mimesis

The every-story, run 1: result

Note, 2026-10-12, after the run: John’s call holds on what it meant. The video is the Feb 1, 2019 meeting itself, word for word wherever the official record prints his words (192 of 195, nothing added), wall and all. What failed was the reference: the official transcript cuts 338 of his 532 words (64%). Grok, asked on X, agrees: “The printed match supports the video as the same event” (post). In set terms: the official transcript’s words are a proper subset of the video’s (all 192 printed are in it, and 338 more are not printed). John’s estimate, the full 2019 telling, is the proper superset. The scored verdict below stays as it was.

Run on 2026-10-12 at 21:12:38 UTC, exactly as pre-registered in PREREGISTRATION.md (published at mimesis 7d3efc6, pushed before the run). The command was the published one, word for word. The script checked all 4 input hashes against inputs.SHA256SUMS and printed “inputs: all sha256 match the pre-registration”. Output: out/every_story.md, out/every_story.json, the video transcript out/video.txt, and the console log out/run1.log.

Result: John’s call does not hold under the pre-registered rules

Test A, elements: 18 of 32 found, not 32 of 32. Fails.

  video deseret foxnews whitehouse
E1 girl kidnapped in Central America yes yes yes yes
E2 11 years old yes no no yes
E3 groomed for two years yes no no yes
E4 boy kidnapped in Mexicali yes yes yes no
E5 a wall in the boy’s story yes no yes no
E6 Calexico port of entry no yes yes no
E7 rescued in San Bernardino yes no no no
E8 the wall was the difference yes no no yes

Test B, word for word (match = 95% or more of the video’s 532 words, in order): no text matches. Fails.

Text Video words found Share
deseret 133 / 532 25.0%
foxnews 91 / 532 17.1%
whitehouse 275 / 532 51.7%

Read after the run (not part of the scoring)

Everything below was looked at only after the result above was written. It does not change the verdict.

1. The 2019 White House transcript leaves out the middle of his remarks. The American Presidency Project prints his opening, the girl story up to “the clientele that’s the big money”, then the bracketed line “[Mr. Ballard continued his remarks, concluding as follows.]”, and picks up again at “The difference between those two cases is two plus two equals four. The wall was the difference.”

That gap is why the transcript has no Mexicali boy (E4–E7) and why Test B stops at 51.7%.

2. Where the transcript does print his words, the video says them word for word. We lined up the transcript’s printed Ballard text (195 words) against the video, with the same word rules as the script. 192 of 195 (98.5%) are in the video, in the same order. The three misses are “I”/”I’ve” and “Sánchez”, which the word rules split into two pieces.

The video holds 254 words exactly where the transcript has the bracket: “They brought this little girl through a part of the southern border where there was no wall…”, down to “…rescued 12 other children in San Bernardino, California.” Its ending (“We are having to do operations in Mexico… fly swatter, which is the wall… Thank you very much. I appreciate it.”) is also not printed.

3. So our “2025 video” is footage of the Feb 1, 2019 meeting itself (the transcript page’s date). The opening (“Thank you so much, Mr. President…”) is the same, and so is “Olga Sánchez, just last month”. “2025” is when the first mimesis script got its copy, or when it was going round again (“as they rolled out the same script”). It is not when it was said. The Deseret op-ed (Feb 4, 2019) came three days after the meeting.

4. The video’s E6 “no” is a transcription miss. This run’s Whisper small heard “the Mexico port of entry”. On 2026-10-10 the same model heard “Calexico Port of Entry” (THE CUT, QUOTES_CHECKED.md). Scored as pre-registered.

5. The op-eds tell the same two cases in other words. Fox has the Mexicali boy and the wall in his story (E4, E5). Deseret has Mexicali and Calexico but not the wall within 400 characters (E5). Neither has “11 years old” or “groomed for two years” in those words, and neither repeats the speech word for word (17–25%).

How to check this yourself

  1. Fetch the three pages from the links in the pre-registration. Pages change over time, so your hashes may differ from ours. Ours are in inputs.SHA256SUMS.
  2. On the transcript page, find “Mr. Ballard continued his remarks, concluding as follows.”
  3. Read out/video.txt beside it. The words before and after the bracket are the same as the video’s, and the Mexicali boy sits in the gap.
  4. Re-run: the command in the pre-registration. It stops if your inputs are not byte for byte ours.

Hashes of the outputs, as written by the run:

3fddf8804b695f376a911399768ff6c512220f1552d397e6940bfbc291ddf17d  out/every_story.json
b4f451b86c9c07681651ec4580d590472919e22ffd38ef0e27c6521f10a070fe  out/every_story.md
34622a0d3639ecb0a2d950acd8ab479e72c8af068cff493e1232562667ed3ad5  out/video.txt