What the game is called, part two: Heimir for a few hours, then Audamir and four realms

Article 18 ended with the claim that nothing had changed and the game was still called Auda Wars. That sentence held only until the morning of 31 October 2026. According to the commit times, the same day brought these decisions: the realm Taflgard is called Rokgard (with the saved id too), skeletons of two more realms, Vitgard and Telgard, were made, the game was renamed Heimir for a few hours and in the afternoon Audamir (until 31 October Auda Wars), and the hero is called Auda. It turned out that the second renaming that day was small, because the first one had set up one place for the name of the game and one test that guards it.

One day by the commit times

The times in the table are from the git log of the game on 31 October 2026. They are the times of the commits, not the moments when a decision was made.

Heimir was in the development version of the game from the merge at 11:22 until the merge of the fix at 16:12, so a little under five hours. In the renaming that was in progress, Audamir had replaced it already at 14:39.

The game and the hero: three variants and the choice against the recommendation

The second round ended with a question: does the hero stay Auda and the game get another name, or will the game and the hero share one name? Section 8.5 of the document docs/NAME.md compared three variants.

The cost of changing the hero was low, and the reason is in the code. Hero.DEFAULT_NAME is a constant, and setName deletes the key hero.name when the name equals the default (js/core/hero.js, checked 1 November 2026). So the default name is never saved: after the constant changes, every player who has not named the hero sees the new name, and whoever named the hero keeps it. The gender of the name broke nothing: gender is a player setting, and the narrator's first-person lines (5 in Slovak, 5 in Czech, 4 in Serbian and 5 in Russian, according to NAME 8.5) follow it, not the name. So the change of the hero meant one constant, one list of names that stay in Latin letters in Serbian Cyrillic, and 47 lines of tests in 15 files. No migration.

Fourth addendum: the family of four realms

Article 18 describes that the third realm (numbers and logic) had the best name in the rough filter, Vitgard. My answer from that morning, however, went its own way: Vitgard appealed to me for Pexeso, for the realm of number games and mazes my proposal was Telgard, and Rokgard seemed easier to remember than Taflgard (NAME 4.4, fourth addendum). The plan with Pexeso supported it: the realm for Pexeso is already reserved in the analysis of the board kit (step B12; see the article on the kit); it only lacked a name.

The distances (Levenshtein) were recalculated while this article was written, and they agree with the document: no two names are closer than three letters, and each has its own first letter and vowel (K-a, R-o, T-e, V-i). With Taflgard the pairs were worse: Telgard differed from it by two letters, Kastgard by three.

Telgard has two close neighbours that the earlier rounds did not search for: Telegard (a program for electronic bulletin boards from the late 1980s and the 1990s, one letter more) and Telengard (a dungeon crawl game, so about a maze, two letters more). The document found them in the search results, and read only the summaries, not the pages; the marks of neither were read. Nothing called Telgard was found.

Taflgard was measured in the document for the first time: all 8 domains were free, and there was no title in the App Store. Steam, itch.io and Google Play were not read for it. Against Rokgard the document states that tafl is the only root in the family that names a genre (board games), that rök means "reasoning, arguments" in Icelandic and "smoke" in Swedish, and that "rok" is a year in Slovak and Czech. The last point is a general judgement in the document, not a verified source. The conclusion of the document: measurements do not decide it; my ear does.

Rokgard: the renaming with the saved id too

The document estimated the cost of changing only the displayed name at **19 lines in js/**, a line of one tool, two tests and two fixtures. The saved id taflgard should have stayed frozen (R3-REALM-ECONOMY, D11). After the work began, my decision cancelled that: the id is to be renamed too, because it is "not yet tied to anything", so the word should remain neither in the code, nor in the keys, nor in the translations (DESIGN §102.1). NAME.md records the addendum at 9:27. The first two commits of this work (10:16 and 10:19) still kept the id, and the third (10:43) changed it. The id had been used in the code since 28 October, so for three days (header of tests/_support/retired-names.js). The real scope was different from the estimate:

The work went in four commits, each of which could be reverted on its own: the tool (9:27), the folders, packages and keys (10:16, without a visible change), the name on the screen (10:19) and the saved id (10:43). One more commit with the tests that the whole suite found (10:54) followed. The tool tools/i18n-rename.js learned the switch --prefix OLD NEW and a map that is read to the end: the key colony.kicker was first renamed to taflgard.colony.kicker and then to rokgard.colony.kicker, so work that still wrote the oldest name received the newest. Proof: in 9 of 9 languages, the merged catalogue after the renaming has the same 3,788 keys and the same texts as the catalogue before the change.

