Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Surviving a Full Hetzner Datacenter Outage: What 3-DC k0s HA Really Costs Versus Single-DC
A 3-datacenter k0s cluster on Hetzner survives a full DC outage for about $103/mo in servers and $0 in extra traffic — but every etcd write pays ~25ms and a third of worker capacity must sit idle. Here is the layer-by-layer HA design, the verified cost math, and when single-DC plus backups wins instead.
MLflow's MCP Registry: Versioned, Governed Tool Dependencies for Production Agents
MLflow 3.15.0's experimental MCP Registry versions MCP servers with semver, promotes them through staging and production aliases, snapshots their tools, and links traces to exact versions — a worked walkthrough of the promotion loop, the registry-vs-gateway boundary, and when a second tenant makes the catalog mandatory.
MCP Won: Why Every Major AI Vendor Shipping One Protocol Makes 'Deploy From Chat' a Safe Roadmap Bet
Anthropic, OpenAI, Google, Microsoft, and Amazon all ship MCP now — 97M monthly SDK downloads, 10,000+ servers, neutral foundation governance. What that convergence changes about betting your platform's deploy/rollback/logs agent surface on one protocol, and the four residual risks to budget for.
Kubernetes 1.36's Sharded List and Watch: the N× Watch Tax It Kills, and the Ceiling It Doesn't Touch
Kubernetes 1.36's alpha sharded list and watch cuts sharded-controller stream cost from N× to 1× — but it shards controller fan-out, not etcd. The math, the mechanism, and the operator checklist for self-hosted fleets.
Rootless Kubelet Goes Beta in Kubernetes 1.37: A Hardening Checklist for Self-Hosted Fleets
Kubernetes 1.37 promotes KubeletInUserNamespace to beta, letting the kubelet and node components run as an unprivileged user so container breakouts stop at a UID instead of host root — plus the nine-item cgroup, runtime, CNI, and storage checklist for making rootless your node-image default.
Three Vendors Pitched You a Heroku Exit in 2026. Here's What All Three Guides Left Out.
Microsoft, DigitalOcean, and Cloud66 each published a Heroku exit guide after the February 2026 sustaining-mode announcement: a totaled reference bill (147 USD/mo) against all three landings, the egress sensitivity math that flips the ranking, and the meters, mappings, and labor the guides omit.
One Heroku Calculator, Three Different Winners — and What Happens When Self-Hosted Is the Fourth Option
A 2026 Heroku cost calculator crowns a different winner per priority — Railway on cost, Render on simplicity, Fly.io on speed — and rerunning the same workload with a $50 flat-rate box as the fourth option puts the honest crossover at 1–2.5x the reference traffic, depending on how you price ops time.
Four 2026 Heroku-Alternative Rankings, Four Different Winners: The Disagreement Is the Data
Back4App, thesoftwarescout, techsy.io, and SaaSOffers each crowned a different Heroku alternative in 2026. A meta-read of why the winners disagree — the seats, egress, and free-tier criteria that move the crown — against the flat Hetzner box that anchors every bill.
Sealed Secrets vs External Secrets Operator in 2026: What Git-Stored Ciphertext Really Costs a Three-Person Platform Team Against Vault Sync
Sealed Secrets vs External Secrets Operator for tenant secrets on a self-hosted multi-tenant fleet: a grounded cost comparison of vault sync against git-stored ciphertext, with tenant-isolation recipes and the thresholds for switching.
Subscribe
New posts land in your reader as soon as they publish. Pick a format — all three carry the same posts.
Current feeds keep roughly two days of posts so daily polling does not miss a burst. Older entries stay reachable from the feed's next-page link in readers that follow it, or from the blog archive.
Following one topic instead? Browse tags