Propose–Solve–Verify (PSV) for Formal Discovery

PSV is a training loop where an autonomous proposer generates formal problem specifications, a solver attempts programs/proofs, and a formal verifier accepts only fully proven solutions; verified wins become high‑quality training data for the solver. By replacing unit‑test rewards with formal verification as the selection mechanism, PSV makes self‑generated, provably correct mathematics and software a scalable outcome. — If PSV generalizes, it changes the landscape of scientific discovery, software assurance, and industrial R&D—creating systems that can autonomously create and verify high‑confidence results and thus shifting regulatory, safety and workforce policy.