From governed evidence to a reviewed recurring report.

What exists now

A source register, initial principle registry, research workflow, illustrative report families, and candidate building blocks. There is no finished schema, generator, certification, validated best-practice profile, or published/installable report atom.

Terminology

Canon

Reviewed body of sources used to derive the house methodology; inclusion is not blanket endorsement.

Principle

A scoped, source-backed proposition with conditions and an evaluation method.

Candidate atom

Working-model term for the smallest independently testable report unit.

Recurring report

A report family repeatedly issued under a stable semantic and presentational contract.

Human and AI responsibilities

Humans remain accountable for purpose, metric governance, domain judgment, consequential inference, approval, access, and decisions. AI may assist with retrieval, transformation, drafting, consistency checks, and surfaced questions only when its sources, uncertainty, and changes remain reviewable. It does not manufacture authority or replace approval.

Report lifecycle

1. Declare purpose, readers, task, cadence, and success criteria.
2. Govern definitions, evidence, comparison frames, and access.
3. Analyze and compose observations, inferences, qualifications, and prompts.
4. Review statistical integrity, argument, accessibility, and provenance.
5. Approve, distribute, record decisions, and preserve the issued version.
6. Measure reader outcomes and revise the next cycle transparently.

Quality dimensions

Truth and traceability; comprehension; statistical discipline; decision utility; uncertainty calibration; accessibility and plain language; semantic consistency; and operational trustworthiness.

Repository pointers

• PRODUCT.md — sole active product definition.
• docs/reporting/README.md — admission rules and workflow.
• docs/reporting/canon.md — source and verification register.
• docs/reporting/principles.md — candidates, boundaries, controversies.

Browse the public method, worked examples, or candidate building blocks.

Roadmap and current status

1. Complete source-level extraction and expand missing domains.
2. Audit representative biai-os report families.
3. Establish reader-centered baselines and run realistic evaluations.
4. Revise principles while preserving contrary evidence.
5. Only then decide whether stable concepts warrant software.

Non-goals

Atomic Bible is not currently a BI platform, chart gallery, template marketplace, generic AI reporting tool, accounting or accessibility certifier, report generator, SDK, package, hosted service, or continuation of the prior installable AI-systems library.