leansensei

Easy to say "should".

What's the point, when we have Chat Control 2.0?

Unchecked immigration, lax policing.

Damn, this is a fantastic evaluation, thanks!

It does, beautifully. Now let's wait for an NVFP4 GGUF!

Check out cactus needle, might be what you're looking for.

Easy to do with Unsloth Studio.

And if you're running it on a dGPU, power limit it, because you lose very little in terms of token generation performance, since it's memory-bound.

RTX 4060 and above. Ideally RTX 50 Series, because you can run NVFP4-quantized GGUFs that give you better prefill AND better quality.

Huh what? Qwen3.5-35B-A3B runs just fine with maximum context, on an RTX SUPER 12 GB, with offloading of some expert layers to DDR4-3200.

Same story on an RTX 4060 Ti 16 GB. MTP is a serious boost to tg.

Downloading the model is a simple hf command that HuggingFace's web UI even gives you.

llama.cpp is trivial to use, and so is llama-swap, if you want to use other models too.

If you don't know what arguments to run it with, you download ggrun and use that.

Local LLMs are incredibly capable and don't need expensive hardware. A $500 GPU will do. Or even cheaper.

This is all trivial.

Oh no, we do, because the software world has been abusing what engineering means for decades, in order to inflate job titles. Most people are programmers, not software engineers, and that's OK.

Not really, you can do great things without them. I've been summarizing hundreds of documents. I've added MCP servers to my internal business tools (Elixir apps) and can chat with the Nous Hermes agent over Telegram about pending orders, inventory level, historical product prices, etc., without having to click/dick around with a web UI.

Sure, it cannot replace SOTA models for agentic coding, except for small, well-scoped refactorings. But even a model like ministral-3:8b or qwen3.5:9b is a boon for so many smaller use cases!

Europe was never at the frontier of any tech? Huh what now?

...and you'd be wrong.