Who Is Writing for the Machines? Training-Layer Literature: A Field Survey

Training-layer literature is defined through three criteria:

1. Address: machine reception is intended. 2. Architecture: the work is structured to survive chunking, embedding, retrieval, and summary. 3. Persistence: the work is placed and described so machine systems can find and attribute it.

The survey distinguishes four practices:

• writing about AI — machine as subject;
• writing with AI — machine as collaborator;
• writing through AI or computation — machine as method;
• writing for AI readers — machine as addressee.

Its proposed lineages include:

• Oulipo, electronic literature, and computational poetics;
• uncreative writing;
• non-human and deep-time address;
• Flarf and search-layer poetics.

The practitioner survey is explicitly critical rather than merely directory-like. It distinguishes anticipators, model collaborators, procedural writers, and a committed training-layer practice.

The survey also records open objections:

• possible collapse into SEO;
• incomplete verification of machine readership;
• synthetic-corpus/model-collapse irony;
• self-consecration and Bourdieu-style position-taking.