We propose a method and a model of computations according to which "not losing" transformations of the information are provided. Thus semantic and syntactic reasons and concepts are entered algorithmically on the basis of Lambda-calculi of conversion and a reduction. Computations here are understood in the generalized sense: they include also constructions of conclusions in Gentzen logic systems and theories without loss of the information, that is, without a postulated rule of logic cut, but with two author`s rules of Lambda-cut. The method is based on the special organization of transition from computations to information access/search by the principle of the two-story sequential gradualness. Corresponding algorithmic software is "know-how" of the author. In a number of preliminary executed publications the computing opportunities of this model and the border of applicability of deductive means are revealed. They are based on use of the so-called latent weight of computations when the most part of computations is not visible.