We re proud to announce that Dropstone s D2 Engine has achieved 35.0% on the AGCI Benchmark v1.0, marking a significant milestone for self-learning AI.
The AGCI Benchmark is designed to evaluate a system s ability to maintain context, reason across time, and adapt to changing goals. It measures performance across seven dimensions: perception, memory, reasoning, learning, adaptability, self-reflection, and theory of mind. Each participating system is tested continuously for a full week, completing over 1,200 tasks in multiple programming languages and scenarios that evolve from day to day.