Dora Noda
Software Engineer
2313 posts · View all authors
Devin Outposts: Cognition Shipped a Kubernetes Operator Instead of a Bigger Cloud
Cognition's Devin Outposts keeps the reasoning in the cloud but moves command execution onto your own machines — and shipped a Kubernetes operator to run it. Here's exactly what devin-outpost-k8s deploys and what a Cluster-API fleet needs to add to host it natively.
Dokploy vs. Coolify in 2026: What Happens the Day You Click 'Add Server'
Coolify and Dokploy both added multi-server features in 2026. Neither provisions a machine, heals a dead node, or wires up a load balancer without a human at the keyboard. Here's the feature-by-feature gap against a Cluster API fleet.
eBPF Made Continuous Profiling Nearly Free — But Someone Still Has to Store It: Parca vs. Pyroscope on a Self-Hosted Fleet
Parca and Pyroscope both run eBPF-based continuous profiling under 1% CPU overhead. Here's what standing either one up as a Cluster API service actually costs in storage and operational surface, worked against Datadog and Grafana Cloud's own profiling bills for the same fleet.
Hetzner's ARM (CAX) Instances Won't Save You Money Until Your Build Layer Can Target Them
Hetzner's Ampere ARM (CAX) instances are pitched as cheaper per core than x86 CX. Current 2026 pricing says otherwise — and the real gate on ARM economics is whether your buildpacks pipeline can produce an arm64 image at all.
One Rack Switch Died in Nuremberg. Here's What a Hetzner Cloud Leaf Fault Actually Costs a Cluster API Fleet
On July 24, 2026, a switch fault in one Nuremberg rack took down Load Balancer, Networks, Cloud Server, and Object Storage together. What that actually costs a running CAPH cluster, why it's not a one-off, and the concrete mitigation CAPH exposes today.
Jellyfin's Leaders Didn't Get Rug-Pulled. They Just Left.
Jellyfin's founder and two core leads quit in one week over burnout and an AI-tooling dispute — no feature was paywalled, no license changed. Here's why that's a different, scarier failure mode than a Plex- or MinIO-style rug-pull, and what Jellyfin's own written succession plan gets right.
CVE-2026-24834: The Kata Containers Bug That Turned a Hardware Isolation Boundary Into a Config Default
A container process with nothing but CAP_MKNOD got root inside a Kata Containers microVM meant to be hardware-isolated. Here's the exploit, the exact audit to run against your fleet, and why the fix was a config default, not a KVM flaw.
kro Just Landed in CNCF's 'Adopt' Tier Next to Helm — While Still Being Alpha Software
CNCF and SlashData's Q1 2026 Platform Engineering Radar put kro in the 'Adopt' tier alongside Helm and Backstage — a maturity signal for a project still shipping a v1alpha1 API. Here's the actual RGD YAML, what the rating measured, and what building a tenant-facing App CRD on kro's pattern would cost a Cluster API fleet.
Kubernetes 1.36's CRI List Streaming Doesn't Cut Kubelet Memory — It Removes a Hard Container-Count Wall Bin-Packed Nodes Actually Hit
KEP-5825's own text admits CRI List Streaming does not reduce kubelet memory usage. The real fix is a hard 11,000-container / 14,000-pod gRPC ceiling that's already bricked production nodes — and it matters more on a fixed Hetzner box than an autoscaling cloud fleet.