The paper develops a new logical system for arithmetic that takes seriously the idea that a statement should only count as true if there is a concrete, terminating process that makes it so. This is called "grounded" logic, inspired by philosopher Saul Kripke's work on truth. In ordinary classical logic, a statement is either true or false even if we have no way to verify which. Grounded logic is more cautious: a statement is true only if a finite chain of reasoning establishes it, and false only if a finite chain refutes it. Many statements end up neither. The paper builds a system called Inductive Grounded Arithmetic (IGA) by adding a key rule: whenever you define something recursively in a well-behaved way, you automatically get an induction principle for it. The whole system has been formally verified using the proof assistant Isabelle, so there are no hidden errors.
With this machinery in place, IGA can prove two famous theorems that classical mathematicians once thought required strong assumptions about infinity. Goodstein's theorem concerns sequences of numbers that, despite growing wildly at first, always eventually reach zero. The Hydra game, introduced by mathematicians Kirby and Paris, involves chopping heads off a mythical creature that regrows them according to a fixed rule, and the theorem says you always win eventually. Both results were known to be unprovable in standard Peano arithmetic and were thought to need substantial set-theoretic assumptions. IGA proves them using grounded derivations at every step, which is a notable achievement for a cautious, constructive-style logic.
The paper then investigates how far this grounded system can reach toward a famous ordinal called Gamma-zero, which represents a well-known boundary in proof theory. IGA can handle the Cantor and Veblen ordinal notation systems that fall below Gamma-zero, but it cannot prove Gamma-zero itself is well-founded without one additional classical assumption: that every position in a certain ordering is either reachable or not, with no middle ground. The paper shows this single instance of excluded middle is exactly equivalent to the well-foundedness of Gamma-zero and to a related claim about a classical computability-theoretic object. This precisely locates Gamma-zero as sitting just one logical step beyond what grounded reasoning alone can achieve.