What happens to the data of the old id. The document did not read this from the code, but measured it. The profile and the games in progress were made by the old code (the state of the game from 31 October at 9:20) and loaded by the new one. A profile with realms.<old id> loads without an error and stays as it was. The code no longer reads it, so the level, experience and gold tokens of the realm start from zero, while the history of games, the games in progress and the backups keep working. The cost is therefore written down: a player who played Colony or Lines before this change would lose the level and the tokens (DESIGN §102.4). The game had no public address or shop at that time (NAME 1.5).

The guard against a return is tests/core/retired-names.test.js: in js/, tests/, tools/, the page, sw.js, the manifest, package.json and the README, the old word must not stand either in the text or in a file name. A helper that knows the word assembles it from two pieces (['taf', 'lgard'].join('')), so that the guard does not find it either. While this article was written, it was checked: grep -rIil over these places found nothing, while the word as history remains in 31 files in docs/.

Heimir: a renaming that could be reverted

The work on the name Heimir began with step N1 of the plan in NAME.md (part 5): the name of the game stands in one place, in Meta.name. Four fallback literals in the code (the subject of the bug report, the information in the game snapshot, the text "Copy result" and the picture title) read it. The guard tests/data/meta.test.js goes through the files js/, tools/ and the root, looks for the current and the former names of the game, and the places where a literal may stand are one table with a count for each. "The next renaming is one value, the texts and the rows of that table," writes DESIGN §103.1.

The text step N2 changed six texts in nine languages, that is 54 values. The name is never declined, so it is written as "the game Heimir" (in Slovak "Vitaj v hre Heimir!", in Russian "Добро пожаловать в игру Heimir!"). Printing the name on the picture raised Picture.VERSION from 3 to 4, and 174 pictures of the golden cases matched byte for byte after the name and the version number had been replaced (174 of 174, 522 occurrences of the name).

The reason for the withdrawal at 14:15 was written in the document in my sentence: Heimir is hard to pronounce, so I am going back to the original Auda and the ending mir (NAME 9.1, 31 October 2026). The document had pointed to the weakness already, but as a judgement, not a measurement: an English, a Slovak and a German reader pronounce the first syllable in three ways (hej-, hay-, high-; NAME 8.4). The criterion from article 18 applies again: the name is tested aloud.

The return was small, and DESIGN §108 says why: because chapter 103 set up one place for the name and one guard, this step is small. What was left: 54 values, a few constants and the rows of one table.

• Meta.name is Audamir and Hero.DEFAULT_NAME is Auda (checked in js/data/meta.js and js/core/hero.js); the test in tests/data/meta.test.js states that they are two separate constants.
• The spelling Audamir was chosen by the main session with the assistant, from two forms in my note (Audamir or AudaMir): in one word it reads as a proper name and does not take a case ending. AudaMir remains possible; it costs one constant, one test and one run of the picture tool.
• Picture.VERSION 4 → 5 (the comment in the code: version 4 is "the name for a few hours that was never published"). The same proof: 174 of 174 pictures were the same after the name was replaced; picture-golden.json has version 5 and 174 cases (checked).
• Entry 0.67 in the changelog was corrected, not added: the renaming replaced the renaming of the same day.
• French could not be replaced blindly: "de Heimir" (with the letter h) becomes d’Audamir. This was found by reading the catalogue, not by a failing test, because until then no test said that French should elide. The guard got a new check.
• The data did not move: the default name of the hero is not saved, so the return to Auda changed no saved byte (DESIGN §108.2 lists five cases for the old and the new tree).
• There were 3,455 tests (3,454 passed, 1 skipped), three were renamed, none disappeared; 11 of 11 mutations (deliberate breakings of a copy of the tree) were caught.

