The problem is in finding the good academic papers. For someone like me who is not affiliated with a university, the best papers are often behind paywalls.
For recent stuff in computer science this is somewhat less of a problem, because most people post PDFs of their own papers on their websites. Citeseer (http://citeseerx.ist.psu.edu/) indexes and caches the self-hosted PDFs as well.
Once you find a good paper or two (and/or the academic literature's relevant buzzwords for your issue), you can span out through the citations for quite a while. "Introduction / Survey of ..." papers (e.g. http://archives.cs.iastate.edu/documents/disk0/00/00/02/04/0...) exist mainly for the bibliography. Also, when you find someone whose name keeps coming up, look for their faculty page, e.g. Andrew Appel's (http://www.cs.princeton.edu/~appel/papers/).
For papers that don't have .pdf or .ps files cached on Citeseer (due to primary hosting behind the ACM's paywall) try searching on Google Scholar and clicking "X other versions". Typically, the ACM is the first result, but faculty pages have publicly visible copies.
Google Scholar is really good at showing you a free PDF link if there is one, usually from Citeseer, but also from .edu sites. If you host under your own domain, like me, you are unfortunately left out. I am not sure how to get round this yet.
There are several ways around that; some depend on the field you are looking at.
At least for Mathematics, Physics and Computer Science the authors often have a working draft on their homepage. Alternatively, and in all fields, you can just write a letter or email to the author asking for an electronic copy.
Then you could befriend someone in a university---just ask around on HN, I am sure somebody will step forward and agree to download you articles from behind paywalls every once in a while. Or often just joining the library of your local university is good enough.
A complementary approach is to find a really recent general one and skip to the end for the citations. Where possible, finding the original paper on a topic (e.g. gelman & gelman on mcmc) always makes me feel like a child with ten pence for sweeets.
Go to the author's/university/lab website. You will usually find a pre-publication version or a closely related paper, months to years ahead of the actual publication. Some of the stuff you find isnt very good (too narrow, for example) but the signal to noise ratio is better than scanning lot of blogs.
The problem is often not to download the papers but to find out what paper is good if you are not in the paper's field. The number of citation is usually a good metric but how about recently published papers or draft works ?
If you really want to read a paper that is behind a paywall with an unreasonable download fee, you might try to find the university websites of the authors. They often make their papers available online from their "publications" webpage.
Alternatively, just emailing and asking the author for a pdf might work (if you're comfortable with that).
I really miss being a student and having wide access to all kinds of journals. The professional societies (like IEEE computer) do provide some access to some journals, but the coverage is so limited. One would have join all of them to get the same access that a university student has.
