OpenCode was the first agent harness I used, and I have always like it. You can configure a wide variety of providers, but it's open source and has a number of core contributors.
The other opinionated option is Pi (the Pi agent harness). This is a great lightweight option and also supports a number of providers. You can also use local model servers.
That's true. Even then, though, you're dealing with backwards-compatibility support as the system updates. A compiled binary might run well for the systems it was compiled for, but what about longer timelines (a decade)? Will the newest system be able to easily run that compiled binary? Not always... and there's always the possibility it might include vulnerabilities that weren't discovered until later.
I was reading about terminal text editors (em, en, vi, vim, neovim, etc.), and it's interesting how some of the software that "lasts" is more like Theseus' Ship. All the original components replaced over time, but the core concepts last.
Really nicely written and quite thought-provoking. I think about when I die, will anyone be able to use or maintain any of the software I've written? Updates and patches are something so entwined with software that I doubt much of my code would be worth using if it suddenly froze.
It puts a beautiful spotlight on OSS communities and what they do to keep software alive through refactoring, iteration, patching. Also, on well-written documentation— perhaps that's even more important than the code for longterm maintenence and value. A good thesis that encourages someone to write it again, and better?
I've been using dotter for 5+ years and it's been an amazing tool. Doesn't go all in like NixOS, but lets you get close to that functionality by allowing templating to reuse a single dotfile repo across different systems. I also like how you can easily conditionally select configurations and scripts to load depending on environment.