Context is king which is why even the biggest models get tied in knots when I try them on my niche coding problems. I’ve been playing a bit with NotebookLM which promises to be interesting with enough reference material but unfortunately when I tried to add the Vulcan specs it complained it couldn’t accept them (copyright maybe?).

We have recently been given clearance to use the Gemini Pro tools with Google office at work. While we are still not using them for code generation I have found the transcription and meeting summary tools very useful and certainly a time saver.

The OSI have had a go: https://opensource.org/ai/open-source-ai-definition

I’ve been using https://containertoolbx.org/ recently to manage my “other distro” requirements. It doesn’t do anything special but works nicely as a wrapper around podman and does all the bind mounts and uid mappings so you can just enter your $HOME as though you have set up your account in a new OS.

Also the PFN page locking patches so device memory can be reliably shared with VMs (used for some of the virtio-gpu modes).

That was a trip down memory lane. I think a lot of the engineers that worked at Transmeta ended up in places like Intel (and maybe Apple?) which tracks with them being an early pioneer in managing power envelopes.

Signal is the non-corporate replacement for What’s App although realistically you end up having both due to network effects.

I can kinda see where the vacuum and (lack of) gravity might help with crystal growth but how do you then return something that sensitive back to earth?

I’ve never really gotten on with the trackpads although they do feel nice and tactile. I’ve now got a dock setup so I can switch my primary monitor across to the steam deck along with the audio and a usb switch for my keyboard and mouse. I’m finally catching up with the RTS and strategy games in my unplayed queue.

I use foot which is Wayland aware and renders Unicode fonts. Honestly I don’t need much from the terminal itself as I’m usually in tmux to deal with all the “tabs” and scrollback.

Does this use the same attention architecture as traditional tokenisation? As far as I understood it each token has a bunch of meaning associated with it encoded in a vector.

Quite. Servers aren’t free and someone needs to pay the bills and increasingly distribute the moderation load. I’m happy with my Mastodon and following a few federated accounts on threads and bsky. But I’m not going to someone they are a bad person for choosing something that is familiar yet a little different while escaping x/itter.

Ah that will be it. Still grey on transparent isn’t super accessible.

I’m not sure why it rendered so poorly in Lemmy. It’s a terrible colour scheme but at least I could make out the bars when I followed the link.

More passion, more energy…

It was a nice couch co-op game to play with my wife and kids.

If you license a design from someone you’ll still be paying something. Sure there are also free implementations but they are aimed at microcontrollers, you won’t get any server class chips for free.

Now you can install uboot and get a property uefi implementation it shouldn’t take too long: https://social.treehouse.systems/@cas/113539953511804908

I need to check the driver situation but I don’t think there was anything particularly windows only on the SoC.

I just want to buy home automation gadgets that don’t need a bloody cloud account to work.

I think car automation peaked at adaptive cruise control. It’s a simple tractable problem that’s generally well confined and improves the drivers ability to concentrate on other road risks.

It depends what they want to do. They can fork and take on the burden of maintaining the whole tree in which case good luck with that, linux is too much of a fire hose to enable a 3rd party to assemble something similar making different choices about what they merge. Otherwise they can maintain a re-based fork that tracks the Torvalds tree and then congratulations you’ve just invented a feature tree that can do contribution with extra steps.

I don’t think algorithms themselves are to blame but what they are tuned for. While engagement/eyeball hours for the adserver is the prime metric the quality of experience will be subservient to it. If the algorithms could better measure your mood and stimulation levels and maximise for that the effect would be less toxic. Ideally if it realised you were just mindlessly consuming it could suggest maybe you’ve done enough today and to try something else. But that I fear that is not something the owners of the various ecosystems want.