In the last few years it has become clear that there are a variety of gluon actions that give accurate results on coarse lattices (a = 0.2 − 0.4 fm); see various contributions to Lattice 96. Such actions have been constructed within the Symanzik improvement program, eliminating O(a) errors using tadpole improvement 2 at treeand at one-loop level, as well as within the MCRG (or “perfect action”) approach. There are of course several open problems that should be addressed, but, generally speaking, the errors of these actions on coarse lattices seem quite small, much smaller than those of any improved quark action proposed so far. Given the dramatic increase in cost of a full QCD simulation as the lattice spacing is decreased, it is very important to find improved quark actions that are accurate on coarse lattices. This is the aim I address in this contribution. Besides the use of improved actions, another tool that seems likely to become a staple of lattice QCD technology, is the use of anisotropic lattices, with smaller temporal than spatial lattice spacing, a0 ≡ at < as ≡ ai (i = 1, 2, 3). [A lattice with ξ = as/at will be referred to as a “ξ : 1 lattice”.] Such lattices have clear advantages in the study of heavy quarks, lattice thermodynamics, and for particles with bad signal/noise properties, like glueballs and P-state mesons. On the classical level anisotropic lattices are as easy to treat as isotropic ones. On the quantum level, however, more coefficients have to be tuned to restore space-time exchange symmetry. Perturbative calculations and preliminary simulations with heavy quarks and glueballs have appeared using improved anisotropic gluon actions; further work is in progress.