Outside Access / Communication
Outside Access: give an unsuccessful request a clear ending
Refusal handling now distinguishes planning from an attempted generation and provides an explanation or suitable alternative instead of an endless retry loop.
Work covered: Sep 30, 2026 – Oct 1, 2026
A failure should not leave the person guessing
When an image request cannot be completed, the surrounding workflow still has work to do. The late-September changes made that outcome clearer, including handling for a request rejected during planning before any provider attempt occurred. That case is different from a generation attempt that was made and then refused.
Keep the request’s accounting aligned with its state
The repair addressed reservations associated with work that had not actually been attempted. Planning decisions could be reconsidered without labeling the request as completed generation. The important product behavior is consistency: a request’s visible status and its accounted work should describe the same event.
Explain the outcome rather than invite repetition
The final refusal policy provides a suitable alternative or an explanation when a request cannot proceed. It does not keep inviting the person to repeat the same unsuccessful request. Unresolved operational failures are surfaced to the operator so that a person is not left to diagnose a hidden problem through repeated submissions.
What the live checks support
The record includes activation and end-to-end verification of the revised failure path. It does not promise that a provider will accept every request or that an alternative will always satisfy the original intention. The milestone is a clearer, finite workflow: distinguish what was attempted, communicate the result and route a genuine operational problem for attention.
Explore the product
Based on dated development and release records from September 25–October 8, 2026. Availability and acceptance are stated for the recorded milestone.