Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
MCP Just Passed 22,000 Servers. Almost None of Them Can Deploy Anything.
MCP has 22,000+ servers and 97M monthly downloads, but almost all of them are read-only lookups. Here's what idempotent deploys, rollback, and audit logging actually require from the ones that aren't.
Kubernetes Turned Off Its Own StatefulSet Speedup: The Real Status of maxUnavailable in 1.35/1.36, and the Quorum Math for 1.37
Kubernetes 1.35 shipped StatefulSet maxUnavailable as beta-and-on-by-default — then a rollout-deadlock bug got it flipped back off within the same release cycle. Here's the verified current status, plus the quorum math for when it's safe to raise it.
Ingress-NGINX Retired in March 2026 — Here's Exactly What Ingress2Gateway Migrates, and What It Doesn't
Ingress-NGINX officially retired in March 2026. Here's the annotation-by-annotation breakdown of what Ingress2Gateway migrates cleanly to Gateway API, what it can't touch, and why the gaps are riskier for a multi-tenant fleet than a single app.
GitHub Wobbled Three Separate Times in Eight Days: What Owning No Runner Capacity Costs
Three unrelated GitHub subsystems failed independently within eight days in July 2026. A concrete accounting of what broke, why that's a different risk shape than one outage repeating, and what actually changes when the build step and deploy target run on owned infrastructure instead.
Your Git-Push PaaS Assumes GitHub. Here's What Wiring It to Forgejo + Woodpecker Actually Takes
A git-push PaaS built GitHub-first assumes a webhook, a payload shape, and a status write-back that don't just carry over to a self-hosted Forgejo + Woodpecker CI stack. Here's the concrete four-piece build-out — webhook receiver, deploy trigger, commit-status write-back, OAuth app — that makes it a first-class integration instead of a race condition.
E2B Grew 375x to 15M Sandbox Runs a Month. Here's What the Sub-500ms Bar Actually Costs to Clear Yourself
E2B's sandbox volume grew 375x in a year, setting a sub-500ms boot bar for AI agents. Here's what that latency actually decomposes into, and how a Kubernetes-native fleet on owned hardware clears it with Kata Containers instead of rebuilding E2B's control plane.
Docker Hub's Unauthenticated Pull Limits Are Still 100/6hr: When a Self-Hosted Registry Actually Pays Off
Docker Hub still caps unauthenticated pulls at 100/6hr and free accounts at 200/6hr in 2026. A concrete threshold for when a CI fleet should authenticate, pay for Pro, add a pull-through cache, or run its own registry.
Docker Hub's 2026 Price Hikes Make the Case for a Self-Hosted Harbor Registry
Docker's 2026 price hikes and pull rate limits push a busy CI pipeline into a $1,800/year seat-priced plan for one feature: unlimited pulls. Here's what a self-hosted Harbor registry costs instead.
Kubernetes 1.36's Declarative Validation Goes GA: Which Webhook a Self-Hosted Cluster API Fleet Can Actually Delete — and Which One It Can't
Kubernetes 1.36's Declarative Validation only touches built-in types, not CRDs — the webhook a Cluster API fleet can actually delete graduated two releases earlier under ValidatingAdmissionPolicy. Here's the exact boundary between the three mechanisms, and the stateful checks none of them can replace.
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