Observed arrival · 2026-10-20
SHARQ: a simulated shark brain for local language models
Credibility concern recorded. The source reference remains available for verification and correction.
SHARQ uses a small simulated shark brain to adjust four sampling settings for a language model running on your machine.
- For
- Local language-model tinkerers and simulation builders
- Worth noticing
- The script is described as one sharq.mjs file for Node 20+, with no packages and four localhost backends.
Field notes
The interactive brain display reports 240 simulated neurons and lets visitors fire individual dots or supply smell, electric-field, ripple, light, and bump inputs. Its outputs are limited to four sampling settings; SHARQ does not write the model's words. The downloadable sharq.mjs is described as requiring Node 20+ with no packages and connecting to local Ollama, LM Studio, llama.cpp, or vLLM endpoints.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue