# Claude Reconstructs P-256’s Public Seed From a 1997 Generator Run By Simon Yoon Canonical URL: https://www.tokenpost.com/news/technology/34391 Published: 2026-10-18T14:10:16.000Z Updated: 2026-10-18T14:10:16.000Z Section: Technology > The 160-bit seed matched candidate 1,842 in a timestamp-initialized `lrand48` run; why that candidate was selected and how the curve’s base point was generated remain unresolved. Anthropic’s Oct. 16 reconstruction ties P-256’s public seed to a timestamp-initialized run of the C library’s lrand48 generator, offering a repeatable explanation for a parameter used in a curve relevant to device-backed Ethereum signing. The run started Aug. 15, 1997, at 9:52:34 p.m. ET (Aug. 16 at 1:52:34 a.m. UTC). P-256’s published 160-bit seed matched candidate 1,842, with each candidate represented by 20 bytes. Also known as secp256r1, P-256 is an elliptic curve standardized by the National Institute of Standards and Technology (NIST) and used in device security and authentication systems. The broader reconstruction examined seeds for five NIST curves. It tested combinations of random-number generators and output segments. A pattern in the upper bits of several seeds was consistent with lrand48, which produces 31-bit outputs from a 48-bit internal state. The result connects to Ethereum through Ethereum Improvement Proposal 7951 (EIP-7951), which is marked Final. It specifies P256VERIFY, a precompile at address 0x100 for verifying P-256 signatures, with a stated cost of 6,900 gas. The proposal describes potential use with Apple Secure Enclave, Android Keystore and FIDO2/WebAuthn devices. The recovered seed is a public input used to generate curve coefficients. It is distinct from private keys used for signing and from passkey credentials. The finding does not indicate a cryptographic break, a new attack or a weakness in the elliptic-curve digital signature algorithm (ECDSA). The reconstruction leaves the selection process unresolved. Candidate 876 from the same run also produced a prime-order curve of the required size using the same parameter-selection approach. The findings do not explain why candidate 1,842 was chosen, how many timestamp-initialized runs were tried or how P-256’s separate base point was generated. The result adds detail to the broader reconstruction of five NIST curve seeds [covered in earlier reporting](), while leaving those questions about candidate selection and the base point open. ## Links in this article - [covered in earlier reporting](https://www.tokenpost.com/news/technology/34151)