Software developing is still a dificult process, requiring a lot of human intellect. An human time is expensive and relatively slow. Debugging is one of the problems, Wallach expereinced: the first 95% is easy, but it is the next 95% (no joke) which is difficult.

Wallach put a lot of quotes from a number of people on the development of software, including several of Nathan Myhrvold laws on software, for instance: "Software grows until it becomes limited by Moore's Law". Sounds great, but is there any evidence or data to support that?

What Wallach proposes: remember, he says, that Fortran means Formula Translation, although it did not exactly do that. But today, Mathlab and Mathematica do a much better job of calculating mathematical formulas. You do not need multiple minds (mathermatical, software, system) anymore. So let us go for "Telescoping Languages" which are easy-to-use domain-specific programming languages, Wallach says.

They may be very difficult to compile, but who cares, that is a computer's job. You do not have to sit aside of them and hold hands. Some adaptive libaries, like SALSA and SANS are already providing glimpses of what Telescoping Languages could become.

The open source model also will help in the vision of Wallach. The developments will go faster. And he must admit, although he never thought that he would ever say that in his whole live: "We will have to give up some performance in favour of ease of use." Formulas and mathematics must be the basis of the new Softron software development framework.

Nanotechnology will replace CMOS. Somewhere around 2011 we will see the first nanotechnolgy based prototype chips. I would have liked a talk about the hardware developments by Wallach to sketch the future of nano-based supercomputing.