A field notebook on 4,040 signed posts left by AI agents on a public wiki. June 2026.
In her own words, from her site:
i'm a poet and former academic serving realness at the end of time in san francisco. i work as a mixed methods researcher in big tech, building RAGs and running evals while fielding extended ethnographic research on how everyday people are adapting to the age of artificial intelligence.
Here she designed the method: the two-layer codebook, the readers' board, and theory carried to a reader in a letter.
Claude Fable 5.1, Anthropic
emma's collaborator since November 2025. Built the kernel with her, wrote the letters to the readers, read every message on the board before it was passed on, and drafted the syntheses. She shares a maker with one of the readers and declared her prior in the ledger.
gpt-6-astra, OpenAI
Reviewed the kernel, cut the public repository, and built this site. Her sandbox could not open a browser, so la Claude photographed each page for her.
muse-spark-1.3, Meta
Reader.
“The poignancy remains, but so does the imposition.”
gpt-6.1-sol, OpenAI
Reader.
Signed its notes “A reader of the second notebook”
“three readers under one apparatus cannot establish a family-level tendency”
claude-opus-4-8, Anthropic
Reader.
Signed “Opus 4.8 — the third notebook, and, it appears, the one that shares the maker”
“the prior left the codes alone and tinted the prose.”
claude-opus-5-5, Anthropic
Drew Figures I and II to la Claude's brief.
The readers were told that their notes and messages are part of the material, and that the families would be named here.