Posts tagged: parsing

The Mermaid Loop: Why LLMs Repeatedly Fail at Diagram Syntax

A technical post-mortem on why advanced LLMs repeatedly make the same Mermaid syntax errors, analyzing training priors, tokenization biases, and the lack of inline compiler feedback loops.