Archive
Long-form posts only — TILs, links, and quotes live on the feed. Sorted newest first. Posts get edited; the originals are linked in each post's footer.
Tags: go · linux · perf · postgres · networking
2026
- apr 09EEVDF: a first look at the new Linux schedulerlinux · perf
- mar 22Go arenas are gone, and that's finego
- feb 14When the Postgres planner lies to you (and how to tell)postgres
- feb 02Tail latency for people who write CRUD appsperf
- jan 18A year on NixOS, mostly happilylinux
2025
- dec 11My homelab, after the MikroTik switchnetworking
- oct 28Go 1.23 range-over-func, in practicego
- aug 14The "SQLite is single-writer" myth, and what WAL actually doespostgres
- jun 02I had high hopes for io_uringlinux · perf
- apr 19QUIC and HTTP/3 in production: the boring reportnetworking
- feb 07eBPF ringbuf vs perfbuf, with numberslinux · perf
2024
- nov 30Eight years of writing Go in productiongo
- sep 12p99 is a liar, and you've been lied toperf
- jul 04gomon: a small tool, slightly opinionatedgo
- may 21VACUUM: the most underappreciated word in databasespostgres
- mar 03Three years on, BBR is still mostly the right callnetworking
2023
- dec 19Most of your Grafana dashboards are wrongperf
- oct 02Go error handling, after eight years, is finego
- jul 16cgroups v2 in angerlinux
2022 and earlier
Older posts have been quietly retired — they were either wrong, or embarrassing, or both. A few of the better ones survive on the wayback machine if you really want to go digging.
Total: about 19 posts that I'd still defend, plus everything on the feed.