I was exploring techniques to teach novel gesture interaction “in the moment”. The alternative is to have the user learn everything interaction up front, or to have a large toolbar that takes up screen real estate. I created a concept video that showed me working on a multi-touch tabletop I was using “for the first time”. The table was non-interactive, so its behaviour was pre-recorded, and I was “playing the mime”. Instruction is afforded with animated hands for the user to mimc. For every interaction context, the most likely interactions were acted out to the user with an animated hand. If the user was touching the interface with no effect, the interface would suggest what it thought the user might be trying to do, and how to do it with another animated hand. This concept was shown to a few people and inspired later work.