Audamir: what was measured about the word

Part 9 of NAME.md measures the word like the other rounds. The table puts the three names of the day side by side; the figures are from the document and from 31 October 2026.

This needs care. The word Audamir is clean everywhere that was asked, but Adamir is registered in class 28, which includes games. The document marks it as a registration that a lawyer should look at. WIPO also found four records of the string audami, which Audamir contains; the document names three: Germany (class 41, ended), the USA (class 8) and Brazil (classes 35 and 25, 2026). The question of the Audi mark remains: Audamir is three letters away from Audi (Auda is one, recalculated), which by the document's account is a judgement, not a legal fact.

Reading in the game's languages is, according to the document, a judgement without native speakers. The ending -mir is the Slavic "mier, world" (Vladimir, Stanimir); the Germanic twin of the root, Audamar, gave rise to the names Ademar and Otmar. A German reader may split the word into "Auda mir", which the document calls harmless. The document's reservation: Audamir is read as a given name (Audemir, Ademir, Adamir), and that is the name of the hero, not of the game.

Vitgard and Telgard: realms without a game

My assignment from 31 October (DESIGN §104): create the basis for two new realms, add to them gradually, and let them work in parallel. A skeleton is a package js/realms/<id>/ with three code files and nine text files (12 files in each of the two folders). The manifest realm.js has 36 and 37 lines, the table of modes modes.js has 23, and the part of the profile and economy slices.js had 122 lines at first (now 126).

The realm manager learned three things: rulesets may be an empty list (before, 1 to 8), home.gate may be null while the realm has no tile, and a card may carry a line home.text. The rule that hides the skeletons is one condition in js/core/modetable.js: visible are the realms whose rulesets.length > 0. Cards, hubs and lists in windows follow from that, so nothing new appears in a player's saved layout. Proof from §104.8: the home screen and the hubs of Kastgard and Rokgard had the same HTML in the old and the new tree, and one snapshot of each was the same pixel by pixel.

The provisional coins are Golden cards (Vitgard) and Counting stones (Telgard), and the realm icons; I am to confirm them (the note in the source of the manifest says so). None of these realms has a wardrobe, a hero voice, an achievement or a row in the reward table: an invented number is worse than a missing one.

The real output of the step is the list of places that the first game of a new realm will touch: 20 lines, from the table of packages through index.html and sw.js to the migration tools, and a checklist of eight steps. Some lines are marked as shared, because every game edits them, and the joining of the edits of two realms is handled by the main session. Check of the step: 3,387 tests, 3,386 passed, 1 skipped, 21 new tests, 9 mutations caught. No game of Vitgard or Telgard has started; both folders contain only those 12 files.

What could not be verified and what remains

• Trademark. The documents contain no assessment of the mark Audamir, of the registration Adamir in class 28, or of the question of Audi, and no lawyer's report; they record it as open work (step N8). TMview and WIPO are not registers and have no legal effect. TMview stopped answering in the second round, so some names were not read there.
• Domains. Free means "not registered on 31 October 2026". The documents do not record a purchase or a reservation, and prices were not checked (step N7).
• Neighbours of Telgard. The pages about Telegard and Telengard were not opened.

Sources and notes

• Decisions and measurements: docs/NAME.md (banner of 31 October 2026, parts 4.4, 8 and 9), docs/DESIGN.md §102 to §108, docs/VITGARD-TELGARD.md, and a note at D11 in docs/R3-REALM-ECONOMY.md. The figures on domains and shops are from 31 October 2026 and were not measured again.
• Times: the git log of the game on 31 October 2026; these are the times of the commits, not of the moment of the decision.
• Checked on 1 November 2026 in the repository (version 0.81.2): names in the code, pictures, the skeletons of the realms, and the counts of lines and files. The other figures are from the documents.
• Earlier articles of the series: What the game is called, From one game, more games, A move that must change nothing, Two games on one board.

What could not be verified is marked in the text above. The legal assessment of the name Audamir, the purchase of a domain and the registration of a mark are neither part of the article nor, according to the documents, settled.