hoover

The idea: Filter podcasts.

1. On an EC2 Linux instance grab every new MP3 and use speech-to-text. trigger on keywords.
2. for mp3s i've kept, filter for compression (duration / file size)
3. feed to me off ec2 by authoring an RSS to the bits.

Optimizing for the fact I'm on a low-bandwidth device! short first, well-compressed next.