I’ve never kept a blog. Every time I considered it, the overhead argument won: a platform to maintain, a hosting bill, and posts that would have to be written from scratch. Not worth it.
Two things changed this year, so the math finally works.
Who I am
I’m David Bensoussan, a software engineer. Most of what I build outside work lives in one monorepo: the apps, the infrastructure that runs them, and the CI in between. The infrastructure part is a homelab — two second-hand Debian machines running about 45 services: photos, media, documents, monitoring, an LLM server on an old GPU, and the CI runners themselves. The full topology and progress log is at homelab.bensoussan.de if you want the inventory. Everything is provisioned with Ansible, and every significant decision gets written down.
Hosting is free now
This blog is Hugo-based, so it’s just a simple, standard static site. No CMS, no database, no bill.
The homelab itself also needs less babysitting: I moved it from Docker to rootless Podman under systemd this year, so no more separate daemon to run and monitor.
Writing is cheap now
Every project in the repo keeps a running progress file. Decisions become short Architecture Decision Records, and incidents leave detailed commit messages. I keep these for the AI coding agents I work with, so they can pick up tasks with context. A convenient side effect: by the time something is worth a blog post, most of it already exists as notes.
And to say it plainly: an AI agent helps me draft these posts, the same one that helps me build the things they describe. It works from the decision records and fact-checks claims against the git history, because left alone it will produce a better anecdote than what actually happened. The opinions and the mistakes are mine.
That setup isn’t risk-free, and I won’t pretend it is. The same agent once read a file it was told not to open, and once deleted directories it shouldn’t have touched. Both incidents, and the guardrails they produced, are coming up as posts.
What to expect
A post roughly every two weeks, on those topics, at least:
- AI agents on real infrastructure — what breaks, and the working rules that came out of it.
- An affordable homelab — two used machines, with real costs and the reasoning behind each choice.
- Architecture decisions — systemd instead of Kubernetes, backup design, and where those choices failed in practice.
- Self-hosted CI — the odd bugs you meet when your GitHub Actions runners are in your own home.
- What I built — the useful and the useless.
No guides, no listicles. Just things that actually happened here, with the commit log to back them up.