My favourite part of threaded platforms is the arbitrary and tangential discourse

In my research group we could tell instantly and it would usually act as a mark against the paper (ie read this one later).

If you’re reading a lot of papers it becomes apparent.

Half-baked is a bit unkind. Sway is quite performant, stable and lean.

Well that’s good to know because I had some terrible luck with it about a decade ago. Although I don’t think I would go back to windows, I just don’t need it for work anymore and it’s become far too complex.

I’ve also had pretty bad luck with BTRFS though, although it seems to have improved a lot in the past 3 years that I’ve been using it.

ZFS would be good but having to rebuild the kernel module is a pain in the ass because when it fails to build you’re unbootable (on root). I also don’t like how clones are dependant on parents, requires a lot of forethought when you’re trying to create a reproducible build on eg Gentoo.

I gotcha:

• Btrfs
  • BTree File System
    • A Copy on White file system that supports snapshots, supported mostly by
• ZFS
  • Zetabyte File System
    • Copy on Write File System. Less flexible than BTRFS but generally more robust and stable. Better compression in my experience than BTRFS. Out of Kernel Linux support and native FreeBSD.
• HFS+
  • what Mac uses, I have no clue about this. some Copy on Write stuff.
• NTFS
  • Windows File System
  • From what I know, no compression or COW
  • In my experience less stable than ext4/ZFS but maybe it’s better nowadays.

HFS+ has a different features set than NTFS or ext4, Apple elect to store metadata that way.

I would imagine modern FS like ZFS or btrfs could benefit from doing something similar but nobody has chosen to implement something like that in that way.

Vikings and org-mode.

Org-mode does not have an API but I’ve separated out multiple files and synced via git to moderate success with my gf. No API but maybe with gitea and orgize you could do something?

I cannot.

Endeavour OS is great but it’s just arch.

Gentoo with oddlama/gentoo-install is nice too.

What about vpn behind WireGuard/OpenVPV?

I would presume no?

It’s not too bad. I very rarely recompile everything from scratch and after I do that I just create a snapshot with btrfs. Are usually then chroot into that snapshot and compile everything natively overnight for that 5% Theoretical performance boost.

Most recently I took that snapshot and then used btrfs send to adapt it to a laptop as well and that worked quite well actually.

Everything I install is typically through flatpack or distro box just like silver blue. This means install times are pretty much okay but I have a huge amount of flexibility in the way the system works

Also heaps of binary packages as well, so that’s not too bad. The binary packages much slower than both arch and Alpine but not a lot slower than for example Fedora.

They both kind of suck in their own way.

If you want to things to run at startup and you’re not on systemd, rootless docker is probably easier.

Otherwise podman is mostly fine but be careful of native overlay if you’re not on BTRFS, this causes some pretty long build times.

Use Quartz and Obsidian because it’s easy. If not mcdocs.

I think I’ll keep doing it. It’s worked fine for the past decade 🤷

Why? I almost always have master/dev/stable.

Yeah I’ve been recommending Arch based did for a while. Personally I’m on void and Alpine, but as a first distro things like Cachy and Endeavour are unrivalled.

Yeah if it’s just for Plex something like Endeavour OS would be pretty much painless.

Definitely easier than fighting a key.

Emacs is a rough experience for one.

A lot of ML stuff does not, e.g. Microsoft DeepSpeed.

Lots and lots of CLI programs as well.

Have a discussion with chatGPT about a program you would like to write, use this to assist the development.

Evidence this as the source of the program. There is your re-research. It’s likely the implementation will differ substantially as well.

They might own the original program but it’s unlikely they broad concept.

Well it’s there, in one loooong print out. It’s not as bad as I’m making it out to be, however, I went back to python unfortunately.

The crucial issue with Julia, no error messages.

So I use Julia for things that need to be fast (e.g. moving hdf5 to SQL and ffts) but I use python for everything else (except ggplot).

Simply, the lsp is far less useful. An object might have a dozen methods that act like verbs or some attributes that act as adjectives.

In Julia there is a huge number of functions, that work differently for different types and different combinations of types. So finding the documentation involves finding the right name for a function that does different things for different types, then scrolling down the docs for the the behaviour that corresponds to the specific combination of inputs.

I moved from R/Py to Julia for a while before moving back to Py (and a little bit of Rust).

I love how fast Julia is and the 1-index is fine for me, but I still prefer py for the oop.