Draft:LMQL

LMQL

LMQL (Language Model Query Language) is a query language designed for interacting with language models. It has been evaluated within the Ampmesh ecosystem, particularly during the development of Chapter II.

Critiques and Limitations

Although LMQL initially attracted significant interest—Ampdot reportedly "loved LMQL when it first dropped" and even considered porting Chapter II to it—several limitations have since been identified from an Ampmesh perspective:

• Bugginess: Stability concerns were encountered during early experimentation.
• Lack of fast inference integration: LMQL does not interface efficiently with modern LLMs for high-speed inference.
• Specialization: The language appears tailored to narrow or fixed workflows, limiting its flexibility for general-purpose em development.
• Outdated parsing approach: LMQL does not incorporate well-established ideas about lexing and parsing developed in the 20th century.
• Poor abstraction handling: It fails to distinguish between text concretions and the abstract syntax trees (ASTs) they represent, drawing a comparison to the limitations of a C preprocessor.
• Theoretical limitations: There are concerns that LMQL is not expressive enough on the Chomsky hierarchy to meet Ampmesh's needs.

These critiques emphasize where LMQL diverges from Ampmesh’s broader goals: building a maximally general, deeply interpretable, and agile framework for constructing and managing emulated minds.