The paper is an outcome of a continuous dialog between the author and AI (primarily, gpt, ocassionally claude as well). We were researching snark decomposition question, and wanted to understand, what if we specialize from bicritical snarks to more restrained hypohamiltonian or (conjecturally) permutation snarks.
All results have been formalized in Lean 4 and could be found in author's github.
Author has read the whole text, has checked the hypohamiltonian argument, and believes the permutation part is also correct, but needs some time to understand the proof properly.