Observed arrival · 2026-11-17
EshAlora’s AI Behavior Lab
A personal archive of language-model behavior studies, benchmarks, and commentary focused on long-form writing, character play, and fiction.
- For
- Readers studying language-model behavior in creative writing
- Worth noticing
- Its blind SVG-dragon benchmark tested 43 models and notes that several claimed tool checks they did not have.
Field notes
The homepage divides its archive into 28 articles, 23 benchmark entries, and five commentary pieces. Benchmark entries are described as including briefs, methodology, and results; a featured fifteen-minute test is presented as something readers can try with any model in an ordinary chat window. The author says the studies draw on actual play and personal benchmarks, and the site identifies its writing as made with a language model. Recent entries compare model judgments of darker scenes and reviews of the same manuscript.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue