What was already known
Hart, Kim and Pillay proposed stable forking in 1996. Several special positive cases were known; the general conjecture remained open.
AI involvement
Freitag and Mutchnik describe giving GPT-5.6 Sol detailed constraints and a Kim–Pillay strategy; the model supplied the candidate counterexample. The human authors proved the needed facts and wrote the paper.
GPT-5.6 Sol supplied the candidate after detailed human prompts specifying constraints and the Kim-Pillay independence strategy. The authors developed the setup and proofs.
Importance
A counterexample settles the general stable-forking conjecture in the negative.
Limitations
The authors supply a mathematical proof, but the source does not document a separate formal check or outside proof audit.
Technical verification
The authors wrote the proofs establishing the candidate counterexample. This validates the model-supplied candidate rather than checking a separate AI-written proof.
Human understanding
The authors explain the construction, independence relation and obstruction to stable forking.
Correction and publication history
Claim-level primary-source review for non-notifying local inclusion.
Read this published revision