'The Order in the Horse's Heart': A Case Study in LLM-Assisted Stylometry for the Discovery of Biblical Allusion in Modern Literary Fiction

The paper's deeper contribution isn't just finding allusions in McCarthy, it's stress-testing whether LLMs can perform interpretive literary reasoning reliably enough to be trusted as research instruments, not just search tools. The validation step against published scholarship is doing real methodological work here, because without it the pipeline would have no external anchor.

This connects directly to the reliability questions raised in our April 16 coverage of 'Diagnosing LLM Judge Reliability,' which found that roughly one-third to two-thirds of documents show logical inconsistencies when LLMs evaluate text pairwise. That paper was about automated scoring pipelines, but the underlying problem is the same: when an LLM is asked to make a judgment call about language, how much do you trust the output? The stylometry paper's two-stage design, combining rule-based rare-word matching with open-ended LLM reading, looks like a practical hedge against exactly that failure mode. The humanities application is largely disconnected from the coding and agentic stories we've covered recently, but it sits squarely in the emerging question of where LLM judgment is reliable enough to substitute for expert human reading.

Watch whether other literary corpora with well-documented source texts (Milton's influence on Romantic poetry, say) get the same pipeline treatment in the next year. Replication across multiple authors and source texts would tell us whether this method generalizes or is tuned specifically to McCarthy's unusually distinctive biblical register.