Alibaba.com President: AI agents can talk, but can they actually do the work? | Fortune

The strongest frontier model we tested successfully completed 61.7% of the tasks — high enough to be useful and low enough to be a warning.

Raccontata dafortune.com