Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Per-Seat Pricing Is the Hidden Team Tax: A Five-Person Receipt from Vercel, Render, Railway, and Netlify
A five-person team pays $100 a month before compute on Vercel Pro — and $20 to $25 on Render, Railway, or Netlify. How 2026 flattened the platform seat, and why AI agents are about to break per-seat billing for good.
One CLI Command to a Full Remote Stack: What Moniepoint's Crossplane-Built Dev Environments Cost to Build Versus Buy
Moniepoint's eshu turns Score YAML into a full remote stack with one CLI command via Crossplane composition functions. The architecture, the two hard lessons (render-cycle steady state, 63-character names), and the build-vs-buy math: building wins past roughly 17 developers.
Nine Years in Beta Is Over: What Kubernetes 1.37's Stable Metrics API Means for Your Autoscaler
Kubernetes 1.37 promotes the metrics.k8s.io API to stable after nearly nine years in beta. What the frozen contract changes for autoscaling and usage metering on self-hosted clusters — plus the HPA gotcha and everything the API still doesn't do.
10,000+ MCP Servers in Production: What the 2026 Numbers Say About Treating MCP as Infrastructure, Not a Toy
MCP has crossed 10,000 public servers and 97 million monthly SDK downloads — but benchmarks show the stdio transport collapsing at 20 concurrent connections while Streamable HTTP holds 100% success. This breakdown turns the 2026 numbers into a production checklist: stateless transport, backpressure, real-path health checks, OAuth, and a load-test recipe for deploy and rollback tools.
MCP Memory Servers: What Persistent Memory-as-a-Tool Means for a Deploy Agent That Forgets Every Session
MCP memory servers give deploy agents persistent memory any client can share — last good deploys, environment preferences, and incident fixes recalled by retrieval instead of replayed history. The catch: multi-tenant isolation and poisoned-memory attacks that turn one bad fact into a standing prompt injection.
Kubernetes 1.36 In-Place Pod Resize Goes Beta: What Restart-Free Vertical Scaling Replaces in a PaaS Autoscaler
Kubernetes 1.36 graduates in-place vertical scaling for pod-level resources to beta: the kubelet can now grow a running pod's CPU and memory budget without a restart. This post shows the resize operation, the before/after for a PaaS autoscaler, and the beta caveats to clear first.
Java 27 Is GA: G1 Everywhere, Quantum-Safe TLS, and Goodbye Intel Macs — Your Buildpack TODO List
Java 27 turned on G1, compact object headers, and hybrid post-quantum TLS by default while ending Intel Mac maintenance. Here is what each change means for a self-hosted PaaS Java buildpack, with container memory math and a build-fleet checklist.
Hetzner's €214 GEX45: What 24 GB of Blackwell VRAM Buys Self-Hosted Inference (and When It Beats Serverless GPUs)
Hetzner's €214/month GEX45 pairs an RTX PRO 4000 Blackwell card with 24 GB of VRAM. The crossover math against serverless GPU rentals, and which models newly fit.
Heroku's Shutdown Isn't Announced — Your Contingency Checklist Should Be Done Anyway
InfoWorld reads Heroku's February 2026 sustaining mode as a prelude to phase-out. No shutdown is announced — but the export-ready audit (generation and stack inventory, data backup drills, container images off the Cedar-only registry, config extraction, add-on disentanglement order) is the same work as scoping your migration.
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