CHANGE REHEARSAL

Software changes can introduce behavioral regressions even when existing tests pass. Developers often have to manually identify affected behavior, reproduce scenarios, compare the previous and changed versions, and collect evidence before deciding whether a change is safe to merge. This process can be time-consuming and makes behavioral regressions difficult to identify early. CHANGE REHEARSAL addresses this problem by providing a rehearsal workflow for software changes before merge. A developer submits an intended change, such as “Add caching to Product API to improve response time.” The system turns that intent into executable journeys and identifies behavior that should be protected. The rehearsal then compares baseline and candidate behavior. Differences are presented through a Behavioral Diff, allowing developers to see which journeys or behaviors changed. Affected journeys can be replayed to show exactly where the candidate diverged from the baseline. An Evidence Capsule packages the relevant reproduction information so the developer has concrete evidence when evaluating the change. The developer then reaches a decision through Developer Actions, such as investigating a potential regression, accepting an intentional change, or requesting a proposed fix. Generate Fix provides a proposal rather than silently modifying the source code. The project is designed around the principle: “AI proposes; deterministic execution verifies.” AI-assisted analysis can help identify intent, behaviors, and scenarios, while deterministic execution and comparison provide evidence of what actually happened. The target users are software developers and engineering teams who need to validate changes before merging them. Instead of relying only on whether tests pass, CHANGE REHEARSAL focuses on whether important existing behavior remains consistent. The result is a structured workflow from Change Submission through Rehearsal Progress, Behavioral Diff, Journey Replay, Evidence Capsule, and Developer Actions, giving developers a clearer way to evaluate behavioral impact before a change reaches the main branch.