A deep (although at the first glance naïve) question which may be addressed to embryonic development is why during this process quite definite and accurately reproduced successions of precise and complicated shapes are taking place, or why, in several cases, the result of development is highly precise in spite of an extensive variability of intermediate stages. This problem can be attacked in two different ways. One of them, up to now just slightly employed, is to formulate robust macroscopic generative laws from which the observed successions of shapes could be derived. Another one, which dominates in modern embryology, regards the development as a succession of highly precise 'micropatterns', each of them arising due to the action of specific factors, having, as a rule, nothing in common with each other. We argue that the latter view contradicts a great bulk of firmly established data and gives no satisfactory answers to the main problems of development. Therefore we intend to follow the first way. By doing this, we regard developing embryos as self-organized systems transpierced by feedbacks among which we pay special attention to those linked with mechanical stresses (MS). We formulate a hypothesis of so-called MS hyper-restoration as a common basis for the developmentally important feedback loops. We present a number of examples confirming this hypothesis and use it for reconstructing prolonged chains of developmental events. Finally, we discuss the application of the same set of assumptions to the first steps of egg development and to the internal differentiation of embryonic cells.