Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
15,465 MCP Servers, 0 Governance: What OX Security's Census Means for Your Deploy Tools
OX Security mapped 15,465 MCP servers and found agents reaching China, Russia, home networks, and six abandoned domains buyable for $4. The four findings with exact numbers, the Always-Allow attack chain, and a control matrix for governing deploy agents' tools.
Outhaul Is a Single-Binary Self-Hosted PaaS on SQLite: What a Deliberately Minimal Dokploy and Coolify Alternative Gets Right About the First Box, and Where It Stops
Outhaul packs push-to-deploy into a 22 MB Go binary and one SQLite file. A parts-count comparison with Dokploy and Coolify, plus what the move costs when one box stops being enough.
OpenShip's 132K-View Launch: What Its Mail Server, One-Click Stacks, and MCP Bet Validate — and Where Docker+SSH Diverges From Kubernetes
OpenShip's July 2026 launch put a built-in mail server, one-click data stacks, and MCP agent ops on servers you own. The numbers behind each bet, and where Docker-over-SSH parts ways with Kubernetes.
One Sandbox API for E2B, Modal, Vercel, and Kubernetes: What Capability Negotiation Actually Buys You
Open Sandbox Router puts one API over E2B, Modal, Vercel Sandbox, Daytona, and your own Kubernetes cluster, placing each sandbox by declared capabilities, cost, and live health. This evaluation walks through the mechanism, the 2026 pricing math, the two failover regimes — and the isolation differences no router can hide.
Northflank Prices the Single-Box Decision: What a Price-Verified Dokploy-vs-Coolify Check Gets Right — and What Neither Panel Answers at Machine Two
Dokploy Cloud is cheaper at exactly one server; Coolify Cloud wins at every size above it — and neither panel provisions, heals, or autoscales machine two. A price-verified walkthrough of Northflank's September 2026 check plus the fleet gap both panels stop at.
Kubernetes 1.37's Node Lifecycle Conditions: Five New Signals Your CAPI Fleet Will Learn to Read
Kubernetes 1.37 reserves five well-known Node conditions for drain, maintenance, and shutdown state — a shared vocabulary no core controller reads yet. Here is the before/after map for a Cluster API fleet's health-check and scale-down signals, and what to publish this quarter.
No More Self-Hosted Git After 15 Years: What AI Actually Changed About the Self-Hosting Calculus
A veteran Linux developer shut down his public git server after 15 years because AI scrapers killed it — then watched the eulogy hit 298 points on Hacker News. His real reason, the two AI pressures behind it, and a keep/move/hedge framework for deciding what stays on your own machines.
Nobody Gives You Always-Awake for Free: September 2026's Paid Floor, From $2 Fly Machines to a $4 Hetzner Box
A September 2026 audit of every hosting free tier finds no always-awake compute for $0: the paid floor runs from $2 Fly machines to $5 Railway Hobby, with managed Postgres adding $5-6 — while a Hetzner box at roughly 4 euros a month holds the whole fleet.
MCP Won the Interface War: What 10,000 Servers and 97M Downloads Mean for Your Self-Hosted Endpoint
MCP passed 10,000 public servers and 97M monthly downloads under neutral Linux Foundation governance, but registries, managed gateways, and vendor defaults are the new lock-in surface. Here is the case for running your deploy-authority endpoint on infrastructure you own.
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