Build Your Own Legal AI Benchmark
Define what good means for your firm, then test every model on your own tasks inside LegalWork.
Product noteLegalWork
Blog.
Short notes and updates from inside the lab.
Why Law Firms Should Train Their Own Models
Sovereign models enable strategic positioning and continual improvements.
Essay
Run LLM Emotion Steering at Scale
Extract, test, and serve Qwen3-8B emotion vectors through a high-throughput vLLM API.
Technical noteCode
How We Measure the Predictiveness of Agents
A map of what is predicted, how it is measured, and how it is validated in LLM-based human simulation.
Working paper