This technical note addresses a fundamental challenge in autonomous computational systems: how to uniquely identify artifacts in ways that reflect their computational origins rather than arbitrary system assignments. The note proposes that artifact identity be derived deterministically from the computation that produced it and the artifacts used as inputs. Under this scheme, identical computations using identical inputs always produce identical artifact identities — making it impossible to have two different artifacts with the same identity, or two instances of the same computation with different identities. Traditional systems identify outputs arbitrarily using file paths, random strings, or database keys, making it impossible to verify computational lineage or validate artifact graphs. Deterministic identity resolves this by enabling verification of computational results through re-execution, artifact reuse across independent distributed systems, and the accumulation of computational work rather than its repetition. Multiple independent systems can contribute to a shared artifact graph, building upon prior computation rather than recreating it. This is Technical Note 06 of the Agent Artifact Availability (AAA) Framework series.
Rich Kopcho (Wed,) studied this question.