Another PhD student, in computational neuroscience (modeling of thalamocortical circuits). The difference between the level of code-sharing and code community between general purpose software and scientific software is just astounding. In computational neuro-modeling at least (as opposed to very intensive bio-analysis like genomics), there's incredibly little sharing of code or interchangeability, outside of [ModelDB](
http://senselab.med.yale.edu/ModelDB/), although I've heard systems biology-modeling (like with CellML) is in a better state.
I love this site for many reasons, e.g. searching "site:ycombinator.com " is one of the best sources of finding modern opinions on languages, libraries, the state of coding in certain places, and most of all some of the best opinions on what can act as good introductory materials on anything software/langs (in addition to stackoverflow).
It's nice to have a finger on different tech scenes and where people, sometimes, are willing to call each other on BS. Oppositely, it's cool to see posts about a product/service and see that the actual company actively responds to comments here in a lively fashion.
One of the biggest reasons for me has been seeing whenever's the occasional open academia article or movement starting to spawn, whether it be open access, open data (e.g. [dat](http://dat-data.com/)), or my favorite, open code. There's an unbelievable amount of reinventing the software wheel that goes on with a ton of PhDs, and the FOSS community proves that doesn't have to be the case. Trying to take the lessons it's learned to the sphere of science is something that an increasing amount of people are thankfully interested in, and this site is one of the only ways I've heard about initiatives in that regard - most notably, [Mozilla Science Lab](http://mozillascience.org/)