Controlled editing system
Let the model propose an edit. Make the system decide whether it can exist.
Allow intelligent editing without permitting a model to write fragile timeline state directly.
A typed, validated boundary between a model proposal and professional timeline mutation — with uncertainty preserved when retrieval could miss useful footage.
A plausible instruction can contain an invalid source, incomplete range, inconsistent time basis, or unsupported destination. Direct model-written files would turn ambiguity into fragile state.
A failed search is also not proof useful footage does not exist. The system had to preserve uncertainty.
Each proposal entered an envelope containing action, source, in, out, and destination. The system validated references and resolved time in frames before committing one timeline state.
The gate had only two outcomes: apply fully, or refuse with a reason.
XMEML, FCPXML, and OTIO were adapters around one frame-based timeline rather than independent editing systems. Each had to resolve the same cut to the same frame boundary.
Search widened through declared boundaries. Output separated status from diagnosis so no result, uncertain retrieval, and a stronger absence signal remained distinct.
A 150-probe field disclosed one dangerous miss. It is not a generalized 149-of-150 accuracy claim: the other probes do not establish that every unreturned item is absent.
Source identity, transformations, model/configuration choices, and index state form a reproducibility boundary.
The client is described only as A creative production studio. This record is limited to the disclosed architecture and evaluation.
Dependable agent action begins where model freedom ends: at the typed operation, the deterministic boundary, the independent check, and the honest statement of uncertainty.