Interesting project! What kind of proving times are you getting? Say a sudoku for argument's sake. Also are you aware of the flaw in the original zkcp demo by bowe and maxwell? (To do with witness indistinguishability). Maybe it doesn't apply because STARKs not SNARKs but just wondering.
Discussion
ahah! I was sure someone must've thought of this idea before me, but I didn't have the right search terms. "ZKCP". This is great background reading, thank you nostr:npub1vadcfln4ugt2h9ruwsuwu5vu5am4xaka7pw6m7axy79aqyhp6u5q9knuu7!
Interesting to learn that the SNARK setup is more sensitive than I had thought. I'll be sure to correct my article to reference the original idea and fix the flaw which I unwittingly duplicated.
As for performance, so far I've only tested the PTLC-bridging program here shown in this section: https://conduition.io/bitcoin/zkpreimage/#Optimizing
With that, it took 3 minutes to generate a proof with the RISC0 STARK prover. Not great but also not completely impractical. I'd like to test other prover programs too
Article fixed, now I'm assuming transparent setup ZKPs only (e.g. STARKs). Thanks for pointing this out 😄