Adam Bosworth almost always makes me think, so I jump at a chance to listen to him or read what he writes. He recently gave a talk in NYC as part of the Google Speaker Series and Darryl Taft wrote up a report at eWeek.
Bosworth talked about how physics and psychology affect which applications fail and which succeed. His examples: AJAX, PDAs, and natural language recognition. His recommendations:
- Keep it simple and stupid--even if that requires more clicks
- Use AJAX where it makes a difference--not just the geewhiz factor
- Make your tools transparent and fast
- Support bottom-up learning
Google's success, of course, can be attributed to these (and other) ideas.