Yep, I use them pretty often now. What I would like to do is create a library for this. Features would include: 1. Choosing arbitrary time estimate to use as a

Yep, I use them pretty often now. What I would like to do is create a library for this. Features would include: 1. Choosing arbitrary time estimate to use as a target. 2. Providing distractions for if the estimate is much too low (show a secondary progress bar or spinner, maybe even misdirect so the primary bar can backtrack without the user noticing) 3. Calibration based on real results. A simple way would be to use linear regression to map the input guesses to real world results. Then the resulting equation should be easily available to plug into a persistence library or LocalStorage, so that you continue to get accurate results moving forward.