Someone reproduced the numbers before agreeing

A post on Ethereum Research from ignotusnemo describes Parano1d, a proof-of-work layer one where blocks carry cryptographic proofs rather than executable transactions. The state model is UTXO rather than EVM, validity proofs are recursive from genesis, and block bodies become prunable after an 18-block reorganisation window, leaving constant-time verification of finalised checkpoints.

The stack is hash-based and needs no trusted setup: a Poseidon2b permutation, traces over GF(2^128) with challenges over GF(2^256), and GKR, batched sumcheck, zerocheck, lincheck and FRI-Binius or BaseFold above them. A reduction the author calls FROST-GKR is reported to give a 10.69× median prover speedup, 14.80× on the verifier, and 51.67% smaller proofs.

Published measurements are specific about hardware. On a 12-thread AVX2 laptop, a B25 query takes 10.734 seconds for a 971,732-byte proof; on 24 threads with AVX-512 the same query takes 6.905 seconds. The heavier B255 class runs 34.938 and 21.053 seconds against proofs of 1,081,108 bytes. Target block interval is 20 seconds.

The part worth copying

The author separates 127 bits provable from 127 bits conjectured rather than rounding both to 128-bit, and states plainly that the claims rest on premises without external audit. That is unusually careful language for a protocol announcement.

A second participant, TMerlini, reproduced every published figure on different hardware and then changed the BaseFold query count from 133 to 132 to see what would happen. The system recalculated its security boundary rather than carrying the published conclusion forward unchanged. That test — does the claim move when the parameter moves — is the one a reader can actually run.

The lessons the author draws for Ethereum's own post-quantum migration are that swapping signatures is not the same as securing a system end to end, that the hash choice determines proof size, recursion cost and bandwidth, and that recursive validity separates historical execution from what a new node must process to start.