Wikis LLM: Why a Pure Python Compiler is More Efficient

Wikis based on large language models (LLMs) are often criticized for their excessive complexity. They typically rely on agents, embeddings, and repeated calls to models to organize local notes. However, a simpler and more deterministic alternative exists: a pure Python compiler.

This compiler transforms disorganized markdown into a structured and verified wiki, using only Python's standard library. This method avoids the unnecessary sophistication of LLM systems and offers a more straightforward solution for the mechanical organization of text.

During the creation of this compiler, two real bugs were fixed, and performance was evaluated on two different operating systems. This experience demonstrated that, for certain tasks, a compiler can outperform an agent in terms of efficiency and simplicity.