Anyone tried using this as a sub-agent for a more capable model like Claude/Codex?

If you’re looking into small models for tiny local tasks, you should try Qwen coder 0,5B. It’s more of an experiment, but it can output decent functions given the right context instructions.

> [Qwen coder 0,5B] can output decent functions given the right context instructions

Can you share a working example?

So… a prompt? I’m not on my laptop but I hooked it to cmp.nvim, gave it a short situational prompt, +- 10 lines, and started typing. Not anywhere near usable but with a little effort you can get something ok for repetitive tasks. Maybe something like spotting one specific code smell pattern. The advantage is the ridiculous T/s you get

The combined (input/output) context window length is 4K. Claude would blow through that even when trying to read and summarize a small file.

With a small/minimalistic harness like Pi maybe it works well?

I was thinking about the other way: Could you use this in front of Claude to summarize inputs and so reduce your token counts?

project started with

trying to run openclaw with it in ultra token saving mode, did totally not work.

great for shell scripts though (my major use case now)

Would be really interested to know what the building is actually for. just a tech demo? or does it have a real use?

The website talks about it being a tech demo and art. It's something designed to be an attraction.

https://www.tor-alva.ch/en/

If you log out/in or reboot it will take effect.

Previously killing SystemUI server was enough to restart the menu bar, but that doesn't seem to do the trick anymore.

> The users who run into issues with menubar space would probably be well served to question if they really need all that stuff

Look, I don't even want half that stuff, but the reality is that a bunch of tools mandate by my employers, and a bunch of utilities that are needed to make modern MacOS work reliably, all live up there.

A couple of VPNs, DropBox, OrbStack/Docker, eqMac to unfuck volume control on external displays, BetterDisplays and BetterSnapTool to unfuck everything else about external displays... I'm already most of the way to the notch on a 14" macbook

Dude. How am I only just learning this? This needs to be plastered loudly over the internet

Are they? The docs say "You can also register any Hugging Face model into your Lemonade Server with the advanced pull command options"

That won't give you NPU support, which relies on https://github.com/FastFlowLM/FastFlowLM . And that says "NPU-accelerated kernels are proprietary binaries", not open source.

Ah, that old soft jazz...

Is there a write up of the underlying approach? The summary on the repo mentioned SIMD, but not a whole lot else.

> I don't think the kind of drone you can build with a lawnmower engine would be likely to do any significant damage to any but the smallest ship

It's not really a lawnmower engine, but the L550E clones used in the Shahed drone are roughly the same scale as a big lawnmower engine (higher power/weight, but similar horsepower), and they've successfully taken out $100 million radar installations.