This video explores OpenDevon, an open-source AI software engineer, and its recent updates. The video highlights the introduction of CodeAct 1.0, a new coding agent that achieves a remarkable 21% solving rate on the Sway Bench Light unassisted benchmark. The video also discusses a simplified evaluation harness for testing coding agents, which aims to improve agent performance over time.
10514 7 месяцев назад 10:00