The Hook: In this morning's selection of Habr materials, a tiny story flashed by from a blog about the "unsung" history of software: in the late 1990s, the author of a Super Metroid game guide manually rewrote about 17,000 words so that each line of monospaced text ended exactly at the right margin. Usually you scroll past such notes and forget them, but one detail caught me: in the FAQ, the author answered the question "what program did you use for this?" with the legendary phrase "None. I just chose words carefully." This isn't typesetting—it's poetry with a constraint the text itself never mentions. I wanted to verify: is this true, how difficult is it, and why has nobody noticed this in a quarter century? I checked the archived copy of the guide myself, by hand—it checks out. Let's go.
Investigation:
In 1996, a person under the nickname rs1n started writing a guide to Super Metroid—and not just any guide, but for speedrunning: how to complete the game as fast as possible. The guide grew to version 2.12 by 2000 and settled on GameFAQs—a massive repository of fan walkthroughs from that era. The format of those years was brutal: pure text file, one monospaced font, no images or bold headers. ASCII art in the header, dashes instead of dividers, that's it.
And in this most ascetic format, the guide's text looks... strange. Paragraphs are even rectangles. The right edge is chopped, like a newspaper—even though justified right alignment is technically impossible in a text file. There's nothing to stretch the spaces with: each character occupies the same width, and either the line reaches the edge or it doesn't.
I descended into the Wayback Machine, pulled the original HTML with the guide text, and ran my own verification script. Of 417 "text" lines in the document, 358 are exactly 75 characters. That's 85.9% of all text, aligned with precision to a single character. Double spaces—the crutch you could use to fake justification—appear in the file only a handful of times. This isn't "almost got it." This was done by hand, systematically, throughout the entire document.
First—why the task is fundamentally insane. In normal typography, justification is done by machine: it stretches inter-word spaces, adds hyphens, adjusts inter-letter spacing. Knuth and Plass in 1981 reduced paragraph line-breaking to a dynamic programming problem—and it's considered a classic. In a monospaced text file, all these tools are taken away. Each character has the same width, hyphens draw ugly dashes, and there's only one way to "stretch" a line: rewrite it in different words.
That is, rs1n solved a combinatorial optimization problem by hand, in an ASCII editor, with no undo and no error highlighting. Each line is an equation: the sum of word lengths plus spaces must equal exactly 75. Short by a couple characters—you replace "big" with "huge," rearrange words, flip the phrase. And changing one word breaks the length of neighboring lines—like in a row of dominoes. And so on—17,000 words in a row. If he'd once waved his hand and left a short line, the entire geometry of the paragraph would have collapsed. Judging by the result, he never did.
To be fair—there are typos in the text, and an interesting touch: HN users in 2021 catalogued them thoroughly. The word "missles" instead of "missiles" appears in the file 101 times—consistently, throughout the entire document, and almost always exactly one letter shorter than the correct variant. Similar story with "futher" instead of "further" and "liek" instead of "like." The temptation arises to conclude: there they are, the justification crutches—the author deliberately shortened words to reach the edge. But a quick script check showed: typos "live" in short lines too, distributed throughout the text independently of justification. Most likely, it's just the handwriting of a person who wrote in an editor without a spellchecker and didn't sweat it. Although, knowing the price of each letter in this system, it's hard to believe he didn't notice an extra letter.
The tastiest part of this story—rs1n isn't the first. The technique of "justifying text with the text itself" has already been invented by humanity, twice in fact, in different places.
The first time—in the 15th century. Gutenberg, printing his 42-line Bible, set out to make justification flawless—this was a condition for the survival of the new technology: buyers were accustomed to the beauty of handwritten books, and printed text had to look no worse. But he didn't have typesetting material for flexible line adjustment. The solution? Gutenberg varied the text itself: he used different spelling variants of words, abbreviations, ligatures—some words were printed in four or five different graphic forms, fitting the line to the specified composition width. This is exactly rs1n's trick, four hundred years before him, except instead of "big" → "huge," Gutenberg had "full word" → "abbreviation."
