augmented & reconfigured
by John Cayley   published by Counterpath Press  Spring 2023  ISBN 9781933996806
afterword by Garrett Stewart [ Programmatology ]

Writing To Be Found

Since 2009 Writing to be Found has been a term of the author’s which has come to embrace various writing practices and to describe extant writing-as-such that is to be found in relation to linguistically implicated internet services. It signals a crucial, not to say catastrophic, moment in the history of language practice, when it was suddenly possible for any internet-connected writer to believe that they were able to know whether a particular sequence of words could or could not be found in the indexed corpus — of Global English — that has a pretended, growing status as the documented domain for ‘all our language.’ Writing to be Found is another name for the Future of Writing or the End of Writing, if, that is, this is writing that may be found but not read. Most of the writing in this book is in dialogue with Writing to be Found. Our relationships, already mentioned above, with entities such as the longest common phrase and the zero-count phrase are crucial in this regard. Today, as of this writing, they delineate human reading and require that we develop a deeply critical understanding of the services that allow us to specify and work with their instances.

The poems of ‘Write Thus’ and ‘Monoverse Selections’ are mostly composed from phrases and lines that have all been assembled quasi-randomly and in accord with various formal constraints and then searched to discover some aspects of their relationship to the internet’s indexed corpus of language. Lines are then either rejected or selected and composed on the basis of what has been discovered concerning, typically, their relative frequency.

The programs making these selections and tests are remarkably simple although they do make use of code libraries — especially Daniel C. Howe’s RiTa — that encapsulate natural linguistic tools and affordances: lexicons with parts of speech, rhyme, word division, and stress patterning information, for example. The effort to produce concise code that is nonetheless able to propose significant and affective language generation is in a specialist tradition and explicitly acknowledges work of this kind done by Nick Montfort, some of it recently published in his book #! [Shebang], Counterpath, 2014, and reviewed by the author. The most concise of the programs used for ‘Monoverse Selections’ — 493 characters of Java source code although referencing RiTa — is given here:

import rita.*;
class M {
     public static void main(String[] a) {
        new M();
    String A = "tienoa", b, c;
    RiLexicon l = new RiLexicon();
    M() {
        p("\\r" + d("vb", "rb").toUpperCase() + "\\r\\r");
        for (int i = 0; i < 6;) {
            p(d("jj", "nn"));
            if (b.indexOf(A.charAt(i)) > -1) {
    void p(String s) {
    String d(String... j) {
            b = r(j[0]);
        while (m(b));
            c = r(j[1]);
        while (m(c));
        return b += " " + c;
    String r(String w) {
        return l.getRandomWord(w);
    boolean m(String w) {
        return ("" + l.getFeatures(w).