A fair number of recent projects involve the Rcpp package for seamless R and C++ integration, and several related packages. The Rcpp website has more about Rcpp, and the Rcpp Gallery presents a number of worked examples.
RcppArmadillo, RcppEigen and RcppGSL extend, respectively, to the Armadillo, Eigen and GSL frameworks. Most of these packages are authored jointly by the Rcpp Core team.
Several few smaller and sometimes still incomplete packages use Rcpp to extend to other interesting package or libraries.
A few examples are RcppAnnoy (fast Nearest Neighbours classifier), RcppAPT (interfacing APT on Debian/Ubuntu/...), RcppBDT (Boost DateTime access), RcppCCTZ (time-zone library), RcppCNPy (read/write NumPy files), RcppDE (differential evolution optmizations), RcppGetconf (system configuration access), RcppKalman (Kalman Filter), RcppRedis (faster Redis read/write access), RcppSMC (Sequential Monte Carlo / Particle Filter), RcppStreams (Event Stream Processing), RcppTOML (TOML Configuration File Reader), RcppXts (xts access at C level), and RcppZiggurat (faster Normal RNGs).
Last but not least RInside uses Rcpp to embed R in a very easy-to-use C++ class for deployment in other programs.
Also, a long list of packages using Rcpp is available.
littler lets R be used in commad-line scripting, pipes and more.
Rblpapi connects R to Bloomberg API (provided you have Bloomberg).
RQuantLib is another relatively large project connecting R to QuantLib.
BH provides Boost C++ headers for use by R.
digest provides hash sums of R objects.
RPushbullet interfaces the PushBullet notification service.
RApiSerialize provides the C-level API for R object serialization.
RVowpalWabbit interface the Vowpal Wabbit on-line learner.
pkgKitten makes creating and writing R packages a breeze.
drat supports lightweight package repositories.
rfoaas interfaces the excellent FOAAS service.
random provides true (hardware) RNG draws.
rmsfact and gaussfacts offer some (lighthearted) quotes in the fortune() tradition.
RDieHarder tests RNGs.
RPostgreSQL connects R to Postgres (now maintained by Tomoaki Nishiyama).
Older projects include the Perl-based BeanCounter (and SMTM) as well as Finance::YahooQuote.
In (approximately) reverse chronological order: