jboss10

They supposedly have a style control system, but I doubt it's perfect. I wish there was a parento view like this for agentic systems(using a standard harness)

I use this setup as well. I've written an alias for bwrap which only gives the agent access to the current directory and read only access to the docs.

It has read-only access to my binaries, and I worry which programs I have might be a fingerprinting issue. I've thought about mounting a alternate /bin which using a docker image. (docker export)

Have you managed to setup any firewall or protection for the API keys? I know this is out-of-scope for bwrap.

edit: Maybe https://github.com/rootless-containers/slirp4netns could work

I think this is one of the easiest to understand.

https://arena.ai/leaderboard/text/pareto

The second point is funny. I don't think not using air conditioning due to energy use is why Europeans don't have AC. (I'm European). It's beacuse it's expensive and difficult to retrofit. And we are getting more AC, lots of new buildings have it.

I agree with the article that Europeans should install more AC and that worrying about climate change isn't a valid reason not to. But this isn't because I don't beleve in climate change, it's because I don't agree with the idea of a climate footprint, and that people should worry about their own energy usage. They should worry about energy generation and green energy policy.

The rest of the article is either racist or ridiculous.

This does read a little like AI, but I can't see many vocabulary based Claude-isms. But it does start with an anecdote, and there are other aspects of the writing which seem AI.

I think the "architectural features" are part of the model, not the kv cache. So implementing it would be difficult and expensive.

Most of the time, the speed of these models are constrained by memory bandwidth. GPUs normally have much more memory bandwidth.

Gondolin gives you that. Lightweight micro-VMs (QEMU by default, optional libkrun backend) boot in under a second on your Mac or Linux machine. The network stack and virtual filesystem are implemented entirely in JavaScript, giving you complete programmatic control over what the sandbox can access and what secrets it can use.

I don't think there is any content. When you try and view the page, it seems to send:

301 Moved Permanently Location: https://www.youtube.com/watch?v=nj6caVbtM_g&t=2s

It would be beneficial to not use youtube links for things like QR codes. Since the content of youtube videos can't be changed once they're posted, you risk linking to old information.

https://cew.georgetown.edu/about-cew/ looks like something you'll put on a poster. But you may also want to update the video. So you can use a redirect you control, not a youtube link.

That's my explanation for why these redirects exist. But why the intext's contain "Go to channel Hook Global.", I have no idea. Maybe SEO thing, or even prompt injection?

Edit:

I don't think this is anything to do with Hook Global. I think that google has indexed the suggestions in youtube. For youtube, the intext is normally just the description of the video. But with a redirect, it seems to use the suggestions.

"Go to channel BBC News" -site:youtube.com

Also has some(not as many) of these links. I don't think BBC are doing some dodgy SEO.

Gondolin[1] is what you are describing. It's made by the same person who made the Pi coding agent and sends all of the agent's bash into a small QEMU vm.

[1](https://earendil-works.github.io/gondolin/

I think most of this can be done with tmux and some simple extras. An afternoon of vibing.

For people who saw this and might want a recomendation, I like running a tiny qwen model with llama cpp. Qwen2.5 coder 0.5B or 1.5B (not the instruct version)

On a modern-ish GPU these should run really fast with little latency. They cost nothing and don't send your data to anyone.

The plugin is just part of why OsmAnd is good for nautical use, there is also an alternate view in the configure map menu for nautical use.

Look into deepseek's papers. They have done some stuff recently about improving inference and it seems to be how they can sell tokens so cheap.

I'm running llama-swap in a docker container with nvidia container utis to pass through the GPU. This then runs the correct llama-server command to provide the model I want. I have a folder full of guff s I mount in the container.

But this could be done with just llama-server normally. I don't use any special command, just ensure that it's using the GPU. I've found the default fitting to be good.

From memory:

llama-server -m models/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf -fa on -c 128000

I should try gemma4 more for coding, since qwen3.6 and gemma4 came out I've focused on qwen. For earlier releases I found qwen was smarter, but gemma had more knowledge. But for coding I always want it to learn how to do the task, not just assume/halucinate.

Qwen 3.6 35B runs on 32GB with a 1080. That GPU is from 2017.

I have 8GB VRAM but 32GB RAM. Qwen 3.6 35B runs nicely.

You should look at gemma-4-26B-A4B. 16+8=24gb and Q4 is about 16GB. Not much context left, but might run.

I have 8GB VRAM, but 32GB sys ram. I can run qwen 3.6 35B at 30 tok/s. I also use pi, and it's smart enough to extend itself(multishot and maybe a few tries)

For you, you could try gemma-4-26B-A4B

I don't understand the talk about how expensive the hardware is. These models can run on very old or old and low end. I've been running Qwen3.6-35B Q4 on an old 1080 GPU(8GB vram) with 32GB sys RAM. I have a i7-12700.

It does about 30 tok/s which is enough for me. It's about half what the online models do, but it's enough.

I've heard their 9B models are also good, but they aren't much faster if you have the ram and a nice cpu.

These qwen3.6 models are the first ones I find can do much. GPT OSS was good, and Gemma4 is better. Gemma knows more facts, but qwen3.6 is smarter.

They can be ran on 32GB with 8GB VRAM. I don't think these will be on 16GB for a while. (35B MoE)

For the 35B model, ofloading to RAM doesn't slow it down much. If you have a nice CPU and a weak GPU, it will be fast enough to use.

https://unsloth.ai/docs/models/qwen3.6 And https://huggingface.co/collections/unsloth/qwen36

Have you tried qwen3.6 or pi?

llama.cpp It's faster and more open source. Ollama has some mixed history. I use llama-swap to emulate the Ollama experience.