For my country is a Living Room, Carlo Zanni used Google’s autocomplete feature at a rather early stage of its development, using the text editor Scribe and its automatic translation service Translate. To mark the 150th anniversary of Italian unification, Zanni typed fourteen strings of characters or words into Scribe and collected the autocompleted stanzas, placing a period after each iteration. The resulting poem is a reflection on the public and private, the national and local, represented by the country and the living room respectively. This political dimension is highlighted by the book cover, which shows the souvenir statuette of Milan Cathedral with which the then Italian prime minister, Silvio Berlusconi, was attacked at a political rally in 2009. The artist also wanted to point to the “sleepy,” somehow politically hypnotized state of Italy at that time (artist in an email to Apod.li).

Accordingly, the first input was “I wok,” which Google autocompleted as “I woke up.” The prompts ranged from “Tod” (for “Today”) or just “H” to more complex clusters such as “Mr.LivingRoom deceives” and “My country dies,” indicating a will to influence the outcome of this generative writing method. At this early stage of Scribe’s development, it was possible to help shape the result, especially for longer sentences, by pressing the spacebar after typing the first few letters, since Scribe suggested a new word each time the spacebar was hit.

The poem, originally in English, was then automatically translated into fifty-seven languages (the ones available on Google Translate at the time), to form the second part of the book. In the first part, each stanza takes up one page of the book, with Zanni’s prompts in bold, giving the impression of sentences rather than verses. The translations in the second part present all stanzas on one page, giving a more traditional impression of a poem.

The website had Scribe’s autocomplete feature implemented so that each time a user visited it, the prompts Zanni used were automatically resent to Scribe, triggering new autocompletions. The outcome changed, as Google’s suggestions are based on filtering and the statistical analysis of millions of everchanging websites. In a sense, “[t]hese transformations [are] somehow ‘suggested’ by the collective writings of billions of people around the world” (Carlo Zanni, project website).

The resulting new versions were stored in an archive accessible through the website and a pay-per-view subscription system. However, this implementation was shut down when Google denied any access to Scribe one year later, leaving the archive with a total of 111 generated poems. With its dual output as book and website, the project provides an example of the fragile liveness of the web and the more stable output of print.