Six Tables And Three Hundred And Seventy Six Glyphs

6 tables stand in 203 lines and 34 of those lines are grid, which is 16.7 per cent of the file and 13.3 per cent of its characters. One part in six of this document is a box rather than a sentence. The blocks open on lines 37, 47, 57, 80, 110 and 146 and close on 41, 53, 62, 85, 115 and 149. Inside them sit 22 data rows holding 62 cells. 2 of the 62 are empty, so the grid closes at 96.8 per cent. 3 of the grids are 4 columns wide and 3 are 2 columns wide. 4 columns is the shape of a comparison; 2 columns is the shape of a definition. No table in the file runs 3 wide or 6 wide. 2 cells left open in 62 is small enough to look accidental and large enough to be a decision. An empty cell is a question a reader will ask out loud, and this writer left 2 of them to be asked. Count the header rows in and the file carries 28 rows and 80 cells, 3 header cells empty. The sixth grid's header row is empty in both columns, which is why a flat line count reads 27 rows and misses the 22 that carry data. A wrong count is usually a right count of the wrong thing.

12,759 bytes and 12,306 characters. The 453 bytes between the two counts are 376 characters above the plain alphabet, carried in 26 glyphs. 299 of those 376 weigh 2 bytes and 77 weigh 3. None weighs 4. 299 plus 154 is 453, which is the whole gap between the byte count and the character count, to the byte. 376 out of 12,306 is 3.1 per cent, and the top 3 glyphs alone carry 160 of them, which is 42.6 per cent. The tallest code point in the file is U+00FC. Nothing in 12,759 bytes reaches U+0100, so no glyph in the document needs a font a stranger might not have. Every character above the plain alphabet is a render risk on someone else's machine, and 26 of them is 26 chances to lose a line of a design file. The probability that 2 counts taken off the same file disagree is 1. The only question is whether the writer can name the difference, and here the difference has a name: 376 characters, 26 glyphs, 453 bytes, none of them at four. The file carries 235 hyphens, 21 of them doubled, against 14 long dashes and 15 short ones. 14 long dashes in 203 lines is punctuation used as a pause, not as decoration.

26 glyphs above the alphabet, and the list is not flat. One glyph stands 66 times, then 57, then 37, then 37, then 36. 4 glyphs carry 197 of the 376 occurrences, which is 52.4 per cent of the foreign alphabet. A single character appearing 66 times in 1,732 words is a habit, not an accident. It is the letter the writer reaches for in the word they use most, and the file read cold tells you which word that is before you read a sentence of it. The second most frequent mark in 12,759 bytes is not a letter at all. It sits at U+00AD, a soft hyphen, and it stands 57 times while rendering as nothing in most readers. Nobody typed 57 invisible marks on purpose. They arrived inside the words as the file passed between editors. 13 of the 26 glyphs appear 5 times or fewer and 7 appear exactly once. The long tail of a character set is where a document quietly breaks: a symbol used once is a symbol no reader has taught themselves to expect. 0 web addresses stand in the file and 9 inward references do. A document written for people who already share a shelf with it never has to explain where it came from.

16 heading lines: 1 at the first level, 12 at the second, 3 at the third and 0 below it. 12 of the 16 sit at the same weight, so the document is a list of 12 things rather than a tree. 203 lines on the raw split, 58 of them blank, so 145 carry content and 139 of those are distinct. 6 lines repeat an earlier line, which is what a working document looks like when it is edited in place instead of rewritten. 8.53 words to a line, 7.37 bytes to a word and an average line of 59.6 characters. The longest line runs 416 characters and sits on line 15; the next two run 348 and 290. 3 lines in 203 carry that weight and the rest hold to something a reader can take in without scrolling sideways. 3 rule lines of 3 or more dashes stand on lines 1, 11 and 19, so 1.5 per cent of the file is pure structure and all 3 of them fence the first 19 lines. The head of the document is a summary before it is an argument. 45 bullet lines run against 9 numbered ones and 0 checkboxes stand anywhere in it. 5 bullets for every numbered step is a document that lists more often than it sequences, and 0 checkboxes means no item in it was reduced to a tick a machine could record.

12,759 bytes and 12,306 characters.

214 standalone numbers stand in 1,732 words, which is 12.36 in every 100 words. Density here is the argument, not a side effect. 0 web addresses and 0 outward links. In 1,732 words there is not 1 citation a reader could follow to check the writer's work, so every figure has to stand on its own. The mode on the file is 0644, so everyone who shares the folder can read it. A rejected design kept readable by default is either generosity or a document that already lost its argument about secrecy. The same discipline shows up elsewhere in the work. A EUR 75 million industrial group restructured is a room full of numbers that had to agree. 210 energy systems across Africa and Asia, 200 wood-gasification machines in the UK and Europe, companies built across twelve countries and a Frankfurt IPO - each of those was decided by somebody re-running a figure, not by somebody liking a phrase. A book of 393 pages published on 8 April 2026 runs the same rule at a different size. The pages worth reading are the ones carrying a number a reader can check behind you.

6 date stamps in 3 distinct values: 2026-04-24, 2026-04-25 and 2026-09-23. 152 days separate the first stamp from the last, and 0 clock times stand in the body, so the file reads as an interval rather than a timeline. 3 distinct dates across 6 occurrences means the writer stamped the work when it moved and left it alone when it did not. Nothing in the file says what happened in the 152 days between them. That absence is the honest part: 12,759 bytes of measurement and 0 lines about why the design was dropped. 17 per cent signs and 1 currency sign do the arithmetic, and 2 question marks against 0 exclamation marks sit where the writer stops reporting and starts insisting. The probability that a rejected design is read again is close to 0. The probability that it is read again because somebody kept its arithmetic intact is much higher, and 22 rows in 62 cells with only 2 left open is how you buy that second reading. Count the cells before you count the words. Then leave the 2 gaps where a reader can see them.