peculiar-ragdoll: language models at home, quantized models on Hugging Face, and notes from the room where it happens.https://peculiar-ragdoll.com/enhttps://peculiar-ragdoll.com/projects/callosum/https://peculiar-ragdoll.com/projects/callosum/Fast zero-token classification probabilities from the local models you're already serving, even while the model is generating or prefilling other requests. Optional budgeted reasoning. Supports llama.cpp for now.Thu, 15 Oct 2026 00:00:00 GMTprojecthttps://peculiar-ragdoll.com/projects/code-drawn-scenes/https://peculiar-ragdoll.com/projects/code-drawn-scenes/An agent skill for building illustrated scenes in code (Canvas2D, SVG, WebGL-2D) that look finished on the first pass and can be developed into interactive websites.Thu, 15 Oct 2026 00:00:00 GMTprojecthttps://peculiar-ragdoll.com/blog/code-drawn-scenes/https://peculiar-ragdoll.com/blog/code-drawn-scenes/The room on this site is drawn entirely in code. What it took to get an agent to draw scenes like it on the first pass, and the skill that came out of it.Wed, 07 Oct 2026 00:00:00 GMThttps://peculiar-ragdoll.com/blog/sharp-system-prompt/https://peculiar-ragdoll.com/blog/sharp-system-prompt/A model running at under a token a second taught me to make every token count. The system prompt I wrote for it now ships in the Sharp chat templates, and it gets Qwen to the answer faster without making it any less smart.Wed, 07 Oct 2026 00:00:00 GMT