Maku — “one utterance, two readings”
407 words. Part of the Conlangs as Tools project, edition v1.2 (2026-09-27).
Brief drafted by Doruga (Pete’s mobile-companion agent), edited by Peter Kaminski, 2026-05-28.
♡ Copying is an act of love. Please copy and share.
© Peter Kaminski · CC-BY 4.0 (Creative Commons Attribution 4.0 International)
Maku brief · Maku deep dive → · project home · all files (.zip)
Languages: Toki Pona · Taso · Kipona · Ren · Maku · Sito · Olu · Vesi · Korē · Drosa · Hesi
What it explores
Toki Pona’s pervasive compounding (tomo telo = bathroom, jan pona = friend) looks, from a distance, suspiciously like what machines do: small primitives composed into larger structures. Maku asks: what if a language were designed so that every utterance is simultaneously natural human speech and a parseable machine call — no translation layer?
The lesson it tests: Lojban tried to be machine-parseable and stayed niche; Toki Pona is machine-friendly almost by accident. Maku looks for the middle — something humans will actually say, that machines can also act on.
Profile
Every Maku content word combines with a short role-marker syllable that names the word’s semantic role (agent, patient, instrument, location, time, manner, source, goal). The role-markers are euphonic — they sound like part of natural speech — but they parse uniquely.
A polite request and a structured tool-call are surface-identical. A rude request is one with missing role-markers — speakers feel ambiguity as incivility.
Maku is meant to be the contact language of an era where humans and machines talk to each other constantly. It is not a programming language; it is a way of speaking that respects both sides of the conversation.
Examples
Role markers:
- -bi (instrument), -ki (patient), -mi (agent), -tu (time), -lo (location), -na (manner), -su (source), -go (goal)
Sentence:
- “i bringnow you-ki book-ki by-hand-bi” → “I, now, am bringing you this book by hand.”
- Parsed:
bring(agent: i, patient: [you, book], time: now, instrument: hand)
Compare:
- “open-mi door-ki gentle-na” → “I open the door gently.”
- Parsed:
open(agent: i, patient: door, manner: gentle)
A user saying “Maku, set timer five minutes” is grammatical Maku and a valid tool call.
Future action research
- LLM-native training. Pretrain a small LLM on Maku-shaped data, then compare its tool-use accuracy and conversational naturalness to English-trained baselines. Does the structured shape help?
- Voice assistant trials. Deploy Maku as the contact language for a home assistant. User-experience study: do speakers adapt? Do error rates drop?
- Safety-critical domains. Pilots, surgeons, and air-traffic controllers already use controlled languages. A Maku adaptation might combine safety and naturalness. Compare incident rates.
- Comparison with Lojban. What does Maku gain by not being pure logic? Run head-to-head learnability studies. Hypothesis: Maku is easier to acquire because it preserves natural prosody.
- Agent-to-agent communication. Two AI agents talking — does Maku reduce ambiguity-budget compared to English? Measure tokens-to- resolution on cooperative tasks.