Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Railway's One-Click Postgres Upgrades Land as Postgres 14's Clock Runs Out: the Button, the CNPG Runbook, and the 6x Cost Gap
Railway's September 11 one-click Postgres major-upgrade button, mapped click-by-click to its CloudNativePG equivalent: what the preflight, backups, and revert banner replace, a six-step runbook for doing it on owned hardware, and the 6x meter-vs-flat cost gap at two sizes.
Railway's $100M Bet: What Per-Second Pricing Really Costs When the Container Never Sleeps
Railway raised $100M to challenge AWS, but the same always-on API, worker, and Postgres costs ~$148/month on its per-second meter versus ~$20 on flat Hetzner boxes — and the '50% under hyperscalers' claim inverts against an actual AWS baseline. A worked three-way recompute plus the three-hour rule for when the meter wins.
Every Pulumi Stack Is Now a Kubernetes CRD: What Operator 2.0 Buys a Cluster API Fleet (and Where Crossplane Already Covers It)
Pulumi Kubernetes Operator 2.0 runs each stack in its own Workspace pod behind the stable Stack CRD. We grade what IaC-as-CRD buys a Cluster API management cluster, where Crossplane already covers the same ground, and which one your agent-operators inherit.
Your Queue Is Retrying Itself Slow: What OpenFaaS Adaptive Concurrency Fixes About Async Batch Work on Kubernetes
OpenFaaS's April 2026 queue-worker update replaces greedy async dispatch with a feedback loop that learns each function's capacity — clearing the same batch about 50% faster with far fewer 429 retries. How it works, what it costs, and whether a git-push PaaS should host functions or just containers.
MCP Breaks Forward Again: Pinning a Shipped Infrastructure Server Through the 2026-07-28 Protocol Churn
MCP revision 2026-07-28 deletes the initialize handshake and session IDs while 10,000-plus servers run in production. What broke, how each client/server era-mismatch fails for deploy, rollback, and log reads, and the pinning checklist — plus why a stable REST API under the MCP layer is the real durability story.
Kubernetes 1.37 Let HPA Scale to Zero: Do You Still Need KEDA?
Kubernetes 1.37 graduates HPA scale-to-zero to beta, on by default. A worked idle-bill example on Hetzner metal shows when native HPA replaces KEDA's HTTP add-on — and why queues and events still belong to KEDA.
Delete Your Validating Webhook: What Kubernetes 1.36 Declarative Validation GA Actually Means for PaaS Operators
Kubernetes 1.36 graduated declarative validation to GA — but CRD CEL rules shipped in 1.29. What each milestone means for platforms defining App and Tenant CRDs, with a before/after webhook deletion guide.
Kubeflow Grew Up in 2026 — Here's What Your PaaS Should Steal (and What to Leave Alone)
Kubeflow graduated CNCF in August 2026 with Kale 2.0 and a rebuilt Trainer for distributed AI and HPC. A borrow-vs-skip guide for PaaS teams running agent sandboxes: take JobSet, Kueue, and DRA-aware placement, leave the platform behind.
Kepler Rewrote Its Power Meter in v0.10: Re-Test Your Per-Pod Energy Numbers Before They Become the Showback Default
Kepler v0.10 replaced its eBPF measurement engine with direct hardware-meter reads. The numbers got cheaper to collect but rest on a different accuracy contract — here is the verdict table and a six-step re-validation runbook before per-pod joules become tenant showback.
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