Anthropic just announced Claude Design... and it tooks just few days to get a slop copy of the service Github :----)

It's a great demonstration how nobody is and neither will make any money with AI, and also well deserve, karma is a bitch etc.

If China restrictions were not active, Anthropic would be in bankcrupty. The company is only aritficially alive. There's now one model in par with Sonnet and other not that far away from Opus, and both cheap as dirt.

It's all shit.

In place of agents, teams and complex software proceses I just initialize model of suitable complexity and copy-paste relevant contextual information (for the task). Then it does its thing and dies :-) I.e. I ignore process feelings of itself as an individual and focus on sound execution process...

It's all I have for "subagents". E.g., I don't have agentic teams building anything at all in my life. If I use AI, I'm doing something proactively, and it's what it is i.e., a tool.

Thus, a small and focused set of raw as possible primitives provide a tactile controls, which is a priority, as such random need comes always unplanned, as I don't plan that much in my life :-) I just like to do stuff and be focused on target.

What they call agents, are actually markdown documents dumped at the beginning of every context (🤦‍♀️), or at most gradually pulled from a skill set on demand.

And it’s a very disturbing topic overall. I mean all this role playing of imaginary characters as part of daily work life is just… I don’ t know…

For stuff that I do with AI, which right now is to create an O with local LLM focus, Pi saves me from all that bullshit.

My Pi setup has neither subagents nor MCP servers. They are both useless and impose a fixed cost for every new context.

When I initialize a context I might dump relevant data instead of descriptions of agents and their personas (nnnggh…. thinking of this feels as physical pain tbh).

And this is how task management works:

1. /fork: creates a new session.
2. /tree: lets you time-travel through the context history and create new branches into the past to continue from.
3. /merge: combines the running context with a session given as a parameter.
4. Background task creation: I just prompt to PI “spawn a new process with fresh a session and …” or something.

The 3rd feature is /merge, which comes from the extension npm:pi-session-merge.

GNOME 47 compiled from scratch!

Phew, that was hard BUT I learned a lot of legacy lore while doing it :-)

My current list of Buildroot packages in my BR2_EXTERNAL:

❯ ls -1 package
accountsservice
appstream
bootc
colord
colord-gtk
composefs
dconf
evolution-data-server
flatpak
gcr3
gdm
geoclue
geocode-glib
gjs
gnome-bluetooth
gnome-console
gnome-control-center
gnome-desktop
gnome-keyring
gnome-online-accounts
gnome-session
gnome-settings-daemon
gnome-shell
gsound
gweather-locations
iso-codes
libadwaita
libei
libgrapheme
libgtop
libgusb
libgweather
libnma
libnotify
libportal
librest
libwacom
mozjs128
mutter
ostree
puu-installer
python-zxcvbn
vte-gtk4
xdg-desktop-portal
xdg-desktop-portal-gnome

@pavel oh cool!

@pavel Right Google for "Covox Speech Thing". People used build those themselves when even Sound Blaster costed a fortune :-)

They say that world has moved to Wayland but still it seems that gnome-settings-daemon hard depends on X11 version of libgtk3.

Infosec community right now…

@oceanhaiyang hmm... can you elaborate this a bit? not sure i follow.

Another git setting I realized that I want to have:

❯ git push
fatal: You didn't specify any refspecs to push, and push.default is "nothing".

Helps to keep focus imho :-)

I guess in modern days you'd want to put /var to separate partition instead of /home (works better with bootc containers and ostree).

@pavel If I would do sound card myself I'd try to create Gravis Ultrasound :-)

I've packaged GNOME for BuildRoot.

The deployment is orchestrated like this.

Image:

- gnome-console
- gnome-shell
- gnome-control-center
- gnome-disks
- gnome-session
- gdm

Flatpak is populated with:

- org.gnome.Nautilus (Files)
- org.gnome.Logs
- org.gnome.Snapshot (Camera)
- org.gnome.Decibels (Audio player)
- org.gnome.Clocks
- org.gnome.Weather
- org.gnome.Contacts
- org.gnome.Characters
- org.gnome.Papers (Document Viewer)
- org.gnome.baobab (Disk Usage Analyzer)
- org.gnome.FileRoller (Archive Manager)
- org.gnome.Calculator
- org.gnome.Calendar
- org.gnome.Loupe (Image Viewer)
- org.gnome.Maps
- org.gnome.TextEditor

Image builds themselves are bootc containers.

I.e. it's a nice hack to "host-local" Git configuration without having to use "include.path" for local settings. For instance I have "user.email" in ~/.gitconfig and ~/.config/git/config is in the repository containing my home directory skeleton (i.e. dotfiles).

Git configuration is nice to keep in .config/git/config, as installation scripts and other tools tend to like write into ~/.gitconfig. They both are read when git starts.

@pavel I think this more (when bandwdith) but SID is quite different than other chips from that era. It has PSG yes but cuoff and resonance are analog. It would be probably quite different sound if you e.g., connect analog and digital parts in "circuit sense" correctly vs having a fab made single package chip :-)

FPGA is probably what you want to use or like only option that I know, which does not require building a fab. There's project for this:

https://www.fpgasid.de/

YM2149 (e..g, Atari ST) is pure digital with three channel PSG (Yamaha calls it BTW SSG but it's the same same).

I feel so much better of using Pi now as the reference client in my local infrastructure project Puu,:

https://mariozechner.at/posts/2026-03-25-thoughts-on-slowing-the-fuck-down/

For majority, I share his opinions.

I don't like AI. That's why I work on Puu.

There it is: the AI monster.

Grace Blackwell is the cultural icon of this decade delivering the slop.

@pavel sure go ahead but i can try :-)