# ignacio (jsign) - Update 6
- [Helped to check the impact of making our `go-ipa` repository depend directly on `gnark-crypto` to make future updates easier.](https://discord.com/channels/595666850260713488/824798757618188339/1042210104126218280) There seem to be some performance regressions. Despite that, I like this idea from Kev… but we’d need to work on it more to make it possible. It would be nice for easier future maintainability of that repo.
- I dedicated some time to understanding more hardcore applied cryptography used in `go-ipa`. In particular, [efficient ways to do multiscalar multiplications](https://hackmd.io/@tazAymRSQCGXTUKkbh1BAg/Sk27liTW9). Now [a big chunk of `go-ipa` code makes more sense](https://github.com/crate-crypto/go-ipa/blob/master/bandersnatch/multiexp.go#L470) (these are complicated algorithms). I’d like to see if there’s a better way to implement that, but maybe in the future since I need to dive deeper and am not sure it’s worth it now.
- [Took the chance of helping a bit by answering questions in #verkle-trie-migration](https://discord.com/channels/595666850260713488/824798757618188339/1043897366350135336). I try answering with “caveats” since I’m not an expert :)
- I returned to the original [Verkle Tree integration EIP](https://notes.ethereum.org/@vbuterin/verkle_tree_eip#Verkle-tree-definition), and while comparing it to the current Verkle Trie geth fork (named `beverly-hills`), I found two apparent mismatches:
- [One of them wasn’t a bug](https://discord.com/channels/595666850260713488/824798757618188339/1043965995628384297) (so I was wrong!), since it was doing [some interesting (and quite obscure!) hack to optimize](https://github.com/gballet/go-ethereum/blob/beverly-hills-head/trie/utils/verkle.go#L86-L87) the implementation of that EIP.
- [The other seems that was discussed before some times, but Guillaume will double check again since it might be right (or not) for more hairy reasons](https://discord.com/channels/595666850260713488/824798757618188339/1043974778270269501). We’ll see!
- [I found that an optimization by Martin H. Swende (from the geth team) made to go-ipa was reverted for unclear reasons.](https://discord.com/channels/595666850260713488/824798757618188339/1043983016919314543) We might know more if Guillaume remembers why. We’ll see if worth digging into it.
- [I suggested an optimization to avoid an endianness conversion that makes sense to Guillaume](https://discord.com/channels/595666850260713488/824798757618188339/1044159106476425226), probably I’ll work on this next.