things i've built and problems i've solved.
Open-source evaluation harness for long-horizon coding tasks: hermetic Docker grading, pluggable agent adapters, and verifiable graders.
Python · Docker · LLM Evaluation · Agents · Pytest