Claims of recursive self-improvement are easy to make and hard to falsify, the authors of a new arXiv preprint argue. The central difficulty is attribution: did the improvement come from re-using designer-provided mechanisms, or from mechanisms the system authored itself? Without a way to separate those cases, such claims remain ambiguous.
To address this, the preprint introduces a pre-registered, two-tier framework. Based on the title, the work reports "first audited operator-genesis events" in a physics-discovery runtime, suggesting the framework has been applied in at least one concrete setting.
Because the abstract is cut off mid-sentence, the specific mechanics of the two tiers are not available in the source. The paper is a single preprint, so there are no other sources to compare or contrast; it has not yet been peer-reviewed.