The Lab
Things I'm building with AI.
In validationQuantization research
EAQ: a quantization framework built around how TensorFold runs models, for coding, agents and tool calling. First target, GLM 5.3 Flash: 41% lower error on what it writes and 54% lower on agent decisions, at about the same speed.
v2.40Monitoring and benchmarks
A real-time dashboard for one or many DGX Sparks and NVIDIA GPU machines, Windows PCs included, with a phone app, and now LM Studio and Ollama detection: live GPU, memory and LLM metrics, and the decode, prefill, quality and tool-calling benchmarks behind the numbers on my recipe pages.