Giant Swarm

Hundreds of production Kubernetes clusters. Enterprise customers. No maintenance window. We decided to migrate them all live anyway.  ¯\_(ツ)_/¯ Running managed platforms for enterprise customers means solving problems most teams only hit once. Writing them up is how we stay honest about what the work actually involves. Joe gave this talk at KCD UK 2025. Full post in the comments.

Writing this from our Giant Swarm Observability Team onsite in Nice. The scenery is stunning, the food is incredible, and the progress we're making is even better. A little perspective on how far we've come: Two years ago, at a team onsite in Slovenia, we were staring at a bunch of Prometheus and Grafana instances that could barely keep up with the data coming in from our cluster fleet. It was an internal tool, and even our own people were frustrated with it. The idea of offering this to customers? Laughable. Fast forward to today: we have a stable, production-grade observability platform with real customers using it and paying for it. And this week in Nice, we built a PoC of our platform deployed on a plain EKS cluster - in a single day! That's a big deal. It means you can run our observability platform independently of the Giant Swarm Runtime. Host it wherever you want, pay only for the observability capability, and nothing else. And since we don't do ingestion-based pricing, that's really, really affordable. But I need to get back to the discussion now — we're preparing for the new Mimir and Loki architecture, coming with the next majors, moving ingestion to Kafka. Quite some work ahead. If you're managing a LGTM Stack on your own… good luck with that. :)) Seriously though, if you need help, reach out and let's talk!

My dog Marshmallow has successfully led a platform team from reactive cost center to strategic asset. And I was able to share her success on the Platform Engineering Day at KubeCon EU this year :D For my talk I built a case study where Marshmallow - yes, my actual dog - runs a platform team and applies product thinking to figure out what her users actually need, learn how to say no, and selling the value of the platform to the people with budget. And honestly? She's doing a pretty great job at it! The recording is now out! If you missed it in Amsterdam or spent your time in the hallway track (honestly, fair), you can watch the full 30 minutes from wherever you are. Fair warning: there are a lot of dog pictures. (But I think not enough yet.) And yes, Marshmallow got fairly compensated for her story - she got a whole cheese wheel with tasty Amsterdam gouda. Now she's telling me new case studies every day in hope for more. Not a video person? I'm turning the talk into a blog post as we speak. Give me about a week and it's yours. So tell me - does your platform team think like a product team? And more importantly: do you also use your pets as case studies? Because I can highly recommend it :D 🎤 Link to the recording in the first comment!

We'll handle it internally" is a strategy. It's also sometimes just what you say when you haven't made a decision yet... 69% of platform teams surveyed at KubeCon EU are planning to solve a staffing and tooling problem with the same staff who named staffing and tooling as their biggest obstacle. Oliver on what two years of data actually shows. Links in the comments.

We surveyed platform teams at KubeCon EU in 2024. Then again in 2026. Same city, same questions. The priorities didn't change. Neither did the obstacles. And the root cause has a name now: the platform assembly tax. Not the cost of building your platform — the cost of keeping it alive. Upgrades, patches, dependency chains, compliance. The work that never makes the roadmap but always takes the time. Oliver wrote up both pieces. The survey data and the structural problem it points to. Links in the comments.

KubeEdge explainers are easy to find. This one is built on talks our own engineers Antonia and Xavier have given, and grounded in what we've actually hit running it in production. Link in the comments.

I tell companies to stop adding shiny tools to their platform and focus on understanding their users first. And now I actually have to do it myself. I'm bootstrapping my 3rd product team at Giant Swarm and I have to say - the new beginnings of a team are always something special. But this one is extra special, because it's my very first real AI product! After all the positive feedback we got with our live demos of our Agentic Orchestration Platform on #KubeCon it was clear: we have something very special at our hands and need to place it in a proper product team that's focusing on getting the value, user experience and tech just right. And this bootstrap comes at a busy time! Two weeks ago I was on stage at KubeCon talking about how product thinking can help platform teams scale. And right now we're in the middle of a Platform Strategy Assessment Workshop with a customer, diving deep into the platform maturity model and platform-as-a-product. And honestly that's the best part. After all that advocating for shifting your focus from technology to capabilities, for putting your platform users into the center and understanding their needs and journey instead of just adding more and more tools because they're shiny - I actually get to put the money where my mouth is. Live what I preach and start from understanding the capabilities, exploring the value, the user journey. That's why I love those workshops and sharing my experience in general - it doesn't just help our customers, but also myself to keep all that theory alive. So when you actually have to build another team, another platform, it comes to you quite naturally :) If you missed our demo at KubeCon but are curious how we can orchestrate a fleet of background agents for any complex work you might want to use an agentic system for - give me a shoutout and we can set up a demo session! Or if you're interested in a Platform Strategy Assessment with us - let's talk!

hola if you missed the cluster api migration talk(s) last year || you're interested in how we at giant swarm migrated from our custom cluster management system to cluster api here's the blog post version of it https://lnkd.in/ea23F9M9 thank you thank you goodbye

We've been saying the assembly tax is real for a while now. So we asked hundreds of platform engineers at KubeCon (twice, two years apart) to describe their biggest obstacles without using our words. Hiring. Too many tools. Operational overload. No time to fix any of it. They didn't use our term. But they described it precisely. Full breakdown by Oliver on the blog — link in comments.

Say goodbye to Giant Swarm. After months of testing new messaging, collecting brutally honest feedback at #KubeCon, and honestly just trying everything to find our final form - I'm pretty sure we found it. We're rebranding. Completely. Say hello to 𝐆𝐢𝐚𝐧𝐭 𝐒𝐰𝐚𝐧. 🦢 Why settle for a swarm when you can be a swan? Elegant. Graceful. Unmistakably poultry. I want to genuinely thank everyone who helped us during the last couple of months. Every honest reaction at the booth, every "have you considered..." moment, every late-night whiteboard session - it all led us here. We couldn't have done this transformation without you. We already got the website up and everything. Check it out here: https://lnkd.in/dGC2yeFu And if you ever need help with poultry-related platform problems - send us a fax. We're ready. :D