I've organized my notes from AGBT, which you can download here. I know a lot of people were not able to attend the conference, so I hope this is helpful in addition to all the other comments from the twitter feed (HT @Massgenomics for the Google Docs link).
There were a number of interesting talks on clinical genomics, sequencing technology, and bioinformatics. There were also a few very interesting metagenomics talks. I don't know a whole lot about this area, but the talks certainly got me interested in learning more about this topic.
Aside from the interesting science, there were some nice features about the conference (catered meals, lots of freebies, etc.), but I also feel like there was some room for improvement. For example, I wish they ended the conference a little earlier on the last day in order to make it easier to catch a flight out of Florida. Also, I personally prefer conferences that take place in major cities (that don't require a 1-hour shuttle to get to the airport), although I'm sure there are varying opinions on this matter.
Finally - all minor complaints aside, I am very grateful to be able to attend this conference and present a poster for my new DNA methylation algorithm (COHCAP - City of Hope CpG Island Analysis Pipeline). For those that couldn't attend the conference, COHCAP is an algorithm that quantifies the consistency of methylation patterns for CpG sites within CpG islands (for Illumina methylation array data and targeted BS-Seq). The paper is currently under review, but the algorithm is currently available to download and I'd be glad to answer any questions that you have about it.