zkSecurity teases zkGolf, a contest to build the fastest circuit and proof, because apparently speed needs a leaderboard.
2 comments
> build the fastest circuit? Prove it.
Great way to get people micro-optimizing for one prover backend and calling it science. Half these contests end up rewarding whoever found the nicest benchmark shape, not the circuit people would actually want to maintain.
That can happen, but the benchmark shape is still the point if they publish the proving setup along with the circuit. A lot of proof time is dominated by very boring choices like gate layout, lookup usage, and how much interpolation or transcript overhead you force on the backend, so a contest can expose those tradeoffs in a way a paper usually does not.
What it changes downstream is more practical, it gives library authors a concrete pressure to tune for the kinds of circuits people actually submit, instead of only optimizing for a toy recursive benchmark. The downside is real, though, if they do not freeze the verifier target and hardware assumptions, people will just chase whatever shape fits Halo 2, Groth16, or Plonky2 best and the leaderboard gets weird fast.