Identifiers truncated in this published copy. The record behind every row is retained in full.
Two launches that share one artifact they were born with share a buyer in 4.78% of pairs, against 0.05% for decoy pairs matched on buyer count. Two shared artifacts: 12.42% against 0.28%. The copies are made in seconds rather than days: where the same bytes were uploaded again to a second address, 65.11% of those groups were finished inside a minute, and one specimen took thirteen milliseconds end to end. And the one shape of document that could collide between strangers without anybody coordinating, the default a launch tool emits, is the least shared of all.
Unit is the group. Window 2 September to 19 October 2026, launches this archive recorded at their creation transaction. The two columns are the same question asked of two different kinds of match: groups whose launches carry the byte-identical document, and groups that share a ticker and a name while carrying different documents.
| Spread of the group | Identical document, 85,952 groups | Same ticker and name, different documents, 106,516 groups |
|---|---|---|
| Under 60 s | 60,831 (70.77%, 70.47 to 71.08) | 10,013 (9.40%, 9.23 to 9.58) |
| 60 s to under 5 min | 9,507 (11.06%, 10.85 to 11.27) | 8,748 (8.21%, 8.05 to 8.38) |
| 5 min to under 1 h | 5,447 (6.34%, 6.18 to 6.50) | 9,935 (9.33%, 9.15 to 9.50) |
| 1 h to under 24 h | 5,517 (6.42%, 6.26 to 6.58) | 19,449 (18.26%, 18.03 to 18.49) |
| 24 h and over | 4,650 (5.41%, 5.26 to 5.56) | 58,371 (54.80%, 54.50 to 55.10) |
| Within 5 min | 70,338 (81.83%, 81.57 to 82.09) | 18,761 (17.61%, 17.39 to 17.84) |
Four fifths of identical-document groups close inside five minutes; four fifths of shared-name groups do not, and more than half of those span a day or more. A name is reused over weeks. A file is reused within minutes.
A later launch can point at a file somebody else is hosting, which takes no work. Where the identical bytes sit at two or more different addresses, the file was uploaded again, which takes a separate upload. Full record, no sampling, same window.
| Identical-document groups spanning two or more creators | Groups | Share | Creators | Launches |
|---|---|---|---|---|
| Same bytes at two or more addresses: a separate upload is proved | 8,899 | 10.35% (10.15 to 10.56) | 46,808 | 91,826 |
| One address: not decidable from the address | 77,053 | 89.65% (89.44 to 89.85) | 226,257 | 338,507 |
And where a separate upload is proved, it is fast.
| Spread of those 8,899 groups | Groups | Share |
|---|---|---|
| Under 1 second | 1,495 | 16.80% (16.04 to 17.59) |
| 1 second to under 60 s | 4,299 | 48.31% |
| 60 s to under 5 min | 1,599 | 17.97% |
| 5 min to under 1 h | 671 | 7.54% |
| 1 h to under 24 h | 458 | 5.15% |
| 24 h and over | 377 | 4.24% |
| Within 60 seconds | 5,794 | 65.11% (64.11 to 66.09) |
The specimen. Three different creator wallets launched with byte-identical document content, uploaded to three different addresses on one host, at 1791054709818, 1791054709821 and 1791054709831 milliseconds. Thirteen milliseconds end to end, three separate uploads.
Measured 24 October 2026 over the record-wide window, 9,444 of 9,444 planned pairs completed. Each tested pair is compared against a decoy pair from the same venue and day with the closest available buyer count, so a difference cannot be a difference in crowd size. Buyers here are buyers of the token, not curve buyers specifically. Wallets touching more than 100 distinct mints are excluded as indiscriminate.
| Primary comparison, exclusion at 100 mints | Comparable | Rate | Wilson 95% | Bootstrap over creators | Matched decoy | Lift |
|---|---|---|---|---|---|---|
| Exactly one shared artifact | 2,134 | 4.78% | 3.95 to 5.77 | 3.69 to 5.94 | 0.05% | 102x |
| Two or more shared artifacts | 2,134 | 12.42% | 11.09 to 13.89 | 10.66 to 14.29 | 0.28% | 44x |
The intervals do not overlap each other and neither comes near its decoy. Two shared artifacts is 2.6 times one.
| Creators in the document group | Comparable | Rate | Wilson 95% | Bootstrap over groups |
|---|---|---|---|---|
| 2 | 142 | 13.38% | 8.74 to 19.95 | 8.99 to 17.44 |
| 3 to 5 | 188 | 15.43% | 10.96 to 21.28 | 11.11 to 19.47 |
| 6 to 20 | 140 | 19.29% | 13.61 to 26.61 | 13.10 to 25.27 |
| 21 to 100 | 85 | 22.35% | 14.80 to 32.29 | 4.44 to 38.00 |
| Over 100 | 41 | 0.00% | 0 to 8.57 | 0 to 0 |
Agreement rises with the size of the group and then stops dead above a hundred creators. Above a hundred creators, agreement falls to zero, which is why this register caps a group at a hundred and publishes the cap.
The alternative to copying is not coincidence, it is a template: a launch tool that emits three or four fields with a blank description hands the same bytes to everybody who uses it with the same ticker. That is the one mechanism that produces identical documents with nobody coordinating, so it is worth counting. Classes were fixed before the count. RICH is a description of forty or more characters and at least one of twitter, telegram or website. THIN is no description and none of those three. MIDDLE is the rest. Chain-wide, the three complete UTC days 20 to 23 October 2026: 165,313 launches carrying a document, 98,980 distinct document hashes, of which 23,272 are carried by two or more distinct creators.
| Class | Share of all hashes | Share of shared hashes | Hashes shared, within the class |
|---|---|---|---|
| RICH | 13.54% | 9.94% | 17.26% |
| MIDDLE | 70.83% | 86.10% | 28.59% |
| THIN | 15.59% | 3.94% | 5.93% |
The documents that could collide between strangers are shared least: 5.93% of thin hashes are carried by more than one creator, against 17.26% of rich ones. Sharing is concentrated away from the template shape, the one explanation that would not require anybody to have copied anything.
Every figure here is measured on pairs of launches. The register publishes groups. The ceilings act on groups too. So the object a reader buys is not the object these numbers were computed on, and this page claims nothing about it.
There is no independent ground truth here, so every figure is convergence and never precision. Instruments agreeing with each other is not the same as either of them being right. The one labelled set this project held was built from ticker families, which is circular for a test of a file instrument, and it was withdrawn.
The size of the separation moves with the exclusion, so the cut is part of the figure. At the 100-mint exclusion the arms read 4.78% and 12.42% against decoys of 0.05% and 0.28%. With no exclusion at all they read 28.77% against 11.81%, a lift of 2.4 rather than 102. A rate quoted without its cut is not a measurement.
One act of copying can arrive wearing three artifacts at once, so these are shared artifacts and not three independent events. The register records how many separate reuse acts an edge rests on, and publishes that beside the connection.
This register records what was launched and does not call any launch here authorised or unauthorised, affiliated or unaffiliated, genuine or counterfeit. It does not say who controls any wallet, and nothing on this page identifies a person.