We released the SEMQ whitepaper

Today we're publishing the first SEMQ whitepaper: a technical overview of the core ideas behind Semantic Quantization and the intermediate representation (IR) that enables stable, symbolic representations of semantic information.

The document explains the problem SEMQ addresses, its high-level architecture, and the specification required to implement the encoder/decoder pipeline consistently. It introduces the symbolic angular domain, the semantic invariants it preserves, and a framework for reasoning about representation quality beyond raw numeric reconstruction of FP32 embeddings.

Rather than focusing on vectors as numerical artifacts, the whitepaper frames SEMQ as a representation layer: one that allows semantic information to be stored, compared, transmitted, and reasoned about in a compact and time-stable form. This makes it relevant not only for retrieval and indexing, but for agent memory, multimodal systems, semantic routing, and long-term AI infrastructure.