Skip to content

Evidence is not truth.

A practical way to keep important shared knowledge useful, traceable, and safe to change.

If your team repeatedly asks “what do we currently believe, why, and who can change it?”, you have the problem this architecture is designed to address.

Truth Machine helps keep a current account that people can responsibly act on when sources disagree, evidence arrives late, or a decision needs to stay understandable after the original conversation is gone.

It is not software to install, a database to adopt, or a claim that every team needs a formal system. Start smaller when a straightforward document, script, or application already gives you the confidence you need.

The model keeps five meanings separate:

  1. Evidence records what a source said or showed.
  2. Reconciliation compares that evidence with current truth under bounded source authority.
  3. A proposed change states the exact delta, rationale, evidence, uncertainty, actor, and expected state.
  4. Review and validation either advance the complete coherent transition or advance nothing.
  5. Current truth and change history move together, preserving why every consequential value stands.

Truth here means the best current account on which the people responsible for the work can act. It does not mean certainty. “Unknown” and “conflicting” can be honest, useful current states.

evidence -> reconciliation -> exact reviewed change -> current truth
-> publication or encounter -> contribution or observation -> evidence

The architecture is independent of storage engine, programming language, interface, AI model, and domain schema. Its conformance requirements govern meaning and transition boundaries.

About this documentation: the specification is a maintained reference. The Git-native records and publication packets are historical examples, not a process you need to adopt in order to use the ideas.