Karl Smith asked for a new Taylor rule suitable for NGDP level targeting. So here's my old suggestion compressed into a single Taylor rule.
The old Taylor rule can be stated like this:
i = p + r + ap * (p - p*) + ay * (y - yp).
My proposal for the new rule would be:
i = p + r + ap * (p - (p* + aL*L)) + ay * (y - yp),
where aL is a constant to be chosen, and L = (NGDP*/NGDP - 1) * 100 (the NGDP gap in %).
Still, does any central bank use the Taylor rule? Do we really need a new one? Svensson's What Is Wrong with Taylor Rules? might be of interest.