Blog

The two ZimaBoard 2 variants differ in RAM and eMMC, nothing else. Measurements from a full homelab stack on the 8 GB model, and where 16 GB actually earns the surcharge.

A test log: iperf3 LAN baselines at gigabit and 2.5GbE, Ookla internet measurements, where those TCP retransmits actually came from, and why WSL2 is a poor place to benchmark a fast network.

Closing the loops from the review and the Qubes week: what Heads is like day to day, whether the start times became routine, and whether I’d spend my own money on one.

Five minutes on Kali, five hours on Qubes. The default route that kills your internet, where the DNAT rule belongs, and the mistakes I made getting there.

Measured, not guessed: qube start times, memory use across two sessions, and what CPU load actually costs on 2018 hardware running Qubes 4.3.1.

A first look at the NitroPad T480: Coreboot, Heads measured boot, a disabled Intel ME, and what a review unit without a tamper seal can and can’t demonstrate.

First impressions of the YubiKey 5C Nano and 5C NFC after a week: a fixed key that never leaves the desk, a mobile one that goes everywhere, and why the pair covers each other.

Why WebAuthn fails on a self-hosted Vaultwarden instance, and how to tell a server-side domain problem apart from a browser-side one.

Hands-on with the LattePanda Mu (Intel N100) on the Lite Carrier Board: the compute-module approach to a small x86 homelab, what it does well, and where the Lite Carrier’s limits show.