According to CryptoSlate, citing the ACE-CaP-0 research published on November 12, researchers conducted 30 physical evaluations for each of the pre-frozen programs across 8 robotic tasks, totaling 241 trials with 221 successes, achieving a success rate of 92.1%. Human guidance and demonstrations were permitted during the program development phase; during evaluation, the programs still utilized perception and control systems, but humans and coding agents no longer participated in real-time decision-making. The results apply only to the tested tasks and do not yet demonstrate that the programs can universally adapt to new environments.