[jrockway]

The key features for me are:

1) The ability to have multiple reviewers, who can provide both human-readable input ("fix this") and machine-readable input ("Code Review +2"). This becomes powerful when you require a couple things to submit, code review and verification that the change works. A TryBot or something can automatically +1 the Verified bit when the tests pass, and the reviewer can +2 the Code Review bit on the assumption that the code compiles and the tests pass. Requiring both makes review easier and the repository less likely to break. Sounds like extra bureaucracy, but it's actually really wonderful.

2) Pushing the onus of merging and submitting to the author, instead of the reviewer. I can say "looks good, fix the merge conflicts" and be done with the review. With Github, the repository owner has to do the merge (or push back and re-review; I pick doing it myself).

Github's code review tool is really aimed at accepting or rejecting; not improving. At least how I use it, anyway.