I lead marketing at De Agostini Libri and teach at four universities. My work sits at the point where publishing meets technology: how a book is sold today, and how you verify what a machine claims.

I have been through some of the turning points that reshaped the industry: the joint book-and-newspaper sales at Repubblica, when the newsstand became a bookshop; the shift of advertising from print to digital at Seat PagineGialle; the Fifty Shades phenomenon at Mondadori, a commercial case without precedent; Rizzoli and its acquisition by Mondadori — market consolidation seen from inside both sides; the transformation of DeA Planeta into De Agostini Libri. Each time, the same question: how is the way content finds its readers changing?

That question calls for an interdisciplinary approach: out of necessity, not fashion. Contemporary publishing cannot be understood through a single framework. The premises I start from are: semiotics, to decode how the grammar of content changes; economics, because producing content without questioning the business models is a way of getting hurt; Kahneman and behavioural economics, to understand how our mind reasons — built for adaptation, not for understanding phenomena; Taleb and his explorations of probability, to navigate uncertainty among geopolitical black swans; Bayes' theorem, to update beliefs about consciousness, mind and the cognitive sciences in an era of infinite AI production.

In my university courses I often problematise Kotler: does global marketing still work in a balkanised internet? How do you build strategy when national digital silos replace the open web? I explore new publishing business models (podcasts, YouTube, digital platforms, the creator economy) and write on my Substack, The Abstract, where I publish analyses that cross current trends with epistemological rigour.

The project I am working on now is called Validating AI. Language models are Bayesian machines: they do not reason, they produce probability distributions over sequences of tokens. Certifying their internal processes is senseless even before it is impossible. The right question is not "how does it work inside?" but "how do you check the output after it has been produced?". Man in the loop is not enough: it introduces cognitive biases, fatigue, unwitting delegation. I am building a post-cognition framework — an infrastructure in which the model monitors its own first output through a second, structured pass, before the human steps in. The goal is a taxonomy of epistemic errors, a test corpus and a methodological proposal intended for academic publication.

My audience: managers, publishers, strategists and creators who want to understand why the rules are changing, not just how to adapt.

I also write The Abstract, a newsletter on content industries, AI and media from a European perspective.