Skip to content
Atomic Bible
Business reporting · research foundation

Make recurring reports worthy of the decisions they shape.

Atomic Bible is building an evidence-graded foundation for truthful, comprehensible, decision-relevant business reports—before deciding what should become software.

Current phase · methodology research · no generator · no certification · no finished report schema

The reporting problem01 / 04

Polish can hide weak evidence.

Recurring reports often mix governed facts, unstable definitions, causal guesses, conventions, and unreviewed AI prose.

  1. Readers cannot trace a metric or comparison frame.
  2. Routine variation becomes a confident story.
  3. Recommendations lose their uncertainty and warrant.
Working product model

Govern evidence. Compose candidates. Review the whole.

“Atom” is only a candidate term for the smallest independently testable report unit. Definitions, snapshots, signals, explanations, decisions, uncertainty, and provenance may compose into a report—but report-level quality cannot be assumed from its parts.

01

Evidence

Govern definitions, sources, transformations, and freshness.

02

Candidate units

Give each report job testable inputs, output, and evidence linkage.

03

Report

Compose around audience, task, cadence, argument, and accessibility.

04

Review

Approve with human judgment and learn from reader outcomes.

Worked preview

A report should expose what is known—and what is not.

This fictional operating review connects sample values to drivers, a decision prompt, and evidence states without implying a real customer or validated template.

Illustrative · Monthly operating review

Northstar Services · June 2026

Decision: capacity plan

Revenue$4.82m+3.2% vs plan
Gross margin41.6%−1.4pp vs plan
On-time delivery92%−3pp vs target
What changed

Revenue finished ahead of plan, while expedited contractor work compressed margin.

DriverImpactEvidence state
Enterprise renewals+$210kReconciled
Contractor overtime−0.9ppObserved
Project mix−0.5ppWorking inference
Evidence & method

No undifferentiated “best practices.”

Normative requirements, empirical findings, academic frameworks, institutional guidance, practitioner methods, exemplars, and house conventions remain distinct.

N / E1 / E2

Trace authority

Verify the source and preserve the scope, task, population, and measured outcome.

F / G / P / X / H

Label interpretation

Do not turn a useful model, method, example, or local choice into a universal rule.

Audit / Evaluate

Test utility

Use biai-os field reports and representative readers to measure comprehension and decisions.

Now / later

Current foundation

  • Initial canon and source verification register
  • Candidate principle registry with controversies
  • Illustrative report anatomy and candidate model
  • Planned dogfood audits and reader evaluation

Possible future—only after evidence

  • Accepted or conditional house methodology
  • Stable report concepts informed by evaluation
  • A documented decision on whether software is warranted

Not promised: a schema, generator, certification, package, or installable atoms.