← ResearchMethod proposal

Implementation and validation pending

When can an agent reuse what it remembers?

Before reusing a remembered claim, an agent would check where it came from, whether it has changed, and whether it has permission to act on it or share it.

Agent memoryThe current record links back to its source and retains the superseded version.Current recordSuperseded versionSource
A current claim keeps a path back to its source.Proposed provenance model.
How the method works
  1. Remember

    Retain the claim and its source.

  2. Believe

    Check whether the claim still has adequate support.

  3. Act

    Check that this use has the right authority.

  4. Transfer

    Check whether the recipient may receive it.

Question

What should an agent check before using a remembered claim to make a decision or share information?

Method

The study distinguishes retaining a claim, believing it, having authority to act on it and having permission to transfer it. The proposed contract keeps each decision alongside source and revision information.

Worked scenario

Imagine a fictional customer address has been superseded. A withdrawal should stop the old address from being used for delivery while retaining enough revision history to explain the correction, where retention permissions allow it. A decision to share the corrected address still needs its own permission.

Limits and next check

The study proposed the design. Implementation and validation were outside its scope. A follow-up should test whether a withdrawal propagates to both retrieval and action after a source changes.

Source study · Durable memory research report, 14 October 2026

Related reading · W3C PROV-O. A standard for representing provenance.

Evidence and next check

Public evidence review is pending. Test the proposal with corrected, disputed and restricted records before making claims about reliability.

← Research
Converze v76