief.com

LMGT Awards: https://lmgt.org