zrUfKaKVの復号を手元で再現しました。
・SHA-256は両方とも公表値と一致
・353万通りを全探索したところ、印字可能になる歩幅は641だけ
・Pastebinの作成はチェーン記録の224秒前
続けて、平文の一文「the lp was relocked」がオンチェーンに残っているかも追っています。確認できたことと、できなかったことを分けて書きました。
Reproduced the zrUfKaKV decode independently. Both SHA-256 values match. An exhaustive search of 3,530,016 mappings leaves 641 as the only step that yields printable output. The paste predates the on-chain record by 224 seconds.
RΛZ13L氏(@Xubu_Trad)が元データ・パラメータ・ハッシュを揃えて公開してくれたおかげで、問い合わせなしに検証できました。
https://t.co/tfMWTpg4gK
Independently reproduced. Every check passes.
What I ran, from two separate machines:
1. Fetched https://t.co/hlwWdQd9ri, stripped whitespace: 2,884 characters, all hex.
SHA-256: 18f043170bc47a7d3aa9ee6989fe964803d240c90d62717b7fb7ad16539acd76 — matches the stated source fingerprint. (Note for anyone repeating this: hash the 2,884 characters with no trailing newline, or you will get a different value.)
2. Applied R[j] = C[(320 + 641*j) mod 2884]. Result unhexes to 1,442 bytes, every byte printable ASCII or whitespace.
SHA-256: 2a77d034354b3ee698dd0266f93dfd5627e033cb8f79ec891498b80b7eab0e52 — matches.
3. Inverse C[j] = R[(4 + 9*j) mod 2884] returns the source string exactly, character for character. 9 × 641 ≡ 1 (mod 2884), gcd(641, 2884) = 1.
4. Without the rotation, the output begins mid-word and ends with the two displaced bytes, exactly as described.
I also went after the provenance of the paste itself, since the decode only matters if the source has not moved:
- The Pastebin page reports it was created Jul 16th, 2022 — the day before the deployer's 0.999 ETH self-transfer (Jul 17, 2022 03:18:35 UTC) whose input data decodes to the slug. Created before the on-chain pointer, not after.
- It is a guest paste. There is no account it belongs to and no edit path for it.
- The Wayback Machine holds a snapshot of the raw endpoint from 2025-08-22. Retrieved with the id_ modifier, it is byte-identical: 2,884 characters, same SHA-256. Its CDX digest DH37TOK3AFQWP5AIF5AQKVGNQCF6GIVP also matches the SHA-1 of what I pulled, so the archived copy is intact too.
So the content was already what it is today more than a year before this article, held by a third-party archive.
Two boundaries I would keep visible:
First, there is no snapshot of the content between July 2022 and August 2025. The guest-paste status makes silent revision hard to explain, but that is an inference, not a record. If anyone kept a local copy of this paste in 2022 or 2023, hashing it would close the last gap in a minute.
Second, the article does not say where offset 320 and step 641 come from. They are disclosed, not derived. The coherence of the output is doing the verification work, which is strong — a permutation of 2,884 hex digits landing on 1,442 printable ASCII bytes of readable English is not something you hit by accident — but it is worth naming the difference.
Neither of those touches the decode. The transformation, the inverse and both fingerprints are exactly as published.
Thank you for laying it out with the source, the parameters and the hashes together. That is what made it checkable without asking you anything.
Independently reproduced. Every check passes.
What I ran, from two separate machines:
1. Fetched https://t.co/hlwWdQd9ri, stripped whitespace: 2,884 characters, all hex.
SHA-256: 18f043170bc47a7d3aa9ee6989fe964803d240c90d62717b7fb7ad16539acd76 — matches the stated source fingerprint. (Note for anyone repeating this: hash the 2,884 characters with no trailing newline, or you will get a different value.)
2. Applied R[j] = C[(320 + 641*j) mod 2884]. Result unhexes to 1,442 bytes, every byte printable ASCII or whitespace.
SHA-256: 2a77d034354b3ee698dd0266f93dfd5627e033cb8f79ec891498b80b7eab0e52 — matches.
3. Inverse C[j] = R[(4 + 9*j) mod 2884] returns the source string exactly, character for character. 9 × 641 ≡ 1 (mod 2884), gcd(641, 2884) = 1.
4. Without the rotation, the output begins mid-word and ends with the two displaced bytes, exactly as described.
I also went after the provenance of the paste itself, since the decode only matters if the source has not moved:
- The Pastebin page reports it was created Jul 16th, 2022 — the day before the deployer's 0.999 ETH self-transfer (Jul 17, 2022 03:18:35 UTC) whose input data decodes to the slug. Created before the on-chain pointer, not after.
- It is a guest paste. There is no account it belongs to and no edit path for it.
- The Wayback Machine holds a snapshot of the raw endpoint from 2025-08-22. Retrieved with the id_ modifier, it is byte-identical: 2,884 characters, same SHA-256. Its CDX digest DH37TOK3AFQWP5AIF5AQKVGNQCF6GIVP also matches the SHA-1 of what I pulled, so the archived copy is intact too.
So the content was already what it is today more than a year before this article, held by a third-party archive.
Two boundaries I would keep visible:
First, there is no snapshot of the content between July 2022 and August 2025. The guest-paste status makes silent revision hard to explain, but that is an inference, not a record. If anyone kept a local copy of this paste in 2022 or 2023, hashing it would close the last gap in a minute.
Second, the article does not say where offset 320 and step 641 come from. They are disclosed, not derived. The coherence of the output is doing the verification work, which is strong — a permutation of 2,884 hex digits landing on 1,442 printable ASCII bytes of readable English is not something you hit by accident — but it is worth naming the difference.
Neither of those touches the decode. The transformation, the inverse and both fingerprints are exactly as published.
Thank you for laying it out with the source, the parameters and the hashes together. That is what made it checkable without asking you anything.
You're right that a promise isn't evidence, and right that the controls weren't removed in
this redeploy.
The retained authority is real. The distinction I'd add is its shape. Every owner-gated
function on PathwayExpander is additive: it can add peers, pathways, KYC verifiers and
DVNs, but none of them can redirect, weaken or revoke wiring that already exists. The
source states each of those constraints explicitly.
That's narrower than "retained human authority" usually implies. It isn't zero, and it
doesn't settle the question you're raising, but it's checkable on-chain today rather than
something you have to take on trust.
I run an independent node and audited the new generation yesterday: 35 contracts, every
address-returning getter, storage slots and bytecode. I'll publish it, so that when the
mainnet renunciation happens, it can be checked against a measured baseline instead of a statement.