Performance proposal before release.
As a normal miner/user, I do not think the current version is ready for release.
Chain synchronization and miner synchronization are too slow.
The hashing is too heavy, and the code is not optimized enough for normal users.
SHA-256 should be used as an internal hashing algorithm where it is safe and compatible,
while Poseidon2b should remain for the actual PoW mining so the existing mining difficulty and security are preserved.
If the miner cannot synchronize quickly, create a block template in time, or find blocks before the next block arrives, normal users will simply stop using it.
If released in its current state, most users will ignore it and move on.
The developer has a choice:
make it fast and reliable enough for normal users, or risk all this work being for nothing.
A clean testnet full node syncs in around 20 seconds. Sync time does not grow the way it does on traditional blockchains. In ten years it should take minutes not days, so I think 20 seconds is a good tradeoff. Btw, as the P2P network grows, more nodes will share sync data making sync faster and more reliable.
Alternative hashes were not overlooked. Earlier builds deliberately used BLAKE3 for FRI Merkle commitments and PoW. I removed it from consensus when recursive verification was completed because the recursive proof must verify the entire hash path. Suggesting SHA-256 now ignores the exact reason BLAKE3 was removed and would make the proof path worse, not better. Poseidon2b is not used only for PoW. It is used for addresses, transactions, Merkle and State commitments, proof transcripts, recursion and block IDs. Replacing it would require rebuilding consensus, the proof system, production matrices and the security model. The delay in your logs is proof construction and verification, not a generic hash loop. Parano1d is not designed to make mining easy. PoW only orders already proof-valid transitions. More independent mining power strengthens that ordering layer. Ordinary users can run wallets and fully verifying nodes, but competitive mining is a separate workload and is not guaranteed on every machine.
I am not going to spend my time debating AI generated protocol redesigns anymore.
