# Nova research & implementation

## better understanding on Nova's cost model

- first, let's start with fully understanding Sangria's cost model
- what is the exact cost of folding circuits with high-degree custom gates (See $3.3 of Sangria paper)
- also, what is the exact cost of verifying Sangria's satisfying witness inside PlonK?

## Nova on-chain verification

- we should verify Sangria's satisfying witness inside Halo2
- (Lev's idea) to reduce the cost of MSM inside circuit, explore ECIP paper(https://eprint.iacr.org/2022/596.pdf)
- implementing ECIP in Halo2 maybe the useful benchmark
- We should verify Grumpkin side IPA in bn254 halo2 for on-chain verification, which causes linear-sized MSM. It would be more performant to apply bucket method in it. (https://github.com/kilic/halo2msm)
- Liam's scheme would be more performant in folding scheme, but not sure.
- Some people mention about Hyrax, should check about it too in our side.

## CCS & HyperNova(Optional)

- we need format to represent CCS
cc
- https://github.com/iden3/r1csfile/blob/master/doc/r1cs_bin_format.md
- https://github.com/thor314/ccs-hack
- convert from R1CS to CCS
- we can fold CCS style format with HyperNova