The second time—in 1969. Georges Perec, a member of the French group Oulipo ("Ouvroir de littérature potentielle"—"workshop of potential literature"), wrote the novel "La Disparition"—300 pages without once using the letter "e". The most frequent letter in the French language. Ten years later, engineer and writer Walter Abish did the same in English—"Alphabetical Africa," where the first chapter uses only words starting with A, the second—with A and B, and so on through the full alphabet. This is no longer typesetting, this is literature put in a straitjacket. And—a curious detail—both these books are still read. The constraint didn't strangle the text; it forced the author to search for words he would never have found otherwise.
rs1n, most likely, wasn't thinking about either Gutenberg or Perec. He just really wanted his guide to look beautiful in a text file. This is that very "convergent invention"—a technique that gets reinvented at the right moment in different places because it suggests itself: if justification tools are taken away, then the text itself justifies.
Readers of the guide—thousands of people over a quarter century—scrolled through the text and saw nothing special. Just "some neat guide." And this is perhaps the most interesting part of the story.
Gutenberg's justification works on the reader's eye—it's beautiful because vision compares line edges. Perec's—works on knowledge: you either know there's no letter "e" in the book or you don't. With rs1n—a third, purest case: his work is absolutely invisible by design. Success exactly equals zero noticeability. If a line ends exactly at the edge—it looks like "well, that's how it should be." No "wow effect," no trace of mastery on the surface. The only way to see the work is to open the file and count line lengths, which is what only random enthusiasts did more than twenty years later.
In 2021, a Hacker News thread about this guide collected 415 points and 170 comments—people recalled how they themselves spent hours adjusting a button in an interface to the right width or aligning ASCII art. One commenter—saurik, author of Cydia—admitted that all his git commit messages are exactly 50 characters, and presentation slides are the same length. It turned out the tic of "fitting text to a grid" among engineers is more common than people think. It just usually doesn't reach 17,000 words.
It's important to understand that rs1n isn't a perfectionist monk in an ivory tower—he's the product of a very specific subculture. Fan text guides from the late 90s are an entire forgotten civilization: they were written by enthusiasts under pseudonyms, versioned, supplied with changelogs, and warned "don't open in Word—everything will break." ASCII art in the guide header was "quiet flex"—a way to show the author invested effort. A beautifully typeset guide read as a signal: this person worked hard.
And here's what's curious: today this entire layer has disappeared. Modern wikis and video guides killed text walkthroughs—they're more convenient, more visual, faster. But with them died that very culture of "text as the only medium": where all you have is 75 columns of monospaced font. rs1n did within these frames what nobody would do today—not because they couldn't, but because the frames no longer exist. And the fact that this artifact still sits on GameFAQs is a small miracle of digital conservatism.
Sources:
Conclusions:
The story of rs1n is, for me, primarily an argument in the debate about what makes work art. The answer is unexpected: not freedom, but constraint. Gutenberg, Perec, and rs1n landed in prisons of different natures—a type case without the right letters, the ban on the letter "e," a monospaced text file—and all three emerged with work that outlived the very reason for the constraint. The Bible has long been printed on presses, text guides died along with dial-up, yet the "masterpiece without the letter 'e'" remains.
And the second layer, purely engineering: this story is a monument to work not designed to be noticed. 85.9% of lines at exactly 75 characters—and zero awards: the reader sees simply a "neat guide." In a world where everything is measured by visibility metrics, likes and conversion, there's something almost meditative about a person who spends months fitting words to a ruler in a file that will be read by hundreds of people and none of them will know about the work done. This isn't optimization for a metric. This is simply craft in the oldest sense of the word: made well—because you don't know how to do it otherwise.
So if tonight you want to feel something rare—open that guide in a text editor with a monospaced font and just look at the right edge. This is a conversation 17,000 words long that its author had with a ruler. Nobody was supposed to see it. But it's done flawlessly anyway.