The Resource Gap

• Too few high-quality biblical resources in most languages
• The need far exceeds what experts can produce
• Western resources often aren’t the right fit

Lost in Translation

• Western systematic theology
• Abstract, academic, inaccessible
• Emotional exegesis, oral cultures, storytelling cultures
• Giving adults resources designed for children

The Resource Gap

• Study Bibles
• Bible dictionaries and commentaries
• Bible engagement resources
  • Group study guides
  • Individual study guides
• Training materials for pastors, church leaders, and teachers

The Resource Gap

• Reader’s guides to the Hebrew Old Testament and Greek New Testament
• Accessible Greek and Hebrew lexicons
• Grammars and discourse analyses

A Wealth of Biblical Data

A Wealth of Biblical Resources?

• Hebrew Reader’s Guide
• Discourse Flow
• Discourse Flow Review Tool
• Narrative and Cultural Commentary
• Exegetical Frameworks for Multiple Audiences

What AI Does Well

• Summarizes multiple large datasets (imperfectly!)
• Quick first drafts
  • For brainstorming
  • Initial versions of resources
• Rapid prototyping of new resource types
• Extracts data from a wide variety of files
• Vibe coding — writes pipeline code

Where AI Falls Short

• Hallucinates — plausible errors, hard to catch
• Drifts — pursues its own goals
• No self-evaluation — confident ≠ accurate
• Always confident. Always plausible. Sometimes true.

Expertise — It Still Matters

• The data is irreplaceable
  • Decades of annotation by domain experts
  • AI reasons from this data; it cannot replace it
• Experts can get it right when AI cannot
  • LLMs go off in dead ends with these datasets
  • They don’t understand the data; experts do
• The goal: datasets that AI and people can trust
  • Not AI-generated slop
  • Grounded, verified, scholar-reviewed

Building Communities of Knowledge

• Knowledge is tested in community
  • Peer review, mentoring, challenge
  • Test everything; hold on to what is good (1 Thess 5:21)
• Creating together
  • Essential for human flourishing
  • How truth is found
  • AI should support communities of knowledge, not replace them

Keeping the Real Goals in View

• Are they engaging the text, or just approving AI output?
• Are they working together, or each alone with AI?
• Do they own what they produce?
• Do they trust their own judgement?
• Are they becoming more capable, or more dependent?

Scripture Pipelines

• A declarative pipeline language for AI-assisted resource production
• Built for complex, heterogeneous data — the kind biblical scholarship runs on
• Grounded in Human at the Helm principles

Scripture Pipelines: How It Works

• Prompt contracts: each step declares its inputs and outputs
• JSON: shapes what the model focuses on and guarantees what it returns
• Intermediate artifacts: every step saved and inspectable
• Debug files: exactly what the model saw and said

The Design Principle

• The person is in charge
• AI follows direction, accepts correction
• Design documents, plans, audits, test suites — the person sets the agenda
• Human at the Helm: scholar commands, AI executes
• The pipeline architect must constantly watch out for AI drift

Who Does This Work?

• Someone who knows the data
• Someone who understands the need and the users
• A developer with good vibe coding skills
• Built with the communities who need them
• Mentoring people globally — in every role

Auditing

• AI synthesizes large amounts of data into an audit summary
• Human engages interactively — questions, pushback, clarification
• AI can make proposals; human decides when they agree
• Human decides what to do about it

You Can’t Hit What You Can’t See

• Review HTML: efficient interface for auditing large amounts of data
• Scholar marks correct / uncertain / incorrect — can also add notes
• Generates JSON for AI or programs to act on

Where We Are

• This is still very much alpha
• We are mentoring our very first teams — and learning how to do this
• We will ramp up slowly and deliberately
• We will interview new teams when we have capacity to mentor another team