Temporal Radio: toward controlled retrocausal communication (two preprints, supplementary material and code)

This record contains two preprints and their supplementary material, in English and Italian, together with the code and LaTeX sources that reproduce every number and figure. Toward controlled retrocausal communication: the successive-rewriting model and its bottleneck (main paper). Can information be sent to the past? With known physics it cannot; the possibility is not, however, excluded a priori, and the Temporal Radio project asks what would be needed. We show that the bottleneck has a single root: making even a single controllable bit appear in the past without a cause in that past. We show it through the author's successive-rewriting model, formulated with declared premises: a message that reaches the past alters the causal chain, reality is perturbed and resettles, overwriting the future, and a protected log numbers the versions and guides the senders. We formalize it as a dynamics over the versions, in which consistency is not an imposed principle but the equilibrium between settling and the senders' choices; a toy model shows that the author's rule of intervening only for a serious reason allows the succession to stop despite a strong butterfly effect. The bottleneck appears in two forms. The first is the channel: the message must reach the past on a physical signal, which we discuss as a hypothesis in terms of the advanced component of the electromagnetic field; two entangled particles supply a key that only the receiver can use, but reduce neither the amount of information to be sent back nor the difficulty. The second is the persistence of information across versions: with a protected log the model makes a conditional prediction that distinguishes it from Novikov consistency, the alternating entries left by the paradoxical protocol. Future boundary conditions and feedback loops: two solvable models that send no signal to the past (companion note). In models with boundary conditions in the future it is easy to mistake a global dependence for a signal to the past. Two elementary, solvable models make the distinction explicit: two coupled linear oscillators with data at both ends, and a scalar field crossing the throat of an Ellis wormhole placed in a feedback loop. An operational criterion in five conditions organizes what each model does and does not show. The supplementary material contains the catalogue of models, the verification register with the evidence level of every number, and a guide to the computation package. The work was carried out with the assistance of an AI model, as declared in the papers.