This is what I would do: Princeton's Integrated Science curriculum. This is how science should be taught:
The course begins with an overview of the different disciplines in the natural sciences (physics, chemistry, biology) and our strategy for providing a unified understanding of them. The rest of the semester is arranged around two major classes of quantitative models: dynamical and probabilistic (see below). The power of these models to describe key phenomena in physics, chemistry and biology will be taught. Through applying these models, students will learn both the most critical facts and fundamentals of these disciplines and the key unifying features that cut across them. In addition, students will dramatically expand their skills in solving quantitative problems.
A key tool in quantitative problem solving is the computer..The computational background built over the first ~ six weeks of COS 126 will be used throughout the Integrated Science sequence, where the utility of computers for testing our mathematical description of the world (even when we cannot find exact solutions to the relevant equations) will be repeatedly emphasized. Additional lectures and problem sets introducing specific computer programming concepts of special utility to emerging scientists (with a focus on computational biology) will also permeate the entire freshman year.
This is what modern scientists need to know. Too many undergraduate science curricula are mired in the past.
If Princeton published this curriculum as set of books, they might rank up there with the Feynman Lectures on Physics.