Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
AI Coding Agents Ship Daily but Break More Often: What the 2026 Data Says a Rollback-First Deploy Pipeline Needs
Heavy AI coding-tool users ship daily but remediate 22% of deploys at over 7 hours each. The 2026 data says pipelines must be rollback-first: one-command rollback, agent-readable deploy health, and automatic canary rollback.
Before Agents Create Their Own Compute: A Capacity Budget for Agent Sandboxes on Kubernetes
Agent-callable sandbox creation turns capacity into a control-plane decision. A worked budget for a three-node Hetzner fleet: warm-pool reserve, 32 claimable sandboxes, TTL defaults, per-tenant quotas, and admission rules that stop one runaway loop.
A2A Hit 1.0 and Moved In With MCP: What Agent-to-Agent Handoff Actually Adds to Deploy-from-Chat
A2A reached 1.0 in March 2026 and joined MCP under the Agentic AI Foundation in August. Here is what agent-to-agent handoff concretely adds on top of a one-agent MCP deploy server, the exact extra surface a self-hosted PaaS must expose to support it, and when that work belongs on the roadmap.
X Ships a Hosted MCP Server, and Agent Access Is Now Table Stakes
X's hosted MCP server joins Slack, Notion, GitHub, and Stripe in making first-party agent access the default. Here is the concrete production-grade bar — auth, transport, scoping, rate limits — and how it re-sequences your own MCP roadmap.
Your CI Server Just Became an Agent Tool: What TeamCity's MCP Server Means for Deploy-From-Chat
TeamCity's MCP servers — the community's 87-tool Full Mode and JetBrains' official 2026.1 endpoint — turn builds, tests, and agent pools into agent-callable tools. The CI half of deploy-from-chat is here; per-tool scoping and human approval decide whether it is safe to use.
1.5TB of RAM and 192 vCPUs for Under €600/Month: What a Hybrid Talos-on-Hetzner Build Validates About Splitting Control Planes From Workers
A third-party builder ran 1.5TB of RAM and 192 vCPUs on hybrid Talos-on-Hetzner iron for under €600/month. Here is what the numbers verify, what broke on the VM-to-bare-metal seam, and where the bare-metal-is-cheaper claim lands.
Suga Ranked 7 Heroku Alternatives for Docker Workloads — I Recomputed the Bill on One Reference Stack
Suga's price-verified ranking compares seven Heroku alternatives for Docker workloads. We reprice the same reference stack on all seven plus owned hardware — and find egress is the line item that decides it.
Running Your First StorageVersionMigration: What Kubernetes v1.37's GA Migration API Changes in Your Cluster API Upgrade Runbook
Kubernetes v1.37 turned storage version migration into a stable, always-on API. Here is how to run your first migration — Secrets re-encryption, CRD version collapse, and per-resource status checks — and where it slots into a Cluster API fleet's upgrade runbook.
Your Pod Can Now Evict Its Neighbors to Grow: Scheduler Preemption for In-Place Resize Goes Alpha in Kubernetes v1.37
Kubernetes v1.37's alpha resize preemption lets the scheduler evict lower-priority pods so a growing pod completes its in-place scale-up instead of stalling in Deferred. Here is how the lifecycle works, a ten-minute kind reproduction, and the PriorityClass policy a self-hosted PaaS needs before trusting it.
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