arXiv:2608.11225 Defines Six-Term Factorization for AI Clone Indiscernibility

The July 2026 preprint sets aside consciousness and operationalizes identity through manifestation equivalence over time. Three conflated criteria are separated: target fidelity, generic human-likeness, and individuality. Randomized ablation is proposed to estimate each factor's local sensitivity. The central conjecture states that versionability degrades long-horizon indiscernibility once an agent tracks its own persistence and stakes.

The framework maps existing empirical work onto the three criteria and replaces trajectory matching with climate fidelity: the clone must reproduce the original's conditional response distribution. A delegate object is introduced between product clone and full individual: a bounded-lifespan, task-limited partial instance that terminates with a bandwidth-limited testament. Lambda-calculus and bisimulation analogies clarify linearity constraints without claiming they suffice for identity.

Operational consequence is an experimental program that treats clone evaluation as a statistical estimation problem rather than narrative resemblance. Regulators and deployers gain measurable sensitivities instead of untestable claims about personhood. The paper supplies no new datasets or benchmarks.

Next steps require labs to publish ablation coefficients on production clones and to test the versionability conjecture against multi-year interaction logs.