A Linux, self-hosting and AI newsletter for people who run their own hardware and want the numbers before the opinions. One email a week, no sponsored placements, and every benchmark in it comes with the script that produced it.
What’s in each issue
Concrete things, with the measurements attached:
- Benchmarks with the raw numbers and the code to rerun them. The runtime comparison found Deno serving 133,093 req/s on a single core against Node.js at 47,734, and Bun installing 585 packages in 5.8 s cold versus npm’s 11.8 s. Every figure came out of a public repo you can clone and run: Bun vs Deno vs Node.js .
- When the boring tool wins, and by how much. DuckDB finished a join in 20 ms on the same file where pandas needed 6,795 ms: DuckDB on CSV and Parquet .
- Local AI on hardware that isn’t a data center. Which quantization and offloading combinations actually fit a 26B mixture-of-experts model into 8 GB of VRAM, and which ones swap themselves to death: Gemma 4 26B on 8GB VRAM .
- Linux tooling compared on Botmonster’s own machines. Latency, memory, and where each one falls over: Alacritty vs Kitty , Hyprland vs Sway vs COSMIC .
- What the people actually running a thing say about it, pulled from live threads with upvote counts and verbatim quotes: how Reddit received Claude Opus 4.7 .
Why subscribe to this Linux and self-hosting newsletter
- Numbers before opinions. Measured on real hardware, with the method stated so you can argue with it.
- No vendor marketing. Nothing here is paid for by the thing being tested.
- One email a week. That is the entire sending schedule.
- One-click unsubscribe, and your address is never shared or sold. See the privacy policy .
How often it sends, and every past issue
One email a week. Nothing else gets sent to the list, so subscribing costs you about 4 emails a month and no surprises.
Every issue stays readable without subscribing: read the full archive on Buttondown . Skim a couple of issues before you decide.
Who writes it
Botmonster is a software engineer and technology writer. Everything published here gets tested on hardware Botmonster owns and pays the electricity bill for.
Every performance claim ships with its benchmark. It goes into a public repository
as a single run.sh that installs pinned tool versions and reproduces the whole run on any Ubuntu box, with the raw output committed alongside it. If a Botmonster number is wrong on your machine, you can prove it.
Botmonster Tech