Skip to content

Methodology

Research Pipeline

Each study in this series was produced through a multi-agent pipeline (bible-study5) with five phases:

  1. Reference Gathering — A reference agent searches for prior studies in the library, reads completed series studies, and queries external corpora (Ellen G. White writings, Secrets Unsealed) for research leads.

  2. Scope Discovery — A scoping agent uses tools (not training knowledge) to discover the study's scope: Nave's Topical Dictionary queries, Strong's concordance searches, and verse reference extraction. The output is a PROMPT.md file listing every verse, topic, and word to investigate.

  3. Library Generation — Missing word studies and grammar references are generated as canonical entries before research begins, ensuring the research agent has comprehensive data.

  4. Data Gathering — A research agent retrieves full verse text with chapter context, runs cross-testament parallels in both directions, parses Hebrew and Greek morphology, and organizes all data into structured files. The research agent does NOT interpret — it only gathers.

  5. Analysis — An analysis agent reads the clean research files and produces verse-by-verse analysis, pattern identification, and a comprehensive conclusion. Every claim requires verse support.

Interpretive Principles

  • Scripture interprets Scripture — meaning is derived from how the Bible itself uses terms, not from external theological frameworks
  • Contextual analysis — expanding circles: verse → chapter → book → same author → cross-testament
  • Historical-grammatical method — Hebrew and Greek grammar inform meaning; context must match before importing meaning across passages
  • Scripture is the only authority for doctrine — external corpora (EGW, church fathers) are research leads, not evidence
  • Consistency — answers must be consistent with ALL analyzed verses, not just selected proof-texts

Series Reference Boundary

Each study's final analysis and conclusion may ONLY cite studies completed earlier in the series. This prevents:

  • Circular reasoning — Study 18 cannot cite Study 23's synthesis to support its own claims
  • Anachronistic reading — each study stands on its own evidence at the time it was written
  • Cherry-picking — the series builds progressively, not retrospectively

The reference agent, scoping agent, and research agent may use the full study library. The restriction applies ONLY to 03-analysis.md and CONCLUSION.md.

Tools Used

All tools are local Python scripts running against local databases:

  • Nave's Topical Dictionary — 5,319 topics with verse references
  • Strong's Concordance — Hebrew, Aramaic, and Greek word database with translation distributions
  • Hebrew Parser (Text-Fabric/BHSA) — morphological parsing, clause structure, construct chains
  • Greek Parser (Text-Fabric/N1904) — morphological parsing, tense/voice/mood analysis
  • Cross-Testament Parallels — hybrid scoring (semantic + keyword + theological phrase matching)
  • Concept Context — finds verses with same theological concepts, prioritized by proximity
  • Grammar Reference Search — semantic search across 10 Hebrew/Greek grammar textbooks
  • Word Studies Library — 500+ canonical word studies by Strong's number
  • Grammar Reference Library — Hebrew stems, Greek tenses, passage analyses

No internet searches were used for biblical evidence. Web resources (Blue Letter Bible) supplement lexicon data only.