Phonology

Taught on the front page and in Lesson 01.

Alphabet

Twenty letters, each with exactly one sound. No digraphs, no diacritics.

c is always the sound of chai, church — as in ca (tea), cok (many), keci (small). Never k, never s. This follows Indonesian, Malay, Hausa and Somali, which all write that sound with a plain c. (Until September 2, 2026 it was written ch; the digraph was dropped because it was the alphabet’s one exception to “one letter, one sound”.)

y is always a consonant glide (ya, yu), never a vowel.

One letter, one sound

There are no silent letters, no letters that change value by position, and no letter that can be read two ways. A word is pronounced exactly as it is written, and written exactly as it is pronounced.

Syllable shape

Syllables stay simple: never more than two consonants in a row, and none of the tongue-twisters other languages allow (strengths, vzglyad). If you can say “banana” and “taksi”, you can say anything in Amadunia.

Two-consonant sequences that occur in current vocabulary:

• tr — tri (3)
• ks — taksi (taxi), across the syllable break tak-si
• kt — doktor, across the syllable break dok-tor
• rb — corba (soup)
• rg — harga (price)
• rk — porke (why), market, across the syllable break por-ke
• rs — mersi (thank you), across the syllable break mer-si
• rd — pardon, across the syllable break par-don
• ns — insan (person), across the syllable break in-san
• pl — plis (please), word-initial like tr
• fr — fruta (fruit), word-initial like tr
• mb — ambil (to take), across the syllable break am-bil
• mr — kamra (room), across the syllable break kam-ra
• nd — anda (to walk), across the syllable break an-da
• ng — angin (wind), cang (long)
• bl — problema
• br — libro (book), across the syllable break li-bro
• gr — grupo (group), word-initial
• hr — cehra (face)
• ht — anahtar (key)
• lc — dulce (sweet)
• fl — flor (flower), word-initial
• ft — hafta (week)
• lb — kalb (heart)
• lt — kultura
• mp — tempo
• nc — pencere (window)
• nk — banka
• nl — yanlis (wrong)
• nt — kanta (to sing)
• ny — nyama (meat), tanya (to ask), punya (to have)
• pr — problema
• rp — korpo (body)
• rt — start
• sf — asfar (yellow)
• sk — eski (old), skola (school), across the syllable break and word-initial
• sm — asman (sky)
• sp — espera (to wait), across the syllable break es-pera
• st — studi (to learn), word-initial like tr and pl

No two words a sound apart

Settled: no two roots of the same length may differ in exactly one letter. Between l and r the ban is absolute and outranks everything.

One letter is one sound here, so a minimal pair is a pair of words a single sound apart. beli and beri would be such a pair, and telling them apart is the hardest contrast on Earth for well over a billion speakers — Japanese, Korean, and much of southern China and East Africa. A world language cannot rest a distinction on it.

This is the rule that kills most candidates. beli (to buy) fell to beri, guru to suru, luna to luma, kota to kita, buku to buka. It also outranks how widely a word travels: ruma, a house, was in use eighty-four times before it was replaced by dom, because ruma stood against luma.

The eighty-four was recounted on September 5, 2026 and it is exact. The repository as it stood at the commit before the rename holds eighty-four occurrences of ruma, spread across twenty-one files — ten of them in Lesson 06 alone.

The twenty-four that were already here

The rule was adopted after the language had started, so twenty-four pairs predate it and are exempt. They are listed in check.py, which rejects any pair that is not among them.

What the exempt pairs have in common is not where they came from but how short they are: 23 of the 24 are under four letters — ba/ca, go/no, hi/mi, kan/pan, pat/rat — and mama/nama is the only exception. Short words cannot avoid each other; there are not enough of them. That is the same fact that closed the space below four letters in CONTRIBUTING rule 2, and the two rules are really one observation seen from two sides.

Not one of the twenty-four is l against r, and check.py tests the exempt list for that too — an exemption may excuse a pair, never that pair.

Vowel sequences

Settled: these five sequences and no others, and never three vowels in a row. Both halves are rules, not observations — the page listed them as “observed in current vocabulary” while check.py rejected anything else, and until September 3, 2026 the checker enforced only the first half. kuai is legal as two pairs, ua then ai, and illegal as a run of three; it is the example CONTRIBUTING rule 4 names.

The ua row read uan, una until September 5, 2026. una has no ua in it — u-n-a — so the table that settles which vowel sequences exist illustrated one of them with a word that does not contain it. It was the only one of the nine in this table that failed, and it had been there since the table was written. Every example on this page — in this table and in the consonant list above — is checked against the sequence it illustrates now.

Open questions

• Whether a vowel pair is one syllable or two — settled in stress.md on September 3, 2026, together with the stress rule that needed it: a vowel pair is one nucleus, so dunia is two syllables and familia three. Counting a syllable means counting vowel groups. It affected 28 of the 300 roots.
• Whether stress is fixed (and if so, on which syllable) — settled in stress.md on September 3, 2026: the second-to-last syllable of every word. This bullet outlived the decision by a day, and the day is the point: the briefing it points at had said Decided at the top of it the whole time.
• Whether the two-consonant sequences above are the complete permitted set, or merely the ones used so far, is not yet decided.