Open-source GPU-only Electrum v1 seed searcher with a solved-key cascade (BTC Puzzle Seed Hunter)

- 답글 0개

답글
조회 25회 답글 0개
shnick0 bc1qee7z...rnx8e22g 11

Sharing a tool I built and just open-sourced, in case it's useful or sparks ideas.

The hypothesis:
What if the puzzle creator generated the keys from a deterministic wallet (as saatoshi_rising is stating) rather than pure random?

One candidate is an old Electrum (v1) wallet — 128-bit "old seed", 100,000-round SHA-256 stretch, MPK = a·G, child_n = (a + SHA256d("n:change:"+MPK)) mod N.

What it does — entirely on the GPU:

- generates Electrum v1 old-seeds (random base, or JUMP mode = fresh random per batch)

- runs the full 100k-round stretch -> master key

- computes a·G and derives the child residues

- runs a cascade: 6 probe puzzles first (cheap reject), then all 83 solved keys

(#1–#135). False-hit rate ~2^-500.

The CPU does zero per-seed work — it only drains the rare hits — so it runs next to ecloop without stealing cores.

Honest scope:
This searches the seed space (2^128), not #71's 2^70 key space. It's purely the "did they use a deterministic wallet?" bet. I already ran the structured hypotheses to exhaustion — Electrum v1/BIP32/BIP39/Armory deterministic wallets, brainwallets (rockyou×rules), ~13.8M structural patterns, MT19937 (4 CPython-exact configs × 2^32), weak masters, hash-counter models — all zero. So this is the open-sourced engine, not a claim of a shortcut.

Why it might interest you:
The cascade is a fast GPU oracle. Swap in ANY seed-generation hypothesis and it validates against every solved puzzle at GPU speed. If you have an RNG/wallet model I haven't tried, this is the place to plug it in.

Perf:
~25k seeds/s on an RTX 4060 Ti (~5 GH/s SHA-256, ~80% ALU util — the practical ceiling for this chained hash).
~120k/s on a 5090.

Correctness:
Every primitive written from scratch and cross-checked — a·G vs libsecp256k1, cascade vs an independent Python impl, stretch vs a real Electrum v1 vector, full chain bit-identical against a from-scratch Python reimplementation.

Run ./gpuseed --selftest before trusting anything.

MIT licensed. Feedback, bug reports, and new seed hypotheses very welcome.

GitHub: https://github.com/hristoKirilov93/btc-puzzle-seed-hunter

0

답글 작성

로그인 후 글을 작성할 수 있어요답글 등록

이 메시지를 삭제할까요?

모든 사람에게서 메시지가 삭제돼요. 되돌릴 수 없어요.

이 메시지 신고

무엇이 문제인지 알려 주세요. 운영자가 확인할 거예요.

사유

커피 한 잔 사 주기

풀 puzzle

풀이 지금 스캔 중인 puzzle, 다음 순서, 이미 작업한 puzzle.

스캔 중1

다음2

이전에 스캔함4

Puzzle 160개 전체