Context that changes.
A record you can trace.
Axiom investigates how an AI system can revise a conclusion when circumstances change while retaining the observations that originally supported it.
Provisional application draft in preparationWhat should change—and what should stay?
An earlier conclusion can become outdated without making its original evidence disappear. Keeping the two distinct makes the change explainable.
The research brings together persistent context, source provenance, and revision across time. A provisional patent application draft is being prepared around this direction; its scope and technical disclosure remain under review.
Application status: in preparation, not filed. No patent-pending status or granted patent is claimed.
What you can inspect now.
The public deterministic example uses synthetic deployment records. Changed, confirming, and missing evidence lead to different treatments of the working conclusion, with the original source retained.
This is a focused context resolver. The broader memory implementation, migration account, and scripted model-swap illustration have their own scope on the Axiom page.
Try the three evidence scenarios ↗Early influences.
Independent questions.
My father’s feed-system patent and the people who introduced me to computing belong to the systems-lineage story. Axiom is my independent research, with its own development history and artifacts.
Read the systems lineage ↗Discuss the mechanism.
Bring a question about memory, revision, or the evidence a public example demonstrates. We can discuss the relevant implementation and the next research question.
Discuss Axiom ↗