The focus of debate on capital theory still is on the macroeconomic aggregate production function, almost seventy years after Joan Robinson attacked this concept. It has turned out that reswitching is rare in large systems. Reswitching and reverse capital deepening once were the most effective arguments against the production function. Later it was shown that an approximate surrogate production function could be constructed, using the approach of random matrices. This seemed to weaken the critique, but a new one has emerged, which shows that the number of effective techniques on the wage curve is small and that the possibilities of substitution between capital and labour are quite restricted in the relevant range or profit. This paper reconstructs the path by which the new results were arrived at and presents a new variant of the proof of zero substitution.