The paper tackles a foundational question in mathematical logic, specifically about a set theory called New Foundations (NF), invented by W.V.O. Quine in 1937. Set theories are formal systems that try to rigorously define what mathematical objects exist. NF is unusual because it avoids certain paradoxes through a syntactic restriction called stratification, which limits how sets can refer to themselves. The version studied here, called NF-0, is a modification that removes the empty set and instead requires that any definable collection must have at least one member before it counts as an object. The central question is whether this theory can prove that infinitely many things exist, which is a non-trivial fact that some set theories need to assume as an explicit axiom.
The proof works by contradiction. The authors assume every object in the theory is finite and then derive an impossibility. Their notion of finiteness is carefully defined from within the theory itself, using well-orderings and the absence of infinite descending chains. Under the all-finite assumption, they analyze how counting is affected by a specific operation called type-raising, which shifts sets up one level in the stratification hierarchy. They show that adding a single point to a finite set flips its parity (odd becomes even and vice versa), using a clever graph-rewiring argument. Since type-raising preserves parity but also increases the count by exactly one, the theory demands a number equal to itself plus one, which is a contradiction. Therefore, not everything can be finite, meaning at least one infinite object must exist.
The authors also connect their work to Lean 4, a modern computer-assisted proof verification system, identifying which parts of the argument have been formally checked by a computer. This adds an extra layer of confidence to the result. Importantly, the paper does not claim to resolve the long-standing open question of whether ordinary NF is consistent with standard mathematics; it only establishes that infinity follows from the axioms of NF-0 specifically. The result is notable because it gives a direct, internally motivated proof rather than relying on model-building or importing results from other systems.