竹竹YZZ

There was a line in the Newton whitepaper that I almost skimmed over at first: Newton compiles the whole Rego policy engine into a program that a general zero-knowledge virtual machine can run. Any strategy written in Rego doesn’t require additional engineering effort and is, by nature, already covered by a mathematical proof of correctness.
Last weekend I happened to look up the actual proving time for these general-purpose zkVMs, and the results changed my interpretation of that sentence by about half.
Most zero-knowledge proof applications on the market today follow a "purpose-built for a specific use case" approach: they design a circuit specifically for a particular computation task. This is efficient, but every time the scenario changes, the circuit has to be redeveloped. Newton takes the opposite direction: compile the entire interpreter into the RISC-V instruction set and run it on a general-purpose zkVM such as SP1 or RISC0. In this case, the object of the proof is no longer whether "the result produced by this strategy is correct," but rather that "this interpreter indeed executed everything honestly and exactly as intended." I agree with this idea. Strategy authors don’t need to understand cryptography—once they write the Rego code, verifiability is automatically obtained.