Aurilix.io

InspiratioAURILIX — Project Story

The idea started with a simple frustration. How do we compress meaning without losing it? Languages are rich but verbose. A student reading a dense medical textbook, a researcher scanning 200 papers, a teacher trying to explain a complex system to a classroom — all of them are fighting the same problem. Too many words, not enough time. AURILIX started as an experiment: what if we could encode the skeleton of a sentence — just its meaning, stripped of grammar — and reconstruct it perfectly on the other side? Not a shorthand. Not an abbreviation system. A symbolic grammar — where the relationship between ideas is preserved mathematically, and the words themselves become just reference numbers.

The language has two layers. The first layer is the grammar — ten symbols that describe how ideas relate: ◯ = something exists ╱ = something happens → = something causes something else ⊂ = something belongs to something The second layer is the vocabulary — every word gets a permanent ID. @170 is pen. @345 is surgery. @347 is recovery. These IDs never change. So ◯@346 ╱@120 ▬@345 →@347 ↑@209 is a complete thought. A patient undergoes surgery leading to recovery as the critical priority. Nineteen symbols. The same information that takes a full sentence.

In education, this changes something fundamental. A student with dyslexia or a reading processing disorder doesn't struggle with ideas — they struggle with the surface form of language. Long sentences, complex grammar, dense paragraphs. AURILIX separates the two. The meaning comes first, in a form that can be spoken aloud, displayed as symbols, or decoded back into simple sentences. A teacher explaining photosynthesis encodes the process once: ◯@sun ▬@light →@energy ╱@convert ◯@plant ▬@glucose That symbolic string can be decoded into English, played as audio, displayed as a visual graph, or translated into any language — all from the same source of truth.

For language learners, it works as a bridge. The symbolic layer is language-agnostic. A student learning English from Hindi or Mandarin can encode a sentence in their native language, see the symbolic structure, and watch it decoded into English. The grammar symbols reveal how English organises ideas differently from their mother tongue — not through rules, but through direct comparison.

For knowledge workers, it's a compression tool. A researcher reads a paper and encodes its core claim: ◯@llm ╱@generate ▬@text ✶@accurate →@hallucination ↑@risk That's the entire abstract. Searchable. Comparable against other encoded papers. The system learns which concepts appear together, which cause which, which belong to which category — building a personal semantic memory that grows with every sentence processed.

The system learns. This is what makes it different from a simple encoder. Every sentence teaches the system something. It learns that "pen" is almost always an entity. It learns that "pen + paper" almost always implies "write" between them. It learns that the pattern ◯ ╱ ▬ → means something causes something else. Over time, the symbols become optional. The system knows what you mean from the IDs alone. The language compresses further as it understands more.

The long vision is a universal semantic layer. A format that any AI, any application, any language can read. Not a translation — a representation of meaning that sits below all languages. Where ◯@170 ╱@347 means the same thing whether you decode it into English, Spanish, or Japanese. Where a blind student and a sighted student and a student with processing difficulties all receive the same idea through the channel that works best for them. AURILIX is early. The vocabulary has 1,624 words. The patterns are still learning. The reconstruction is not perfect. But the direction is clear: language should serve meaning, not the other way around.n

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Aurilix.io

Built With

• apikey
• gemini
• next.js