The Mathematics Search Engine

14+ min ago   (464+ words) When building autonomous agents, we often treat tool calling as a black box. We provide a set of definitions to a Large Language Model (LLM) and assume that if the prompt is good enough, the model will pick the right function…...

21+ min ago   (597+ words) Is it deployable? Yes. The Apache 2.0 weights run today on a 16 GB laptop or a single 24 GB GPU. You need PrismML’s llama.cpp fork or its MLX runtime. The model keeps the Qwen3.8 27B architecture unchanged. It has 27.36B parameters. That splits into…...

7+ hour, 33+ min ago   (865+ words) Rusty is a quantum science nerd. He's been into academic science all his life, but spent his formative years doing less academic things. Now he turns his attention to write about his passion, the quantum realm. He loves all things…...

1+ hour, 1+ min ago   (931+ words) Apple's built-in Sound Analysis model can label cat_meow and cat_purr. It has no cat_hiss class. Point a mic at a real hiss and you often get silence, a generic cat hit, or — worse — snake_hiss. Energy metering only proves something crossed a loudness line. Neither…...

1+ hour, 37+ min ago   (1549+ words) Data science has become one of the most important technology-driven career fields, but the skills expected from data professionals are changing quickly. A few years ago, learning Python, statistics, SQL and machine learning could form the core of a data…...

1+ hour, 10+ min ago   (1260+ words) The newly created company, as yet unnamed, looks to be a force in the exploding AI data infrastructure arena with its combined data transformation and data integration technologies. The new company created through the recent merger of dbt Labs and…...

1+ hour, 19+ min ago   (242+ words) Follow-up to: “LLMs are (still) mostly powered by imitative learning, not RL” A common take I’ve been hearing is: “LLMs are especially good at math[1] because math is easy to verify”. But that story doesn’t make much sense. So here’s…...

1+ hour, 27+ min ago   (894+ words) Bad synthetic data and flawed rules broke model evaluations, but high overall scores hid it. Test by category using realistic, production-style data. A detector I built was scoring 0.067 recall on temporal errors, meaning it caught about one in fifteen of…...