Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Interview Kubernetes vs Production Kubernetes: What a Viral Essay Gets Right About Running Clusters on Owned Machines
A viral June 2026 essay argues Kubernetes interviews test what you can define, not what you can operate — mapped here onto owned hardware: PDBs, memory limits, DNS failure modes, and five scenario questions to hire by.
From Heroku to Render at 2,800 RPS: What a Rails Migration's Latency Trade-Off Reveals Before You Self-Host
A 2,800-RPS Rails app makes the Heroku-to-Render decision concrete: routing, timeouts, queueing math, and incident fate, plus a checklist for stopping at Render or self-hosting instead.
Heroku to Kubernetes in a Weekend: The $140-to-€11 Migration Playbook
Salesforce put Heroku in sustaining mode in February 2026. This weekend runbook moves a $140/month web + worker + Postgres + Redis stack to a €11.37/month Cluster API fleet on Hetzner — commands, cutover sequence, both invoices, and the rollback plan.
Heroku Is Frozen and Fir Can't Run Your Dockerfile: Which Workloads to Move First
Salesforce froze Heroku in February 2026 and the Fir generation still can't run Docker images, leaving containerized apps with no on-platform future. A concrete sequencing plan: move the Docker workloads first onto a Dockerfile-native target, and let buildpack apps ride out Cedar.
Heroku Dyno Math in 2026: What $50 Standard-2X and $250 Performance-M Bills Buy Against Owned Hetzner Hardware
A line-by-line pricing of typical $50 Standard-2X and $250 Performance-M Heroku stacks against a $55 owned Hetzner server, with exit sketches for buildpacks, databases, and review apps — and the month each move pays for itself.
GitHub Purges Node 20 From Actions Runners on September 23: What Breaks, What to Bump, and Why Your Own Metal Sets Its Own Deadline
On September 23 GitHub removes Node 20 from every hosted Actions runner, and workflows pinned to node20-era actions fail outright instead of warning. The exact majors to bump, the runner floor for self-hosted fleets, and why owning your build plane turns the next vendor deadline into a planned window.
The Free Tier Is Dead in 2026 Except Render: What Zero-to-Free Now Costs a Side Project on Each PaaS
Heroku killed free in 2022, Fly.io and Railway followed — only Render still offers a free-forever tier. A platform-by-platform accounting of trial credits, minimum paid floors, sleep schedules, and 12-month totals for a typical side project, against the flat-rate Hetzner box that never sleeps.
Fly.io Spent Three September Weeks on CLI Polish, Not Regions: What Pure DX Signals
Fly.io's September 1-19 flyctl history holds zero regions and eight CLI papercut fixes. That ratio is the strategy: once compute is a commodity, retention is manufactured in the deploy interface — and it hands open CLIs a nine-item checklist.
Fly.io Turned Off Its GPUs on August 1: Where GPU-Backed Tenants Go Now (With Real Numbers)
Fly.io shut off GPU Machines on August 1, 2026 after cutting its free tier and 17 regions. A workload-by-workload cost comparison of the five destinations — Modal, RunPod, Koyeb, hyperscaler serverless GPUs, and owned Hetzner hardware — with the bin-packing math that decides between them.
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