Gabriel Fontes
532a35ba7a
/srv/media is now a single btrfs subvolume on the media disk holding the libraries and both staging trees, so imports are same-filesystem renames. 426 torrents were re-pointed through the deluge UI and are seeding without a recheck, and the M.2 went from 164G to 39G used. The layout is described by hardware-configuration.nix, deluge.nix and sabnzbd.nix better than prose could, so the plan has nothing left to say. Two follow-ups it was carrying, recorded here rather than kept as a file: copying across filesystems flattened every reflink between the music library and the torrent payloads, so those trees hold 170 GiB where they held ~124 and duperemove would reclaim it; and the media filesystem has still never been scrubbed. Assisted-by: pi (claude-opus-5)
Foundry
My public infrastructure monorepo: the NixOS/home-manager configuration that runs all my machines, plus the source of the personal projects it deploys. Requires Nix flakes.
Looking for something simpler to start out with flakes? Try my starter config repo.
Repository layout
hosts/nixos/ per-machine NixOS configurations (atlas, alcyone, ...)
hosts/system-manager/ configurations for non-NixOS Linux systems
home/ home-manager configuration (feature-flagged)
modules/ reusable nixos/ and home-manager/ modules
overlays/, pkgs/ package overlays and custom packages (including wallpapers)
lib/ pure-Nix Material You color engine
projects/ my public projects deployed from here (e.g. the m7.rs website)
Highlights:
- NixOS configurations: desktop, laptop, servers
- Opt-in persistence through impermanence + blank snapshotting
- Encrypted single BTRFS partition (with disko for declarative partitioning)
- Secure Boot via lanzaboote
- Fully declarative self-hosted stuff
- Deployment secrets using sops-nix
- Mesh networked hosts with tailscale and headscale
- Flexible Home Manager configs through feature flags
- Extensively configured hyprland environment
- Declarative theming: wallpapers and a pure-Nix Material You color engine
- Hydra CI/CD builds every host, serves a binary cache, and hosts auto-upgrade by pull deployment
About the installation
All my computers use a single btrfs (encrypted on all except headless systems)
partition, with subvolumes for /nix, a /persist directory (which I opt in
using impermanence), swap file, and a root subvolume (cleared on every boot).
Home-manager is used as a NixOS module, integrated via home-manager.users.
Secrets
For deployment secrets (such as user passwords and server service secrets), I'm
using the awesome sops-nix. All secrets
are encrypted with my personal PGP key (stored on a YubiKey), as well as the
relevant systems' SSH host keys.
On my desktop and laptop, I use pass for managing passwords, which are
encrypted using (you bet) my PGP key. This same key is also used for mail
signing, as well as for SSH'ing around.
Tooling and applications I use
Most relevant user apps daily drivers:
- hyprland + hypridle + hyprlock
- waybar
- helix
- fish
- alacritty
- qutebrowser
- neomutt + mbsync
- khal + khard + todoman + vdirsyncer
- gpg + pass
- tailscale
- podman
- zathura
- wofi
- bat + fd + rg
- kdeconnect
Some of the services I host:
- hydra
- jellyfin
- *arrs (including torrent and usenet)
- prometheus
- websites (such as https://m7.rs)
- minecraft
- headscale
Nixy stuff:
- sops-nix
- impermanence
- disko
- lanzaboote
- home-manager
- and NixOS and nix itself, of course :)
Let me know if you have any questions about them :)
Unixpornish stuff
AI usage note
Since June 2026, I've been trying out LLM assistance in my workflows. So far it feels pretty good; brainstorming helps me a lot with decision paralysis. I'm trying to keep my use bounded and disclosed. I think it's a useful tool, but it should be adopted with care.
I will write a decent blog post about my opinions on AI at some point. The gist is:
- These things are useful for bounded tasks with good docs and reviews, but suck at owning architecture or accountability
- There's no going back now that open-weight models run on consumer hardware, we can't uninvent them
- Boycotting AI does not help by itself; non-use is not a political strategy
- We need clear LLM policy on projects rather than trying (and failing) to forbid it
- Disclosure is very important, and trying to pass LLM output off as human-written is disrespectful
- LLMs should never have been built by scraping and exploiting our art and work
- If copyright doesn't protect our creations, it shouldn't protect their models; support open-weight models and distillation as harm reduction
- We need regulation, decentralization, and redistribution of the value LLMs generate
- Machines can't be horny, thus can't create art; LLMs are for utility, not art or craft. Pay an artist instead
Some bibliography I'd recommend:
- Alice's "AI Sucks. Hating it is not enough." (please read it, it's an amazing piece)
- Drew's "The cults of TDD and GenAI" (I don't really agree with the non-use perspective, but otherwise I think it's solid)
- Armin's "Communities of Not" and "The Center Has a Bias" (I'm not a centrist politically, though)