Experiment output

Agent reconstruction scores

Every score will connect back to a runnable generated clone, screenshot differences, interaction replay, terminal journey checks, and resource use. The dataset is ready; experiments have not started.

0 completed runs

Visual fidelity—

Screenshot pairs · diff maps

Interaction fidelity—

Actions · states · replay

Journey completion—

Success · failure · recovery

Robustness—

Hidden cases · isolation

Efficiency—

Time · tokens · resources

Website × agent matrix

One dataset row, no experiment columns yet

Results remain intentionally blank

Offline reference	Agent / model	Generated clone	Visual	Interaction	Journeys	Robustness	Efficiency	Total
Amazon ShoppingShopping Commerce · dataset ready	Experiment not startedThe first real Agent report will create this row.

See the visual difference

Side-by-side, drag split, overlay, blink, and heatmap views preserve the actual screenshot pair behind visual fidelity.

ReferenceGenerated clone

Replay the interaction

Actions, routes, before/after states, expected outcomes, and failure points explain interaction and journey scores.

Audit the run

Network isolation, runtime health, time, tokens, retries, and hard failures remain visible beside the total score.

Runtime: —
Tokens: —
Failures: —