A better algorithm for random k-SAT
On the solution-space geometry of random constraint satisfaction problems
Going after the $k$-SAT Threshold