the results? the best models in automate mode are not always the best models in augment mode.
for example - Opus and Sonnet are great on performing tasks on their own, but not in proving assistance, while GPT-5-Mini was great on both dimensions.
Gemini models were stronger as assistants than as automators.