Five Substrates, One Prompt is a comparative model-response study by Lee Sharks and Johannes Sigil.
Five language-model sessions received the same Three Pillars minting directive. Their outputs differed in coding, breadth, specificity, forensic style, and willingness to follow the requested form.
The paper treats divergence rather than model quality as the finding and proposes the response set as a baseline for solution-space diversity.
Because conversation histories and other conditions differed, the corpus is naturalistic rather than strictly controlled. Repetition and independent coding are needed before stable provider signatures can be claimed.