I started working in General Relativity and with Mathematica simultaneously in 1996. But it was in the beginning of 2002, while I was a postdoc with Carsten Gundlach in the GR group at Southampton, that I realized that my Mathematica scripts could be useful for other people. I'm deeply indebted to the people in that group. The needs of Paul Lambert (the "nice printing" np command) were the seed of the whole project. Paolo Matteucci helped me to shape the mathematical structure of the system, and suggested the use of the algorithms which had been developed a few months earlier by R. Portugal to canonicalize indices. Reinhard Prix showed me the advantages of releasing the system as free software and suggested the name xTensor, instead of my original choice exTensor.
UPM Madrid 2003
During my year as a lecturer at the Polytechnic University of Madrid I studied Computational Group Theory intensively (Butler's book) and I coded up xPerm in pure Mathematica language. It became immediately clear that it was too slow, so that I rewrote it in C.
CSIC Madrid 2004 - 2008
The pair xPerm/xTensor became public with their first public release (version 0.7.0) in March 2004. I decided to spend quite some time testing the system in some nontrivial problems. The first one was my work on hyperbolic formulations of the Einstein equations (with C. Gundlach, 2003-06); the second one is my work on high-order perturbation theory in GR (with D. Brizuela and G. Mena Marugán, 2005-08); the third one is my work on the polynomials of the Riemann tensor (with R. Portugal, L. Manssur and D. Yllanes, 2005-08). Other authors have tested the system in different areas of GR and field theory, and in particular I thank Alfonso García-Parrado for showing me how many things xTensor can do! I also thank Guillaume Faye for showing that xTensor performs better than MathTensor in the tests so far in the very difficult area on post-Newtonian computations.
Version 0.8.0 was released in July 2006, for the Marcel Grossmann meeting at Berlin, and version 0.9.0 in December 2006.
With David Yllanes, in 2006 we introduced a preliminary version of xCoba (a pun on "escoba", the Spanish for "broom"), a package for computations with components and bases. We are very ambitious in this project and we want this type of computations to be performed with the abstract tools of xTensor, which is being harder than expected. A first, but almost complete, working version has been released in August 2007 (xAct_0.9.2).
Paris 2008 - 2010
Two more years of testing and extensions. The xAct google group has been a continuous source of ideas and information, and I thank everybody there for their support and interest in xAct. Finally, version 1.0.0 of xTensor was released on September 30, 2010.
Champaign 2010 - 2015
I do not currently develop new functionality for xAct, but other authors are doing it instead, including the development of many other xAct packages. As of August 2015 there are already 16 packages! The activity in the google group is steadily increasing, and reported bugs are fixed as soon as we can.