Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Your Deploy Agent Is a Script With an API Key: What Agent Lifecycle Managers Actually Gate
A deploy agent with a scoped API key is a script with credentials that can improvise. How SPIFFE identity, RFC 8693 delegation, and Kagenti's agent registry gate every deploy call — and what scoped-keys-plus-audit-logs can't see.
K3k Puts a Whole Kubernetes Control Plane in Every Tenant's Hands: What Virtual Clusters Cost and Buy on Shared Hardware
Rancher's K3k runs a full K3s control plane per tenant as pods on one shared host cluster. Shared mode reflects workloads onto host nodes for near-namespace density; virtual mode nests whole clusters at two to three times the control-plane cost. The mode table, the 100-tenant overhead math, and where the isolation actually stops.
Durable Execution Without the $5B Price Tag: What Self-Hosted Inngest Actually Covers That a Queue Can't
Temporal's $300M raise prices crash-proof workflows as premium infrastructure. What Inngest's self-hostable step functions cover for tenant background jobs — and the honest limits of the community-supported path.
Your Load Balancer Finally Tells You Why the Target Is Down: Hetzner's Health-Check Reason Codes, Decoded Into a Runbook
Hetzner's August 2026 API change gives every unhealthy load balancer target a reason code. What each of the six codes means, who owns the fix, and how to feed them into Prometheus so pages arrive pre-routed.
Heroku Is Gone: Where Ex-Heroku Workloads Landed in 2026 and How to Pick Your Last Migration
Heroku entered maintenance mode in February 2026. One reference app priced on Railway, Render, Fly.io, and a self-hosted box at two usage levels, plus a decision tree keyed on dyno count, Postgres entanglement, and ops capacity.
Headlamp's Cluster API Plugin: The Free Fleet Dashboard the Old Kubernetes Dashboard Never Got
Headlamp shipped a free Cluster API plugin that puts Machines, MachineDeployments, and fleet health in a browser UI. A task-by-task verdict on what it replaces, what still needs kubectl, and why a PaaS should deploy it for operators but not tenants.
Gateway API 1.6 Puts v1alpha2 Routes on Death Row and Moves Experiments to x-k8s.io: A Migration Guide
Gateway API 1.6 graduates TCPRoute and UDPRoute to v1, deprecates v1alpha2, and moves experiments to x-k8s.io — with one controller already broken today and TLSRoute showing what happens to stragglers. A six-step runbook for fleets still on the old channel.
From $3K Kubernetes to $200 Fly.io: What a Six-Month Reverse-Migration Receipt Actually Proves
A small team moved from self-run Kubernetes at $3,000 a month to Fly.io at $200, with six months of outage history to show for it. The line-by-line ledger, where the $200 breaks, and a six-point checklist for trusting any migration receipt.
Fly.io's September Report Card: What a 6-to-12-Month Credit Runway Buys Before the Edge-vs-PaaS Bill Comes Due
SaaSOffers' September 2026 Fly.io review prices $500 in startup credits at a 6-to-12-month runway and calls the PaaS choice three ways: Render for simplicity, Railway for deploy velocity, Fly.io for global edge. The worked monthly math behind each claim, when the edge earns its keep, and what the bill looks like after the credits run out.
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