Jonathan R Shewchuk

1) I'm gonna partly contradict everyone else. In Berkeley CS, it helps a lot to know what area you want to work with, because then professors in that area may choose to fund you. We do have "general admits" who aren't associated on entry with any particular area, but they're a small minority. 2) Not much. A strong student who did solid research in area X in CS and now wants to do area Y in CS will be just as exciting to a prof in area Y. Solid research talent is rare; ability to learn a new field is commonplace.

Answers to your questions: 1) A lot. I'm not sure I understand what "workshop publication at a top-tier conference" means, but if it's a workshop associated with a conference, it's the reputation of the workshop itself that matters, much more than the reputation of the conference. Having said that, what your letter-writers say about your work matters more than the quality of the venue. The quality of the venue loses its importance if your contribution is genuinely wonderful enough that (1) your letter-writers rave about it or (2) someone on the committee reads the paper and sees that it is good. 2) Yes, it will hurt.

MS programs in CS differ greatly. Some universities have cash-cow MS programs that are relatively easy to get into...but are unlikely to help you pay the tuition. I don't think a 3.3 GPA rules out MS programs at mid-tier schools. It probably will make funding hard to come by.

I'm a CS prof at Berkeley. I would estimate that our median admittee has one publication. We do still admit some students every year with no pubs. When I see someone who never got a lower grade than A (well, maybe one A-) and almost maxed out the GREs (with a perfect Q score), I tend to assume the raw intellectual firepower will make up for a lack of research. Don't underestimate the matchmaking aspect of admissions...if there's a professor with funding who wants a student with your interests, you just need to beat the other applicants with the same interests. Thus, there's more variability in quality than you might think. I'm pretty doubtful about the suggestion that Stanford doesn't read the SOP. The SOP is the most important part of the matchmaking process, because it tells us specifically what research the student wants to pursue. I strongly recommend naming specific profs in the SOP and saying how you'll fit with them.