What are you doing this weekend?

TL;DR: Unholy union of Unison, Bevy and Lisp tied to a graphics engine.

I'm finishing an initial release of Yabu a natively-compiled Lisp with a static strong type system, type inference, persistent data structures and content-addressable version control system made for Game Jams. What does it mean made for game jams? I've taken my Rekishi idea of a database backed version control system and implemented it natively for this language. You can open a session and your friends can connect to it. From then on you are building the codebase interactively with them. If they create a function that function appears in your system as well. There is a concept of branches/namespaces where you can isolate yourself for a bit to work on something, but you can collaborate freely inside each namespace.

A branch/namespace is nothing else than a mapping from symbols -> implementation hashes, when you create a new version of a function I compile the new version, generate a hash and change the symbol pointer to it but the old version still exists. This plays nicely with the next feature:

Playing the game records all inputs as well as setting a randomness seed for each frame, this creates fully replayable game sessions that also exist inside the project shared database, allowing anyone connected to replay that session and introspect each frame to see what happened. Each frame also keeps track of symbol -> function hash mapping. When you replay a session the system pulls the necessary functions from the shared database so you have all the necessary functions available. Keeping track of this per frame instead of per session allows you to fork the execution of a session at a particular frame to interactively fix bugs. If you have a bug at frame 100, you can navigate there, then change the functions. The previous frames will keep running with the old functions allowing deterministic reproduction of the state up to frame 100 without having to save all state. Then from frame 100 onwards it will run with the new functions. These forks also exist in the shared database so you can see all branches and changes done to fix the issue.

I've also been playing quite a bit with how a static typing lisp may work with a repl-based workflow and I've come to what I think is a fairly cool system. You can modify functions and they are allowed not to pass typechecking, the new definition will be kept but in a kind of "staging area". This allows you to iterate on fixing type issues in an interactive manner. Once everything type checks that is merged into the codebase making it availabe to everyone. So while the code evolves for everyone the shared codebase will always typecheck.

All of this can also extend to things like assets.

With all this together I can realize my dream of fully collaborative, interactive, repl-driven, fast, introspectable, experimentable game development in game jams.

Mostly spending time with my daughter. However, in between, I'm working on migrating my handful of public self-hosted services from a rented Hetzner server to a DMZ VM in my home lab, which is sat behind a cheap-as-chips bastion VPS. I've been brushing up on my knowledge of VLANs, firewalls, routing, and security as I begin to migrate services. It's been fun!

I'm playing with "generative art" and animation, and trying to sync procedural animation with audio.

The first step was this basic FFT and waveform visualization.

Right now it uses Cairo to write PNG images and then ffmpeg to create a video file, but I'd like to add support for realtime animation using X, as well, so I can preview in realtime, and also create video files.

Decluttering the house. So far we're up to 7 car trips to goodwill and the recycling centre.

Same thing different week, I just implemented parsing PRAGMAS for sqleibniz and now I’m starting to work on INSERT and it seems like I’m in for a long red green cycle

Finishing a home NAS setup on FreeBSD. Actually, I’ve been doing that over the last few weeks.

Well - why not share some thoughts around that story here?

FreeBSD was my very first Unix system - the one that introduced me to this whole world and genuinely changed my life.

It was around 2007 when I first built a kernel with custom parameters. FreeBSD 6… or maybe 7 - I can’t quite remember. I also don’t remember the exact CPU (Pentium II? Celeron?), but I clearly remember having 176 MB of RAM. The kernel build took about 8 hours.

After that, I used FreeBSD for years on VPS servers for my blogs and learned system administration with it. The very first Unix-related book I read was Absolute FreeBSD, 3rd Edition by Michael W. Lucas.

Around 2012, I switched to Linux - CentOS. That happened around the time FreeBSD moved to pkgng. I ran into quite a few issues during that transition and got a bit frustrated. From then on, my servers were running CentOS. Later, when the CentOS project was effectively closed (2021?), I migrated to Debian. As for desktops, I’ve been using Arch Linux (btw) since 2016.

Recently, in December 2025, I bought an old ThinkPad X200 and decided to try FreeBSD on it. An old laptop - why not try an OS from my younger years? 🙂

That installation experience was originally described on my blog and later shared on lobste.rs by @terryfilch: FreeBSD: Installation on a ThinkPad X200 in 2025

And when I used FreeBSD again after so many years - gosh, it felt really good. After years of Linux with all the systemd complexity, FreeBSD felt simple, clean, and well-organized. Almost sweet. And very nostalgic.

Using FreeBSD again brought back memories of when I was just starting to learn sysadmin work - Unix, networking, security.

That’s what eventually pushed me to build a home NAS. I bought a Lenovo ThinkCentre M720s SFF and three SSDs (one for the system, two as a ZFS mirror), and started playing with FreeBSD, ZFS, NFS, Samba, and related tooling.

That said, while I think FreeBSD is a great fit for a home NAS, I’ll still stick with Arch Linux on the desktop - mainly because of hardware support and day-to-day convenience. Another important factor for me is Docker: on FreeBSD it’s only realistically available via virtual machines, while for my daily work (I’m in DevOps) I often need to spin up containers locally to quickly test and validate things.

For servers, I prefer Debian (yet) - mainly because it’s very much a “set it and forget it” system: install it once, enable unattended-upgrades for security patches, and it takes care of itself. FreeBSD, in comparison, tends to require a bit more hands-on attention, which is perfectly fine - and even enjoyable - but not always what I want from production servers for personal use.

Apologies for the long comment - this just brought back a lot of memories.

I'm going to write a simple profiling display using only HTML grid. I did a PoC on yesterday and I'm confident.

Why? In 2025, I made traceboot to trace process creation starting at early boot. The output was something to feed to Perfetto. The results where good enough but Perfetto isn't designed for me and simple changes can break the final output.

Two things planned:

• Adding two-tier discovery to epithet to simplify client setup. Epithet is an SSH auth management system built around short lived certificates and managed via a bespoke ssh agent.

• Figure out alternate sources of data for Ook to cover books which Open Library doesn't know about. Generally these are indie or self published works. Ook is a MacOS desktop app for fast local search of my kindle library.

I'll be going through Introduction to Zig, focusing mostly on the projects. I'm well-enough acquainted with C that I expect that I can skim most of the sections related to how Zig itself flows as a language; I find that I learn best by doing!

Baking it seems:

• Brown butter cake with a maple buttercream frosting
• Another go at trying to make Detroit-style pizza at home

(I really wish dairy wasn’t so expensive in SEA)