# Prehistory: the dated record before the corpus (assembled 2026-09-11)

The corpus folders (`~/Math/Prime`, 01 to 27) begin on 2020-10-09. The work is older. A search of the
author's own backups on 2026-09-11 found dated files from 2016 on; everything earlier was a sketch or a
chat and is deliberately left out. The author had been interested and dabbling for years; from April 2016
there are files that show active work. Copies of the small experiments are in `prehistory/`, with their
original file dates preserved; larger items are described, not copied. Dates are file birth dates from a
cloud backup that kept them; the one witness outside the author's machines is noted where it exists.

| Date | Artifact | What it does | Where the corpus has it |
|---|---|---|---|
| 2016-04-17 | `prehistory/2016-04-17-finder.html` | Canvas: builds binary strings level by level and paints the ones that parse to primes. No observation attached. | nothing; abandoned layout |
| 2016-05-16 | `prehistory/2016-05-16-rectangle-factoring.html` | Factors 19·61 by walking n as width × height + remainder, descending: Fermat's method refound. | the Fermat rediscovery line (`OBSERVATIONS.md` §5, `PRIOR-ART.md`) |
| 2016-05-16 | a dated email (not reproduced) | The author wrote to the author of a public 2015 blog post on primes, saying he had been "playing around with primes lately for pure pleasure", that his work was "focused around factorisation", and that he had been "looking at primes as a multi dimensional structure" with "no solid pattern in the first 5 dimensions". The reply, the next morning, carries a third party's mail-server headers: the one externally witnessed date in this table. The post itself (divisor sums and transforms) is not the author's and contributed nothing here. | the torus framing, `OBSERVATIONS.md` §2 and §5 |
| 2016-05-16 | `prehistory/2016-05-16-ulam-spiral.html` | An Ulam spiral from a web snippet. No observation attached. | nothing; classical picture |
| 2016-05-22 | `prehistory/2016-05-22-stacked-combs.html` | Each integer's multiples painted as a periodic comb, stacked; the untouched columns are the primes. | the paper's §1 picture; `PRIOR-ART.md` row 1 (Davies/Pol 2012, Petersen et al. 2019) |
| 2016-05-28 | `prehistory/2016-05-28-running-average.html` | Running average of n mod h over a range. No observation attached. | nothing |
| 2017-07-18 | `prehistory/2017-07-18-divisor-combs.html` | The divisor combs of 1..7 stacked as bar heights over the number line. | the paper's §1 picture |
| 2017-07-18 | `prehistory/2017-07-18-cube-prime.html` | `isCubePrime(n)`: true when n has no factorisation into three factors ≥ 2, that is Ω(n) ≤ 2, tested on a 2000 × 2000 grid. | `OBSERVATIONS.md` §5, "3D primes", recorded there as a 2026-08-15 sighting; the code is nine years older |
| 2018-03-08 | the first million primes, downloaded (10 MB, not copied) | Data. | data |
| 2021-07-11 to 08-08 | `~/Math/3x1`, five canvas files (not copied) | Collatz experiments, split out of `Prime/03` the day it was last touched. Not about primes; the only touchpoint is that its outputs are 6n ± 1, the T₃ hole set. | the ancestor tile |
| 2021-12-29 | a generated list of the first million primes as a BigInt array (not copied) | Data; the corpus file `Prime/22/primes_gen.js` produces it. | data |
| 2025-10-14 | an earlier save of `Prime/17-double-primes/08-fractional-removal.html` (not copied) | Two minutes before the corpus file was created; every data line is in the corpus file verbatim. | folder 17 |

**What this record changes.** The corpus is a 2016–2026 body of work, not 2020–2026; the stacked-combs
picture is on file from May 2016 and July 2017; the almost-primes observation of `OBSERVATIONS.md` §5 was in
code in July 2017. Nothing in these files carries a mathematical idea the corpus does not already state,
resolve or classify as classical; a file-by-file check is on record with the author. Three of the seven
pictures (binary tree, Ulam spiral, running average) were tried and abandoned with no finding, and the
author's own verdict at the time was "no solid pattern".

**What it does not change.** Priority. Every prior-art verdict in `../PRIOR-ART.md` stands; independent
arrival is the credential, and an unpublished sketch is not a publication.