HomePostsProjectsUsesAbout
16 min read

ZFS vs Btrfs vs ext4: What Should You Use in 2026?

ZFS, Btrfs, and ext4 have all matured by 2026, but they answer a different question: who is on the hook when a disk starts returning garbage? A sysadmin comparison of copy-on-write, checksums, RAID, snapshots, and out-of-tree kernel coupling, and which volume each one belongs on.

Latest

View all
16 min

BackupDrill Review: Supabase Backups That Prove They Restore

Supabase's built-in backups never leave the project, never cover Storage files, and are never restore-tested for you, but they do cover the Auth schema, which third-party dumps cannot reach. BackupDrill writes standard pg_dump copies of your schemas (plus Storage files) into your own bucket and drills them weekly on paid plans. A sysadmin review of what the drill checks, what it costs next to PITR once egress is counted, and where the young vendor falls short.

Business
10 min

FreeScout vs LibreDesk: Migrating a Self-Hosted Helpdesk in 2026

FreeScout and LibreDesk are the two self-hosted helpdesks sysadmins keep comparing in 2026: a mature PHP/Laravel app whose saved replies, custom fields, API, and SAML SSO are paid one-time modules, and a single Go binary on PostgreSQL and Redis that includes all of it free. A high-level comparison of stacks, deployment, and what actually migrates when you switch.

Development
15 min

Codex vs Claude Code vs OpenCode vs Pi vs Oh My Pi vs Hermes: Which Agent Fits Whom

Six coding agents and one personal agent, ranked light to heavy: Pi is a minimal terminal loop, Oh My Pi wires an IDE into it, OpenCode and Codex cover the middle ground, Claude Code is the polished Anthropic default, and Hermes is a persistent personal agent you can reach from Telegram or Signal. A comparison of capability, models, remote use, and who each one is actually for.

News
9 min

I Replaced My Alert Thresholds With a Forecasting Model

Every alert config has "page me if CPU is above 90 percent" rules that fire during normal peaks and stay silent during real anomalies at quiet hours. This post makes the case for swapping them for forecast-based alerting, made practical by Mistral's new Timestral time-series foundation models, which forecast zero-shot on a CPU with quantile output and schedule-aware covariates.

DevOps
16 min

Talos Linux vs Flatcar vs Fedora CoreOS: Three Ways to Run Immutable Kubernetes Nodes

Talos Linux, Flatcar, and Fedora CoreOS all promise an immutable, self-updating Linux for Kubernetes nodes, but they answer one question differently: how much of being a Kubernetes node does the OS decide for you? Talos is a shell-less appliance where Kubernetes is part of the OS, Flatcar is a minimal immutable host where you bring your own cluster, and Fedora CoreOS is the rpm-ostree Fedora underneath the Red Hat CoreOS world. A sysadmin comparison of lineage, governance, update mechanics, and which to pick for a fleet.

DevOps
17 min

IncusOS vs Talos Linux: Two Immutable Operating Systems, Two Different Philosophies

IncusOS and Talos Linux share one pitch: a locked down, immutable, API-managed operating system with no shell to log into. But they bet differently: Talos turns hardware into a Kubernetes appliance (now gaining a hypervisor in alpha), while IncusOS turns hardware into a virtualization host that runs everything else. A concept-level comparison for new users: how each philosophy works, what you give up, and which to pick and when.