For anyone relying on a proof system, the practical question is what its verification actually covers. In an article published on May 21, 2026, the author examined Succinct Labs and Nethermind’s formal-verification work on the SP1 Hypercube RISC-V zkVM, including its assumptions, axioms and chip-level theorems.
The work addressed the correctness of 51 opcodes, but the author identified cases that were uncovered or incorrectly specified. The article reports that testing found a JALR instruction bug: its lowest bit was not cleared. It also says an LLM-assisted audit found a vacuous proof for SLTI and an incorrect specification for load instructions.
The distinction matters: verifying stated properties is not the same as establishing that every instruction behaves as intended. The author calls for clearer specifications, reproducible results and further verification.

