having labs open their research: what harness system they used, what types of problems they tackled, which problems success and which fail, how they success and fail so we have a better idea of what tasks LLM are currently good at