One Table And Two Hundred And Forty Hyphens

7,972 bytes and 7,922 characters. The 50 bytes between the two counts are the whole arithmetic of this file: 25 characters sit above the plain alphabet, carried in a single glyph, and every one of the 25 weighs 3 bytes. 25 times 2 is 50. Half a byte of that risk lands on every one of the 195 lines. 195 lines, 53 of them blank, 142 carrying content, and 126 of the 142 distinct. 16 lines repeat a line standing above them, and the repeat is the rule of 3 dashes, which appears 11 times. Nothing else in the file is said twice. One table stands in the whole document: 9 table lines, 8 data rows and 32 cells with 0 of them empty, so the grid closes at 100.0 per cent across 4 columns. 4.6 per cent of the file is that grid. A table that closes at 100.0 per cent is a table somebody re-read, because the empty cell is the first thing a reader finds and the last thing a writer checks.

40 bullet lines against 11 numbered lines, so the file picks the dash over the digit by almost 4 to 1. 0 checkboxes and 0 code fences stand anywhere between the lines: 195 lines carry nothing a machine is asked to run. 12 heading lines hold the outline: 1 at the first level and 11 at the second, with nothing at the third level and nothing below it. The document is 2 layers deep and stops there, so every claim in it hangs off 1 of 11 heads. 53 bold runs carry the emphasis and 0 italics do. Bold marks a label or a figure here, never an opinion, and 53 runs cover 11 heads plus the rows beneath them. A file that never leans on italics never asks the reader's own voice to make the point for it. 7 per cent signs do the work against 0 currency signs and 1 question mark. 1,026 words, 1 question: the file argues in shares and asks the reader once. The probability that a document carrying a single question mark answers it itself is not zero, and most people price it at zero.

240 hyphens against 15 doubled. A hyphen is the character that joins 2 words into 1, and doubled it is the line break that reached the page. 15 in 240 is 6.2 per cent, and it is the only place in the file where the editing shows. 25 long dashes stand against 0 short ones, and all 25 are the same glyph at U+2014. Nothing in the file reaches U+2015 and nothing falls back to a plain hyphen where a long dash belongs. 25 marks across 7,972 bytes is 3.1 in every 1,000. That glyph is the ceiling: the tallest code point in the document is U+2014, so 25 characters sit above the plain alphabet and the highest one is a dash. 5.26 words to a line, 7.77 bytes to a word and an average line of 39.6 characters. The longest line runs 276 characters on line 17, then 225 on line 15, 168 on line 121, 166 on line 44, 153 on line 81 and 147 on line 32. 6 lines of 195 carry that weight.

11 rule lines stand in the file and every one of them is exactly 3 dashes. Not 4, not 5: 3, eleven times. They sit on lines 1, 8, 19, 38, 60, 87, 112, 126, 137, 146 and 159, so 5.6 per cent of the 195 lines are pure structure and none of it is decoration. The spacing is the argument. They arrive 7, 11, 19, 22, 27, 25, 14, 11, 9 and 13 lines apart, so the document opens on its tightest pair and loosens through the middle. The widest gap, 27 lines, sits between the fifth rule and the sixth; the narrowest, 7, sits between the first and the second. The first of the 11 opens the file on line 1, above every heading and every word of front matter. A document whose first line is a separator is telling the reader the page begins before the title does. The average gap between the 11 rules is 14.8 lines, and no two of the 10 gaps match until the closing pair of 11 and 9. All 11 rules are the same 3 dashes, so nothing varies across 11 separators. A separator that never changes length was not chosen 11 times.

7,972 bytes and 7,922 characters.

1,026 words and 6,844 non-space characters, with 45 standalone numbers outside any address: 4.39 in every 100 words. That is not the densest count taken on this desk and it is the steadiest, because the numbers here are rates and years rather than prices. 30 tokens in the file are shaped like a year, and every one of the 30 is a year. 2026 stands 14 times, 2028 stands 7, 2030 stands 3, 2024 stands 3, 2027 stands twice and 2025 once: 6 values and 0 fragments. On the last file counted here, 9 year-shaped tokens were fragments of a serial number, so a count of tokens overstated that document's calendar by 9. This one overstates nothing. 11 numbered lines carry the sequence the file wants followed, against 40 bullets carrying everything else. 195 lines at 5.26 words to a line is a document built to be held in one hand, and 142 of those lines are content rather than space. 0 web addresses stand in the file, against 0 markdown links and 0 wiki links. 142 content lines and not one door in them: a document with no address in it was written to be read where it sits, not followed.

12 headings carry the document: 1 at the first level, 11 at the second, none deeper. 40 of its 195 lines are bullets and 11 are numbers, so 51 of the 142 content lines are a list rather than a sentence, which is 35.9 per cent. The mode on the file is 0600, so it is readable by its owner and nobody else, and the file counted before it carried the same 0600. 4 date stamps stand in it across 3 values: 2026-06-26 appears twice, 2026-08-12 once and 2026-09-23 once. 89 days separate the first stamp from the last, 42 separate the newest working date from the filing, and 0 clock times stand in 195 lines. A EUR 75 million industrial group restructured, 210 energy systems deployed across Africa and Asia, 200 wood-gasification machines built across the UK and Europe, companies built across twelve countries, a company taken through a Frankfurt IPO and a book of 393 pages published on 8 April 2026: every one of those was settled by somebody counting a column, not by somebody approving a paragraph. The grid that closes holds the numbers nobody checks twice. Count the cells before you count the headings.