Start with the decision being reviewed
The right evidence depends on the claim. A single image may show a vehicle occupying a region, but a queue-cut decision can depend on an earlier queue, a passing movement, a directed insertion and continued travel. If the record only preserves the final image, a reviewer cannot reconstruct those prerequisites. A visually sharp screenshot can therefore be a poor explanation of a temporal decision.
LaneSplitter's saved maneuver path retains supporting source history through confirmation. Its clips and manifests are bounded by resource limits, and incomplete coverage must remain visible. A record should tell the reader which part of the event it contains. The absence of footage before or after a decision should not be disguised by looping the few frames that happen to exist. The visual treatment should honor the available timeline.
Keep identities with the record
A system evolves. The same input may be processed by a different model, calibration or rule version later. A record that omits those identities makes comparison difficult and can turn a retrospective explanation into guesswork. The local evidence path retains configuration, model and rule identities alongside its event. Those fields are not decorative technical metadata; they tell a reviewer which decision-making system produced the output.
The interface can translate this information without hiding it. A concise explanation can tell a nontechnical reader why configuration matters, while an expandable record exposes the exact identifier to a technical reviewer. Copy actions should copy the intended value and report completion clearly. They should not copy a truncated presentation string or replace the field with an animation that makes it difficult to compare two records.
Integrity has a specific meaning
An artifact hash allows a later comparison with the bytes that were recorded at verification time. If the bytes change, the hash changes. That is useful for identifying files and detecting accidental or intentional modification relative to a known value. It does not by itself prove who created the file, when it existed, whether its original content was true, or whether the collection process was lawful.
Calling a hash a cryptographic chain of custody would therefore say more than the mechanism establishes. A trustworthy product should name the exact guarantee and leave space for additional controls where the operating context requires them. Signatures, trusted timestamps, access history and organizational procedures solve different parts of that problem. A website should not collapse them into one reassuring shield icon.
Design the review around questions
A reviewer needs to move between an overview and the original material. What happened? Which observations support it? Where did continuity break? Which conditions were configured? What was recorded locally, and what was later delivered? Organizing the page around these questions produces a more useful interface than displaying every database field with equal weight. It also makes missing information easier to identify.
The saved replay should preserve source time when scrubbing, identify the sensor view and show actual measured overlays separately from explanatory graphics. The plain-language outcome must remain available if video cannot load. If a receipt is downloadable, its identity should match the replay on screen. A polished media component and a correct evidence binding are both necessary; neither compensates for the absence of the other.
A record can be useful and still incomplete
An evidence packet should include its limitations in the same place as its result. That is not an apology or a disclaimer bolted onto the footer. A known missing prefix, approximate road contact or synthetic source changes how the record can be interpreted. Hiding it forces the reviewer to reconstruct the boundary from scattered notes. Stating it directly helps the next engineering decision.
This approach applies well beyond roadside systems. Operational approvals, inventory corrections, incident handling and financial workflows all benefit from records that preserve the reason for a decision. The aim is to make future review possible without relying on the memory of whoever happened to be present. Good architecture leaves a system understandable after the moment has passed.
