The blog post Wave Bot: Annoty includes short demonstrations videos of two wave robots that annotate blips as you type.  The basic idea is useful--augment wave conversations with relevant information. This is the basic idea behind smart tags, intellisense, code sense, and various other text augmentation technologies.

But there is a fundamental weakness with wave robots due to a limitation of the wave API.  At least for the time being, a wave robot has no way to traverse the entire blip hierarchy within a wave.  Instead, it must rely on robot event notifications for visibility into wave metadata.  Since event notifications contain details only about the blip that triggered the notification event (typically a blip being edited), the robot can only augment blips as they are updated by users.

This limitation means that, in order to be maximally useful, a robot must be added very early in the life of a new wave, otherwise it misses the opportunity to augment blips created before the robot is added.